Southern Cross University

ePublications@SCU
Theses

2006

Factors influencing unit trust performance

Cheong Sing Tng
Southern Cross University

Publication details
Tng, CS 2006, 'Factors influencing unit trust performance', DBA thesis, Southern Cross University, Lismore, NSW. Copyright CS Tng 2006

ePublications@SCU is an electronic repository administered by Southern Cross University Library. Its goal is to capture and preserve the intellectual output of Southern Cross University authors and researchers, and to increase visibility and impact through open access to researchers around the world. For further information please contact epubs@scu.edu.au.

FACTORS INFLUENCING UNIT TRUST PERFORMANCE

Tng Cheong Sing B.Sc. (Hons) Lond, MFM CQU A thesis submitted in partial fulfilment of requirements for the Degree of

Doctor of Business Administration

Southern Cross University

Graduate College of Management

July 2006

Statement of original authorship

This thesis has been prepared in accordance with the rules set out for the Degree of Doctor of Business Administration at Southern Cross University. The work presented in this thesis is, to the best of my knowledge and belief, original, except as acknowledged in the text. The material has not been submitted, either in whole or in part, for a degree at this or any other university.

Tng Cheong Sing July 2006

ii

Acknowledgements

Special thanks to Dr. Geoffrey Meredith, AM (Emeritus Professor, Southern Cross University) for his professional supervision of this research. Dr. Julia Sawicki (Assistant Professor, Nanyang Technological University), Dr. Koh Seng Khee (Associate Professor, Singapore Management University) and Dr. Teo Cheng Swee (Adjunct Professor, Southern Cross University) provided helpful comments on initial drafts of this thesis. Mr Andrew Kwek (executive director, Investment Management Association of Singapore) and Ms Teo Jing Ling (administrative officer, Central Provident Fund Board) answered queries about unit trust data. For helpful comments and suggestions, I gratefully acknowledge Southern Cross Universitys DBA workshop participants and University of Tasmanias research seminar participants, as well as participants at the Academy of International Business Southeast Asia Regional Conference, Australasian Finance and Banking Conference, International Business Research Conference and Waikato Management School Student Research Conference. Mercer Investment Consulting and S&P Fund Services Asia provided secondary data. Data entry assistants were funded by the School of Accounting and Finance, University of Tasmania.

iii

Abstract

Bank-managed equity funds are not inferior to their non-bank counterparts. Previous research reporting relative underperformance of bank-managed funds ignored their differing fiduciary standards. To evaluate bank and non-bank funds facing similar fiduciary responsibilities, domestic retail funds approved for Singapores Central Provident Fund Investment Scheme were examined, as they meet the same standard for managing social security savings. Returns from these funds correlate highly with market performance. Even though these fund returns exceeded guaranteed interest rates, they did not outperform their market index. With financial market deregulation in Southeast Asia, local banks in small economies withstand erosion of business by foreign competitors. Banks, in order to increase profits, compete with local as well as foreign insurance and investment companies by offering mutual fund products. To remain competitive, banks need to shed their reputation for not being able to generate impressive fund returns, as their funds are not inferior to those from insurance and investment companies in terms of assets under management, expenditures, returns and risk. To gain competitive advantage, banks can differentiate their fund characteristics and reduce portfolio management costs. Mutual fund characteristics can affect expected returns or transaction costs. Factors affecting expected returns include asset allocation and systematic risk, while transaction costs include explicit and implicit ones, which can be measured by expense ratios and size of funds respectively. Insignificance of transaction cost determinants in affecting actual returns can be attributable to dominance of factors affecting expected returns.

iv

Keywords

Assets under management Bank-managed fund Expense ratio Financial institution Fund performance evaluation Mutual fund management Unit trust

Presentations and publications

Tng, C 2005, 'Sustainable development of bank-managed mutual funds: evidence from Singapore's retail funds', paper presented at 9th Waikato Management School Student Research Conference, University of Waikato, 25 October (Appendix B, pp 194-206). Tng, C 2005, 'Performance of approved equity funds: evidence from Singapore's retail funds', paper presented at 18th Australasian Finance and Banking Conference, Sydney, 14-16 December (Appendix C, pp. 207-221). Drafts presented at Graduate College of Management DBA workshop, Southern Cross University (13-15 May) and School of Accounting and Finance Research Seminar, University of Tasmania (19 August). Tng, C 2007, Effects of expenditures and size on mutual fund performance, Singapore Management Review/Asia-Pacific Journal of Management Theory and Practice, vol. 29, no. 1 (Appendix D, pp. 222-232). Paper presented at School of Business, Monash University Malaysia campus, 10 July 2006. Draft presented at 2nd International Business Research Conference, University of Technology Sydney, 58 December 2005. Tng, C 2007, Bank fund management challenges and opportunities being reviewed by Singapore Management Review/Asia-Pacific Journal of Management Theory and Practice (Appendix E, pp. 233-243). Draft presented at Academy of International Business Southeast Asia Regional Conference, Manila, 24-26 November 2005.

vi

Table of contents
Statement of original authorship ........................................................................................................ii Acknowledgements ...........................................................................................................................iii Abstract .............................................................................................................................................iv Keywords ........................................................................................................................................... v Presentations and publications ..........................................................................................................vi Table of contents ..............................................................................................................................vii List of tables....................................................................................................................................... x List of figures ....................................................................................................................................xi List of equations...............................................................................................................................xii Abbreviations ..................................................................................................................................xiii Glossary...........................................................................................................................................xiv Chapter 1 Research overview............................................................................................................. 1 1.1 Introduction.............................................................................................................................. 1 1.2 Background to the research...................................................................................................... 3 1.2.1 Nature of mutual funds ..................................................................................................... 3 1.2.2 Challenges facing mutual funds ....................................................................................... 4 1.3 Research problem and questions.............................................................................................. 5 1.4 Definitions ............................................................................................................................... 7 1.5 Methodology............................................................................................................................ 8 1.5.1 Data collection.................................................................................................................. 8 1.5.2 Statistical analysis............................................................................................................. 9 1.5.3 Model validation............................................................................................................... 9 1.6 Assumptions and delimitations .............................................................................................. 10 1.6.1 Assumptions ................................................................................................................... 10 1.6.2 Delimitations .................................................................................................................. 13 1.7 Thesis plan ............................................................................................................................. 14 1.8 Conclusion ............................................................................................................................. 14 Chapter 2 Background and research justification............................................................................. 16 2.1 Introduction............................................................................................................................ 16 2.2 Background to the research problem ..................................................................................... 17 2.2.1 Singapores demographic characteristics ....................................................................... 17 2.2.2 Demography of Singapores asset management industry............................................... 18 2.2.3 Equity, bond, balanced and money market funds........................................................... 19 2.2.4 CPF-approved and non-CPF-approved unit trusts.......................................................... 21 2.2.5 Risk classification of unit trusts...................................................................................... 22 2.2.6 Fee structure of unit trusts .............................................................................................. 24 2.2.7 Regulation of unit trusts ................................................................................................. 25 2.3 Performance of unit trusts in Singapore................................................................................. 27 2.3.1 Fund performance from 1976 to 1994............................................................................ 28 2.3.2 Fund performance from 1999 to 2003............................................................................ 30 2.3.3 Performance of CPF-approved unit trust investors ........................................................ 31 2.4 Contribution of the research................................................................................................... 32 2.4.1 Mutual fund performance model .................................................................................... 32 2.4.2 Conference presentations and article publications ......................................................... 32 2.4.3 Research outcomes for finance teachers and researchers ............................................... 32

vii

2.5 Justification for the research .................................................................................................. 33 2.5.1 Gaps in the mutual fund literature .................................................................................. 33 2.5.2 Research approach.......................................................................................................... 34 2.5.3 Benefits for Singapores unit trust industry.................................................................... 34 2.6 Conclusion ............................................................................................................................. 34 Chapter 3 Literature review.............................................................................................................. 36 3.1 Introduction............................................................................................................................ 36 3.2 General theories relating to mutual fund performance .......................................................... 37 3.2.1 Efficient market theory................................................................................................... 37 3.2.2 Mutual fund performance ............................................................................................... 38 3.2.3 Measurement of mutual fund performance..................................................................... 41 3.2.4 Asset pricing theories ..................................................................................................... 45 3.3 Determinants of mutual fund performance ............................................................................ 48 3.3.1 Asset allocation............................................................................................................... 50 3.3.2 Investment style.............................................................................................................. 51 3.3.3 Risk................................................................................................................................. 52 3.3.4 Past performance and performance persistence.............................................................. 52 3.3.5 Flow of funds and assets under management ................................................................. 54 3.3.6 Research and trading costs ............................................................................................. 56 3.3.7 Type of fund management company .............................................................................. 57 3.4 Theoretical framework........................................................................................................... 59 3.4.1 Mutual fund performance models................................................................................... 60 3.4.2 Research issues and propositions.................................................................................... 63 3.5 Conclusion ............................................................................................................................. 67 Chapter 4 Research methodology .................................................................................................... 70 4.1 Introduction............................................................................................................................ 70 4.2 Research approaches.............................................................................................................. 71 4.2.1 Case research .................................................................................................................. 72 4.2.2 Survey research............................................................................................................... 75 4.2.3 Secondary data research ................................................................................................. 77 4.2.4 Justification for research using secondary data .............................................................. 79 4.3 Research design ..................................................................................................................... 81 4.3.1 Hypothesis testing........................................................................................................... 81 4.3.2 Non-causal investigation ................................................................................................ 81 4.3.3 Minimal researcher interference..................................................................................... 82 4.3.4 Non-contrived setting ..................................................................................................... 82 4.3.5 Fund management institutional group as unit of analysis .............................................. 82 4.3.6 Longitudinal time horizon .............................................................................................. 82 4.4 Data collection ....................................................................................................................... 83 4.4.1 Data collection methods ................................................................................................. 83 4.4.2 Downloading of financial data from online sources....................................................... 83 4.5 Data analysis .......................................................................................................................... 86 4.5.1 Regression analysis......................................................................................................... 86 4.5.2 Hypothesis testing........................................................................................................... 87 4.6 Research quality..................................................................................................................... 88 4.6.1 Internal validity............................................................................................................... 89 4.6.2 External validity ............................................................................................................. 92 4.7 Ethical considerations ............................................................................................................ 93 4.7.1 General ethical issues ..................................................................................................... 93 4.7.2 Specific ethical issues..................................................................................................... 94 4.7.3 Independent academic research to avoid ethical lapses.................................................. 94 4.8 Conclusion ............................................................................................................................. 95

viii

Chapter 5 Data analysis.................................................................................................................... 96 5.1 Introduction............................................................................................................................ 96 5.2 Characteristics of domestic equity funds ............................................................................... 97 5.2.1 Fund management company........................................................................................... 97 5.2.2 Returns............................................................................................................................ 98 5.2.3 Beta............................................................................................................................... 100 5.2.4 Expense ratio ................................................................................................................ 101 5.2.5 Fund size....................................................................................................................... 103 5.3 Regression analysis.............................................................................................................. 104 5.4 Hypothesis testing................................................................................................................ 105 5.4.1 Performance of domestic equity funds ......................................................................... 105 5.4.2 Performance comparison of bank and non-bank funds ................................................ 110 5.4.3 Expenditures, size and performance of domestic equity funds .................................... 115 5.4.4 Comparison of bank and non-bank fund characteristics............................................... 117 5.5 Conclusion ........................................................................................................................... 120 Chapter 6 Conclusions ................................................................................................................... 122 6.1 Introduction.......................................................................................................................... 122 6.2 Conclusions about the research questions............................................................................ 124 6.2.1 Overall performance of domestic equity funds ............................................................ 124 6.2.2 Performance of bank and non-bank domestic equity funds.......................................... 125 6.2.3 Factors affecting performance of domestic equity funds.............................................. 126 6.2.4 Factors differentiating bank and non-bank fund performance...................................... 127 6.3 Conclusions about the research problem ............................................................................. 128 6.4 Implications for finance theory............................................................................................ 131 6.4.1 Performance of domestic equity funds ......................................................................... 131 6.4.2 Performance comparison of bank and non-bank funds ................................................ 132 6.4.3 Expense ratio, size and fund performance.................................................................... 133 6.4.4 Comparison of bank and non-bank fund characteristics............................................... 134 6.5 Implications for policy and practice..................................................................................... 135 6.5.1 Implications for financial institutions........................................................................... 135 6.5.2 Implications for individual investors ............................................................................ 136 6.5.3 Implications for government policies ........................................................................... 137 6.6 Research limitations............................................................................................................. 137 6.7 Implications for further research.......................................................................................... 138 6.8 Implications for research methodology................................................................................ 138 References ...................................................................................................................................... 140 Appendices..................................................................................................................................... 146 Appendix A Data, computation and regression outputs................................................................. 147 Appendix B Conference paper 1: Sustainable bank-managed mutual funds ................................. 194 Appendix C Conference paper 2: Performance of approved equity funds..................................... 207 Appendix D Journal article 1: Effects of expenditures and size on mutual fund performance ...... 222 Appendix E Journal article 2: Bank fund management challenges and opportunities ................... 233

ix

List of tables
Table 1.1 Summary of model validation techniques........................................................................ 10 Table 1.2 Thesis chapter outline ...................................................................................................... 14 Table 2.1 Growth of Singapore's unit trust industry ........................................................................ 19 Table 2.2 Contribution of various types of unit trusts...................................................................... 20 Table 2.3 Holding period returns of unit trusts in Singapore from 1976 to 1994 ............................ 28 Table 2.4 Distribution of profit and loss for CPF investors ............................................................. 31 Table 3.1 Selection of mutual fund performance determinants ....................................................... 59 Table 3.2 Measurement of fund characteristics................................................................................ 60 Table 3.3 Research issues and hypothesis........................................................................................ 68 Table 4.1 Summary of plausible approaches for financial research................................................. 79 Table 4.2 Research sample............................................................................................................... 84 Table 5.1 Characteristics of domestic equity funds from banks and non-banks .............................. 97 Table 5.2 Fund performance rankings for 1999-2002 and 2003-2004............................................. 98 Table 5.3 Fund beta rankings for 1999-2002 and 2003-2004 ........................................................ 101 Table 5.4 Fund expense ratio rankings for 1999-2002 and 2003-2004.......................................... 102 Table 5.5 Fund size rankings for 1999-2002 and 2003-2004......................................................... 103 Table 5.6 Regression of equity fund and market index risk premiums.......................................... 106 Table 5.7 Information ratios for domestic equity funds................................................................. 107 Table 5.8 Sharpe and Treynor ratios for domestic equity funds .................................................... 109 Table 5.9 Performance measures of bank and non-bank domestic equity funds ........................... 110 Table 5.10 Two-sample t-test for bank and non-bank fund returns ............................................... 111 Table 5.11 Two-sample t-test for information ratios of bank and non-bank funds........................ 112 Table 5.12 Two-sample t-test for Jensen alphas of bank and non-bank funds............................... 113 Table 5.13 Two-sample t-test for Sharpe ratios of bank and non-bank funds................................ 114 Table 5.14 Two-sample t-test for Treynor ratios of bank and non-bank funds.............................. 114 Table 5.15 Two-sample t-test for returns of big and small funds .................................................. 115 Table 5.16 Two-sample t-test for returns of high and low expense ratio funds ............................. 116 Table 5.17 Two-sample t-test for expense ratios of big and small funds....................................... 117 Table 5.18 Two-sample t-test for size of bank and non-bank funds .............................................. 118 Table 5.19 Two-sample t-test for beta of bank and non-bank funds.............................................. 119 Table 5.20 Two-sample t-test for expense ratio of bank and non-bank funds ............................... 120 Table 5.21 Results of hypothesis testing........................................................................................ 121 Table 6.1 Performance comparison of bank and non-bank domestic equity funds........................ 126 Table 6.2 Relation between fund expense ratio, size and performance ......................................... 127 Table 6.3 Comparison of bank and non-bank fund characteristics ................................................ 128 Table 6.4 Literature contribution on fund performance ................................................................. 132 Table 6.5 Literature contribution on performance of bank and non-bank funds............................ 133 Table 6.6 Literature contribution on fund characteristics and performance................................... 134

List of figures

Figure 1.1 Overview map................................................................................................................... 2 Figure 2.1 Background map ............................................................................................................. 16 Figure 2.2 Singapore's geographical location................................................................................... 18 Figure 2.3 Growth of CPF-approved and non-CPF-approved unit trusts ........................................ 21 Figure 2.4 Risk classification of CPF-approved unit trusts.............................................................. 22 Figure 3.1 Literature map................................................................................................................. 36 Figure 3.2 Weak, semi-strong and strong forms of market efficiency............................................. 38 Figure 3.3 Derivation of research issue 1 on unit trust performance ............................................... 40 Figure 3.4 Derivation of research issue 2 on relative performance of FMCs .................................. 45 Figure 3.5 Derivation of research issue 3 on mutual fund performance determinants..................... 48 Figure 3.6 Derivation of research issue 4 on bank and non-bank fund characteristics .................... 58 Figure 3.7 Conceptual model of mutual fund performance ............................................................. 61 Figure 4.1 Methodology map ........................................................................................................... 70 Figure 4.2 Fund performance model with variables and hypotheses labelled.................................. 85 Figure 4.3 Framework for assessing statistical studies .................................................................... 88 Figure 5.1 Data analysis map ........................................................................................................... 96 Figure 5.2 Daily Straits Times Index from 1987 to 2005 ................................................................ 99 Figure 5.3 Quarterly STI returns from 1988 to 2004 ..................................................................... 100 Figure 5.4 Time series regression of a domestic equity fund......................................................... 104 Figure 5.5 Normal probability plot of a domestic equity fund....................................................... 105 Figure 6.1 Conclusions map........................................................................................................... 123 Figure 6.2 Conceptual model of mutual fund performance determinants...................................... 129

xi

List of equations

Equation 2.1 Calculation of breakeven return for a mutual fund ..................................................... 24 Equation 3.1 Information ratio for Singapores domestic equity funds ........................................... 41 Equation 3.2 Jensen alpha for Singapores domestic equity funds .................................................. 42 Equation 3.3 Sharpe ratio for Singapores domestic equity funds ................................................... 42 Equation 3.4 Treynor ratio for Singapores domestic equity funds ................................................. 43 Equation 3.5 Single-index model for domestic equity fund returns................................................. 62 Equation 4.1 Linear regression model for domestic equity fund returns ......................................... 86 Equation 4.2 t-statistic for comparing a characteristic for two groups of funds .............................. 87

xii

Abbreviations
AIMR APT CAPM CML CPF DBS DJIA EMH EDB ETF FMC FTA HPR ICI ILP IMAS MAS MOM MSCI NAV NYSE OCBC OUB P/B P/E S&P SML STI UOB USA Association for Investment Management and Research Arbitrage pricing theory Capital asset pricing model Capital market line Central Provident Fund Development Bank of Singapore Dow Jones Industrial Average Efficient market hypothesis Economic Development Board Exchange-traded fund Fund management company Free Trade Agreement Holding period return Investment Company Institute Investment-linked insurance product Investment Management Association of Singapore Monetary Authority of Singapore Ministry of Manpower Morgan Stanley Capital International Net asset value New York Stock Exchange Overseas Chinese Banking Corporation Overseas Union Bank Price-to-book Price-to-earnings Standard and Poors Security market line Straits Times index United Overseas Bank United States of America

xiii

Glossary

Balanced fund: Mutual fund investing in a combination of bonds and stocks. Bond fund: Mutual fund investing mainly in government and corporate bonds. Equity fund: Mutual fund investing generally in common stocks. Financial institution: Intermediaries (such as banks, insurance companies and mutual funds) that borrow funds from lenders and make loans to borrowers. Index fund: Passively managed mutual fund designed to mimic an index of securities. Investment company: Corporation, trust or partnership investing pooled shareholder dollars in securities appropriate to the organizations objective. Money market fund: Mutual fund investing in high-quality short-term securities Mutual fund: Investment company purchasing a portfolio of securities chosen by a professional investment adviser to meet a specific financial goal for investors buying shares from the company. Net asset value (NAV): Total market value of assets divided by number of shares outstanding. Risk: Uncertainty associated with an assets return. Unit investment trust: Investment company buying and holding fixed number of shares until a termination date. Unit trust: Financial institution (similar to mutual fund) inviting the public to subscribe in funds invested by the company in assets specified by its trust deed.

xiv

Chapter 1 Research overview

1.1 Introduction
This study concerns an area of interest to investment analysts and fund managersfactors influencing mutual fund performance differences, based on type of financial institution managing the fund. The purpose of this study is to examine relationships between unit trust performance and these factors in Singapores fund management industry: (1) risk, (2) transaction costs, (3) fund size and (4) type of fund management company. Objectives for conducting this research are to: 1. clarify conflicting results from existing literature comparing performance of funds managed by different types of financial institutions; 2. determine whether factors influencing performance of funds managed by banks differ from their non-bank counterparts; and 3. establish a conceptual model relating characteristics of funds to their performance. Existing literature on performance comparison of mutual funds managed by different types of financial institutions reported conflicting results. Until the 1990s, finance literature reported bank-managed funds being inferior to their non-bank counterparts (Bauman & Miller 1995; Bogle & Twardowski 1980), but a later study showed bank-managed bond funds were not inferior to their counterparts from other institutions (Frye 2001). However, Frye did not research how equity funds managed by these institution groups compare. Literature on fund performance differences focused on fund manager characteristics, including age, education (Chevalier & Ellison 1999) and gender (Atkinson, Baird & Frye 2003). This research proposes an empirical study of how factors influencing fund performance differ with type of fund management institutionbanks as well as insurance and investment companies. Studying historical performance of unit trusts invested using

Chapter 1 Singapores Central Provident Fund (CPF) reveals effects of such factors on funds managed by banks and non-banks.1 This chapter provides a preview of research documented in this thesis on an empirical study of factors influencing unit trust performance in Singapore. Figure 1.1 presents a map outlining this overview. Figure 1.1 Overview map below

1 Research overview 1.1 Introduction 1.2 Background to the research 1.3 Research problem and questions 1.4 Definitions 1.5 Methodology 1.5.1 Data collection 1.5.2 Statistical analysis 1.5.3 Model validation 1.6.1 Assumptions 1.6.2 Delimitations 1.2.1 Nature of mutual funds 1.2.2 Challenges facing mutual funds

1.6 Assumptions and delimitations 1.7 Thesis plan 1.8 Conclusion

Source: developed for this research. In this chapter, following section 1.2s background information on nature of mutual funds and challenges facing them, section 1.3 presents research problem and questions concerning fund performance comparison and determination of factors affecting fund performance, before section 1.4 complements these questions with definitions of terms applied in this research. Section 1.5 justifies methodology used while assumptions and delimitations placed on this research are explained in section 1.6. This is followed by section 1.7s outline of thesis chapters before section 1.8 concludes the chapter.
1

CPF is a public defined-contribution pension plan for employees in Singapore, who decide how their CPF accounts are invested in CPF-approved securities, as explained in the following chapter on research background.

Research overview The next section presents background information on mutual funds in a global context.

1.2 Background to the research

Information concerning nature and management of mutual funds are presented before discussing challenges to mutual funds from alternative investment products.

1.2.1 Nature of mutual funds

A mutual fund is an investment vehicle that pools capital from clients purchasing its shares to invest in a portfolio of securities, with purchasing and selling of securities being decided by a fund manager (Reilly & Brown 2003, p. G11). Three parties are involved in a mutual fund: board of directors, a fund management company (FMC) and shareholders. Independent from the FMC, the board of directors is responsible for safeguarding interests of client shareholders by ensuring the FMC complies with contractual regulations regarding duties and compensation. Duties of an FMC include investment research, portfolio management and issuing dividends. As for compensation, management fee is stated as a percentage of total fund value. Shareholders are investors seeking dividend income and capital gains from shares of the fund.2 By purchasing new issues of shares at launch price or secondary issues at asking price and subsequently selling at a higher price, shareholders can earn a capital gain. Alternatively, shareholders may incur capital loss with selling price lower than purchase price (Koh 1999). The FMC appoints a fund manager to generate returns for shareholders while satisfying the funds investment objectives. To achieve economies of scale and appeal to investors with different risk-return preferences, each FMC manages a family of funds with different characteristics, promoting flexibility by letting shareholders switch funds in response to

Specifically, investors return for fund f RETf = (EPf BPf + DIVf) / BPf where BPf, EPf and DIVf are the beginning price, ending price and dividend paid for fund f during the investment horizon.

Chapter 1 different financial conditions (Reilly & Brown 2003, p. 1074). However, mutual funds face competition from alternative investment products.

1.2.2 Challenges facing mutual funds

Mutual funds face competition from various alternatives, as fund performance was generally not impressive. Studies done by Jensen (1968), Malkiel (1995) and Sharpe (1966) reported most funds did not match performance of comparable market indexes. According to these studies, slightly more than 50 percent of mutual funds outperformed their targeted markets before considering transaction costs. After considering such costs, more than 60 percent of funds did not match their market performance, with the remainder performing inconsistently. Transaction costs are therefore considered in Chapter 3s theoretical framework for fund performance determinants. Unimpressive mutual funds are facing competition from newer alternatives, including exchange-traded funds (ETFs), folios and separately managed accounts. These alternatives offer certain advantages over mutual funds. For example, ETFs combine features of index funds with low expenses for trading stocks; folios let investors customize diversified stock portfolios; while separately managed accounts give investors access to professional managers who choose stocks for multiple accounts to achieve diversification (Jones 2003, pp. 144-5). Mutual fund information presented in this section is generally applicable to developed financial markets around the world. The following section presents research problem and questions while the next chapter presents additional information in Singapores context. This research derives from the mutual fund performance parent discipline and contributes to the immediate discipline of fund performance determinants, which are reviewed in Chapter 3.

Research overview

1.3 Research problem and questions

Background information from the previous section introduces the research problem: Factors influencing mutual fund performance differences, based on type of financial institution managing the funds. Factors differentiating performance of funds can be their observable characteristics, among them size, risk and transaction costs, while financial institutions managing funds are classified as banks and non-banks. This research problem leads to questions concerning fund performance comparisons and determination of factors differentiating fund performance. Research question 1: How do Singapores unit trusts vary in terms of performance? Comparing capital markets in Europe and the United States of America (USA) with those in the Asia Pacific region, as Asia Pacific markets were less efficient with more profitable opportunities, a higher proportion of funds in Singapore can outperform their relevant indexes, compared to more developed European and USA markets (Wong 2004). Unit trusts approved for the CPF Investment Scheme are examined. CPF account holders who choose not to participate in the scheme earn guaranteed minimum interest rates of 2.5 and 4.0 percent for Ordinary and Special accounts respectively (CPF Investment Scheme 2005).3 Rational CPF members may leave their contributions in these accounts to earn guaranteed interest rates if returns from CPF-approved unit trusts cannot outperform guaranteed interest rates. Information gathered from government surveys and press releases provide some answers to this question in the next chapter. Research question 2: How does performance of funds managed by banks compare with their non-bank counterparts?

Information about CPF Ordinary and Special accounts are presented in the next chapter.

Chapter 1 Frye (2001) reported bank-managed bond funds were not inferior to their counterparts from other institutions facing similar fiduciary standards, but she did not research how equity funds managed by these institutions compare. Examining CPF-approved funds provides performance comparison of equity funds managed by various institution groups following similar fiduciary responsibilities. Chapter 3s literature review presents conflicting results from studies on performance of funds managed by different types of financial institutions. Research question 3: What are important characteristics of funds affecting their performance? Asset allocation, expenses, risk and size are some important characteristics of funds affecting their performance (Peterson et al. 2002), as funds positive excess returns are intuitively associated with low-cost investments in equities with high level of systematic risk. To confirm these factors, a mutual fund performance model is developed in Chapter 3s theoretical framework. Research question 4: How do differences in fund characteristics account for performance differences among funds managed by various types of financial institutions? As portfolio managers in banks had a reputation for risk-averse investment strategies (McTague 1994), their conservative investment style may incur less transaction costs than their non-bank counterparts. Besides, bank funds inferior performance (Bauman & Miller 1995; Bogle & Twardowski 1980) may result in less popular and smaller funds than their non-bank counterparts. Chapter 3 develops a mutual fund performance model to differentiate characteristics affecting fund performance for various institution groups. To clarify terms used in research questions as well as remaining chapters, the following section explains key terms used for this research.

Research overview

1.4 Definitions
As this thesis uses various financial terms whose definitions are often not uniform among researchers, key and controversial terms are defined in this section to clarify research: balanced fund, bond fund, equity fund, financial institution, index fund, investment company, money market fund, mutual fund, net asset value (NAV), risk, unit investment trust and unit trust.4 In any countrys financial system, there are various types of financial institutions, including banks as well as insurance and investment companies, which borrow funds from lenders and make loans to borrowers (Mishkin & Eakins 2003, p. 8). According to the Investment Company Institute (ICI), the national association of investment companies in the USA, an investment company is a corporation, trust or partnership that invests pooled shareholder dollars in securities appropriate to the organizations objective (ICI 2004). Among investment companies, mutual funds purchase portfolios of securities chosen by professional investment advisers to meet specific financial goals for investors buying shares from these companies (ICI 2004). In Australia and Singapore, instead of mutual fund, the alternative name unit trust is used.5 Unit trusts are financial institutions that invite the public to subscribe in funds invested by the company in assets specified by its trust deed (McGrath & Viney 1998, p. 29). Unit trusts in Singapore should not be confused with unit investment trusts in the USA, which are investment companies that buy and hold fixed number of shares until a termination date (ICI 2004). This thesis uses the terms unit trusts and mutual funds synonymously to refer to funds which are managed by banks as well as insurance and investment firms. Common types of funds managed by these institutions include bond funds that invest generally in long-term

4 5

The Glossary on page xiv provides complete listing of definitions. Even though unit trusts are operationally similar to mutual funds, they are legally different, as explained in the following chapter.

Chapter 1 government and corporate bonds, equity funds investing generally in common stocks, balanced funds investing in a combination of bonds and stocks, index funds investing in securities that make up a market index and money market funds that invest in short-term securities (Reilly & Brown 2003, pp. 86-7). Such funds are risky assets as they have uncertain future returns. A funds market value can be measured by computing its pershare value, or NAV, which is calculated by dividing total market value of assets in the fund by its number of shares outstanding (Reilly & Brown 2003, p. 1074).

1.5 Methodology
Following clarification of financial terms, this section outlines procedures for secondary data collection, data analysis and model validation. These procedures and their results are elaborated in Chapters 4 and 5 respectively.

1.5.1 Data collection

For data collection, quarterly returns, fund sizes and expense ratios for CPF-approved domestic equity funds managed by banks, insurance companies and investment firms, as well as levels of the local stock market Straits Times index (STI) from 1999 to 2004 were downloaded from the CPF Board and Yahoo Finance websites to perform a five-year quantitative analysis.6 From 1999 to 2004, quarterly data were recorded for the following variables: 1. EXP: expense ratio recorded as a percentage; 2. FMC: type of financial institution managing the fund (bank or non-bank); 3. RET: unit trusts percentage rate of return;

Data for CPF-approved unit trusts were taken from quarterly Performance and Risk Monitoring Reports downloaded from the CPF Board web site at http://www.cpf.gov.sg. Reports are available from 1999 after liberalization of CPF rules governing unit trust investments. The Yahoo Finance website is http://finance.yahoo.com.

Research overview 4. RFR: risk-free percentage rate of return, corresponding to prevailing guaranteed interest rate earned by CPF accounts; 5. RSK: risk classification reflecting funds equity and focus risks (see section 2.2.5 for details); 6. STI: 7. SZE: STIs percentage rate of return; and funds net assets under management, measured in Singapore dollars.

Data collected were used for statistical analysis conducted for this research.

1.5.2 Statistical analysis

Secondary financial data analysis involved preliminary classification of funds according to FMC and RSK variables. For each fund, regression analysis of its returns on market returns was carried out based on Sharpes (1964) capital asset pricing model (CAPM) before performance measures developed by Goodwin (1998), Jensen (1968), Sharpe (1966) and Treynor (1965) were computed. Hypothesis testing of fund characteristics was mainly carried out using two-tail pooled-variance t-test for differences in two means.

1.5.3 Model validation

To assess how fitted regression models perform in practice, examining predicted values can identify an invalid model. For example, unreasonable predicted values indicate an incorrect form of the model or badly estimated coefficients. Very huge or very small coefficients with large standard errors or signs opposite of what were expected indicate a poorly performing model. Also, collecting new data for other periods and splitting existing data for cross-validation of the fitted models predictive ability help to validate the model (Mendenhall & Sincich 1996, pp. 489-91). These techniques are listed in Table 1.1 on the following page.

Chapter 1 Table 1.1 Summary of model validation techniques

Technique Examination of predicted values Examination of estimated model parameters Prediction using collection of new data Cross validation Objective Reveal incorrect form of model or poorly estimated coefficients Indicate whether model may perform poorly when applied with new data Determine accuracy of prediction to assess how well model performs in practice Estimate model parameters and assess fitted model's predictive ability using different groups of collected data

Source: based on Mendenhall and Sincich (1996, pp. 489-91). The next section presents key assumptions and delimitations placed on this research.

1.6 Assumptions and delimitations

This section presents theoretical and statistical assumptions, followed by delimitations imposed by these assumptions.

1.6.1 Assumptions
Several major investment theories underlie this research: Famas (1970) efficient market theory, Markowitzs (1952) portfolio theory, Ross (1976) arbitrage pricing theory (APT), and Sharpes (1964) CAPM. After highlighting theoretical assumptions, this subsection presents statistical assumptions. For theoretical assumptions, efficient market theory is presented first, followed by asset pricing theories in chronological order. Efficient market theory Famas (1970) assumptions for an efficient capital market include: large number of independent profit-maximizing participants analysing and valuing securities; new information on securities arriving at the market in a random and independent manner; and profit-maximizing investors adjusting security prices rapidly to reflect effect of new information.

10

Research overview These assumptions underlie the efficient market hypothesis (EMH) reviewed in Chapter 3, which referred to literature supporting the notion of efficiency in Singapores fund market. Asset pricing theories Sharpes (1964) CAPM is the basis for risk-adjusted portfolio performance measures from Goodwin (1998), Jensen (1968), Sharpe (1966) and Treynor (1965) applied in this research. As CAPM was built on portfolio theory, Markowitzs (1952) assumptions about investors are presented first: representing each investment alternative as a probability distribution of expected returns for a holding period; maximizing one-period expected utility on utility curves demonstrating diminishing marginal utility of wealth; estimating portfolio risk from variability of expected returns; making investment decisions based on expected return and risk, so that utility curve is a function of expected return and expected variance of returns; and preferring higher returns to lower returns for a given risk level and preferring less risk to more risk for a given level of expected return. Five additional assumptions imposed by CAPM (Sharpe 1964) on investors are: 1. being efficient by making portfolio selection based on risk-return utility function; 2. borrowing or lending money at risk-free rate of return; 3. having homogeneous expectation for distribution of future returns; 4. having the same time horizon; and 5. buying or selling infinitely divisible assets in capital markets that are in equilibrium, with no taxes, transaction costs, inflation or change in interest rates. Imposing these CAPM assumptions does not make the research unrealistic, because relaxing them does not change the main CAPM implications (Reilly & Brown 2003, p.

11

Chapter 1 239). Besides, this research fulfilled some CAPM assumptions, as guaranteed CPF interest rates were constant and CPF-approved unit trusts were generally tax-exempted from 1999 to 2004. As an alternative to CAPM, APT (Ross 1976) imposes fewer assumptions: capital markets being perfectly competitive; investors preferring more wealth to less wealth; and stochastic process generating asset returns being expressed as a linear function of a set of risk factors. By examining fund characteristics, this research supports APTs notion of various risk factors affecting returns. Besides, regression analysis is performed for each fund to compute its required rate of return based on CAPMs risk-return assumptions. Statistical assumptions Standard least square assumptions apply for the linear regression model based on CAPM: 1. normal distribution of error terms with zero mean and equal variance; and 2. independent errors associated with any pair of observations (Mendenhall & Sincich 1996, pp. 115-6). Besides least square assumptions, assumptions for hypothesis testing of difference between two population means are required: 1. both sampled populations having approximately normal frequency distributions with equal variances; and 2. random samples being independently selected from their populations (Mendenhall & Sincich 1996, p. 63). Violation of statistical assumptions can result in problems identified in the following subsection.

12

Research overview

1.6.2 Delimitations
By applying a regression model, its assumptions impose delimitations on this research. Four such problems are explained (Mendenhall & Sincich 1996, pp. 367-8): 1. Establishing cause and effect relationship. Since data used in regression were uncontrolled rather than experimental, it is inappropriate to deduce cause-and-effect relationship for fund returns. 2. Departure from assumptions. Violation of regression assumptions specified in the previous subsection may lead to unreliable results. However, it is unlikely assumptions for error terms are satisfied exactly. When departures from assumptions are slight, the model remains valid. 3. Multicollinearity. Variables used in the model may be highly correlated resulting in erroneous computation of coefficients as well as t tests on coefficients being insignificant and F test of overall model showing significance. Residual analysis is applied to resolve this problem in Chapter 5. 4. Extrapolation. It is risky to use the model to predict outside the range of collected data because unusual changes, whether economic or political, may make the model inappropriate for such future prediction. As adequacy of model for extrapolation is unknown, reliability of such inference will be less than the 95 percent level of confidence applied for this research. As the research problem focused on explaining factors differentiating fund performance for various types of institutions, prediction about future returns was not contemplated. Forecasting without recognizing delimitations results in regression abuse.

13

Chapter 1

1.7 Thesis plan

After stating research assumptions and delimitations, an outline of thesis chapters in Table 1.2 previews contents in each chapter. Chapter 1 introduces purposes, objectives, research problem and methodology while Chapter 2 provides background information in Singapores context as well as supporting justification for this research. In Chapter 3, a literature review identifies major contributions in mutual fund research to develop a theoretical framework for factors influencing fund performance. Table 1.2 Thesis chapter outline
Chapter Topic Contents 1 Research overview Purposes and objectives Summary of research problem Methodology General background to the research 2 Background and research justification 3 Literature review 4 Research methodology 5 Data analysis 6 Conclusions Demographics of Singapore's fund industry Demographics of mutual fund research in Singapore Importance of the research for Singapore Justification for the research Identification of key mutual fund researchers and their contributions Theoretical framework Secondary data collection Data analysis approach Analysis of secondary financial data Summary of findings Interpretation of analysis Implication of the research Plausible future research

Source: adapted from Meredith (n.d, p. 2). Research methodology is presented in Chapter 4, covering secondary data collection, regression analysis and hypothesis testing for producing findings reported in Chapter 5. Chapter 6 concludes with interpretation of Chapter 5s analysis to suggest implications as well as future research opportunities.

1.8 Conclusion
This overview presented a linear regression approach to modelling performance of unit trusts observed for Singapores CPF Investment Scheme. Even though ETFs, folios and 14

Research overview separately managed accounts offer various advantages, mutual funds remain relevant as they facilitate ownership of diversified securities by individual investors and occupy a significant share of retirement plan assets. The following chapter provides information in Singapores context while Chapter 3 reviews literature to identify factors influencing fund performance. This research on characteristics of mutual funds affecting their performance differences based on type of FMC is important as it can contribute to finance knowledge and fill a perceived gap in the literature on performance of equity funds facing similar fiduciary standards. Most importantly, this research benefits the mutual fund industry by revealing factors differentiating performance of funds managed by various institution groups. These groups can build on identified factors to produce better performing funds that can compete more effectively with alternative products for investors money. Further research justification is presented in the next chapter.

15

Chapter 2 Background and research justification

2.1 Introduction
Chapter 1 introduced the research problem: determining factors influencing fund performance differences based on type of institution managing the fund, and associated questions: (1) how Singapores unit trust performance vary; (2) how performance of funds managed by banks compare with non-bank funds; (3) which fund characteristics determine performance and (4) how differences in characteristics account for performance differences among funds managed by various institution types. In this chapter, background information and research justification are presented in Singapores context. Background information sets the scene for research while justification is targeted mainly at Singapore. A background map outlining sections in this chapter is set out in Figure 2.1. Figure 2.1 Background map
2 Background and research justification 2.1 Introduction 2.2 Background to the research problem 2.2.1 Singapore's demographic characteristics 2.2.2 Demography of Singapore's asset management industry 2.2.3 Equity, bond, balanced and money market funds 2.2.4 CPF-approved and non-CPF-approved unit trusts 2.2.5 Risk classification of unit trusts 2.2.6 Fee structure of unit trusts 2.2.7 Regulation of unit trusts 2.3.1 Fund performance from 1976 to 1994 2.3.2 Fund performance from 1999 to 2003 2.3.3 Performance of CPF-approved unit trust investors 2.4.1 Mutual fund multifactor model 2.4.2 Conference presentations and article publications 2.4.3 Research outcomes for finance teachers and researchers 2.5.1 Gaps in the mutual fund literature 2.5.2 Research approach 2.5.3 Benefits for Singapore's unit trust industry

2.3 Performance of unit trusts in Singapore 2.4 Contribution of the research 2.5 Justification for the research 2.6 Conclusion

Source: developed for this research.

16

Background and research justification In section 2.2, after a summary of Singapores demographic characteristics, background information on Singapores unit trust industry is presented using secondary data from press releases and government surveys. Such data provide some answers to the first research question in section 2.3, while research contribution and justification are presented in sections 2.4 and 2.5 respectively in the context of a deregulating asset management industry facing increasing competition, before section 2.6 concludes this chapter.

2.2 Background to the research problem

Unit trust is a popular form of investment in Singapore, as for mutual funds in the USA. Singapores unit trust industry can be divided into two segments: retail and wholesale segments accepting investments from individuals and institutions respectively. This study focused on CPF-approved unit trusts after liberalization of the CPF Investment Scheme in 1999, which belongs to the retail segment. Majority of findings in this section were derived from information released by the Monetary Authority of Singapore (MAS), Singapores central bank.1

2.2.1 Singapores demographic characteristics

Among the smallest countries in the world, Singapore is an equatorial island located at the southern tip of the Malaysia peninsula in Southeast Asia, as shown in Figure 2.2 on the next page. According to Singapores Economic Development Board (EDB), the country houses more than four million people with a majority of Chinese, Malays and Indians in that order. While English is the language of administration, Chinese, Malay and Tamil are commonly used (EDB 2004).
1

Financial institutions surveyed by MAS included investment advisers, fund managers, finance and treasury centres, operational headquarters and banks; while total assets under management included unit trusts, funds under advisory service, funds contracted by financial institutions in Singapore as well as funds from individual and institutional clients (MAS 1998-2004).

17

Chapter 2 Figure 2.2 Singapore's geographical location

Location: Southern tip of Malaysia Peninsula, Southeast Asia Land area: 685.4 km sq

Source: statistics from EDB (2004). On the economic front, the countrys money supply and foreign reserves totalled about S$37 billion and S$143 billion respectively in 2002. For that year, with S$208 billion of imports and exports reaching S$224 billion, total trade amounted to S$432 billion. Supporting the countrys economy is a labour force of about two million, with unemployment rate less than five percent. As a financial hub in the Asia Pacific region, the country hosts five local banks and more than 100 foreign banks participating in its asset management industry (EDB 2004).

2.2.2 Demography of Singapores asset management industry

Singapores asset management industry reported close to S$465 billion of total assets managed by financial institutions at the end of 2003, an increase of more than S$30 billion per year since 2000, charting more than 11 percent of annual growth, as shown in Table 2.1 on the next page. Among asset types under management, unit trusts represented S$14 billion or 4 percent of total assets in 2003. Funds invested in unit trusts grew close to 500 percent from 1997 to 2003, representing $16 billion of increase from S$3 billion at the end of 1997. Except for 1998, which suffered a slight dip in asset value due to the Asian

18

Background and research justification financial crisis, unit trusts assets under management grew in excess of 15 percent each year, with highest growth registered in 1999, coinciding with recovery of Asian financial markets as well as liberalization of the CPF Investment Scheme (MAS 1998-2004). Table 2.1 Growth of Singapore's unit trust industry
Year No. of FMCs No. of funds Net assets of funds (S$ bil) Proportion of total assets (%) 1997 1998 1999 2000 2001 2002 2003 22 23 25 31 32 32 34 101 127 187 265 319 382 401 3.3 3.2 6.8 7.8 10.5 14.1 19.2 2.7 2.1 2.5 2.8 3.4 4.1 4.1 Total assets under management (S$ bil) 124.1 150.6 273.7 276.2 307.0 343.8 465.2

Source: derived from MAS (1998-2004). Accompanying growth in net assets was an increase in number of FMCs and their funds. As shown in the table above, from 22 companies managing 101 funds in 1997 to 34 managing 401 funds in 2003, the industry witnessed introduction of about 50 new funds per year, with number of companies stabilizing around 30 since 2000. To organize information on Singapores unit trust industry for further analysis, FMCs are classified into banks and non-banks, with non-banks comprising insurance and investment companies. Besides, unit trusts are classified in the next three subsections according to type of securities held, whether they are CPF-approved as well as risk characteristics.

2.2.3 Equity, bond, balanced and money market funds

Actively managed unit trusts can be classified into four basic types: (1) equity funds, (2) bond funds, (3) balanced funds and (4) money market funds, as explained in section 1.4. Besides showing contribution to net assets from these fund types, Table 2.2 on the following page introduces capital-protected fundsbond funds that guarantee investors their capital and a percentage of profits after staying invested for a number of years.

19

Chapter 2 Table 2.2 Contribution of various types of unit trusts

Type of fund Qty 1999 2000 2001 2002 2003 Qty Net assets Qty Net Qty Net Qty Net Net (S$ mil) assets assets assets assets (S$ mil) (S$ mil) (S$ mil) (S$ mil) 134 5,658 189 6,226 206 6,200 203 5,642 186 6,994 24 651 26 709 32 1,133 36 1,389 46 2,415 17 410 28 568 27 986 36 1,112 36 1,521 11 44 12 87 15 423 16 570 10 541 1 0 187 38 10 253 39 1,782 82 5,423 119 11 4 14,147 401 7,635 88 19,195

Equity Bond Balanced Money market Capital guaranteed Others Total

0 0 6,801 265

0 0 7,843 319

0 9 10,524 382

Source: derived from MAS (1998-2004). Proportion of equity funds accounted for more than 70 percent of funds in 1999 and still accounted for about 40 percent in 2003. During the same period, proportion of money invested in bond funds and balanced funds remained relatively constant at about 20 percent; while capital-protected and money market funds grew from a negligible proportion in 1999 to almost 40 percent in 2003 at the expense of equity funds, which experienced a bearish global stock market during that period. ETFs, hedge funds invested in derivatives, as well as real estate investment trusts invested in buildings, represented a very small portion of total funds. For this research, only equity funds invested in the local stock market were considered. Other funds based on non-STI benchmarks were excluded as benchmarks have unique market cycles. The STI, Singapores principal stock market index, is an unweighted index of about 30 stock issues (Ibbotson & Brinson 1993). To fulfil the common fiduciary standard criteria, the research approach, elaborated in section 2.5.2, excluded non-CPF approved unit trusts, which are a minority. Specifically, data are collected and analysed for CPF-approved domestic equity unit trusts.

20

Background and research justification

2.2.4 CPF-approved and non-CPF-approved unit trusts

In 1999, the CPF Board engaged an independent firm, Mercer Investment Consulting, to advise whether proposed unit trusts were suitable for investment of retirement savings. Unit trusts considered suitable are CPF-approved. Before 2000, CPF-approved unit trusts were less popular than their non-CPF approved counterparts. As shown in Figure 2.3 below, number of CPF-approved funds and their assets under management were both less than their non-CPF-approved counterparts before 2000. Figure 2.3 Growth of CPF-approved and non-CPF-approved unit trusts

(a) Number of funds

450 400 350 300 250
1 0, 000 20, 000 1 8, 000 1 6, 000 1 4, 000 1 2, 000

(b) Net assets (S$ million)

200
8, 000

1 50 1 00 50 0 1 997 1 998 1 999 2000 Y ear 2001 2002 2003

6, 000 4, 000 2, 000 0 1 997 1 998 1 999 2000 Year 2001 2002 2003

CPFappr oved

NotCPFappr oved

CPFapproved

NotCPF-appr oved

Source: derived from MAS (1998-2004). Since 2000, CPF-approved funds outgrew their non-CPF-approved counterparts. From 2000, net assets managed by CPF-approved funds exceeded their non-CPF-approved counterparts, while number of CPF-approved funds became more than non-CPF-approved ones by 2002. By 2003, CPF-approved funds represented more than 70 percent of assets under management in Singapores unit trust industry. Suitability assessment of CPF-approved unit trusts is probably conducted using imperfect informationperformance history and limited knowledge of FMCs. This measurement 21

Chapter 2 problem results in information asymmetry, which is reduced by regulation, contracting, monitoring and security design (Sawicki & Thomson 2000). Mercer is an example of private agency contracted by government to monitor offering of CPF-approved unit trusts meeting investment guidelines pertaining to disclosure of information on investments and prevention of excessive investment in few companies or very risky assets.

2.2.5 Risk classification of unit trusts

Besides approving unit trusts for CPF investments, Mercer developed a risk classification system to group them according to risk-return characteristics, which assists CPF members when choosing funds to meet their needs and investment objectives (CPF Investment Scheme Risk Classification System 2004). This system, shown in Figure 2.4, represented investment risk using two dimensions: equity risk for exposure to stocks and focus risk reflecting diversification in various markets. Figure 2.4 Risk classification of CPF-approved unit trusts
Focus risk Equity risk Higher risk (equity fund) Medium to high risk (balanced fund) Low to medium risk (bond fund) Lower risk (money market fund) Low risk Domestic equity funds High risk Broadly diversified Regional Narrowly focused Sector Country

Source: derived from CPF Investment Scheme Risk Classification System (2004). Equity risk dimension was divided into four categories: (1) lower risk, (2) low to medium risk, (3) medium to high risk and (4) higher risk, corresponding to money market funds, bond funds, balanced funds and equity funds respectively. Focus risk dimension was divided into two categories: (1) broadly diversified for lowrisk unit trusts investing in many geographical regions, countries as well as industries; and 22

Background and research justification (2) narrowly focused for high-risk funds investing in particular geographical regions, countries or industries. Within the narrowly focused category were three sub-categories arranged with increasing level of risk: (1) regional funds investing in emerging markets or particular continents, (2) sector funds investing in particular industries and (3) country funds investing in particular countries. Examining distribution of funds using this classification reveals various risk alternatives being represented by CPF-approved funds. There are regional funds invested in Asia, Europe and North America as well as emerging markets. Sector funds are invested in various industries, including biotechnology, finance, healthcare, information technology and small-capitalization companies. As for country funds, China, Japan, Singapore, the United Kingdom and others are included. Non-existence of funds in lower equity risk, narrowly focused categories is expected, as it is practically impossible to offer very lowrisk regional, country or sector funds. Even though this system simplifies a full range of risk possibilities, it fulfils its objectives by assisting investors to visualize risk dimensions. However, investors need to supplement fund classification with insight from quantifying risk-return characteristics and factors differentiating performance, which are discussed in this research. Specifically, to answer the first research question, fund performance is compared with STI index returns and CPF interest rates. The second research question requires analysis of risk and return characteristics to determine relative performance of funds managed by different institution types, while factors differentiating performance among bank and non-bank funds are identified and tested for research questions 3 and 4 respectively. For this research, as only CPF-approved domestic equity funds are considered, the research sample is classified as high-risk narrowly focused country funds, as illustrated in Figure 2.4 on the previous page. After 2002, Mercer ceased to report performance of

23

Chapter 2 CPF-approved funds and was replaced by Standard and Poors (S&P) Fund Services Asia. Before gathering some answers for the first research question in section 2.3, remaining subsections provide information about fee structure and unit trust regulations.

2.2.6 Fee structure of unit trusts

Various charges are associated with unit trust investments. Besides an initial minimum investment of S$1,000 or S$5,000, a front-end fee is payable to offset marketing cost. Noload funds have no front-end fees, but most funds have front-end fees ranging from one to five percent of NAV. After a fund purchase, management fee paid to the FMC ranges from 0.5 to 3 percent of NAV per annum. To sell a unit trust, the investor pays redemption fee ranging from one to five percent of bid price, or an exchange fee about one percent of asking price for switching to another fund managed by the same FMC (Koh & Fong 2003). Ignoring management fees and CPF interest rates, a simplified breakeven rate of return calculation, given front-end and redemption fees, is illustrated below: Equation 2.1 Calculation of breakeven return for a mutual fund 1 + FEf = (1 + BEf) (1 RDf) or BE f = where BEf = breakeven rate of return for fund f; FEf = percentage cost in front-end fee for fund f; and RDf = percentage cost in redemption fee for fund f. Source: derived from Koh and Fong (2003) Performing a scenario analysis, with the best-case no-load fund with FE = 0% and RD = 1%, lowest breakeven rate of return BEmin = 1 / (1 0.01) 1 or 1 percent approximately. For the worst-case load fund with FE = 5% and RD = 5%, highest breakeven return BEmax = (1 + 0.05) / (1 0.05) 1 or around 10.5 percent. With management fee included, breakeven return is more than the calculated range of 1 to 10.5 percent. No research supported getting better returns by investing in load funds with higher fees. In fact, 1 + FE f 1 RD f 1

24

Background and research justification efficient markets literature reviewed in the next chapter confirmed load funds underperforming an efficient market, lending support for passive investment strategies using ETFs and index funds to mimic equity indexes. As expensive fees reduce fund returns, cost reduction is necessary to improve profitability. Competition within the industry and from alternative products (ETFs, folios and separately managed accounts) lead to reduced costs. Direct modes of marketing with electronic commerce further reduce broker commissions. Unit trust legislation can lower costs too.

2.2.7 Regulation of unit trusts

Unit trust legislation protects investors. In Singapore, unit trusts are established by trust deeds under Section 114 of the Companies Act (Chapter 50), spelling out investment objectives, methods for calculating NAVs and prices, responsibilities of FMCs and trustees as well as shareholder rights. Even though unit trusts are operationally similar to mutual funds, they are legally different as mutual funds are investment companies while unit trusts are created through trust deeds (Koh & Fong 2003). Since liberalization of the CPF Investment Scheme, surge in demand for unit trusts is due to more accommodating regulations and investors increased knowledge of such investment vehicles (Koh 1999). According to Singapores Ministry of Manpower (MOM), CPF investments by Singaporeans commenced in 1986 with the Approved Investment Scheme allowing a portion of CPF to be invested in approved financial instruments (gold, bonds as well as approved stocks and unit trusts) to groom retirement income. In 1993, Basic and Enhanced Investment Schemes were introduced, later integrated to form the CPF Investment Scheme in 1997 (MOM 1998). In 1998, the CPF Investment Scheme was liberalized to broaden the range of unit trusts available for investments. Revised investment guidelines gave fund managers more flexibility in diversifying portfolios for reducing risks, while raising disclosure standards 25

Chapter 2 to inform CPF investors about returns, risks and fund performance comparison. During the same year, MAS licensed 24 FMCs meeting quality evaluation criteria (Applications for Admission of Fund Management Companies into CPF Investment Scheme 2001). Since 2001, Special and Ordinary accounts can fully invest in approved financial instruments. While Special account is used for old age contingency purposes and retirement-related financial products, Ordinary account pays for education, housing, insurance and investments (CPF Handbook: Building Our Future 2005). Among other changes, the CPF Investment Scheme allowed ETF investments from 2001 (CPF Savings can now be Invested in Exchange Traded Funds 2001) while unit trusts, investment-linked insurance products (ILPs), ETFs and fund management accounts denominated in foreign currencies were permitted in 2002 (Changes to the CPF Investment Scheme 2002). Unit trust investors use various selection criteria. One criterion is to choose products that earn higher returns than guaranteed CPF interest rates, which are revised quarterly based on local bank rates on deposits and savings accounts. However, the CPF Act guarantees a minimum rate of 2.5 and 4.0 percent for Ordinary and Special accounts respectively (CPF Handbook: Building Our Future 2005).2 As CPF interest rates are guaranteed if members choose not to invest in securities, these rates are risk-free and should therefore be based on the economys real rate of growth, capital market conditions and expected inflation rate (Reilly & Brown 2003, p. 28). If these factors are not taken into consideration, prevailing CPF interest rates may be unrealistic. In fact, Douglas (2003) commented 2.5 and 4 percent interest rates on Ordinary and Special accounts from 2002 to 2003 were unsustainable and hard to outperform on a risk-adjusted basis. However, evaluating unit trust performance with respect to CPF interest rates may be unfair compared to evaluation against market indexes.
2

From 2002 to 2004, annual interest rates for CPF Ordinary and Special accounts were 2.5 and 4.0 percent respectively, while bank savings rate was less than 1 percent.

26

Background and research justification In the following section on fund research in Singapore, evaluation of fund performance with respect to CPF interest rates and market indexes are considered to provide some answers to the first research question.

2.3 Performance of unit trusts in Singapore

Evident by literature reviewed in the following chapter, mutual fund performance was studied extensively in the USA, but little academic research was published on Singapores unit trusts, other than Koh (1999), Koh and Fong (2003) and Tan (2001). This is due to a relatively small and young fund management industry that started with less than ten unit trusts in 1976 (Koh 1999) and grew to a few hundred funds by 2004, compared to thousands of funds in many developed markets. Research indicated short-term negative returns and long-term positive returns with majority of unit trusts under-performing their benchmark indexes and CPF interest rates before 2000 (Koh 1999). From 2000 to 2002, over 80 percent of CPF-approved unit trusts turned in negative absolute returns for the three-year period, according to data released by S&P Fund Services (Ng 2003). Answering partially the first research question on unit trust performance, research by Koh (1999) and Ng (2003) indicated most unit trusts underperforming market indexes, implying efficiency in Singapores unit trust industry. As majority of CPF-approved unit trusts generated negative returns from 2000 to 2002, earning guaranteed interest rates was a better alternative during that period. Poor equity fund performance from 2000 to 2002 coincided with bearish equity markets globally. For this research, examining performance data for a five-year timeframe from 1999 to 2004 spans a business cycle, including strong equity decline during 20002002, a difficult period for fund managers but appropriate for identifying superior managers and institution groups. Chapter 5 reconsidered the first research question by analysing data for the five-year timeframe. 27

Chapter 2 Besides reporting similar results for Sharpe and Treynor performance measures of unit trusts, Koh (1999) found their risk-return profiles congruent with stated investment objectives, and asset diversification comparable to market indexes.

2.3.1 Fund performance from 1976 to 1994

Unit trust performance during the 18-year period from 1976 to 1994 is presented by holding period returns (HPRs) in Table 2.3 for periods ranging from the last six months to 18 years.3 Two perspectives are presented: (1) minimum, average and maximum returns earned by unit trusts over the holding periods; and (2) proportion of funds earning positive returns above suitable benchmarksCPF interest rates and relevant stock market indexes. Table 2.3 Holding period returns of unit trusts in Singapore from 1976 to 1994
Holding period (years) Number of funds Minimum HPR (%) Average HPR (%) Maximum HPR (%) Proportion of funds with HPR > benchmark stock indexes (%) Proportion of funds with HPR > CPF interest rate (%) Proportion of funds with HPR > 0 (%) 0.5 1 2 5 10 15 18 35 33 28 20 9 7 6 -18.46 -33.05 -10.40 -27.22 -35.68 -26.41 17.19 -4.72 -16.27 37.65 42.31 60.93 55.69 143.69 6.61 3.88 97.95 136.86 217.51 305.93 511.45 33 37 46 71 33 14 17 9 14 3 3 82 96 65 85 33 56 14 57 33 100

Source: Koh (1999). Examining fund returns, average HPRs were negative in the short term (up to one year) but increased positively beyond 30 percent in the long run, implying most short-term investments incurred losses while majority of investments in the long term earned healthy profits. This is confirmed by examining proportion of funds earning returns above benchmarksless than 15 percent of funds earned positive returns in the short term, while

HPR is the rate of return earned by remaining invested for the entire holding period. Specifically, the Tperiod HPR for fund f HPRfT = (BIDfT ASKf0 + DIVt) / ASKf0 where BIDfT is the funds bid price at the end of the holding period, ASKf0 is the funds ask price at the beginning of the holding period and DIVt is dividend declared during each period t (Koh 1999).

28

Background and research justification majority of funds reported positive returns in the long run. All 6 funds for the 18-year period reported positive returns. Investment risk is revealed by difference between maximum and minimum HPRs, which increased as the holding period lengthened, with negative minimum HPRs prevailing across most holding periods signifying risk of selecting loss-making funds. However, compared to bank deposits earning sub-zero interest rates in Singapore, the average unit trust generated better returns in the long term. For investments using CPF savings, a fund earning less than CPF interest rates is considered poorly performing. Examining proportion of funds outperforming CPF rates revealed majority of funds not outperforming CPF rates for most holding periods, except for two-year and five-year periods. Therefore, investors may be better off earning guaranteed interest rates with their CPF accounts. This is confirmed in section 2.3.3 reporting minority of CPF-investors outperforming CPF rates. For non-CPF investors, fund performance can be compared with relevant stock market indexes.4 Stock indexes outperformed most funds for all holding periods, except for the five-year period from 1989 to 1994. Therefore, a minority of funds outperformed the market, confirming studies done in the USA reviewed in the next chapter. Summarizing this subsection, Koh (1999) reported unit trusts earning healthy returns in the long term, but only a minority outperformed CPF interest rates and market indexes. The following subsection discusses later research by Wong (2004).

As index funds were not available, fund performance was compared with HPRs of relevant stock market indexes (Koh 1999).

29

Chapter 2

2.3.2 Fund performance from 1999 to 2003

According to Wongs (2004) study of Asia Pacific, European and USA stock markets, the USA market is very efficient, as equity funds from Singapore invested in USA only marginally exceeded S&P 500 index returnseven though some 40 percent of funds outperformed the index in 2003, none of them achieved this for holding period of five years, again confirming earlier studies in the USA. As for equity funds from Singapore invested in Europe, short-term results were similar to their USA counterparts, as more than 50 percent of European equity funds failed to exceed Morgan Stanley Capital International (MSCI) European index returns, but for a five-year period, over 80 percent of them outperformed the MSCI index. Compared to European and USA equity funds, Asia Pacific funds in Singapore produced more impressive results. About half the Japanese equity funds in Singapore exceeded Nikkei 225 index returns during 2003 while every Japanese equity fund outperformed the index over a five-year period. As for Asia-ex-Japan funds in Singapore, over 60 percent of them exceeded MSCI Asia-ex-Japan index returns each year from 1998 to 2003 while more than 80 percent outperformed the index over the five-year period. Investigating Singapore equity funds, results were similar to Asia-ex-Japan funds as about 60 percent of local funds outperformed the STI for each year while 80 percent did the same for the five-year period. Wong (2004) attributed Asian funds better performance to greater market inefficiency in the region, compared to European and USA markets. As market information in Asia did not flow as freely as their European and USA counterparts, fund managers conducting rigorous research can profit from Asias market inefficiency. The next chapter developed research issues and hypotheses to evaluate performance of various groups of unit trusts.

30

Background and research justification

2.3.3 Performance of CPF-approved unit trust investors

While the previous subsections considered fund performance, this sub-section examines performance of retail fund investors. Even though liberalization of the CPF Investment Scheme removed various restrictions and allowed more investment choices, more than 60 percent of CPF monies remained not invested. At end of 2003 first quarter, CPF Ordinary and Special accounts invested about S$28 billion, leaving S$68 billion available (Wong 2003). The larger portion of CPF money untouched may be due to majority of investors not able to outperform guaranteed CPF interest rates. As shown in Table 2.4, even though number of CPF investors increased from about 494 thousand in 2000 to 680 thousand in 2002, proportion of investors outperforming CPF interest rates decreased from 42 to 22 percent during that period (CPF Investment Scheme: Realized Profit/Loss for the Financial Year Ended 30 September 1999-2003). Table 2.4 Distribution of profit and loss for CPF investors
Profit/loss 1999 2000 2001 2002 No. of % No. of % No. of % No. of % investors investors investors investors 119,000 25 207,000 42 216,000 35 148,000 22 7,800 2 9,700 2 23,000 4 23,000 3 259,000 54 183,000 37 239,000 39 284,000 42 90,000 19 94,000 19 133,000 22 225,000 33 475,800 100 493,700 100 611,000 100 680,000 100

Net profit > CPF interest Net profit = CPF interest Net profit < CPF interest Net loss Total

Note: number of investors tabulated for each financial year ended 30 September. Source: CPF Investment Scheme: Realized Profit/Loss for the Financial Year Ended 30 September (1999-2003). During the same period, proportion of investors making outright losses increased from 19 to 33 percent. During the 9-year period from 1993 to 2002, less than 40 percent of CPF investors made profits in excess of guaranteed rates (Wong 2003). Lacklustre performance may be due to poor fund returns, expensive fee structure and imposed regulations, highlighting importance of academic research on how to improve fund performance.

31

Chapter 2

2.4 Contribution of the research

Outcomes from this research make several contributions to the investments field: (1) derivation of a mutual fund performance model; (2) conference presentations and article publications; as well as (3) research outcomes useful to finance teachers, researchers and practitioners. The following subsections discuss these three forms of contributions.

2.4.1 Mutual fund performance model

The next chapter develops a conceptual mutual fund performance model relating type of financial institution managing the fund and its characteristics to performance. Asset allocation, expenditures, risk and size are fund characteristics considered in this model while financial institutions are banks and non-banks.

2.4.2 Conference presentations and article publications

To network with academics interested in mutual fund research, research findings were presented at four international conferences in 2005. For 2006, one article was accepted for publication in 2007 and another was being reviewed by a refereed journal.5

2.4.3 Research outcomes for finance teachers and researchers

This research provides finance professors surveying Singapores unit trust industry with updated information on fund performance and characteristics. The project goes further by providing insight on determinants of performance differences, encouraging further research in other fund industries to uncover more differentiating factors. When comparing fund performance, this research conforms to recommended practice by considering similar fiduciary standards among institution groups. In addition, the research extends findings from bond funds to equity funds. When identifying determinants of fund

Page vi provides a listing of all conference papers and journal articles based on this research.

32

Background and research justification performance, knowledge is contributed towards factors differentiating fund returns among institution groups, especially in Singapores context. This research, by confirming and extending existing knowledge on mutual fund performance, is justified by its contribution to knowledge on factors influencing institution groups fund performance.

2.5 Justification for the research

In this section, research justification is presented from three dimensionsgaps in the literature, research approach and benefits for Singapores unit trust industry.

2.5.1 Gaps in the mutual fund literature

Bauman and Miller (1995) reported bank-managed pension funds under-performing funds from other institutions. Supporting them were Bogle and Twardowski (1980), who reported bank-managed equity funds under-performing their non-bank counterparts. However, Fryes (2001) study found banks not under-performing when comparing bankmanaged bond funds with non-bank counterparts. Previous research reporting underperformance of bank-managed funds relative to nonbank ones may have ignored their differing fiduciary standards (Frye 2001). In contrast to banks focus on short-term investments to avoid interest rate risk and maintain liquidity, investment firms have a wider variety of investment objectives, ranging from high-risk capital appreciation to low-risk money market income (Reilly & Brown 2003, pp. 63-7). Therefore, perceived underperformance of bank-managed funds relative to non-bank counterparts may be due to bank managers prudent avoidance of risks. Comparison of funds managed by institution groups facing similar fiduciary standards, as recommended by Frye (2001), is necessary to give an unbiased picture of their relative performances. Research on performance of equity funds from institutions facing similar fiduciary responsibilities was conducted to fill the perceived gap in mutual fund literature.

33

Chapter 2

2.5.2 Research approach

Examining performance of CPF-approved unit trusts is appropriate as they face similar fiduciary responsibilities for retirement income, effectively controlling for financial institutions differing fiduciary standards. This approach follows Fryes (2001) recommendation for comparing funds with similar fiduciary standards. By evaluating performance of funds meeting similar fiduciary responsibilities, influence of factors on performance of funds managed by banks and non-banks is determined.

2.5.3 Benefits for Singapores unit trust industry

Identifying factors differentiating performance of funds managed by various institution groups is an appropriate research area, as it has important implications for management of local banks facing increasing competition from foreign financial institutions while Singapores financial sector is gradually deregulated. As functional distinction between banks and non-banks become indistinguishable, the mutual fund market is strategically important for financial institutions. Gallo, Apilado and Kolari (1996) showed mutual fund products improve banks profitability. Profitability of banks can be significantly improved if they consistently deliver top-performing funds.

2.6 Conclusion
Based on this chapters background information, Singapores unit trust industry has grown more than 15 percent per annum since 1998. This coincides with penetration of household financial assets in Singapore reaching levels of developed countries (Douglas 2003). In Singapore, there are about 400 unit trusts offering access to most types of assets globally in the form of equity funds, bond funds, balanced funds and capital-guaranteed funds. Among these funds, about 300 unit trusts managed by about 30 FMCs are CPF-approved, with majority of them being narrowly focused regional, sector or country funds rather than

34

Background and research justification broadly diversified. These CPF-approved funds were valued at more than S$13 billion at the end of 2003, with majority of them invested in equity funds. Most CPF-approved unit trusts under-performed CPF interest rates by recording negative absolute returns during the global 20002002 stock market decline, resulting in majority of 700 thousand CPF investors suffering losses and being worse off than members earning guaranteed interest rates by not investing. Disappointing performance of these funds may be attributable to their expensive fee structure, but it may not be fair to compare fund performance with risk-free CPF interest rates. Even though the CPF rates are based on interest rates paid by local banks on deposits and savings accounts, CPF rates are substantially higher and its minimum rate, as guaranteed by legislation, may not reflect the countrys economic conditions. Performance is therefore compared with market indexes. Singapore has signed Free Trade Agreements (FTAs) with Australia, Japan, Europe, New Zealand and the USA (Teo 2003). FTAs enhance Singapores attractiveness to foreign financial institutions by giving them better access to its banking sector. Among responses from local banks are increased level of competition, consolidation and production of a Code of Consumer Banking Practice (Code of Consumer Banking Practice 2002) to preserve and enhance customer relationships for challenging times ahead. Banks may not under-perform their competitors in managing funds. Results from this research will conclude whether banks under-performed non-banks. Most importantly, by understanding factors differentiating fund performance among various institution groups, competitive advantage can be sustained to face challenges from financial deregulation. To better understand the foundation of these factors, a theoretical framework is presented in the next chapter based on literature pertaining capital market efficiency, asset pricing and multifactor models as well as past studies on mutual fund performance.

35

Chapter 3 Literature review

3.1 Introduction
Chapter 2 provided background information and justification for researching Singapores unit trust industry. In this chapter, literature on mutual fund performance determinants is reviewed to identify a concise set of publicly available variables affecting performance of funds managed by banks, insurance and investment firms. A literature map outlining sections of this chapter is presented in Figure 3.1. Figure 3.1 Literature map
3 Literature review

3.1 Introduction 3.2.1 Efficient market theory 3.2.2 Mutual fund performance 3.2.3 Measurement of mutual fund performance 3.2.4 Asset pricing theories 3.3.1 Asset allocation 3.3.2 Investment style 3.3.3 Risk 3.3.4 Past performance and performance persistence 3.3.5 Flow of funds and assets under management 3.3.6 Research and trading costs 3.3.7 Type of fund management company 3.4.1 Mutual fund multifactor model 3.4.2 Research issues and propositions

3.2 Theories on mutual fund performance

3.3 Determinants of mutual fund performance

3.4 Theoretical framework

3.5 Conclusion

Source: developed for this research. Following a broad literature review on the mutual fund performance parent discipline in section 3.2, section 3.3 focuses on the immediate discipline of fund performance determinants to identify suitable factors. Factors to be considered in separate subsections include asset allocation, investment style, risk, past performance, flow of funds and fund

36

Literature review size, research and trading costs as well as type of FMC in this order. These factors are used to build a theoretical framework for fund performance in section 3.4. A conceptual model is built in section 3.4.1 before section 3.4.2 presents research issues and hypotheses to be investigated. Section 3.5 concludes this chapter by giving a summary of performance factors to be tested in the following methodology chapter.

3.2 General theories relating to mutual fund performance

Reviewing the parent discipline of mutual fund research acknowledges development of several major theories in finance, including APT (Ross 1976), efficient market theory (Fama 1970, 1991), portfolio theory (Markowitz 1952), as well as capital market theory, which was independently developed by Lintner (1965), Mossin (1966) and Sharpe (1964). These theories are reviewed in the following subsections.

3.2.1 Efficient market theory

Studies on fund manager performance generally provided support for the strong-form EMH (Fama 1970, 1991). Based on a fair game model, efficient market theory suggested current market prices fully reflecting available information about securities, with expected returns based on current market prices being consistent with their risks. The overall EMH comprised of three sub-hypotheses: (1) weak, (2) semi-strong and (3) strong forms, with each sub-hypothesis encompassing the previous one, as illustrated in Figure 3.2 on the next page. Accordingly, the weak form EMH assumed current security prices fully reflect information from security markets while the semi-strong form EMH asserted security prices adjust rapidly to all public information. As for the strong form EMH, security prices fully reflect all private and public information, implying no investor groups (including financial institutions) can have access to private information to consistently experience above-

37

Chapter 3 average profits. It is normal for an actual fund return to exceed benchmark market return during a particular time period, resulting in positive abnormal return and profit for some investors. Grossman and Stiglitz (1980) showed existence of abnormal returns with costs of information gathering and processing. Such an occurrence does not imply market inefficiency or violation of market efficiency assumptions, as overall market is efficient in the long term. Figure 3.2 Weak, semi-strong and strong forms of market efficiency
reflecting Strong form market efficiency Semi-strong form market efficiency Weak form market efficiency Private information Public information Market information

Source: adapted from Jones (1998, p. 258). Supporting the strong form EMH, numerous early studies revealed most equity funds in the USA not matching aggregate market performance (Carlson 1970; Elton et al. 1993; Ippolito 1989; Jensen 1968; Malkiel 1995; Sharpe 1966). These studies are reported in the following subsection.

3.2.2 Mutual fund performance

Sharpes (1966) evaluation of overall fund performance reported about 30 percent of funds outperformed the Dow Jones Industrial Average (DJIA). To test whether past performance can predict future performance, after dividing the sample period into two halves, comparison of fund performance rankings for the two sub-periods showed past performance was not a good predictor of future performance.1 Besides, examination of any

More recent studies of performance persistence are discussed in section 3.3.4.

38

Literature review relationship between performance and expense ratio revealed an association between good performance and low expenditures.2 When expenses were added to returns to derive gross returns, slightly less than 60 percent of funds outperformed the DJIA, implying average fund performance was as good as the market index. But deducting operating costs resulted in majority of net returns under-performing the index. In a related study, Jensen (1968) found on a yearly basis, funds on average earned about one percent less than expected returns, given the level of risk. Adding expenses to derive gross risk-adjusted returns showed slightly more than 40 percent outperformed the market while deriving net returns revealed one-third of funds outperforming the market. As less than 50 percent of funds outperformed the market on a risk-adjusted basis, funds on average could not beat the market. Extending analysis to broader market indexes, Carlson (1970) reported worse fund performance. Specifically, a few groups of funds had gross returns better than the S&P 500 and New York Stock Exchange (NYSE) composite, but all groups had net returns underperforming market indexes. Corroborating previous studies, consistency in risk-adjusted performance was lacking as less than one-third of funds managed to perform above average for two sub-periods. Therefore, fund performance can differ for different market indexes, but the average fund was consistently inferior to the overall market. Extending range of funds to international ones, Cumby and Glen (1990) found fund performance did not outperform the Morgan Stanley world equity index. As for bond funds, Blake, Elton and Gruber (1993) confirmed these funds under-performed their indexes. On a country basis, Cai, Chan and Yamada (1997) showed Japanese mutual funds generally under-performed their benchmarks. As reported in the previous chapter, research

Section 3.3.6 presents more recent studies on relationship between performance and expenditures.

39

Chapter 3 showed Singapores unit trusts generally under-performed market indexes during 1976 to 1994 (Koh 1999), but outperformed the market from 1999 to 2003 (Wong 2004). A summary of market efficiency literature leading to hypotheses for the first research issue is illustrated in Figure 3.3. Figure 3.3 Derivation of research issue 1 on unit trust performance
Efficient market theory (Fama 1970) Weak form hypothesis Semi-strong form hypothesis Strong form hypothesis Composite portfolio performance measures (Goodwin 1998; Jensen 1968; Sharpe 1966; Treynor 1965)

H1.1 Domestic equity funds do not outperform benchmark stock market index

Research issue 1 Performance of Singapore's unit trust H1.3 No significant difference between returns from domestic equity funds and guaranteed interest rates

H1.2 Positive relation between returns from domestic equity funds and benchmark stock market index

Opposed by Koh (1999)

Supported by: Brinson, Hood & Beebower (1997); Brinson, Singer & Beebower (1991); Ibbotson & Kaplan (2000); Jensen (1968) Supported by: Blake, Elton & Gruber (1993); Cai, Chan & Yamada (1997); Carlson (1970); Cumby & Glen (1990); Koh (1999); Ippolito (1989); Jensen (1968); Malkiel (1995); Sharpe (1966). Opposed by: Wong (2004)

Source: developed for this research. As shown in the figure, performance was measured using four risk-adjusted criteria derived by Goodwin (1998), Jensen (1968), Sharpe (1966) and Treynor (1965), to be discussed in the next subsection.

40

Literature review

3.2.3 Measurement of mutual fund performance

Peer group comparison, the most common method for evaluating fund managers, displays returns from a representative sample of investors over a specific time period in a box-plot (Kritzman 1990). To compare fund performance, the sample is divided into percentiles, indicating each funds relative ranking. This method, by not explicitly adjusting for each funds risk level, is problematic. Risk is implicitly considered by selecting funds with same level of risk in the sample, which is not possible with funds having different investment styles (Reilly & Brown 2003, pp. 1108-9). Solving this problem requires composite portfolio performance measures combining risk and return values. Composite performance measures developed by Goodwin (1998), Jensen (1968), Sharpe (1966) and Treynor (1965) were applied to measure risk-adjusted fund performance. Information ratio portfolio performance measure Information ratio IRf for an equity fund f is computed as the arithmetic average of excess return divided by standard deviation of the excess return (Goodwin 1998): Equation 3.1 Information ratio for Singapores domestic equity funds IR f = ER f ER

where ER f = average excess return for fund f during a specified time period, which is equal to average return for fund f during the period minus average return for benchmark STI index during that period; and = standard deviation of excess return during the period. ER Source: adapted from Goodwin (1998). IR is interpreted as mean excess return per unit of unsystematic risk. Reasonable IR values range from 0.5 to 1.0, corresponding from good to exceptionally good performance (Grinold & Kahn 1995). As for its relation with Jensen and Sharpe measures, IR can be expressed in terms of Jensen alpha while Sharpe ratio is a special case of IR.

41

Chapter 3 Jensen portfolio performance measure Jensen alpha ft for an equity fund f during time period t is shown in Equation 3.2. As ft is difference between actual and expected returns, the measure represents proportion of fund returns attributable to managers ability to derive above-average returns adjusted for risk. Equation 3.2 Jensen alpha for Singapores domestic equity funds

ft = RET ft RFRt f (STI t RFRt )

where RETft = return from fund f for time period t; = beta coefficient for fund f; f RFRt = rate of return on risk-free assets during time period t; and STIt = return from the STI during time period t. Source: adapted from Jensen (1968). Significantly positive or negative values of ft represent superior or inferior managers in market timing as well as stock selection, while a near-zero value matches market performance on a risk-adjusted basis. Sharpe portfolio performance measure Sharpe ratio Sf for an equity fund f is defined as: Equation 3.3 Sharpe ratio for Singapores domestic equity funds Sf = where RET f RET f RFR f

= average rate of return for fund f during a specified time period;

= average rate of return on risk-free assets during the time period; and RFR = standard deviation of return for fund f. f Source: adapted from Sharpe (1966). Sf considers total risk of portfolio by including standard deviation of returns in its denominator, resulting in a measure of risk premium earned per unit of total risk. Treynor portfolio performance measure Similar to the Sharpe (1966) measure, Treynor ratio Tf for an equity fund f is computed as the funds risk premium divided by its beta coefficient, as shown in Equation 3.4 on the following page.

42

Literature review Equation 3.4 Treynor ratio for Singapores domestic equity funds Tf = where RET f RET fi RFR f

= average rate of return for fund f during a specified time period;

= average rate of return on risk-free assets during the time period; and RFR = beta coefficient for fund f. f Source: adapted from Treynor (1965). Comparing Sharpe and Treynor measures, the Sharpe ratio uses total portfolio risk, represented by standard deviation of returns on the capital market lines (CML) horizontal axis, while the Treynor measure uses systematic risk, summarized by on horizontal axis of the security market line (SML). Therefore, in terms of capital market theory, the Sharpe and Treynor measures compare funds with the CML and SML respectively. Applying these four measures to the fund performance hypotheses (introduced in Figure 3.3 on page 40), as Jensen alpha and information ratio compare fund performance with market index, they were used for the first hypothesis to compare performance of domestic equity funds with the STI. Sharpe and Treynor ratios incorporate risk-free returns and were used for the third hypothesis to determine existence of excess returns over guaranteed CPF interest rates. As for the second hypothesis, regression for computing Jensen alphas revealed correlation between fund and index returns, as reported by Brinson, Hood and Beebower (1986), Brinson, Singer and Beebower (1991), Ibbotson and Kaplan (2000) as well as Jensen (1968). Even though composite performance measures improve upon comparisons using peer group and mere returns, some researchers identified bias in these composite measures. For example, Friend and Blume (1970) reported risk-adjusted performance measures of lowrisk portfolios better than their high-risk counterparts; Klemkosky (1973) found positive relation between risk and composite performance measures; while Leland (1999) demonstrated Jensen alpha being biased downward for portfolios limiting downside risk.

43

Chapter 3 Following Jensen (1968), Sharpe (1966) and Treynor (1965), Fama (1972) proposed a more detailed breakdown of portfolio performance into its risk and selectivity components.3 For this research, section 3.4s theoretical framework provides a detailed analysis of performance determinants in terms of fund characteristics. While composite performance measures are not without problems, in the absence of alternatives, all four measures were applied for this research to minimize errors from relying on one measure, as each measure ranks performance differently (Reilly & Brown 2003, p. 1122) and can yield different rankings (Corrado & Jordan 2005, p. 434). These measures use arithmetic mean return as measure of central tendency for performance, and a measure of dispersion for risk, which is variance (or its square root, the standard deviation) for total risk or beta for relative systematic risk. Geometric mean return is an alternative performance measure combining risk and return (Seitz & Ellison 1999, p. 340). If probability distribution of HPRs is the same for each time period and investors objective is to maximize long-term growth of wealth, the fund with highest geometric mean return provides highest long-run growth rate to maximize wealth (Seitz & Ellison 1999, p. 420). However, investors are also concerned with shortterm returns as they select fund managers based on recent performance. Therefore, the more popular mean-variance model is used, even though the geometric mean model has some academic supporters, including Latane and Tuttle (1967), Markowitz (1976) and Rubinstein (1976). Summarizing this review of fund performance measurements, four risk-adjusted measures were used for this research as each criterion measures fund performance differently. While information ratio and Jensen alpha were used to compare fund and market performance,

Portfolio performance = excess return = investors risk + managers risk + security selection skill (Fama 1972).

44

Literature review Sharpe and Treynor ratios were computed to determine excess returns over guaranteed interest rates. Linkage between fund performance literature and the second research issue on relative performance of funds managed by various institution groups is illustrated in Figure 3.4. Figure 3.4 Derivation of research issue 2 on relative performance of FMCs
Portfolio theory (Markowitz 1952) Capital asset pricing model (Sharpe 1964) Composite portfolio performance measures Information ratio Jensen (1968) Sharpe (1966) (Goodwin 1998) alpha ratio Treynor (1965) ratio Critique of composite performance measures: Friend & Blume (1970); Klemkosky (1973); Leland (1999)

Research issue 2 Relative performance of unit trusts managed by various institutional groups H2 Bank-managed unit trusts do not under-perform funds managed by insurance & investment companies Supported by Frye (2001) Opposed by Bauman & Miller (1995), Bogle & Twardowski (1980)

Source: developed for this research. Not shown in this figure, the research hypothesis is broken into five sub-hypotheses testing performance in terms of returns and the four risk-adjusted performance measures reviewed in this sub-section. As shown in the figure, these performance measures were compared for bank and non-bank funds. These measures were based on the CAPM model (Sharpe 1964) reviewed in the next subsection.

3.2.4 Asset pricing theories

This subsection traces chronologically development of theories for measuring risk-return characteristics of mutual funds, starting with Markowitzs (1952) portfolio theory.

45

Chapter 3 Portfolio theory (Markowitz 1952) derived an asset portfolios expected return, with standard deviation of returns as its measure of risk.4 To minimize a portfolios total risk, diversification of investments among imperfectly correlated assets produces an efficient frontier of investment portfolios. Mutual funds represent such portfolios. Based on portfolio theory (Markowitz 1952), Sharpe (1964) developed CAPM to determine required return for any risky asset.5 CAPM introduced risk-free return, which for this research corresponds to guaranteed interest rates earned by investors not investing in mutual funds. Besides, CAPM introduced beta as standardized measure of systematic risk, relating covariance of returns between an asset and a market portfolio to variance of returns from the market portfolio. Beta measurements for portfolios and by implication, mutual fund betas, are generally stable for long time periods, assuming sufficient trading (Reilly & Brown 2003). Expanding CAPMs preoccupation with asset beta as the factor influencing returns, the APT (Ross 1976) contended there are many factors influencing asset returns.6 Comparing CAPM and APT, CAPMs advantage is its theoretical simplicity in specifying its risk factor in terms of market portfolios excess return (RM - RFR), which has an empirical disadvantage when identifying the investment universe M and estimating market portfolio returns RM. Using market index as proxy for M resulted in benchmark error (Roll 1978,

Specifically, expected return for portfolio P

E (RP ) = wi E (Ri ) where wi is proportion of portfolio in

i =1

asset i and E(Ri) is expected return for asset i. Standard deviation of returns for portfolio
n n n P = wi2 i2 + wi w j Covij i =1 j =1i j i =1
5

1/ 2

where

i2 is variance of returns for asset i and Covij is

covariance between returns for assets i and j (Markowitz 1952). Expected return on risky asset i E (Ri ) = RFR + i (R M RFR ) where RFR is risk-free return, RM is
2 i = Covi , M M (Sharpe (1964).

return for market portfolio M and

6

Return on asset i

Ri = Ei + bi11 + bi 2 2 + ... + bik k + i where Ei is expected return for asset i, bik is

reaction from asset is returns to movements in common factor k and i is unique effect on asset is return (Ross 1976).

46

Literature review 1980, 1981), but most academics ignored the error as research evidence generally supported CAPM (Reilly & Brown 2003). Even though CAPMs benchmark error is not resolved in this research, it supports the APT by considering a few performance factors. APT imposes fewer assumptions than CAPM, but does not identify risk factors (Shanken 1982). To implement APT, pioneer attempts analysed multiple periods of realized returns for various securities using multivariate statistical techniques (Chen 1983; Roll & Ross 1980) to reveal three or four statistically significant factors. Multifactor model is an alternative with two approachesidentification of macroeconomic or microeconomic risk factors (Reilly & Brown 2003): Macroeconomic-based risk factor models. Macroeconomic factors identified to capture variations in an assets cash flows and investment returns over time include returns on an index of stocks, growth rates in industrial production, consumer price index measure of change in inflation, difference between actual and expected inflation levels, change in bond credit spread and term structure shift, as hypothesized by Chen, Roll and Ross (1986). For a later study, Burmeister, Roll and Ross (1994) used confidence, time horizon, inflation, business cycle and market timing risks. Microeconomic-based risk factor models. Microeconomic factors identified as relevant security characteristics include excess return of a stock market portfolio, return differential between small-cap and large-cap stocks and return differential between high book-to-market and low book-to-market value stocks (Fama & French 1993). Carhart (1997) added a momentum factor. Elton, Gruber and Blake (1996) used S&P 500, Lehman Brothers aggregate bond index and Prudential Bache indexes for difference between large-cap and small-cap stocks as well as value and growth stocks; while Ferson and Schadt (1996) used additional public information variables, including

47

Chapter 3 shape of yield curve and dividend payouts. For a more extensive microeconomic-based model, an investment consultancy firm used 13 security characteristics and 50 industry indexes (Grinold & Kahn 1994). Figure 3.5 illustrates linkage of asset pricing literature to the third research issue on fund performance determinants. Figure 3.5 Derivation of research issue 3 on mutual fund performance determinants
Asset pricing theories Capital asset pricing model (Sharpe 1964) Implementation Multivariate statistical technique (Chen 1983; Roll & Ross 1980) Multifactor model Arbitrage pricing theory (Ross 1976)

Macroeconomic-based risk factor model (Burnmeister, Roll & Ross 1994; Chen, Roll & Ross 1986) Microeconomic-based risk factor model (Carhart 1997; Elton, Gruber & Blake 1996; Fama & French 1993; Ferson & Schadt 1996; Grinold & Kahn 1994) Research issue 3 Factors affecting mutual fund performance H3 Significant relation between expense ratio, size & performance Supported by Indro et al. (1999)

Source: developed for this research. Before developing a suitable framework for these factors in section 3.4, the following section reviews literature to identify suitable factors.

3.3 Determinants of mutual fund performance

In this section, characteristics of funds that can be important determinants of returns are identified. This research builds on determinants of mutual fund performance identified in existing literature, which is classified by fund types. 48

Literature review For closed-end country funds, Anderson et al. (2001) reported returns were related to targeted countrys market index, discount in funds selling price, exchange rates and returns in targeted markets. For balanced mutual funds and pension funds, Ibbotson and Kaplan (2000) confirmed importance of asset allocation in determining variability of returns, while Peterson et al. (2002) found factors determining equity fund performance. As closed-end funds constituted a minority of the fund management industry (Jones 2003), factors from Anderson et al. (2001) were not used for this research. Asset allocation, according to Ibbotson and Kaplan (2000), can explain around 40 to 90 percent of fund return variation. This factor is therefore the most important determinant and first to be reviewed in this section. Peterson et al. (2002) classified performance determinants into those affecting funds pre-tax or post-tax returns. Factors affecting pre-tax performance include (1) fund expenses, (2) investment style, (3) past pre-tax performance, (4) risk and (5) turnover. Post-tax factors are important as after-tax fund returns are much less than before-tax returns for investors in high tax brackets. Examples of post-tax fund performance determinants are past pre-tax performance, expenses, risk, style, past tax efficiency and recent occurrence of large net redemption (Peterson et al. 2002). Comparison of these two groups of factors reveals risk, style, past pre-tax performance and expenses affecting performance before and after consideration of taxes. Still, fund performance studies focused mainly on pre-tax factors. While many funds are taxable, studies on pre-tax factors are appropriate for funds with non-taxable profits and interests.7 By studying non-taxable funds, this thesis focused on pre-tax determinants. A reason for not considering taxes is the complication arising from differing taxes for investors in different tax brackets. Besides, taxation laws pertaining mutual funds differ
7

Unit trusts approved for Singapores CPF Investment Scheme are examples of non-taxable funds (CPF Investment Scheme 2005).

49

Chapter 3 from country to country. For example, in the USA, while dividends and interest paid by a mutual fund were taxed as ordinary income for the shareholder, short-term and long-term capital gains were taxed at investors marginal tax rate and a 20 percent long-term capital gains rate respectively (Jones 2003). In Australia, income and capital gains derived by several types of superannuation funds were taxable at 15 percent (Veltman 2000). Therefore, the following discussion focuses on pre-tax factors. Research did not agree on factors affecting fund returns. For example, Peterson et al. (2002) did not consider assets under management as an important factor while Indro et al. (1999) reported fund size as a performance determinant. The following subsections review these factors, starting with asset allocation.

3.3.1 Asset allocation

Lacklustre fund performance (discussed in section 3.2.2) led some fund managers to represent asset classes using index funds, as index funds incurred less research and commission costs (Schramn 1998; Williams 1996). Such practices acknowledge asset allocation as a significant performance factor. Asset allocation is the most important factor determining fund performance. According to Brinson, Singer and Beebower (1991), asset allocation can determine at least 90 percent of variability in returns across time for a typical fund. Extending their findings, Ibbotson and Kaplan (2000) noted asset allocation policy explaining approximately 40 percent of variation in returns among funds and even 100 percent of returns from a fund on average. According to Ibbotson and Kaplan (2000), a funds total return can be decomposed into two components, asset allocation return and active return. While asset allocation return is attributable to the targeted market, active return is due to other factors. As this research considers Singapores domestic equity funds, asset allocation return corresponds to returns from the STI index. The following subsections identify factors affecting active returns. 50

Literature review

3.3.2 Investment style

Besides asset allocation, funds can be classified according to investment style. For example, funds are grouped into value, blend or growth funds according to average priceto-book (P/B) and price-to-earnings (P/E) ratios of stocks invested.8 Fama and French (1992) justified this classification after finding P/B and P/E ratios, together with market capitalization, explained cross-sectional variation in returns better than CAPMs beta. However, style classification is subjective and mostly self-defined by information providers (Sawicki 2000).9 This subjectivity was confirmed by DiBartolomeo and Witkowski (1997), who used multifactor decomposition of securities held by funds to show about 40 percent of 700 equity funds examined were classified wrongly. To improve upon previous classifications, Brown and Goetzmann (1997) developed a system based on style factors for predicting fund returns. Even though style classification is a rough representation of funds expected risk-return trade off, Malkiel (1995) found a relation between fund performance and classification. Bogle (1998) reported risk-adjusted returns of index funds, which were low-cost and passively managed, generally outperformed other fund groups examined, confirming actively managed funds generating less returns than their market indexes. Comparing performance of growth and value funds, Chan, Chen and Lakonishok (2002) found growth funds outperforming value funds on average after controlling for style. As a funds manager and investment style change over time, Peterson et al. (2002) considered a three-year timeframe and advised not using too long a time period. This
8

While growth funds invest in companies that have better than average prospects for earnings growth (stocks with high P/B and P/E ratios) and value funds focus on under-valued companies (stocks with low P/B and P/E ratios), blend funds fall between growth and value categories. Morningstar, a fund information provider in the USA, used this classification in the 1980s. In the 1990s, Morningstar added a market capitalization dimension for funds investing in small, medium or large-capitalization stocks, resulting in nine categories. 9 For example, in Australia, FPG Research classified funds according to type of assets held as well as their growth and income objectives (Sawicki 2000). In Singapore, Mercers classification used equity risk and focus risk dimensions, as described in the previous chapter. These two systems clearly differ from Morningstars.

51

Chapter 3 research is for a 5-year period from 1999 to 2004. As a precautionary measure, style classification data for this research were verified using fund sheets detailing investment objectives and asset allocation. As this research targets Singapores domestic equity funds, these funds were verified as belonging to Mercers high-risk narrowly focused country funds classification (CPF Investment Scheme Risk Classification System 2004). Besides, fund betas and standard deviations were calculated using returns data, as discussed in the next subsection.

3.3.3 Risk
According to Sharpes (1964) CAPM model, investors are not rewarded for bearing unsystematic risk, expecting high returns from securities with high systematic risk. This thesis considers fund beta, CAPMs measure of systematic risk, as a performance factor associated with market returns. In terms of existing literature comparing risk of bank and non-bank funds, Frye (2001) found bank funds were less risky then their non-bank counterparts in the USA. Besides beta, standard deviation measurements of unsystematic risk were tabulated for this research using past performance data. Past performance is examined in the following subsection.

3.3.4 Past performance and performance persistence

Whether past performance contains information about future performance is debatable as research in the 1990s reported fund performance persistence (Brown & Goetzmann 1995; Christopherson, Ferson & Glassman 1998; Elton, Gruber & Blake 1996; Grinblatt & Titman 1992; Hendricks, Patel & Zeckhauser 1993) but earlier studies indicated performance similar to random outcomes (Dunn & Theinsen 1983; Jensen 1969).

52

Literature review Therefore, market efficiency theory (reviewed in section 3.2.1) was consulted to reconcile the contradictory results. Efficient market theory (Fama 1970) modelled investment as a fair game and negated possibility of earning consistently superior returns. According to Famas (1970) weakform EMH, there is no relation between past and future returns as current market prices reflect all market information, including past returns. Jain and Wu (2000) confirmed this by noting no performance persistence for advertised funds with one year of good preadvertising performance even though they attracted 20 percent more investor money than non-advertised funds. Lack of performance persistence was also supported by McKeon (2001) who reported majority of award-winning funds declining substantially in value during the year after their award-winning performance, some even declining as much as 50 percent. Apparent contradiction between earlier and later studies can be explained as Jensen (1969) reported a lack of persistence in superior performance while Brown and Goetzmann (1995), Christopherson, Ferson and Glassman (1998) as well as Hendricks, Patel and Zeckhauser (1993) found persistence in poor performance. As for Elton, Gruber and Blake (1996) as well as Grinblatt and Titman (1992), even though they found past risk-adjusted performance predictive of future performance for as long as three years, their findings depended on sampling period. Supporting Famas (1970) weak-form EMH, this thesis rejects past performance as a determinant of future performance. Investors seeking quality funds cannot rely merely on performance record or research companies recommendation of approved funds (Sawicki & Thomson 2000). Even though past and future performances are independent, past performance influences flow of monies into funds, considered in the next subsection.

53

Chapter 3

3.3.5 Flow of funds and assets under management

Fund size, measured by amount of assets under management, changes with flow of monies into and out of the fund. According to Indro et al. (1999), fund size reflects implicit transaction costs of active investment strategieshuge transactions on market prices, opportunity cost of not trading, cost of being scrutinized by market participants and administrative stress due to investment style deviation for large funds. Studies examining relationship between performance and flow of funds revealed investor dollars were placed in funds with superior recent performance as investors looked for quality funds based on performance record (Guercio & Tkac 2002b; Ippolito 1992; Sawicki 2000; Sirri & Tufano 1998). While Ippolito (1992) as well as Sirri and Tufano (1998) found recent poor performance did not lead to outflows from retail funds in the USA, Sawicki (2000) reported investors moving monies out of poorly performing Australian wholesale funds.10 Supporting Sawicki (2000), Guercio and Tkac (2002a) found fund rating changes reflecting past performance influencing flow of money into and out of funds, but suggested asymmetry for a pension fund study (2002b) as clients withdrew from poorly performing funds but did not invest in best performing funds. Explanations for such investor inertia include switching costs (Ippolito 1992), lack of information on poorly performing funds for investors (Sirri & Tufano 1998), investors perception bias in past performance (Goetzman & Peles 1997) as well as disconnection of past and future performances by fund managers changing unsuccessful investment strategies (Lynch & Musto 1997). Sawicki (2001) suggested young funds that were small abandoned unsuccessful strategies for more successful ones to convince investors not to withdraw. In a later study, Sawicki and Finn (2002) found small funds were represented

10

In the mutual fund industry, retail and wholesale segments accept investments from individuals and institutions respectively, as explained in Chapter 2.

54

Literature review disproportionately among top performers but underrepresented among worst performers, indicating fund size may influence performance. Net assets under management can affect performance, as funds need to attain a minimum size to achieve returns net of research expenses and other costs. However, a large fund incurring excessive costs results in diminishing or even negative marginal returns. Initially, growth in fund size provides cost advantages, as brokerage costs for larger transactions are lower while research expenses increase less than proportionately with fund size. After exceeding an optimal size, too large a fund can lead to deviation from original objectives by investing with some lower quality assets, as well as increased administrative costs for additional coordination among staff to manage sub-funds (Indro et al. 1999). Flow of monies into funds with recent good performance is due to investors seeking maximum risk-adjusted returns using asymmetric information. After injection of monies, funds with good recent returns can hardly sustain their performance (Carlson 1970; Dunn & Theinsen 1983; Jensen 1969). As fund managers are compensated based on amount of assets under management, they are rewarded or penalized by clients based on recent performance. Even though past performance and flow of funds based on past performance may not be useful determinants of future performance, amount of assets under management may affect performance. Literature on relation between fund size and performance has mixed findings. Cicotello and Grant (1996), Droms and Walker (1994) as well as Grinblatt and Titman (1994) reported absence of such relation for funds in the USA. The relation was also absent in Australia (Bird, Chin & McCrae 1983; Gallagher 2003; Gallagher & Martin 2005) and Sweden (Dahlquist, Engstrom & Soderlind 2000). However, Indro et al. (1999) reported fund size as a performance determinant in the USA. In Singapore, fund size may be a performance determinant when larger funds achieve economies of scale compared to

55

Chapter 3 smaller funds. As Singapore is relatively small among developed equity markets, its domestic equity funds may not experience diminishing marginal returns with large fund size, which has implication for research on small fund markets. Funds need to attain a minimum size to earn returns net of transaction costs. The following section examines such costs.

3.3.6 Research and trading costs

Passively managed funds incurred lower costs and outperformed actively managed funds (Bogle 1998). Actively managed funds incur various costs, including operating and research expenses, which are measured by the expense ratio. According to ICI, a funds expense ratio is its cost of doing business, expressed as percentage of assets (ICI 2004). Indro et al. (1999) defined expense ratio as the proportion of assets paid for operating expenses and management fees, including administration fees and other costs, but excluding brokerage costs. Even though various costs are included in the ratio, most of the expenses can be associated with financial market research, as Indro et al. (1999) considered explicit cost of research to be reflected by the ratio, which is the price paid by uninformed investors to be informed. As active management incurring research expenses can be compensated by better returns from trading based on research information (Grossman & Stiglitz 1980), fund managers efficiently incurring research expenses may earn positive risk-adjusted returns net of expenses. Otherwise, inefficiency will lead to managers income (proportionate to amount of assets under management) being penalized by investors withdrawal from underperforming funds with excessive expenses. Research on fund expenses and performance reported conflicting results in the USA. While Sharpe (1966) noticed funds with higher reward-to-variability ratios incurring lower expenses generally, Friend et al. (1970) reported insignificant negative correlation between 56

Literature review risk-adjusted returns and expenses. Furthermore, Ippolito (1989) found risk-adjusted returns not related with expenses, while Berkowitz and Qiu (2003) confirmed importance of fee expenses as a fund performance determinant. For large equity markets, high research expenses can be justified by better fund performance with useful information on many investment choices available. However, such research does not lead automatically to superior returns in an efficient market (Fama 1970). For small markets, high research expenses may not be justified due to limited investment choices. As Singapores equity market is small, a passive investment strategy is justified when funds with higher expenses cannot outperform their counterparts with lower expenses. To summarize literature on fund size and expenses, mutual fund performance is related to efficiency of investment strategy, which is characterized by expense ratio and fund size measures of explicit and implicit costs respectively. Expense ratio and fund size are the performance determinants for research issue 3 (introduced in Figure 3.5 on page 48), while research issue 4, illustrated in Figure 3.6 on the following page, compares these determinants and beta for different types of FMCs. As shown in this figure, little literature existed for comparison of beta, expense ratio as well as size for bank and non-bank funds, except for Fryes (2001) observation of bank funds being less risky than non-bank ones. The next subsection considers type of FMC as the last performance determinant for this research.

3.3.7 Type of fund management company

Literature on performance comparison of different fund groups discussed differences in fund managers characteristics, including age, education (Chevalier & Ellison 1999) and gender (Atkinson, Baird & Frye 2003). While Chevalier and Ellison (1999) found

managers with higher level of education generating higher risk-adjusted returns, Atkinson, 57

Chapter 3 Baird and Frye (2003) found no significant difference in performance and risk characteristics among female and male fund managers, but lower asset flows for female managers funds. Figure 3.6 Derivation of research issue 4 on bank and non-bank fund characteristics
Research issue 4 Characteristics of bank and non-bank funds

H4.1 No significant difference in size of bank and non-bank funds

H4.3 No significant difference in expense ratio of bank and non-bank funds

H4.2 No significant difference in systematic risk of bank and non-bank funds

Opposed by Frye (2001)

Source: developed for this research. As for differences between FMCs, various researchers compared performance of bankmanaged funds with their non-bank counterparts (Bauman & Miller 1995; Bogle & Twardowski 1980; Frye 2001) while Berkowitz and Qiu (2003) compared performance of publicly traded FMCs with private counterparts. Berkowitz and Qiu (2003) reported public FMCs in Canada investing in riskier assets and charging higher management fees compared to private management companies, but risk-adjusted returns of funds managed by public companies did not outperform private counterparts. While research till the 1990s indicated underperformance of bank funds compared to nonbank counterparts (Bauman & Miller 1995; Bogle & Twardowski 1980), later research did not detect underperformance (Frye 2001). Frye (2001) suggested earlier research on underperformance of bank-managed funds relative to non-bank ones ignoring their differing fiduciary standards. In contrast to banks mostly focusing on short-term investments to avoid interest rate risk and maintain liquidity, investment firms have wider 58

Literature review variety of investment objectives (Reilly & Brown 2003, pp. 63-6), ranging from high-risk capital appreciation to low-risk money market income. Perceived underperformance of bank-managed funds relative to non-bank ones can be due to bank managers prudent avoidance of risks. Research issue 2 (Figure 3.4 performance of bank and non-bank funds. Table 3.1 determinants discussed in this section. Table 3.1 Selection of mutual fund performance determinants
Selected factor Asset allocation Supporting literature Brinson et al. (1991, 1996); Ibbotson and Kaplan (2000); Bogle (1998) Investment style Chan, Chen and Lakonishok (2002) Malkiel (1995) Systematic risk Sharpe (1964) Fund size Sawicki and Finn (2002); Indro et al. (1999) Research & trading costs Bogle (1998); Sharpe (1966); Berkowitz and Qiu (2003) Type of fund management company Bauman and Miller (1995); Bogle and Twardowski (1980) Rejected factor Reason Past performance Contradict efficient market theory (Fama 1970) Flow of funds Dependent on past performance, which is not considered a performance determinant

on page 45) evaluates relative summarizes fund performance

Source: developed for this research The table tabulates selected factors (asset allocation, investment style, systematic risk, fund size, research and trading costs as well as type of FMC) with their supporting literature. Other factors (past performance and flow of funds) are listed with reasons for not selecting them. A mutual fund performance model is developed in the next section using the selected factors.

3.4 Theoretical framework

Plausible factors to be incorporated in a fund performance conceptual model are: (1) asset allocation, (2) investment style, (3) systematic risk, (4) fund size, (5) research and trading costs as well as (6) type of FMC. Table 3.2 on the next page tabulates these variables and their corresponding measures for this research. In this table, names in brackets refer to data collection variables (listed in section 1.5.1, pp. 8-9).

59

Chapter 3 Table 3.2 Measurement of fund characteristics

Characteristic Performance Measurement Excess return (RET-RFR) as well as computed information, Jensen, Sharpe and Treynor ratios Asset allocation and investment style Targeted market return (STI) Net assets under management (SZE) Fund size Systematic risk Calculated beta Expense ratio (EXP) Research and trading costs Bank or non-bank (FMC) Institution type

Source: developed for this research. Fund performance is measured by returns in excess of risk-free rate, suggested by Fama (1972), as well as computed composite measures by Goodwin (1998), Jensen (1968), Sharpe (1966) and Treynor (1965). Contribution of asset allocation and investment style to performance is measured by targeted market returns while fund size is recorded by amount of net assets under management. Computed fund beta measures systematic risk while expense ratio measures research and trading costs.11 Lastly, type of institution managing the fund is either bank or non-bank. The following subsections build a conceptual model before elaborating on research issues and hypotheses.

3.4.1 Mutual fund performance models

Based on selected measurements of performance determinants from the previous section, an initial mutual fund performance model is derived in Figure 3.7 on the following page. In this model, fund characteristics and their measurements are shown in rectangles and ovals respectively. According to this model, type of FMC (whether bank or non-bank) can influence performance of funds via their systematic risk and transaction cost characteristics. In terms of transaction costs, FMCs can incur explicit and implicit costs to affect fund performance. This model is used for hypothesis testing in Chapter 4 and enhanced in Chapter 6 by incorporating conclusions from this research. More factors can

11

While trading costs can be more accurately measured using turnover ratio, amount of purchases or sales divided by average assets (Indro et al. 1999), turnover data are not available in the dataset collected.

60

Literature review be incorporated into this model, but factors presented here are believed to be major ones influencing fund performance. Figure 3.7 Conceptual model of mutual fund performance
Return Nonbank

Bank

Information ratio Jensen alpha Market return Sharpe ratio Treynor ratio

Financial institution

Risk-free return Asset allocation & investment style Systematic risk

Expected return

Fund performance

Beta

Size

Implicit cost Explicit cost

Transaction costs

Expense ratio

Source: developed for this research. With this conceptual model in place, regression analysis can be conducted using Sharpes (1964) CAPM model. To examine equity fund performance, quarterly fund returns were modelled using Equation 3.5 on the next page. Equation 3.5 allows each funds expected return to depend on three factors: (1) pure time value of money, (2) amount of systematic risk and (3) reward for bearing it.

61

Chapter 3 Equation 3.5 Single-index model for domestic equity fund returns RET ft = RFRt + f (STI t RFRt ) + ft where RETft = return from fund f at time t; f = Covf,STI / 2STI, covariance between returns for fund f and STI divided by variance of STI returns; RFRt = risk-free rate of return at time t; STIt = STI return at time t; and = error term for fund f at time t. ft Source: adapted from Sharpe (1964). First, time value of money was measured by risk-free rate RFRt, reward for not taking risk. Since quarterly returns for CPF-approved funds were used, RFRt corresponds to CPF accounts guaranteed interest rates. For CPF Ordinary and Special accounts, RFRt corresponds to RFR1 = 2.5% / 4 or 0.625 percent per quarter; and RFR2 = 4.0% / 4 or 1.0 percent per quarter respectively.12 Second, amount of systematic risk was measured by f for systematic risk present in fund f relative to systematic risk in average asset from Singapores stock market. Third, reward for bearing systematic risk was measured by STI risk premium (STIt RFRt) for average amount of systematic risk in Singapores stock market. The single-index model therefore measures behaviour of fund returns using beta, market risk premium and risk-free return. While there may be leads or lags in market returns relative to fund returns, use of quarterly data can negate these effects. Lagging quarterly data will imply a fund taking as long as three months to respond to changes in the market, which is not consistent with capital market efficiency (Fama 1970). Studies on Singapores stock market showed its efficiency strengthened with increases in time interval (Wong 1988). Even though Singapores market was not efficient using daily or weekly data (Lim 1985; Saw & Tan 1986), Ariff (1986) used monthly data to show the market was comparable to New York, London and Australian stock markets in adjusting

12

Information about the CPF Ordinary and Special accounts were presented in Chapter 2.

62

Literature review prices efficiently to reflect new information. Efficiency testing of Singapores stock market using more recent data at various time intervals can be carried out for further research. When using quarterly instead of monthly returns data, this research considers the Singapore market to be at least weak form efficient. Assuming weak-form efficiency implies past and future returns are independent (Fama 1970).

3.4.2 Research issues and propositions

Models developed in the previous sub-section were used to test research issues and hypotheses. Four research issues are identified, with corresponding hypotheses. Research issue 1: Performance of Singapore's domestic equity funds. Returns from CPF-approved domestic equity funds are compared with domestic market indexes and CPF interest rates. Hypothesis 1.1: Domestic equity funds do not outperform their benchmark stock market index. Wong (2004) reported majority of domestic funds outperformed their market indexes from 1999 to 2003 while Koh (1999) reported underperformance from 1976 to 1994. Hypothesis 1.2: There is positive correlation between domestic equity fund returns and returns to the domestic stock market index. High positive correlation was found between fund and market returns (Brinson, Hood & Beebower 1986; Brinson, Singer & Beebower 1991; Ibbotson & Kaplan 2000; Jensen 1968). If CPF-approved domestic equity funds could not outperform their market index, as stated in the previous hypothesis, and benchmark market returns determine as much as 90 percent of variability in returns for a typical fund across time, the majority of funds should not be managed actively. Instead, they should be managed passively as index funds to match market performance and minimize research and trading costs. Index funds duplicate

63

Chapter 3 composition and performance of targeted markets while mostly providing comparable or better performance than actively managed funds (Reilly & Brown 2003). Hypothesis 1.3: There is no significant difference between returns from investing savings in domestic equity funds and guaranteed interest rates. Assuming fund investors are rational, they will leave their monies to earn guaranteed interest rates if returns from unit trusts do not outperform guaranteed interest rates. Sirri and Tufano (1998) agreed with Gruber (1996) on fund selection decision being influenced by past performance. The large amount of monies invested in unit trusts and their growth may not support this hypothesis, as funds need to have good performance in order to attract investors. However, Holiday (1994) reasoned novice investors with little financial sophistication relied more on marketing than performance information. These disadvantaged investors fund purchasing decision may be based on advertisement or advice from brokers, which can be reason why poorly performing funds are still popular, as Gruber (1996) hypothesized. In Singapore, Koh (1999) reported majority of unit trusts earning returns lower than CPF interest rates. As objectives and constraints of banks and non-banks differ, it is reasonable to expect differences in their determinants of fund performance, which leads to the second research issue. Research issue 2: Relative performance of bank and non-bank domestic equity funds. Risk-adjusted performances of banks and non-bank funds are compared using composite portfolio performance measures. As alternative measures can rank performance differently, they provide insights otherwise not possible with one measure. Hypothesis 2.1: There is no significant difference in returns between domestic equity funds managed by banks and non-banks.

64

Literature review Hypothesis 2.2: There is no significant difference in information ratios between domestic equity funds managed by banks and non-banks. Hypothesis 2.3: There is no significant difference in Jensen alphas between domestic equity funds managed by banks and non-banks. Hypothesis 2.4: There is no significant difference in Sharpe ratios between domestic equity funds managed by banks and non-banks. Hypothesis 2.5: There is no significant difference in Treynor ratios between domestic equity funds managed by banks and non-banks. Previous research controlling for differing fiduciary standards of institution groups considered only bond funds. Examining Singapores CPF-approved funds allows for performance comparison of equity funds managed by banks and non-banks facing the same fiduciary standard. According to Berkowitz and Qiu (2003), technology usage in the fund management industry is quite homogenous across companies. This may lead to no significant performance difference between equity funds from banks and non-banks. The third research issue explores suitable fund performance determinants besides asset allocation and systematic risk. Research issue 3: Factors affecting domestic equity fund performance. Factors explaining fund performance may differ in terms of degree or importance for banks and non-banks. The following hypotheses explored importance of size and expenditures in affecting domestic equity fund performance. Hypothesis 3.1: There is no significant difference in returns between small and large funds. As Singapore is relatively small among developed equity markets, domestic equity funds may not suffer from diminishing marginal returns due to excessive fund size, as hinted in section 3.3.5.

65

Chapter 3 Hypothesis 3.2: There is no significant difference in returns between funds with high expense ratios and those with low expense ratios. Section 3.3.6 suggested high research expenses in small markets are wasteful with limited investment choices. Passive investment strategy is justified in Singapore when funds with higher research expenses cannot outperform their counterparts with lower expenses. Hypothesis 3.3: There is no significant difference in expense ratios of big and small funds. Large funds may have lower expense ratios than small funds when initial growth in size provides cost advantages in terms of brokerage costs and research expenses, as explained in section 3.3.5. The forth and last research issue compared performance determinants of funds managed by banks and non-banks. Research issue 4: Factors differentiating bank and non-bank fund performance. The following hypotheses focus on size, systematic risk and expenses as plausible determinants differentiating performance of bank and non-bank funds. Hypothesis 4.1: There is no significant difference in size between domestic equity funds managed by banks and non-banks. If there is no significant performance difference between bank and non-bank funds (research issue 2), both groups of funds will be equally popular among sophisticated retail investors, who contribute similar amounts of monies into these funds, resulting in no significant difference in size for both fund groups. However, bank funds may be more popular with unsophisticated investors. As they are not knowledgeable in financial planning, they conveniently choose asset management services of banks as they already have savings accounts with these familiar institutions. This is confirmed by Holiday (1994) who reported unsophisticated investors going for bank funds. Therefore, non-banks may

66

Literature review attract more sophisticated investors than banks. Nevertheless, net assets under management of both fund groups reflect generally their popularity among investors. Hypothesis 4.2: There is no significant difference in levels of systematic risk between domestic equity funds managed by banks and non-banks. If there is no significant performance difference between bank and non-bank funds, it may be related to their similar systematic risk level, as measured by CAPMs (Sharpe 1964) beta. However, McTague (1994) claimed portfolio managers in banks have a reputation for being risk-averse. Their investment style may be more conservative than their non-bank counterparts, resulting in less transaction costs, which leads to the next hypothesis. Hypothesis 4.3: There is no significant difference in expense ratios of bank and non-bank funds. If there were no differences between both groups of funds in expenditures, performance, size and risk, reputation of bank funds being under-performers compared to their non-bank counterparts is unjustified. Bank funds will be more competitive if they can shed their image as under-performers, leading to improved profitability. A summary of research issues and hypotheses are presented in Table 3.3 on the following page. These hypotheses are tested using methodology developed in the next chapter. Test results are presented in Chapter 5. Before proceeding to the methodology chapter, the following section summarizes fund performance determinants.

3.5 Conclusion
Mutual fund characteristics influencing performance are asset allocation, investment style, systematic risk, size, type of FMC as well as expenses. These determinants are useful as investors can observe them before making decisions. If indeed these factors are important, investors should take them into account when making fund decisions.

67

Chapter 3 Table 3.3 Research issues and hypothesis

Research issue Research hypothesis 1 Performance 1.1 Domestic equity funds do not outperform their benchmark stock market index of Singapore's domestic equity 1.2 There is positive correlation between domestic equity fund returns and returns to the domestic stock market index funds 1.3 There is no significant difference between returns from investing savings in domestic equity funds and guaranteed interest rates 2.1 There is no significant difference in returns between domestic equity funds 2 Relative managed by banks and non-banks performance of bank and non- 2.2 There is no significant difference in information ratios between domestic equity funds managed by banks and non-banks bank domestic 2.3 There is no significant difference in Jensen alphas between domestic equity equity funds funds managed by banks and non-banks 2.4 There is no significant difference in Sharpe ratios between domestic equity funds managed by banks and non-banks 2.5 There is no significant difference in Treynor ratios between domestic equity funds managed by banks and non-banks 3.1 There is no significant difference in returns between small and large funds 3 Factors 3.2 There is no significant difference in returns between funds with high affecting expense ratios and those with low expense ratios domestic equity 3.3 There is no significant difference in expense ratios of big and small funds fund performance 4.1 There is no significant difference in size between domestic equity funds 4 Factors managed by banks and non-banks differentiating bank and non- 4.2 There is no significant difference in levels of systematic risk between domestic equity funds managed by banks and non-banks bank fund 4.3 There is no significant difference in expense ratios of bank and non-bank performance funds

Source: developed for this research. Regulatory restrictions lead to investment style differences in banks and non-banks. As funds managed by investment firms have greater risk than their bank counterparts, besides increasing clients investments risk, they also increase fund managers income risk, as income is based on recent performance. As fund performance improves, investors increase their holdings of the fund, enlarging fund size and managers remuneration, since remuneration is proportionate to fund size. However, investing in riskier assets by investment firms require greater research effort from portfolio managers compared to their bank counterparts. Therefore, investment firms have to incur more research expenses and other costs, due to greater risk and effort required. Consequently, investment firms may charge higher management fees than banks. Higher fees may lower risk-adjusted returns. If investment firms are not able to provide

68

Literature review better returns to compensate investors for more expensive fees, investors will avoid funds managed by investment firms. If funds managed by investment firms do not achieve acceptable performance but charge higher fees than their bank counterparts, likely explanation for investors not withdrawing monies will be withdrawal cost or ignorance of fees and risk-adjusted returns.

69

Chapter 4 Research methodology

4.1 Introduction
Following Chapter 3s review of mutual fund literature to build a theoretical framework for this research, this chapter derives methodology for hypothesis testing. A methodology map outlining sections of this chapter is presented in Figure 4.1. Figure 4.1 Methodology map
4 Research methodology

4.1 Introduction 4.2.1 Case research 4.2.2 Survey research 4.2.3 Secondary data research 4.2.4 Justification for research using secondary data 4.3.1 Hypothesis testing 4.3.2 Non-causal investigation 4.3.3 Minimal research interference 4.3.4 Non-contrived setting 4.3.5 Fund management institutional group as unit of analysis 4.3.6 Longitudinal time horizon 4.4.1 Data collection methods 4.4.2 Downloading of financial data from online sources 4.5.1 Regression analysis 4.5.2 Hypothesis testing 4.6.1 Internal validity 4.6.2 External validity 4.7.1 General ethical issues 4.7.2 Specific ethical issues

4.2 Research approaches

4.3 Research design

4.4 Data collection

4.5 Data analysis 4.6 Research quality 4.7 Ethical considerations 4.8 Conclusion

Source: developed for this research. Following this introduction, three research approaches are suggested in section 4.2 before providing justification for best approach. Section 4.3 designs research around selected

70

Research methodology approach before section 4.4 describes data collection. To analyse data, section 4.5 explains statistical methods applied in this research. For assessing research quality, section 4.6 presents a framework based on internal and external validity, while section 4.7 discusses ethical issues relevant to financial research. Lastly, section 4.8 concludes by summarizing methodology used for producing findings.

4.2 Research approaches

Three alternative approaches are proposed in the following subsections: research using (1) case study, (2) survey or (3) secondary data. This presentation order corresponds to exploratory, descriptive and hypothesis testing stages of knowledge advancement. After presenting these methods, the last subsection provides justification for secondary data research as the approach applied in this study. Even though case and survey methods were not used for this research, discussion of alternative methods demonstrates applicability of qualitative and quantitative approaches for finance research. Among these three approaches, case research explores factors leading to superior performance of unit trusts managed by banks or non-banks.1 By presenting a financial context, usage of case research in a quantitative discipline is demonstrated, showing similarity between Yins (1994) most commonly used criteria for judging case research quality and realism-based criteria from Healy and Perry (2000). Survey research describes actual performance of fund managers in Singapore, while secondary data research offers better understanding of relationships among variables explaining fund manager performance. Differentiation between realism and post-positivism paradigms based on qualitative or quantitative methodology emphasis is suggested, even though these two

The research problem driving this case research can be expressed as How and why do unit trusts managed by banks (or non-banks) in Singapore outperform their counterparts?

71

Chapter 4 paradigms were formally considered the same (Healy & Perry 2000; Perry, Riege & Brown 1999). For each approach, its paradigm, data collection method, quality issues and research outcome are examined, with corresponding paradigm selected from positivism, postpositivism (or realism), critical theory and constructivism worldviews (Guba & Lincoln 1994). For paradigm selected, its ontology (reality to be investigated by researcher), epistemology (relationship between researcher and reality) and methodology (researchers technique for investigating reality) elements (Healy & Perry 2000) are presented in the context of this research. Details of data collection are explained, followed by discussion of quality tests for assessing reliability and validity, before presenting study outcomes for answering the research problem. To effectively control for differing fiduciary standards of financial institutions, only CPF-approved unit trusts are considered for case, survey and secondary data research.

4.2.1 Case research

Even though case study method was often used in marketing research (Perry 2001), it is suitable for consulting (Yin 1994) and evaluation research (Patton 1990). Case research may be suitable at the exploratory stage for a consulting firms project to study fund performance. Realism paradigm Case research is carried out within the realism paradigm. As realisms ontology concerns external reality that is probabilistically apprehended, triangulation from various sources of evidence is used to create knowledge that is probably true (Perry, Riege & Brown 1999) about fund manager performance. Using case research for triangulation, the researcher has to be objective in producing findings, since epistemology of realism considers researcher who is gathering facts from interviews as part of the research process. 72

Research methodology Data collection for case research Since case research is based on interviews, a general starting question may be used for fund managers: What is the story of your experiences in outperforming your counterparts in Singapores unit trust industry? Besides, probe questions can be prepared to ensure the interviewee will mention the research issues during the unstructured interview. For example, How do fund managers in your company outperform their competitors? To summarize the interviewees perception towards issues addressed, Likert-scale questions may be included in the interviewing guide (Yin 1994). For a doctoral research project to be credible, about 35 interviews will be conducted (Perry 1998) in 12 financial institutions, ranging from local banks to foreign investment firms for theoretical replication, with each company providing two fund managers and one research analyst as interviewees for literal replication. The institutions can be selected from a list of CPF-approved FMCs (CPF Investment Scheme 2005) and comprise of 3 local banks, 3 foreign banks, 3 local investment firms and 3 foreign investment firms. Quality of case research To judge validity and reliability, Yins (1994) four quality tests of construct, internal and external validity as well as reliability are adopted, as they were the most commonly used criteria for establishing case research quality (Perry 2001). Among these criteria, test for internal validity is concerned with soundness of cause-and-effect relationship but is not applicable to this exploratory study, which is not concerned with making causal statements. The following paragraphs discuss issues relating to the remaining three tests before examining a more recent set of criteria from Healy and Perry (2000) to highlight their similarities with Yins (1994).

73

Chapter 4 Construct validity of case research. For this research, the type of change to be examined is performance of fund managers from 1999 to 2004. The selected measure of change may be number of award-winning unit trusts from fund managers. To ensure selection of the correct operational measure, triangulation of multiple sources of evidence is used to encourage convergent validity during data collection. Specifically, interviews with fund managers and investment analysts from different types of FMCs are conducted. Secondary financial data collected from investment firms web sites are useful for validating fund managers performance claims. Besides, fund management awards given by reputable organizations are considered.2 Lastly, a research analyst from a financial institution can review the draft case study report. External validity of case research. As generalization of findings from a study is dependent on careful selection of interviewees and cases, only experienced fund managers and investment analysts are chosen as interviewees. Selecting foreign as well as local banks and investment firms achieves literal and theoretical replication in choice of cases. Appropriate choice of cases and interviewees will ensure sufficient external validity for analytical generalization of case research (Perry 2001). Reliability of case research. By providing a summary of the case study database and using a protocol to document procedures used for case selection and data analysis, the study can be audited or repeated by another researcher to arrive at equivalent results. As this study is built upon the realism paradigm, the more recent criteria derived by Healy and Perry (2000) may be used to judge research quality. Consideration of their six criteria reveals similarities to those mentioned above: (1) ontological appropriateness as the study dealt with a complex financial phenomenon mostly unknown and whose research problem is expressible in a how and why form (as footnoted at beginning of this section); (2)

Examples include S&P Singapore Investment Funds Performance Awards.

74

Research methodology contingent validity being satisfied by literal and theoretical replication, similar to external validity criterion previously mentioned; (3) epistemological criterions emphasis on multiple perceptions from participants and peer researchers, which parallels Yins (1994) construct validity; (4) methodological trustworthiness using case study database and protocol for fulfilling Yins (1994) reliability criteria; (5) analytic generalization being the outcome of external validity test above mentioned; and (6) construct validity reinforcing Yins (1994) construct validity with prior literature review of relevant finance research. Outcome of case research As the primary purpose of this exploratory research is to build a theory about relative performance of funds managed by banks and their competitors, rather than testing the theorys applicability in various countries, the final outcome is the theory, to be confirmed by cases collected for Singapore. This theory can be statistically tested using survey data collected from other countries to provide an international outcome.

4.2.2 Survey research

Surveys provide quick and inexpensive means to measure unit trust performance from a representative sample of fund managers. The peer group comparison method, previously the most common approach for evaluating fund managers (Kritzman 1990), surveys a representative sample of fund managers to collect fund returns data for the past ten years. Choice of ten years was recommended by Singer (1996) to extend a number of years covering at least one market cycle for identifying performance differences during rising and declining market conditions (Kon 1983). However, for majority of funds with history less than ten years, performance can be evaluated for shorter time periods. To facilitate performance comparison, the sample is divided into percentiles to present a box plot for fund managers relative rankings.

75

Chapter 4 Positivism paradigm Surveys fall within the positivism paradigm, whose ontology perceives discovery of an apprehensible reality (Guba & Lincoln 1994) about performance of funds managed by various institutions. In this paradigm, epistemology views researcher separate from survey. Therefore, value-free findings are generalized to entire population of fund managers. Survey research is used to verify hypotheses about fund manager performance. Data collection for survey research Mail survey sending questions to CPF-approved FMCs can ask each company to select at least three fund managers to provide answers to the following questions: Which unit trust did you manage for your company? and What is the unit trusts annual rate of return for the past ten years? Stratified sampling will group fund managers according to companies they belong to, taking a sample from each company. With about 33 companies in Singapore, each providing at least 3 survey participants, sample size required of 96 participants is obtainable.3 To support returns claimed by fund managers, each manager will be asked to attach existing publications reporting fund performance. Quality of survey research Assessing quality of survey measurements involves evaluating their reliability and validity (Zikmund 2000, p. 280). Reliability of survey research. Ensuring error-free measurements requires assessment of response consistency. Among methods available to assess reliability, the test-retest approach is the most popular (Burns & Bush 2000). Survey is repeated with the same respondents three months later, after release of next quarters fund performance. Consistency of data from consecutive surveys indicates high level of reliability.

To calculate sample size at 95% confidence level with 10% accuracy and expecting great variability (50%), sample size = (1.962)(50x50)/102 96.

76

Research methodology Validity of survey research. To assess surveys ability to measure fund manager performance, face validity of each survey question is evaluated and the question revised until the researcher agrees it helps to measure performance. As this method is subjective, a research analyst critiques the questions to strengthen them before considering convergent validity. Convergent validity is assessed by comparing returns claimed by participants with those stated in publications from participating companies and online investment databases. Outcome of survey research The outcome of survey research is a performance description of fund managers from banks as well as investment firms, which is used to confirm previous research on their relative performance. For a more comprehensive study of factors influencing relative performance, secondary data research is considered.

4.2.3 Secondary data research

Using secondary data is appropriate for a fund performance study, as financial data are readily available from various institutions in printed publications and online databases. Post-positivism paradigm Even though realism and post-positivism paradigms were considered the same (Healy & Perry 2000; Perry, Riege & Brown 1999), this subsection differentiates between them. Post-positivism is transformed from positivism (Guba & Lincoln 1994) and inherits positivisms quantitative roots. In fact, post-positivism has similar characteristics as positivism in its inquiry, knowledge accumulation, quality criteria, values, ethics, inquirers voice, accommodation and hegemony (Guba & Lincoln 1994). However, post-positivisms nature of knowledge is not positivisms establishment of facts from verified hypotheses, but realists probable facts derived from non-falsified hypotheses. Even though the most appropriate training for post-positivist inquirers is qualitative methods in addition to positivisms quantitative approach, a principally 77

Chapter 4 quantitative flavour is retained. This flavour of post-positivism is adopted for secondary data research on fund manager performance, which is only understood probabilistically by performing statistical analysis of financial data. Realism paradigm is differentiated by a more significant contribution of qualitative methodology, which was demonstrated for case research with interviews. Secondary data collection On December 2004, there were some 300 CPF-approved unit trusts in Singapore, managed by more than 30 FMCs (S&P 2003-2004). Quarterly returns, expense ratios and size data for all CPF-approved unit trusts as well as levels of benchmark indexes from 1999 were downloaded from the CPF Board web site and other web-based data sources to perform a five-year quantitative analysis.4 Additional fund data were available from quarterly and half-yearly performance reports published by FMCs. Analysis involved classification of funds according to risk characteristics, regression of unit trust returns on market returns using Sharpes (1964) CAPM model, computing riskadjusted performance measures developed by Goodwin (1998), Jensen (1968), Sharpe (1966) and Treynor (1965) as well as hypothesis testing of fund characteristics using twotail pooled-variance t-tests. Quality of secondary data research Appropriate criteria for judging quality of post-positivist inquiry are internal and external validity, reliability as well as objectivity (Guba & Lincoln 1994). Using Sharpes (1964) single-index model substantiated internal validity of this research, giving credibility to return relationships between unit trust and market index variables. For external validity, comparing results with similar studies conducted for foreign fund industries assesses

Performance data of CPF-approved unit trusts were taken from quarterly Performance and Risk Monitoring Reports downloaded from the CPF Board web site at www.cpf.gov.sg. Reports were only available from 1999 after liberalization of CPF rules governing unit trust investments. Another web-based data source is Yahoo Finance at finance.yahoo.com.

78

Research methodology generalization of this research to its target population of fund managers in general. Reliability of measurements from performance reports was audited by comparing a systematic sample of performance data with those from other publications and online sources. Finally, objectivity of study is assured, as the researcher was a neutral observer of the fund industry, not related professionally to any financial institution. Secondary data research outcome By conducting secondary data research, extensive financial data were gathered to explore factors influencing performance of unit trusts from banks, insurance and investment companies. Besides returns, fund characteristics such as expenditures, risk and size gave a detailed picture of fund performance, which is more effective than survey research.

4.2.4 Justification for research using secondary data

Previous subsections demonstrated alternative research approaches using case study, survey and secondary data to study unit trust performance, as summarized in Table 4.1 below. Action research, with its associated critical theory paradigm for transforming values, is considered inappropriate in this subsection, which examines suitability of each approach to identify the best method. Table 4.1 Summary of plausible approaches for financial research
Case research Research approach Stage of Exploratory knowledge advancement Paradigm Realism Data collection Interviews Research quality Research outcome Survey research Descriptive Research using secondary data Hypothesis testing

Positivism Mail survey

Post-positivism

Online downloading of data Construct and external Test-retest for reliability; face Internal and external validity; reliability and convergent validity validity; reliability and objectivity Theory about relative Description of fund manager Factors influencing performance of funds performance from local and performance of unit managed by banks versus foreign banks as well as trusts from banks and their competitors investment firms non-banks

Source: developed for this research.

79

Chapter 4 Morgan and Smircich (1980, p. 491) suggested suitability of approach was dependent on nature of phenomena being researched. The amount of quantitative material in the finance discipline encourages positivist research using secondary data. Therefore, majority of finance research focuses on measurable financial data, which were independent of their researchers. This independent relationship between researchers and financial data disallowed learning about the phenomena by being involved in changing it, rendering action research unsuitable for this study.5 Indeed, Susman and Evered (1978) indicated action research being inappropriate from a positivist viewpoint. However, it became evident in the 1990s positivist ideal was not always appropriate in financial research. Non-positivist approaches existed for financial research (EasterbySmith, Thrope & Lowe 1991, p. 42). For example, after interpreting economic indicators and company performance, investors psychology can affect security prices. This section demonstrated positivist approach using survey, realist approach using case study and postpositivist approach using secondary data for financial research. Therefore, nature of research question determines methodology used (Brownell & Trotman 1988). Using a variety of suitable methods results in richer understanding of the research problem. The peer group comparison survey approach is problematic for several reasons (Reilly & Brown 2003). Box plots produced from the study do not consider each fund managers portfolio risk level and assume portfolios have the same volatility. Besides, there is difficulty forming a peer group large enough to make rankings valid. Conducting a census for all fund managers is too costly, while letting FMCs select survey participants may reveal results from good performing fund managers only. Most importantly, a quantitative returns comparison of fund managers does not consider differing investment objectives and constraints, even though such qualitative information can be recorded in the survey.
5

In fact, it is unethical to manipulate variables affecting fund performance, changing investors profitability outcomes.

80

Research methodology For this research, evaluation of mutual fund performance was carried out using secondary data. Secondary data research can be enlightened by theories built from case research. For more complete understanding from various perspectives, this thesis recommends future research projects using alternative approaches, including case and survey research.

4.3 Research design

Design for secondary data research is formulated in this section by finalizing the purpose, type of investigation, researcher interference level, study setting, unit of analysis and time horizon.

4.3.1 Hypothesis testing

As research does not always go through exploratory, descriptive and hypothesis testing stages, the purpose of this study is identified. This study is not in the exploratory stage, as it was not undertaken to gain background information about fund performance. Neither is this study in the descriptive stage as it involved more than measuring unit trust performance. This research is in the hypothesis testing stage because it intended to understand relationships between type of financial institution, fund characteristics and their performance.

4.3.2 Non-causal investigation

This investigation is non-causal, as it did not try to find direct cause-and-effect relationship between institution groups and their fund performance. There is difficulty in understanding the relationship causally, as the four conditions of covariation, time sequence, systematic elimination and experimental design must be present before causality is ascertained (Burns & Bush 2000, pp. 139-40). However, a correlation can be found between institution groups and their fund characteristics.

81

Chapter 4

4.3.3 Minimal researcher interference

Extent of researcher interference for this study is minimal as only past financial data were collected. Ethically, the position is unacceptable if variables affecting fund performance can be manipulated in this study, controlling investments profitability outcomes, as elaborated in section 4.7.

4.3.4 Non-contrived setting

Conducting experiments by simulation or field study is not appropriate, as there is no determination of causal relationship in this study. A non-contrived setting is therefore sufficient for collecting secondary financial data.

4.3.5 Fund management institutional group as unit of analysis

Even though performance of each unit trust was recorded, the funds were aggregated to infer performance of banks and non-banks. Therefore, the unit of analysis is each institution group.

4.3.6 Longitudinal time horizon

Conducting a cross-sectional study of funds is not sufficient as it only measures their fluctuating performance at one point in time. Instead, longitudinal time horizon over a five-year period is required to identify long-term trends. To summarize, research design involved evaluating performance of funds managed by banks and non-banks using a non-causal study testing hypotheses about the link between type of FMC, fund characteristics and fund performance. There was no researcher interference in this non-contrived study, which was conducted for a longitudinal time horizon of five years, with each institution group as main unit of analysis. Following secondary data research design, data collection methods are considered.

82

Research methodology

4.4 Data collection

In this section, various data collection methods are presented to identify the most appropriate method for this project.

4.4.1 Data collection methods

Data collection methods can be classified as primary or secondary. To collect primary data, qualitative and quantitative research approaches are possible. While quantitative research involves structured questionnaire surveys on a large sample, qualitative research incorporates observation and asking a small group of people open-ended questions (Burns & Bush 2000). Examples of common qualitative methods include focus groups, interviews and observation. These methods are not used for this research, as they are more costly and less reliable than using secondary data. Using secondary data is popular in finance as quantitative analysis is performed on audited data freely available from companies and institutions.

4.4.2 Downloading of financial data from online sources

To carry out this research, five years of quarterly data from 1999 to 2004 for 19 retail funds approved for Singapores CPF Investment Scheme were downloaded from the CPF Board web site. These funds were invested in shares from the Singapore Stock Exchange. Table 4.2 on the next page identifies funds used for this research. Quarterly data collected were fund returns (RET), risk classifications (RSK), expense ratios (EXP), sizes (SZE) and management organization type (FMC) as well as STI returns (STI) and guaranteed interest rates (RFR).6 The information provider computed returns and sizes as percentages and market value of net assets under management respectively, while expense ratio data

Variable names in brackets refer to data items listed in section 1.5.1 (pp. 8-9).

83

Chapter 4 followed Investment Management Association of Singapore (IMAS) guidelines (S&P 2003-2004).7 Table 4.2 Research sample
Organization type Bank Fund DBS Horizon Singapore Equity Fund DBS Shenton Thrift Fund OCBC Savers Singapore Trust Fund OUB Union Singapore Equity Fund UOB Optimix Singapore Equity Fund UOB Unifund UOB United Growth Fund AXA Life-Fortress Fund AXA Life-Fortress Fund A GE Greatlink Singapore Equities Fund Keppel Managed Fund NTUC Income Singapore Equity Fund OUB Manulife Golden Singapore Growth Fund UOB Life FOF-Unifund UOB Life FOF-United Growth Fund UOB LifeLink Growth Fund Aberdeen Singapore Equity Fund CMG First State Singapore Growth Fund Schroder Singapore Trust Symbol 1999- 20022002 2004 ! ! DHSE ! ! DST ! ! OSST ! OUSE ! ! UOSE ! UU ! ! UUG ! ! ALF ! ALFA ! GGSE ! KM ! NISE ! OMGSG ! ULFU ! ! ULFUG ! ULG ! ! ASE ! CFSSG ! ! SST

Insurance company

Investment company

Note: funds with less than three quarterly periods of complete data were excluded, as they were insufficient for performing regression analysis. Source: funds identified from Mercer (1999-2002) and S&P (2003-2004). For this research, only CPF-approved funds were considered as they followed the same fiduciary standard for managing social security savings, so as to control for differing fiduciary standards. Failure to control for such standards leads to biased test results, as highlighted by Frye (2001). Among these funds, those investing only in the local stock market were selected. As each benchmark has a unique market cycle, funds based on benchmarks other than the STI were excluded. These CPF-approved domestic equity funds were classified according to type of organization managing the fund: (1) ILPs managed by insurance companies; (2) unit trusts
7

According to IMAS guidelines, expense ratio was computed as operating costs (including but not limited to administration fee, amortized expenses, audit fees, custodian and depository fees, goods and services tax on expenses, management fee, printing and distribution costs, registrar fees and trustee fee) expressed as percentage of funds average net assets for given time period. IMAS guidelines required the ratio to be calculated by taking average of annualised expense ratios for two previous six-month periods. For feeder funds, the guidelines required the ratio to be calculated as sum of annualised expense ratios of Singapore feeder fund and parent fund to facilitate comparability with direct investment funds (S&P 2003-2004).

84

Research methodology managed by investment firms; or (3) bank-managed unit trusts. Each type of organization differs in terms of operational structure, priorities and benefits for fund managers, which may influence portfolio returns (Bauman & Miller 1995). To link data variables with the fund performance conceptual model introduced in the previous chapter (Figure 3.7 on page 61), Figure 4.2 labels the model with variable names in brackets. Figure 4.2 Fund performance model with variables and hypotheses labelled
H2.1 Return (RET) Information ratio Jensen alpha Sharpe ratio Treynor ratio

Bank

Nonbank H2

H2.2

Financial institution (FMC) Market return (STI) Risk-free return (RFR) Asset allocation & investment style (RSK) H4.2 Systematic risk Size (SZE) Implicit cost H3.3 Explicit cost

H2.3

H2.4 H1.1 & H1.2 H2.5 H1.3

H4

Expected return

H1 Fund performance

Beta

H3.1 Transaction costs H3.2 H3

H4.1

H4.3

Expense ratio (EXP)

Source: developed for this research.

85

Chapter 4 In this figure, measurements with no variable names labelled were computed using other variables. For example, beta and composite performance measures were calculated using quarterly returns. This figure also shows linkage between the conceptual model and hypotheses listed in the previous chapter (Table 3.3 on page 68). Hypothesis testing is discussed in the following section.

4.5 Data analysis

After data collection, besides computing composite performance measures reviewed in the previous chapter, regression analysis and hypothesis testing were conducted, as explained in the following subsections.

4.5.1 Regression analysis

Equation 3.5 is repeated in Equation 4.1 below to show the linear regression model based on CAPM (Sharpe 1964). Equation 4.1 Linear regression model for domestic equity fund returns RET ft = RFRt + f (STI t RFRt ) + ft where RETft = return from fund f at time t; f = Covf,STI / 2STI, covariance between returns for fund f and STI divided by variance of STI returns; RFRt = risk-free rate of return at time t; STIt = STI return at time t; and = error term for fund f at time t. ft Source: adapted from Sharpe (1964). For each fund, linear regression was performed using its quarterly risk premium (RETft RFRt) as dependent variable and STI quarterly risk premium (STIt RFRt) as independent variable. There were two risk-free rates: RFR1 = 0.625 percent per quarter and RFR2 = 1 percent per quarter for CPF Ordinary and Special accounts respectively; as well as two holding periods for data collected from Mercer (1999-2002) and S&P (2003-2004). Therefore, four sets of linear regression were performed for each fund using a combination

86

Research methodology of risk-free rate and holding period. Hypothesis testing followed regression analysis and computation of risk-adjusted performance measures.

4.5.2 Hypothesis testing

To compare characteristics of two groups of funds, the usual two-tail pooled-variance t-test for difference in two means was conducted with five percent significance level during each of the time periods 1999-2002 and 2003-2004. The hypothesis of no difference between an average characteristic of one group of funds (1) and another group of funds (2) was tested by the following t value: Equation 4.2 t-statistic for comparing a characteristic for two groups of funds t= X 1 X 2 ( 1 2 ) 1 2 1 Sp n + n 2 1

where S and 2 Sp = pooled variance;

2 P

( n1 1)S12 + (n2 1)S 22 = (n1 1) + (n2 1)

X 1 = mean characteristic of sample taken from first group of funds; = variance of characteristic for sample taken from first group of funds; S12 n1 = sample size of first group of funds; X 2 = mean characteristic of sample taken from second group of funds; = variance of characteristic for sample taken from second group of funds; and S 22 n2 = sample size of second group of funds. Source: adapted from Levine et al. (2002, p. 375) The t-statistic was compared with critical t value t( = .05, df = n1 + n2 2) to determine whether difference between means was significantly not equal to zero. A t value greater than t( = .05, df = n1 + n2 2) leads to rejection of hypothesis. This test was used for research issues two to four. For the second and fourth issues, the two groups of funds are bank and non-bank funds. For the third issue, funds are sorted according to size or expense

87

Chapter 4 ratio in descending order, with the top and bottom half being large and small funds or high and low expense ratio funds. Regression analysis and hypothesis testing were performed without compromising research quality by ensuring statistical assumptions in section 1.6.1 (page 12) were not violated.

4.6 Research quality

Even though quality criteria for secondary data research developed by Guba and Lincoln (1994) are primarily for qualitative research, section 4.2.3 showed they are applicable to quantitative research. This section presents a more specialized framework based on internal and external validity for assessing statistical studies, suggested by Stock and Watson (2003, pp. 241-54). Figure 4.3 illustrates this framework. Figure 4.3 Framework for assessing statistical studies
Omitted variable Functional form misspecification Unbiased and consistent coefficient estimator Errors in variables Sample selection Internal validity Simultaneous causality Consistent standard errors Differences in populations Differences in settings Heteroskedasticity Serial correlation

Validity

External validity

Source: adapted from Stock and Watson (2003, pp. 242-54). This framework is used in the following subsections to assess internal and external validity of statistical analysis conducted for this research.

88

Research methodology

4.6.1 Internal validity

Stock and Watson (2003, p. 245) argued a statistical study is internally valid if (1) estimated coefficients are unbiased and consistent; and (2) standard errors of coefficients yield confidence intervals with desired confidence level. Therefore, similar results for different fund markets do not ensure their internal validity. Five potential threats to internal validity that induce bias in estimated effects of fund characteristics on performance are (1) omitted variables, (2) functional form misspecification, (3) errors in variables, (4) sample selection, and (5) simultaneous causality. Each of these five sources of bias can lead to correlation between regressor and error term during regression, violating the first least squares assumption in section 1.6.1 for expected zero value of error term, given a set of regressor values. Omitted variable bias Possible omitted variables remained, such as non-quantifiable characteristics of funds and their management companies, which may cause omitted variable bias. For example, if type of FMC was correlated with quality of portfolio managers (due to better managers being attracted to investment firms offering more risky funds) and if manager quality affected fund performance, then omission of manager quality may bias the effect of institution type on performance. However, this bias on institution type should be minimal, as the previous chapters literature review reported no superior institution group after controlling for differing fiduciary responsibilities (Frye 2001). One way to eliminate omitted variable bias, at least in theory, is to conduct an experiment. For example, funds are randomly assigned to different management companies, and their subsequent performances compared. Such a study is plausible in a simulated fund investment laboratory, but was not reported in the literature.

89

Chapter 4 Functional form misspecification For this research, analysis explored a linear functional form based on Sharpes (1964) CAPM model. If findings were insensitive to non-linear regression specifications, nonlinearity will not be statistically significant to substantially affect results. Non-linear regression can be carried out for future research. Errors in variables Transcription errors during data entry occur at times. Cross checking of data by research assistants were carried out to correct erroneous data. Sample selection As data collection covered all domestic equity funds invested with CPF money, effectively controlling for differing fiduciary responsibilities and market indexes, there was no reason to believe sample selection was problematic. Simultaneous causality Simultaneous causality can arise when fund performance affected amount of net assets under management, when investors withdrew from poorly performing funds to place monies in better performing funds. However, as explained in section 2.2.6, costs incurred when switching among funds discouraged such simultaneous causality. Inconsistent standard errors Heteroskedastic regression error results in computation of unreliable standard errors for hypothesis testing when using homoskedasticity-only formulas. Besides, correlation of error terms across observations threatens consistency of standard errors because simple random sampling was not used. Therefore, all reported results used formulas for computing standard errors robust to both heteroskedasticity and serial correlation. Besides these general factors assessing internal validity of statistical studies, survivorship bias is considered for fund performance studies.

90

Research methodology Survivorship bias Survival bias affected many studies working on performance data for funds that were still available. As funds that did not survive were usually the worst performing ones, not considering evidence from non-survivors results in overstating average performance for each fund group. In a study of fund survivorship, Carhart et al. (2002) demonstrated effect of a multi-period survival criterion for a non-surviving fund with past performance less than some threshold for a specific number of periods. This criterion interacted with survivor conditioning imposed on sample data to generate survivor bias in test statistics. Two forms of survivor conditionings were identified in the study: (1) end-of-sample conditioning that included only funds existing at the end of a sample period; and (2) look-ahead conditioning requiring funds to survive minimum period of time after a specific date. Reviewing empirical studies on mutual funds showed these two types of conditioning being used by many researchers. Besides, empirical testing by Carhart et al. (2002) reported survival criterion can cause annual bias to rise at a declining rate with sample period, approximately ranging from 0.07 percent for a one-year sample period to 1 percent for timeframe longer than 15 years. Survivor bias can therefore affect validity of relationships researched for this thesis pertaining fund performance and characteristics. According to Carhart et al. (2002), for a survivor-only sample, cross-sectional relationship can be biased if a fund characteristic was related to survivor bias in performance. Even with a fund dataset that controlled for survival bias, there may still be a small degree of survivor conditioning due to errors and missing data (Elton, Gruber & Blake 2001). Derived relationship between performance and fund characteristics is affected by using a survivor-only sample as Carhart et al. (2002) demonstrated large magnitudes of bias in slope coefficients for fund size, expenses,

91

Chapter 4 turnover and load fees, in comparison to a sample with survivors and non-survivors, which showed significant negative relationship between performance and expenses, marginal negative relationship with load fees and no relation with fund size and turnover. Collecting data for a survivor-only sample therefore requires characterizing of survivorship bias based on survival criterion and sample period. Assuming better performing funds survive, data for non-surviving funds being unavailable may result in overstating average performance of some institution groups. Specifically, if non-surviving funds were mostly non-bank funds, results of this research may be biased against bank funds outperforming their non-bank counterparts. Alternatively, if banks usually managed non-surviving funds, results may be biased towards bank funds outperforming the rest. If non-surviving funds were CPF-approved ones, the results are biased towards CPFapproved funds outperforming CPF interest rates. But if the non-surviving funds were not CPF-approved, results can give an accurate picture for relative performance of CPFapproved funds compared to CPF rates. Besides, factors differentiating fund performance may or may not be present in non-surviving funds. Omitting data on non-surviving funds therefore leads to biased results. An example of omission in 2003 is Invesco, which shut down ten funds worth S$80 million, including CPF-approved ones, due to prolonged investor apathy from declining stock markets (Chow 2003). To minimize survival bias, this study used data for surviving and non-surviving funds collected from quarterly reports. However, for performing regression analysis, only funds with at least three quarters of data were included.

4.6.2 External validity

After assessment of internal validity in the previous subsection, this subsection assesses external validity. Whether this research was externally valid depended on the population and setting to which generalization was made. With more than one study on similar 92

Research methodology research, comparing their results assesses external validity. Future research projects may compare results of studies from different mutual fund markets. Other than internal and external validity, ethical issues were not compromised.

4.7 Ethical considerations

Various ethical issues have to be addressed in any research study. In this section, general issues are presented with respect to subjects, researchers and sponsors, followed by specific issues relevant to this research.

4.7.1 General ethical issues

Ethical issues to be addressed in any research are classified under philosophies of deontology and teleology (Burns & Bush 2000). Deontology asserts behaviour is not ethical if it violates individual rights (Hunt, Chonko & Wilcox 1984). Research subjects freedom of choice, privacy and informed consent must be assured by the researcher before behaviours are recorded, giving subjects the right to decide whether to participate in research. While deontology focuses on individual rights, teleology emphasizes trade-off between individual costs and group benefits. Behaviour is considered ethical if group benefits outweigh individual costs (Hunt, Chonko & Wilcox 1984). Therefore, even though there is violation of informants rights, the research sponsor can learn from the study to become more efficient. As benefit is accrued to clients of the sponsor, who outnumber participants whose rights have been violated, the research practice is considered ethical from a costbenefit point of view.

93

Chapter 4

4.7.2 Specific ethical issues

Investors are aware of financial scandals that made companies such as Merrill Lynch and WorldCom household names. Evidence presented to regulators in the USA revealed some large securities firms, among them Merrill Lynch, pressured analysts to promote clients (Straits Times 2002a). Among the most stunning financial scandals of 2002, WorldComs estimate of inflated earnings was more than US$9 billion (Straits Times 2002b). Such transgressions are testimonies for regulations that can punish people violating the law, but cannot prevent abuses of financial data. Prevention requires self-regulation from accountants, auditors, fund managers and research analysts in the form of strict ethical standards. The Association for Investment Management and Researchs (AIMR) Code of Ethics and Standards of Professional Conduct was developed to precisely define acceptable conduct and actions for finance professionals (Code of Ethics 2002), based on Averas (1994) principles for standards of conduct in the industry.

4.7.3 Independent academic research to avoid ethical lapses

Some of the above-mentioned ethical problems are due to relationships between research analysts and securities firms they work for, as well as their research being mainly concerned with current investment environment and immediate buy or sell

recommendations. Research for this thesis was conducted as a project independent of financial institutions. As the research was based on theoretical and empirical studies, it has longer-term implications in the field of investments than brokerage firm reports for current investors and was therefore targeted at academic journals. Independent research submitted for blind-review academic journals avoids such ethical lapses.

94

Research methodology

4.8 Conclusion
Chapter 4 described secondary data research for collecting quarterly financial data to analyse relationships between fund management groups, performance and fund characteristics. The research sample comprised of all CPF-approved domestic equity funds from 1999 to 2004 to control for differing fiduciary responsibilities and market indexes. For performing linear regression of fund and market returns based on Sharpes (1964) CAPM model, funds with less than three quarters of data were excluded. Besides returns, four composite risk-adjusted measures were used for measuring performance to minimize bias from relying on one measure. For hypothesis testing, the sample was divided into bank and non-bank funds, large and small funds as well as high and low expense ratio funds before conducting two-tail pooled-variance t-tests for differences in two means. Consideration of survivorship bias and adopting a framework for assessing internal and external validity emphasized research quality. Lastly, ethical issues relevant to financial research were discussed, including AIMRs code of ethics. This sets the stage for reporting findings in the next chapter.

95

Chapter 5 Data analysis

5.1 Introduction
Chapter 4 explained collection of secondary data for regression analysis and hypothesis testing. This chapter presents analysis and test results before Chapter 6 discusses findings from finance literature context. An outline of this chapter is presented in Figure 5.1 below. Figure 5.1 Data analysis map
5 Data analysis

5.1 Introduction 5.2.1 Fund management company 5.2.2 Returns 5.2.3 Beta 5.2.4 Expense ratio 5.2.5 Fund size 5.4.1 Performance of domestic equity funds 5.4.2 Performance comparison of bank and nonbank funds 5.4.3 Expenditures, size and performance of domestic equity funds 5.4.4 Comparison of bank and non-bank fund characteristics

5.2 Characteristics of domestic equity funds 5.3 Regression analysis

5.4 Hypothesis testing

5.5 Conclusion

Source: developed for this research. After this introduction, section 5.2 summarizes characteristics of CPF-approved domestic equity funds in this order: (1) FMC, (2) performance, (3) systematic risk, (4) expenditures and (5) size. This is followed by results from linear regression of fund returns based on Sharpes (1964) CAPM model in section 5.3. In section 5.4, results of hypothesis testing are presented following sequence of research questions developed in Chapter 3: (1) overall performance of domestic equity funds; (2) performance comparison of bank and non-bank funds; (3) mutual fund characteristics affecting performance and (4) fund characteristics

96

Chapter 5 differentiating performance of bank and non-bank funds. Section 5.5 concludes with summary of test results.

5.2 Characteristics of domestic equity funds

Table 5.1 below summarizes average quarterly returns, beta, expense ratio and size characteristics of domestic equity funds managed by banks and non-banks. Table 5.1 Characteristics of domestic equity funds from banks and non-banks
Fund 1999:Q2--2002:Q1 2003:Q1--2004:Q3 Avg qtr ret Size (S$ Expense Beta Avg qtr ret Size (S$ Expense Beta million) ratio (%) million) ratio (%) (%) (%) Bank -0.46 75.31 1.50 0.95 7.28 109.26 1.49 1.19 Bank 4.33 43.96 0.96 1.11 12.30 59.50 1.16 1.51 Bank 4.53 17.03 1.49 1.07 7.49 19.06 1.52 1.19 Bank -6.13 3.00 3.32 0.32 Bank 2.57 2.57 3.83 0.78 4.93 0.85 5.02 1.00 Bank 0.23 35.96 1.57 0.85 Bank 4.51 103.33 1.36 1.01 6.39 125.20 1.18 0.90 Insurance 2.05 17.72 1.45 1.01 6.11 62.20 1.24 0.40 Insurance 4.71 45.19 1.67 0.61 Insurance 4.88 30.31 1.20 0.82 Insurance 0.12 10.58 1.16 1.03 Insurance 7.45 1.69 0.43 1.07 Insurance -0.66 19.56 2.04 1.14 Insurance 7.00 35.05 1.55 0.71 Insurance 6.03 91.65 1.35 0.86 6.11 84.23 1.18 0.90 Insurance -0.64 0.41 1.51 0.87 Investment 5.84 6.97 2.58 1.00 5.59 21.11 2.12 0.71 Investment 5.36 86.33 2.35 1.18 Investment 7.32 113.42 1.71 1.14 6.01 238.16 1.52 0.95 Type 2.63 1.37 3.60 2.94 0.63 1.00 41.43 40.17 42.41 1.86 2.00 1.74 0.94 0.87 0.99 1.00 6.60 7.68 5.84 5.97 0.63 1.00 66.40 62.77 68.98 1.64 2.07 1.34 0.94 1.16 0.78 1.00

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST

Average fund Average bank fund Average non-bank fund Straits Times Index Ord a/c interest rate Sp a/c interest rate

Note: refer to Table 4.2 (p. 84) for a list of funds used for this research. Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Type of fund manager (bank or non-bank), performance, systematic risk, expenditures and net assets under management are summarized in this order in the following subsections.

5.2.1 Fund management company

Among the 19 funds in Table 5.1 above, 7 were bank-managed, 9 were from insurance companies and 3 from investment firms. As these bank funds were from local banks, 97

Data analysis results in this chapter revealed relative competitiveness of local banks and their competitors in Singapores fund management industry.1 Among the non-banks, all investment firms were foreign competitors, while insurance companies were local, with the exception of AXA, a French insurance group. Therefore, non-banks in this research sample represented a balanced mix of local and foreign competitors. The following subsections compare quantitative characteristics of funds managed by banks and non-banks, starting with returns then followed by beta, expense ratio and size.

5.2.2 Returns
Table 5.2 compiles fund performance rankings in terms of quarterly returns for 1999-2002 and 2003-2004. Table 5.2 Fund performance rankings for 1999-2002 and 2003-2004
Rank 1999:Q2--2002:Q1 Fund Type Average quarterly return (%) 1 SST Investment 7.32 2 ULFU Insurance 7.00 3 ULFUG Insurance 6.03 5.84 4 ASE Investment 5.36 5 CFSSG Investment 6 OSST Bank 4.53 7 UUG Bank 4.51 8 DST Bank 4.33 9 UOSE Bank 2.57 10 ALF Insurance 2.05 11 UU Bank 0.23 12 KM Insurance 0.12 13 DHSE Bank -0.46 14 ULG Insurance -0.64 15 OMGSG Insurance -0.66 16 OUSE Bank -6.13 2.63 1.37 3.60 Rank 2003:Q1--2004:Q3 Fund Type Average quarterly return (%) 1 DST Bank 12.30 2 OSST Bank 7.49 3 NISE Insurance 7.45 4 DHSE Bank 7.28 5 UUG Bank 6.39 6 ALF Insurance 6.11 7 ULFUG Insurance 6.11 8 SST Investment 6.01 9 ASE Investment 5.59 10 UOSE Bank 4.93 4.88 11 GGSE Insurance 4.71 12 ALFA Insurance

Average fund Average bank fund Average non-bank fund

6.60 7.68 5.84

Note: returns are ranked in descending order. Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data.
1

The four domestic banks were Development Bank of Singapore (DBS), Overseas Chinese Banking Corporation (OCBC), Overseas Union Bank (OUB) and United Overseas Bank (UOB). OUB was acquired by UOB in 2003.

98

Chapter 5 During 1999-2002, the five best performing funds were from non-banks (ASE, CFSSG, ULFU, ULFUG and SST), each earning more than five percent average quarterly returns, compared with the best performing bank funds earning less than five percent, while the worst performing fund was bank-managed (OUSE). During 2003-2004, the reverse seemed to be true as the two best performing funds were from banks (DST and OSST) earning average quarterly returns in excess of seven percent, while the two worst performing ones were from non-banks (ALFA and GGSE). This observation was confirmed by average return statistics showing returns of 3.60 percent for non-bank funds outperforming bank funds 1.37 percent during the earlier period before bank funds (7.68 percent) outperformed their non-bank counterparts (5.84 percent) during the later period. For the benchmark index to compare fund returns, Figure 5.2 illustrates daily STI levels from 1987 to 2005. Figure 5.2 Daily Straits Times Index from 1987 to 2005 Straits Times Index
3000 2500 Index 2000 1500 1000 500 0 28-12-87 28-12-89 28-12-91 28-12-93 28-12-95 28-12-97 28-12-99 28-12-01 28-12-03 28-12-05 Close

Date

Source: developed for this research from Yahoo Finance data. Quarterly STI returns were computed from daily levels and illustrated in Figure 5.3 on the next page. Observation of both stock market charts revealed a bearish domestic equity market from 2000 to 2002 followed by steady recovery. STI and quarterly fund returns

99

Data analysis from 1999 to 2004 were used for linear regression and hypothesis testing of performance comparison. Figure 5.3 Quarterly STI returns from 1988 to 2004 Straits Times Index Quarterly Returns
40 30 20 10 0 -10 -20 -30 -40 Apr-88 Apr-90 Apr-92 Apr-94 Apr-96 Apr-98 Apr-00 Apr-02 Apr-04

Return (%)

Return

Date

Source: developed for this research from Yahoo Finance data. Results from comparison of fund performance with benchmark index and guaranteed interest rates are presented later in section 5.4.1; while section 5.4.2 provides results for comparing bank and non-bank fund performance. In the following subsections, statistics on beta, expense ratio and size determinants of fund performance are presented.

5.2.3 Beta
Examining fund beta rankings compiled in Table 5.3 on the following page, top three funds with the highest beta during 1999-2002 were from non-banks (CFSSG, OMGSG and SST), while three of the four funds with the lowest beta during that period were from banks (OUSE, UOSE and UU). During 2003-2004, the reverse seemed to be true as the three funds with highest beta were from banks (DHSE, DST and OSST) while the four funds with lowest beta were from non-banks (ALF, ALFA, GGSE and ASE).

100

Chapter 5 Initial observation of fund betas seemed to indicate bank funds were initially less risky than their non-bank counterparts, but had become more risky. This was consistent with average bank fund beta of 0.87 being lower than its non-bank counterpart at 0.99 during 1999-2002, but had grown higher to 1.16 during 2003-2004, while its non-bank counterpart decreased to 0.78. Table 5.3 Fund beta rankings for 1999-2002 and 2003-2004
1999:Q22002:Q1 Rank Fund Type 1 CFSSG Investment 2 OMGSG Insurance 3 SST Investment 4 DST Bank 5 OSST Bank 6 KM Insurance 7 ALF Insurance 8 UUG Bank 9 ASE Investment 10 DHSE Bank 11 ULG Insurance 12 ULFUG Insurance 13 UU Bank 14 UOSE Bank 15 ULFU Insurance 16 OUSE Bank Average fund Average bank fund Average non-bank fund Beta 1.18 1.14 1.14 1.11 1.07 1.03 1.01 1.01 1.00 0.95 0.87 0.86 0.85 0.78 0.71 0.32 0.94 0.87 0.99 2003:Q1--2004:Q3 Rank Fund Type 1 DST Bank 2 OSST Bank 3 DHSE Bank 4 NISE Insurance 5 UOSE Bank 6 SST Investment 7 UUG Bank 8 ULFUG Insurance 9 GGSE Insurance 10 ASE Investment 11 ALFA Insurance 12 ALF Insurance Beta 1.51 1.19 1.19 1.07 1.00 0.95 0.90 0.90 0.82 0.71 0.61 0.40

0.94 1.16 0.78

Note: betas are ranked in descending order. Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. More detailed analysis for betas of bank and non-bank funds is presented in section 5.4.4. In the next two subsections, statistics on expense ratio and fund size are presented separately.

5.2.4 Expense ratio

Comparing expense ratios of funds managed by banks and non-banks in Table 5.4 on the next page, banks managed the two funds with largest expense ratio during 1999-2002 (OUSE and UOSE with expense ratio higher than 3 percent). At the opposite extreme, the

101

Data analysis fund with lowest expense ratio was also bank-managed (DST at 0.96 percent). During 2003-2004, even though a bank managed the fund with highest expense ratio (UOSE with expense ratio larger than 5 percent), a non-bank managed the one with smallest expense ratio (NISE at 0.43 percent). Table 5.4 Fund expense ratio rankings for 1999-2002 and 2003-2004
1999:Q2--2002:Q1 Rank Fund Type 1 UOSE 2 OUSE 3 ASE 4 CFSSG 5 OMGSG 6 SST 7 UU 8 ULFU 9 ULG 10 DHSE 11 OSST 12 ALF 13 UUG 14 ULFUG 15 KM 16 DST Bank Bank Investment Investment Insurance Investment Bank Insurance Insurance Bank Bank Insurance Bank Insurance Insurance Bank Expense ratio 3.83 3.32 2.58 2.35 2.04 1.71 1.57 1.55 1.51 1.50 1.49 1.45 1.36 1.35 1.16 0.96 1.86 2.00 1.74 2003:Q1--2004:Q3 Rank Fund Type Expense ratio 1 UOSE Bank 5.02 2 ASE Investment 2.12 3 ALFA Insurance 1.67 4 SST Investment 1.52 5 OSST Bank 1.52 6 DHSE Bank 1.49 7 ALF Insurance 1.24 8 GGSE Insurance 1.20 9 UUG Bank 1.18 10 ULFUG Insurance 1.18 11 DST Bank 1.16 12 NISE Insurance 0.43

Average fund Average bank fund Average non-bank fund

1.64 2.07 1.34

Note: expense ratios are ranked in descending order. Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Observation of average expense ratio statistics showed bank fund expense ratios were generally higher than their non-bank counterparts during both periods. However, bank fund average expense ratio increased from 2.00 percent during 1999-2002 to 2.07 percent during 2003-2004 while its non-bank counterpart decreased from 1.74 to 1.34 percent. While section 5.4.3 tests relations between expense ratio, size and performance, more detailed expense ratio analysis for bank and non-bank funds are presented in section 5.4.4.

102

Chapter 5

5.2.5 Fund size

Reviewing size statistics in Table 5.5 below, the largest and smallest funds during 19992002 were from non-banks (SST and ULG at S$113.42 and S$0.41 million). While SST more than doubled its size to $238.16 million during 2003-2004 to remain the largest, smallest fund during that period was bank-managed (UOSE at S$0.85 million). Table 5.5 Fund size rankings for 1999-2002 and 2003-2004
1999:Q2--2002:Q1 Rank Fund Type Size ($m) 1 SST Investment 113.42 2 UUG Bank 103.33 3 ULFUG Insurance 91.65 4 CFSSG Investment 86.33 5 DHSE Bank 75.31 6 DST Bank 43.96 7 UU Bank 35.96 8 ULFU Insurance 35.05 9 OMGSG Insurance 19.56 10 ALF Insurance 17.72 11 OSST Bank 17.03 12 KM Insurance 10.58 13 ASE Investment 6.97 14 OUSE Bank 3.00 15 UOSE Bank 2.57 16 ULG Insurance 0.41 Average fund Average bank fund Average non-bank fund 41.43 40.17 42.41 2003:Q1--2004:Q3 Rank Fund Type Size ($m) 1 SST Investment 238.16 2 UUG Bank 125.20 3 DHSE Bank 109.26 4 ULFUG Insurance 84.23 5 ALF Insurance 62.20 6 DST Bank 59.50 7 ALFA Insurance 45.19 8 GGSE Insurance 30.31 9 ASE Investment 21.11 10 OSST Bank 19.06 11 NISE Insurance 1.69 12 UOSE Bank 0.85

66.40 62.77 68.98

Note: Fund sizes are ranked in descending order. Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Confirming this observation with the average size statistics (bottom of table), while the funds had mostly grown in size from 1999-2002 to 2003-2004, bank funds were generally smaller than their non-bank counterparts. The average bank fund grew from S$40.17 to S$62.77 million from 1999-2002 to 2003-2004 while its non-bank counterpart grew from $42.41 to $68.98 million during the same period. While section 5.4.3 presents relations between performance, expense ratio and size, more detailed analysis for bank and nonbank fund sizes is presented in section 5.4.4.

103

Data analysis This section summarized the financial institution, performance, risk, expenditures and size characteristics of funds. The following section presents results from regression of fund returns.

5.3 Regression analysis

Figure 5.4 below illustrates graphically time-series regression of returns for a domestic equity fundthe Aberdeen Singapore Equity Fund (ASE). Based on Sharpes (1964) CAPM model, regression used quarterly risk premiums of the fund and its benchmark STI market index. Figure 5.4 Time series regression of a domestic equity fund

Time series regression of quarterly fund risk premium versus market risk premium for a fund from 1999:Q2 to 2002:Q1
R sq = 0.91 80.00 Investment firm: 60.00 Aberdeen Singapore 40.00 Equity Fund (ASE) 20.00 0.00 Linear (Investment -20.00 firm: Aberdeen -40.00 Singapore Equity -40.00 -20.00 0.00 20.00 40.00 60.00 Fund (ASE)) STIt - RFRt

Source: developed for this research from Mercer (1999-2002) data. In this example, 12 quarterly risk premiums were regressed against corresponding STI risk premiums. As most data points clustered around the fitted regression line, coefficient of determination R2 (0.91) was high. The corresponding normal probability plot is shown in Figure 5.5 on the following page. For residual analysis, mostly linear trend on this plot suggested normality assumption for linear regression was satisfied (Mendenhall & Sincich 1996).

RETft - RFRt

104

Chapter 5 Figure 5.5 Normal probability plot of a domestic equity fund

Normal probability plot for ASE

RETft - RFR1 100 50 0 -50 0 20 40 60 80 100 120 Sample percentile

Source: developed for this research from Mercer (1999-2002) data. Data, computation and regression outputs for all 19 funds were tabulated in Appendix A. The following section tests relationships among fund characteristics summarized in the previous section.

5.4 Hypothesis testing

In this section, results of statistical testing for hypotheses are reported in the same order as they were developed in Chapter 3s literature review.

5.4.1 Performance of domestic equity funds

The first set of hypotheses concern overall performance of domestic equity funds: H1.1: domestic equity funds do not outperform benchmark stock market index; H1.2: there is positive relation between domestic equity fund returns and returns to the benchmark stock market index; and H1.3: there is no significant difference between returns from investing savings in domestic equity funds and guaranteed interest rates. This subsection reports findings using Singapores CPF-approved domestic equity funds. Comparison of domestic equity funds and stock market index returns Linear regression results of each funds quarterly risk premium (RETft RFRt) on the STI quarterly risk premium (STIt RFRt) are shown in Table 5.6 on the next page. Results

105

Data analysis were generally consistent with findings by Jensen (1968) for funds in the USA, confirming average fund cannot beat the market. Table 5.6 Regression of equity fund and market index risk premiums
Fund 1999:Q2--2002:Q1 Avg Beta R sq Jensen alpha 0.625% 1% RFR qtr ret RFR (%) -0.46 0.95 0.98 0.19 0.17 4.33 1.11 0.94 1.13 1.17 4.53 1.07 0.97 1.42 1.45 -6.13 0.32 0.19 -4.73 -4.99 2.57 0.78 0.99 -0.32 -0.40 0.23 0.85 0.83 0.65 0.57 4.51 1.01 0.97 1.55 1.55 2.05 1.01 0.79 2.14 2.14 Avg qtr ret (%) 7.28 12.30 7.49 2003:Q1--2004:Q3 Jensen alpha Beta R sq 0.625% 1% RFR RFR 1.19 0.90 0.32 0.39 1.51 0.67 3.60 3.79 1.19 0.77 0.48 0.55 0.90 0.93 0.77 0.79 0.95 0.96 -1.04 1.04 3.35 ** 1.97 -0.13 0.40 -1.04 0.93 3.13 * 1.82 -0.20 0.43

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST Avg fund STI

4.93 1.00 6.39 6.11 4.71 4.88 0.90 0.40 0.61 0.82

0.12 -0.66 7.00 6.03 -0.64 5.84 5.36 7.32 2.63 2.94

1.03 1.14 0.71 0.86 0.87 1.00 1.18 1.14 0.94 1.00

0.90 0.94 0.98 0.98 0.95 0.91 0.84 0.97 0.88 1.00

0.22 0.25 4.41 3.00 -0.66 2.90 2.01 4.05 ** 1.14 0.00

0.23 7.45 1.07 0.30 4.30 2.95 -0.71 2.90 2.08 4.10 ** 1.11 0.00

6.11 0.90 5.59 0.71 6.01 0.95 6.60 0.94 5.97 1.00

0.88 0.87 0.91 0.86 1.00

0.71 1.19 0.29 1.01 0.00

0.67 1.08 0.27 0.99 0.00

*Significant excess return at the 5% level. **Significant excess return at the 1% level.

Note: refer to Table 4.2 for a list of funds used for this research. Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. From 1999 to 2002, average quarterly return for CPF-approved domestic equity funds was below STI return (2.63 versus 2.94 percent respectively). However, from 2003 to 2004, average fund return of 6.60 percent outperformed 5.97 percent from STI index. Jensen alpha. Examining Jensen alphas in the table above, insignificance of positive alphas downplayed possibility of average fund outperforming the market. According to the Jensen alpha criterion, even though majority of funds registered abnormal returns above expectation with positive alphas for both holding periods (13 out of 16 during 1999-2002; 10 out of 12 during 2003-2004), only one was statistically significant in each period (SST and ALF during 1999-2002 and 2003-2004 respectively). As the SST fund registering

106

Chapter 5 significant abnormal return during 1999-2002 became one of the worst performers during 2003-2004, performance consistency was lacking, supporting Carlsons (1970) observation. Information ratio. Table 5.7 below reports fund information ratios. When evaluating fund performance, reasonable information ratio values should range from 0.5 to 1.0 for good to exceptionally good performance (Grinold & Kahn 1995). Table 5.7 Information ratios for domestic equity funds
Fund 1999:Q2--2002:Q1 Info No. of qtr t-statistic ratio periods 0.12 7 0.31 0.25 12 0.85 0.41 12 1.42 -0.08 6 -0.18 -0.16 3 -0.28 0.15 7 0.39 0.45 12 1.57 0.26 10 0.83 2003:Q1--2004:Q3 Info No. of qtr t-statistic ratio periods 0.42 7 1.11 0.76 7 2.00 0.32 7 0.84 -0.43 0.22 0.03 0.27 -0.59 7 7 7 3 7 5 -1.14 0.58 0.08 0.47 -1.56 1.67

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST Avg fund STI

0.04 0.01 0.57 0.68 -0.15 0.49 0.24 0.92 0.26 0.00

10 7 4 4 10 12 12 12

0.12 0.75 0.03 1.15 1.35 0.06 -0.48 1.71 -0.13 0.82 3.17 * 0.02 0.14 0.00

7 7 7

0.15 -0.35 0.06

*Significant excess return at the 5% level. **Significant excess return at the 1% level.

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. To interpret findings for this study, funds having positive Jensen alpha values also had positive information ratios. Mean information ratio for the sample was 0.26 during 19992002 and deteriorated to 0.14 during 2003-2004, well below the 0.5 standard for good performance recommended by Grinold and Kahn (1995). Agreeing with Goodwins (1998) findings in the USA, average fund in the sample added value to its investments, but its

107

Data analysis performance did not qualify as good. In fact, none of the funds delivered an excellent information ratio greater than 1.0, even though there were a few good performers for each holding period (SST, ULFU and ULFUG during 1999-2002; DST and NISE during 20032004). Again, good performers during the first period cannot sustain their achievement for second period, confirming Carlsons (1970) observation. Relation between returns from domestic equity funds and the stock market index Returning to Table 5.6 (page 106), average fund R2 was quite high at 0.88 during 19992002, with individual R2 values greater than 0.75, except for OUSE with less than 0.20. This indicates well-diversified funds within the local equity market, except for OUSE. In fact, majority of funds had R2 greater than 0.90, supporting Jensens (1968) finding for fund returns typically correlating with market returns at rates higher than 0.90. As benchmark market reflects each funds asset allocation policy, this result also supported Brinson, Hood and Beebower (1986), Brinson, Singer and Beebower (1991), as well as Ibbotson and Kaplan (2000), who reported about 90 percent of variability in returns of a typical fund across time can be explained by its asset allocation policy. From 2003 to 2004, average R2 dropped slightly to 0.86, with less funds registering R2 greater than 0.90. Still, the findings supported existing literature in reporting high correlation between fund and market returns. Excess returns from domestic equity fund investment over guaranteed interest rates Referring to Table 5.8 on the next page, the average fund had positive Sharpe and Treynor ratios for both risk-free rates (0.625 and 1 percent per quarter) during each holding period, implying returns exceeding guaranteed interest rates. Overall, returns from CPF-approved domestic equity funds were higher than guaranteed interest rates of Ordinary and Special accounts for both holding periods, refuting Kohs (1999) earlier finding for funds underperforming interest rates in Singapore.

108

Chapter 5 Comparing performance results obtained for these composite performance measures showed identification of best performing funds dependent on choice of performance measure. For example, during 1999-2002, the best performing fund according to average quarterly returns, information ratio and Jensen alpha criteria was SST, but it was ULFU using Sharpe and Treynor ratios. During 2003-2004, best performing fund was ALF in terms of Jensen alpha as well as Sharpe and Treynor ratios, but it was DST according to average quarterly returns and information ratio. Therefore, various performance measures should be used to minimize bias from relying solely on one measure. Table 5.8 Sharpe and Treynor ratios for domestic equity funds
Fund 1999:Q2--2002:Q1 Sharpe ratio Treynor ratio 0.625% 1% RFR 0.625% 1% RFR RFR RFR -0.07 -0.10 -1.14 -1.53 0.17 0.15 3.33 2.99 0.19 0.17 3.64 3.29 -1.37 -1.45 -20.95 -22.12 0.10 0.08 2.48 2.00 -0.03 -0.06 -0.47 -0.91 0.20 0.18 3.85 3.48 0.08 0.06 1.41 1.04 2003:Q1 to 2004:Q3 Sharpe ratio Treynor ratio 0.625% 1% RFR 0.625% 1% RFR RFR RFR 0.75 0.71 5.61 5.30 0.89 0.86 7.72 7.48 0.71 0.67 5.75 5.43 0.57 0.87 1.70 1.26 0.71 1.24 -0.07 0.45 0.31 -0.09 0.26 0.20 0.31 0.04 0.12 -0.09 0.43 0.29 -0.12 0.25 0.18 0.29 0.01 0.10 -1.12 9.04 6.29 -1.46 5.20 4.02 5.87 1.22 2.31 -1.45 8.51 5.85 -1.89 4.83 3.71 5.54 0.78 1.94 0.52 0.82 1.58 1.15 0.65 1.17 4.31 6.43 13.77 6.65 5.18 6.36 3.93 6.01 12.83 6.04 4.73 6.01

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST Avg fund STI

-0.03

-0.05

-0.49

-0.85

0.81 0.92 0.76 0.93 0.75

0.75 0.85 0.71 0.87 0.70

6.13 7.03 5.65 6.71 5.34

5.72 6.50 5.25 6.27 4.97

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. This subsections performance comparison of domestic equity funds with market index and guaranteed interest rates revealed funds generating higher returns than guaranteed interest rates, but was generally not impressive, as elaborated in section 6.2.1 in the following chapter. The next subsection compares performance of funds managed by 109

Data analysis different institution groups, namely banks versus non-banks (insurance and investment companies).

5.4.2 Performance comparison of bank and non-bank funds

Table 5.9 tabulates computation of quarterly returns and various composite performance measures for bank and non-bank funds. Table 5.9 Performance measures of bank and non-bank domestic equity funds
Fund Type Avg qtr ret (%) -0.46 4.33 4.53 -6.13 2.57 0.23 4.51 2.05 1999:Q2--2002:Q1 Info Jensen Sharpe Treynor Avg qtr ratio ratio ratio alpha ret (%) 0.12 0.19 -0.07 -1.14 7.28 0.25 1.13 0.17 3.33 12.30 0.41 1.42 0.19 3.64 7.49 -0.08 -4.73 -1.37 -20.95 -0.16 -0.32 0.10 2.48 4.93 0.15 0.65 -0.03 -0.47 0.45 1.55 0.20 3.85 6.39 0.26 2.14 0.08 1.41 6.11 4.71 4.88 0.04 0.22 -0.03 -0.49 7.45 0.01 0.25 -0.07 -1.12 0.57 4.41 0.45 9.04 0.68 3.00 0.31 6.29 6.11 -0.15 -0.66 -0.09 -1.46 0.49 2.90 0.26 5.20 5.59 0.24 2.01 0.20 4.02 0.92 4.05 0.31 5.87 6.01 0.26 0.16 0.34 0.00 1.14 -0.02 2.04 0.00 0.04 -0.12 0.16 0.12 1.22 -1.32 3.20 2.31 6.60 7.68 5.84 5.97 0.63 1.00 2003:Q1--2004:Q3 Info Jensen Sharpe Treynor ratio ratio ratio alpha 0.42 0.32 0.75 5.61 0.76 3.60 0.89 7.72 0.32 0.48 0.71 5.75 -0.43 0.22 0.03 0.27 -0.59 0.75 -1.04 1.04 3.35 1.97 -0.13 0.40 0.57 0.87 1.70 1.26 0.71 1.24 4.31 6.43 13.77 6.65 5.18 6.36

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST

Bank Bank Bank Bank Bank Bank Bank Ins Ins Ins Ins Ins Ins Ins Ins Ins Inv Inv Inv

0.12 -0.66 7.00 6.03 -0.64 5.84 5.36 7.32 2.63 1.37 3.60 2.94 0.63 1.00

0.06 -0.13 0.02 0.14 0.26 0.06 0.00

0.71 1.19 0.29 1.01 0.88 1.11 0.00

0.81 0.92 0.76 0.93 0.76 1.06 0.75

6.13 7.03 5.65 6.71 5.96 7.25 5.34

Avg Avg bank Avg non-bank STI CPF ord a/c CPF sp a/c

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. The second set of hypotheses compare performance of domestic equity funds managed by banks and non-banks: H2.1: there is no significant difference in returns between domestic equity funds managed by banks and non-banks; H2.2: there is no significant difference in information ratios between domestic equity funds managed by banks and non-banks; 110

Chapter 5 H2.3: there is no significant difference in Jensen alphas between domestic equity funds managed by banks and non-banks; H2.4: there is no significant difference in Sharpe ratios between domestic equity funds managed by banks and non-banks; and H2.5: there is no significant difference in Treynor ratios between domestic equity funds managed by banks and non-banks. Referring to Table 5.9 on the previous page, when STI index posted an average quarterly return of 2.94 percent during 1999-2002, average bank fund under-performed the market at 1.37 percent, while average non-bank fund managed to outperform the market at 3.60 percent. For 2003-2004, with an STI average quarterly return of 5.97 percent, the reverse seemed to be true as bank funds outperformed the market at 7.68 percent while non-bank funds under-performed the market at 5.84 percent. Returns Table 5.10 performed a two-tail t-test assuming unequal variances for returns of bank and non-bank funds. Table 5.10 Two-sample t-test for bank and non-bank fund returns
Ho: no difference between mean returns of bank and non-bank funds Ha: mean returns of bank and non-bank funds differ 1999:Q22002:Q1 2003:Q1--2004:Q3 =0.05 Bank Non-bank Bank Non-bank Mean return (%) 1.369 3.602 7.678 5.837 Variance 15.158 11.257 7.692 0.838 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 12 5 t Stat -1.209 1.430 P(Tt) one-tail 0.125 0.106 t Critical one-tail 1.782 2.015 P(Tt) two-tail 0.250 0.212 t Critical two-tail 2.179 2.571

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. According to the results, there was no significant difference between returns of bank and non-bank domestic equity funds during each holding period (t = -1.209, p = 0.250 during

111

Data analysis 1999-2002; t = 1.430, p = 0.212 during 2003-2004), supporting Fryes (2001) result for bond funds. This result was confirmed by conducting additional tests for bank and non-bank funds using risk-adjusted performance measures. The following subsections present results using composite performance measures. Information ratio Performing a two-tail t-test assuming unequal variances for information ratios of bank and non-bank funds in Table 5.11 below showed no significant difference in ratio for bank and non-bank equity funds during each holding period (t = -1.215, p = 0.245 during 19992002; t = 0.801, p = 0.446 during 2003-2004). Table 5.11 Two-sample t-test for information ratios of bank and non-bank funds
Ho: no difference between mean information ratios of bank and non-bank funds Ha: mean information ratios of bank and non-bank funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 =0.05 Bank Non-bank Bank Non-bank Mean information ratio 0.162 0.339 0.256 0.058 Variance 0.053 0.123 0.188 0.164 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 14 8 t Stat -1.215 0.801 P(Tt) one-tail 0.122 0.223 t Critical one-tail 1.761 1.860 P(Tt) two-tail 0.245 0.446 t Critical two-tail 2.145 2.306

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. This result supported Fryes (2001) finding for bond funds. Jensen alpha Performing a two-sample t-test assuming unequal variances for Jensen alphas of bank and non-bank funds in Table 5.12 on the next page showed bank funds under-performing nonbank funds significantly during 1999-2002 (t = -2.019, p = 0.034), but during 2003-2004, no significant performance difference was detected (t = -0.263, p = 0.8).

112

Chapter 5 Table 5.12 Two-sample t-test for Jensen alphas of bank and non-bank funds
Ho: no difference between mean Jensen alphas of bank and non-bank funds Ha: mean Jensen alphas of bank and non-bank funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 =0.05 Bank Non-bank Bank Non-bank Mean Jensen alpha -0.015 2.035 0.878 1.111 Variance 4.775 3.141 2.891 1.444 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 11 7 t Stat -2.019 -0.263 P(T<=t) one-tail 0.034 0.400 t Critical one-tail 1.796 1.895 P(T<=t) two-tail 0.068 0.800 t Critical two-tail 2.201 2.365

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Thus, bank funds may have improved their performance to be comparable to their nonbank counterparts. Sharpe and Treynor ratios Returning to Table 5.9 (page 110), while the average fund had positive Sharpe and Treynor ratios during each holding period (Sharpe ratios 0.04 and 0.93; Treynor ratios 1.22 and 6.71), implying returns exceeding guaranteed interest rates, average bank fund actually registered negative Sharpe and Treynor ratios during 1999-2002 (Sharpe ratio -0.12, Treynor ratio -1.32), implying earning guaranteed interest rates in Ordinary and Special accounts were better than investing in bank funds during that holding period. Performing a two-sample t-test assuming unequal variances for Sharpe ratios of bank and non-bank funds in Table 5.13 on the next page showed no significant performance

difference during 1999-2002 (t = -1.228, p = 0.259), but significant underperformance of bank funds during 2003-2004 (t = -2.015, p = 0.039). Table 5.14 s two-sample t-test assuming unequal variances for Treynor ratios of bank and non-bank funds on the following page showed no significant performance difference during each time period (t = -1.261, p = 0.243 during 1999-2002; t = -1.037, p = 0.327 during 2003-2004).

113

Data analysis Table 5.13 Two-sample t-test for Sharpe ratios of bank and non-bank funds
Ho: no difference between mean Sharpe ratios of bank and non-bank funds Ha: mean Sharpe ratios of bank and non-bank funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 =0.05 Bank Non-bank Bank Non-bank Mean Sharpe ratio -0.116 0.158 0.759 1.057 Variance 0.318 0.038 0.017 0.129 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 7 8 t Stat -1.228 -2.015 P(Tt) one-tail 0.130 0.039 t Critical one-tail 1.895 1.860 P(Tt) two-tail 0.259 0.079 t Critical two-tail 2.365 2.306

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. These results indicated higher level of non-systematic risk in bank funds during 2003-2004 than their non-bank counterparts, corresponding to significant underperformance of bank funds according to the Sharpe ratio. Table 5.14 Two-sample t-test for Treynor ratios of bank and non-bank funds
Ho: no difference between mean Treynor ratios of bank and non-bank funds Ha: mean Treynor ratios of bank and non-bank funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 = 0.05 Bank Non-bank Bank Non-bank Mean Treynor ratio -1.323 3.196 5.963 7.252 Variance 78.928 14.077 1.559 8.630 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 8 9 t Stat -1.261 -1.037 P(Tt) one-tail 0.121 0.163 t Critical one-tail 1.860 1.833 P(Tt) two-tail 0.243 0.327 t Critical two-tail 2.306 2.262

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Overall, these test results of performance measures indicated bank funds were comparable to their non-bank counterparts in terms of performance, as concluded in section 6.2.2 in the following chapter. The next subsection explores expenditures and size characteristics as determinants of fund performance.

114

Chapter 5

5.4.3 Expenditures, size and performance of domestic equity funds

The third set of hypotheses concern importance of size and expenditures in determining domestic equity fund performance: H3.1: there is no significant difference in returns between small and large funds; H3.2: there is no significant difference in returns between funds with high expense ratios and those with low expense ratios; and H3.3: there is no significant difference in expense ratios of big and small funds. This subsection tests plausible relations between these characteristics. Performance of large and small funds Referring to Table 5.15 below, during 1999-2002, average return of big funds was 4.29 percent, versus 0.96 percent for small funds. For 2003-2004, large funds return of 7.37 percent continued to outperform small funds 5.84 percent. Table 5.15 Two-sample t-test for returns of big and small funds
Ho: no difference between mean returns of large and small funds Ha: mean returns of large and small funds differ 1999:Q2--2002:Q1 =0.05 Large fund Small fund Mean return (%) 4.290 0.960 Variance 8.542 13.785 Observations 8 8 Hypothesized mean difference 0 Df 13 t Stat 1.993 P(Tt) one-tail 0.034 t Critical one-tail 1.771 P(Tt) two-tail 0.068 t Critical two-tail 2.160

2003:Q1--2004:Q3 Large fund Small fund 7.367 5.842 6.060 1.681 6 6 0 8 1.343 0.108 1.860 0.216 2.306

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Performing a two-sample t-test assuming unequal variances for returns of large and small funds showed significant out-performance of small funds by large funds during 1999-2002 (t = 1.993, p = 0.334), but no significant performance difference during 2003-2004 (t = 1.343, p = 0.216).

115

Data analysis Performance of high and low expense ratio funds Table 5.16 below tests returns from high and low expense ratio funds. During 1999-2002, average return of funds with low expense ratios was 2.56 percent, which was lower than 2.69 percent average return of high expense ratio funds. For 2003-2004, the reverse happened as average returns of low and high expense ratio funds were 7.21 and 6.00 percent respectively. Table 5.16 Two-sample t-test for returns of high and low expense ratio funds
Ho: no difference between mean returns of high and low expense ratio funds Ha: mean returns of high and low expense ratio funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 =0.05 Expense ratio Expense ratio Low High Low High Mean return (%) 2.559 2.691 7.207 6.002 Variance 6.919 21.735 6.897 1.367 Observations 8 8 6 6 Hypothesized mean difference 0 0 Df 11 7 t Stat -0.070 1.027 P(Tt) one-tail 0.473 0.169 t Critical one-tail 1.796 1.895 P(Tt) two-tail 0.945 0.339 t Critical two-tail 2.201 2.365

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Performing a two-sample t-test assuming unequal variances for returns of high and low expense ratio funds showed no significant performance difference during the two periods (t = -0.070, p = 0.945 during 1999-2002; t = 1.027, p = 0.339 during 2003-2004). Expense ratio of large and small funds From Table 5.17 on the next page, during 1999-2002, average expense ratio of large funds at 1.54 percent was lower than that of small funds at 2.17 percent. For 2003-2004, average expense ratio of large funds at 1.30 percent continued to be lower than its small fund counterpart at 1.99 percent. Performing a two-sample t-test assuming unequal variances for expense ratios of large and small funds showed no significant expense ratio differences for large and small funds

116

Chapter 5 during both periods (t = -1.686, p = 0.126 during 1999-2002; t =-1.072, p = 0.333 during 2003-2004). Table 5.17 Two-sample t-test for expense ratios of big and small funds
Ho: no difference between mean expense ratios of large and small funds Ha: mean expense ratios of large and small funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 =0.05 Big fund Small fund Big fund Small fund Mean expense ratio 1.544 2.173 1.295 1.993 Variance 0.156 0.957 0.027 2.516 Observations 8 8 6 6 Hypothesized mean difference 0 0 Df 9 5 t Stat -1.686 -1.072 P(Tt) one-tail 0.063 0.166 t Critical one-tail 1.833 2.015 P(Tt) two-tail 0.126 0.333 t Critical two-tail 2.262 2.571

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Overall, these test results indicated expense ratio and size were not strong performance determinants, as elaborated in section 6.2.3. The next subsection compares these characteristics for bank and non-bank funds.

5.4.4 Comparison of bank and non-bank fund characteristics

The forth and last set of hypotheses compare size, systematic risk and expenditures characteristics of domestic equity funds managed by banks and non-banks: H4.1: there is no significant difference in size between domestic equity funds managed by banks and non-banks; H4.2: there is no significant difference in levels of systematic risk between domestic equity funds managed by banks and non-banks; and H4.3: there is no significant difference in expense ratios of bank and non-bank funds. This subsection compares characteristics for bank and non-bank funds.

117

Data analysis Size of bank and non-bank domestic equity funds Referring to Table 5.18 below, the average bank equity fund managed about S$40 million of net assets during 1999-2002, while its non-bank counterpart managed about $42 million. For 2003-2004, average bank fund managed about S$63 million, while its nonbank counterpart managed around $69 million. Table 5.18 Two-sample t-test for size of bank and non-bank funds
Ho: no difference between mean size of bank and non-bank funds Ha: mean size of bank and non-bank funds differ 1999:Q2--2002:Q1 =0.05 Bank Non-bank Mean size (S$ million) 40.165 42.410 Variance 1431.152 1826.874 Observations 7 9 Hypothesized mean difference 0 Df 14 t Stat -0.111 P(Tt) one-tail 0.457 t Critical one-tail 1.761 P(Tt) two-tail 0.913 t Critical two-tail 2.145

2003:Q1--2004:Q3 Bank Non-bank 62.773 68.984 2953.549 6296.870 5 7 0 10 -0.161 0.438 1.812 0.875 2.228

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. While both fund groups had experienced significant growth in size, bank equity funds seemed to be smaller than their non-bank counterparts. Performing a two-sample t-test assuming unequal variances for sizes of bank and non-bank funds showed no significant difference in size between bank and non-bank equity funds during each holding period (t = -0.111, p = 0.913 during 1999-2002; t = -0.161, p = 0.875 during 2003-2004). This result may imply both groups of funds were equally popular when there was no significant difference in performance. Risk of bank and non-bank domestic equity funds Referring to Table 5.19 on the following page, during 1999-2002, average bank equity fund beta was 0.87, while its non-bank counterpart was 0.99. For 2003-2004, average bank fund beta was 1.16, while its non-bank counterpart was 0.78. With a beta less than one, non-bank equity funds maintained their systematic risk below that of the market. Even

118

Chapter 5 though bank funds initially had less systematic risk than their non-bank counterparts, systematic risk of bank equity funds seemed to have exceeded both the market and nonbank funds during the later period. Table 5.19 Two-sample t-test for beta of bank and non-bank funds
Ho: no difference between mean betas of bank and non-bank funds Ha: mean betas of bank and non-bank funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 =0.05 Bank Non-bank Bank Non-bank Mean beta 0.872 0.993 1.158 0.780 Variance 0.072 0.025 0.055 0.052 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 9 9 t Stat -1.056 2.781 P(Tt) one-tail 0.159 0.011 t Critical one-tail 1.833 1.833 P(Tt) two-tail 0.318 0.021 t Critical two-tail 2.262 2.262

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Performing a two-sample t-test assuming unequal variances for betas of bank and nonbank funds showed no significant difference in betas between bank and non-bank equity funds during 1999-2002 (t = -1.056, p = 0.318), but during 2003-2004, average bank fund beta was significantly higher than its non-bank counterpart (t = 2.781, p = 0.011), providing evidence that bank equity funds were systematically more risky than their nonbank counterparts during the later period. As measured by variance of quarterly returns in Table 5.10 (page 111), bank-managed funds had higher total risk than non-bank counterparts during both periods. This differed from Fryes (2001) results, which found bank-managed bond funds more conservative than non-bank counterparts in the USA. Expense ratio of bank and non-bank domestic equity funds Referring to Table 5.20 on the next page, during 1999-2002, average expense ratio for bank equity funds was 2.0 percent, while its non-bank counterpart was 1.74 percent. For 2003-2004, average bank fund expense ratio was 2.07 percent, while expense ratio for

119

Data analysis non-bank funds was 1.34 percent. Expense ratio for bank equity funds was generally higher than non-bank counterparts. While proportionate expenditures were reduced for non-bank funds, they were increased for bank funds. Table 5.20 Two-sample t-test for expense ratio of bank and non-bank funds
Ho: no difference between mean expense ratios of bank and non-bank funds Ha: mean expense ratios of bank and non-bank funds differ 1999:Q2--2002:Q1 2003:Q1--2004:Q3 =0.05 Bank Non-bank Bank Non-bank Mean expense ratio 2.004 1.744 2.075 1.337 Variance 1.213 0.229 2.739 0.272 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 8 5 t Stat 0.582 0.963 P(T<=t) one-tail 0.288 0.190 t Critical one-tail 1.860 2.015 P(T<=t) two-tail 0.576 0.380 t Critical two-tail 2.306 2.571

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Performing a two-sample t-test assuming unequal variances for expense ratios of bank and non-bank funds showed no significant difference in proportionate expenditures between the two groups during each holding period (t = 0.582, p = 0.576 during 1999-2002; t = 0.963, p = 0.380 during 2003-2004). To summarize, this subsection showed that expenditures, risk and size characteristics of bank funds were generally not inferior to non-bank funds, as explained in section 6.2.4.

5.5 Conclusion
This chapter presented results for testing four sets of hypotheses on mutual fund performance. First, domestic equity fund performance was generally unimpressive, even though their returns were better than guaranteed interest rates. Second, bank funds were comparable to non-bank funds in terms of performance. Third, expense ratio and size characteristics were not strong determinants of fund performance. Last, risk, expense ratio

120

Chapter 5 and size characteristics of bank funds were not inferior to their non-bank counterparts. Table 5.21 provides a summary of hypothesis test results. Table 5.21 Results of hypothesis testing
Hypothesis Result 1.1 Domestic equity funds do not outperform their benchmark stock market index 1.2 High positive correlation between domestic equity fund returns and returns to the domestic stock market index 1.3 Returns from investing savings in domestic equity funds are higher than guaranteed interest rates 2.1 No significant difference in returns between domestic equity funds managed by banks and non-banks 2.2 No significant difference in information ratios between domestic equity funds managed by banks and non-banks 2.3 No significant difference in Jensen alphas between domestic equity funds managed by banks and non-banks during 2003-2004, but significantly lower Jensen alpha of bank funds relative to non-bank ones during 1999-2002 2.4 No significant difference in Sharpe ratios between domestic equity funds managed by banks and non-banks during 1999-2002 but significantly lower Sharpe ratio of bank funds relative to non-bank ones during 2003-2004 2.5 No significant difference in Treynor ratios between domestic equity funds managed by banks and non-banks 3.1 No significant difference in returns between small and large funds during 2003-2004 but significant out-performance of small funds by large funds during 1999-2002 3.2 No significant difference in returns between funds with high expense ratios and those with low expense ratios 3.3 No significant difference in expense ratios of big and small funds 4.1 No significant difference in size between domestic equity funds managed by banks and non-banks 4.2 No significant difference in betas between domestic equity funds managed by banks and non-banks during 1999-2002 but significantly higher beta of bank funds relative to non-bank ones during 2003-2004 4.3 No significant difference in expense ratios of bank and non-bank funds

Source: developed for this research. The next chapter continues the analysis by integrating results from each set of hypotheses with existing literature to form conclusions and recommendations.

121

Chapter 6 Conclusions
6.1 Introduction
Results from the previous chapter were derived for this research problem: Factors influencing mutual fund performance differences, based on type of financial institution managing the fund. Research questions arising from this problem were: 1. How do Singapores unit trusts vary in terms of performance? 2. How does performance of funds managed by banks compare with their non-bank counterparts? 3. What are important factors influencing performance of mutual funds? 4. How do differences in fund performance determinants account for performance differences among funds managed by various types of financial institutions? Answering these questions involved collecting five years of secondary data from 1999 to 2004 for quarterly returns, net assets under management and expense ratios of 19 retail domestic equity funds approved for Singapores CPF Investment Scheme. Regression analysis was conducted for quarterly risk premiums of each fund and domestic stock market index, followed by computation of risk-adjusted performance measures before hypothesis testing was carried out using two-tail pooled-variance t-test for difference in two means. This chapter draws upon results from the previous chapter to make generalizations about mutual fund performance. More importantly, this chapter provides justification for conducting this researchknowledge contribution as well as benefits to finance theory and practice. Figure 6.1 on the following page provides an outline of this chapter.

122

Chapter 6 Following this introduction, results from the previous chapter are integrated to form conclusions for each research question in section 6.2. Conclusions from section 6.2 are then combined to answer the research problem of this thesis in section 6.3. Figure 6.1 Conclusions map
6 Conclusions 6.1 Introduction 6.2 Conclusions about the research questions 6.3 Conclusions about the research problem 6.2.1 Overall performance of domestic equity funds 6.2.2 Performance of bank and non-bank domestic equity funds 6.2.3 Factors affecting performance of domestic equity funds 6.2.4 factors differentiating bank and non-bank fund performance 6.4.1 Performance of domestic equity funds 6.4.2 Performance comparison of bank and nonbank funds 6.4.3 Expense ratio, size and fund performance 6.4.4 Comparison of bank and non-bank fund characteristics 6.5.1 Implications for financial institutions 6.5.2 Implications for individual investors 6.5.3 Implications for government policies

6.4 Implications for finance theory

6.5 Implications for policy and practice 6.6 Research limitations 6.7 Implications for further research 6.8 Implications for research methodology

Source: developed for this research. After answering the research problem, contributions from this thesis to finance literature are examined in section 6.4 in terms of the research questions. Benefits from this research for financial practice are presented in section 6.5, before section 6.6 acknowledges limitations in conducting this research. Lastly, section 6.7 and 6.8 provide suggestions for further research and statistical methodology.

123

Conclusions

6.2 Conclusions about the research questions

In the previous chapter, section 5.4 presented separately test results for four sets of hypotheses concerning (1) domestic equity fund performance, (2) performance comparison of bank and non-bank funds, (3) size and expenditures characteristics as determinants of fund performance, and (4) comparison of characteristics for bank and non-bank funds. This section concludes each research question by integrating findings for each set of hypotheses, in the same order as questions were presented.

6.2.1 Overall performance of domestic equity funds

For the first set of hypotheses, overall performance of domestic equity funds was compared with local stock market index and guaranteed interest rates. Performance comparisons for these funds with the index were carried out using quarterly returns as well as the Jensen alpha and information ratio criteria (H1.1), while comparisons with guaranteed interest rates used the Sharpe and Treynor ratios (H1.3). While comparing with the market index, correlation between fund and index returns was ascertained (H1.2). Concluding results from section 5.4.1, comparison with the market using quarterly returns yielded mixed results, as these funds on average under-performed and outperformed the market during the first and second period respectively. Using Jensen alpha revealed the underlying pattern of such mixed performancemajority of funds registered abnormal returns above expectation for both periods, but only one of them was statistically significant for each period. The information ratio was consistent with Jensen alpha as funds having positive Jensen alphas also had positive information ratios. Therefore, even though the average fund in the sample can add value to its investment, its performance was probably not acceptable.

124

Chapter 6 According to Sharpe and Treynor ratios, returns from fund investments were better than earning guaranteed interest rates during the two periods. Besides, the funds assets were generally well diversified in the local equity market, as returns correlated highly with market index, even though consistency was lacking according to all the performance criteria.

6.2.2 Performance of bank and non-bank domestic equity funds

For the second set of hypotheses, performance comparison was carried out for domestic equity funds of banks and non-banks using five criteria: (1) quarterly returns, (2) information ratio, (3) Jensen alpha, (4) Sharpe ratio, and (5) Treynor ratio, corresponding to the set of hypotheses H2.1 to H2.5. Concluding results from section 5.4.2, there was no significant difference between returns from bank and non-bank domestic equity funds during each period. This result was confirmed using information and Treynor ratios, which showed no significant ratio differences for the two fund groups during each period. However, test results for Jensen alpha and Sharpe ratio were contradictory, as underperformance of bank funds were detected using Jensen alpha and Sharpe ratio for 1999-2002 and 2003-2004 respectively, while no significant performance difference was detected for 1999-2002 and 2003-2004 using Sharpe ratio and Jensen alpha respectively. This difference in results using Jensen alpha and Sharpe ratio is attributable to their different measurement of risk: Jensen alphas emphasis on systematic risk as measured by beta in Sharpes (1964) CAPM model, in contrast to Sharpe ratios emphasis on total risk as measured by standard deviation of returns (Reilly & Brown 2003). Table 6.1 on the next page summarizes findings for performance comparison of bank and non-bank domestic equity funds.

125

Conclusions Table 6.1 Performance comparison of bank and non-bank domestic equity funds
Performance measure Quarterly returns Information ratio Jensen alpha Sharpe ratio Treynor ratio 1999-2002 No significant difference No significant difference Significant bank fund underperformance No significant difference No significant difference 2003-2004 No significant difference No significant difference No significant difference Significant bank fund underperformance No significant difference

Source: developed for this research. As four out of five criteria reported no significant performance difference between bank and non-bank domestic equity funds during each period, overall results for the second set of hypotheses showed bank funds were comparable to their non-bank counterparts in terms of performance.

6.2.3 Factors affecting performance of domestic equity funds

For the third set of hypotheses, performance comparison of large and small funds (H3.1) as well as funds with high and low expense ratios (H3.2) were carried out before expense ratios of large and small funds were compared (H3.3). These hypotheses tested the importance of expense ratio and size characteristics in determining fund performance. Summarizing results from section 5.4.3, first there was significant out-performance of small funds by large funds during 1999-2002, but no significant performance difference during 2003-2004. Second, there was no significant performance difference for funds with high and low expense ratios during both periods. Third, no significant expense ratio difference for large and small funds were reported during both periods. Table 6.2 on the following page summarizes findings for plausible relations between expense ratio, size and fund performance. Even though there was evidence of economies of scale as large funds outperformed small funds during 1999-2002, the phenomenon was not repeated for the other period, indicating size was not a strong determinant of performance.

126

Chapter 6 Table 6.2 Relation between fund expense ratio, size and performance
Comparison Performance of large and small funds Performance of high and low expense ratio funds Expense ratio of large and small funds 1999-2002 Significant out-performance of small funds by big funds No significant difference No significant difference 2003-2004 No significant difference No significant difference No significant difference

Source: developed for this research. Expense ratio was a weaker performance determinant than size as there was no significant performance difference between high and low expense ratio funds for both periods. Besides, there was no relation between expense ratio and size as no significant difference was reported in the ratio between large and small funds for the two periods, suggesting negligible or no economies of scale. Generally, expense ratio and size were not strong determinants of fund performance.

6.2.4 Factors differentiating bank and non-bank fund performance

Even though section 6.2.2 concluded there was no significant performance difference between bank and non-bank funds, while section 6.2.3 reported expense ratio and fund size characteristics as weak performance determinants, there may still be differences in fund characteristics affecting performance for bank and non-bank funds. For the fourth and last set of hypotheses, size (H4.1), beta (H4.2) and expense ratio (H4.3) of bank and non-bank funds were compared. Summarizing results from section 5.4.4, there was no significant difference in size between bank and non-bank domestic equity funds for both periods. In terms of beta, even though there was no significant difference between bank and non-bank funds during the first period, there was significant evidence that bank fund betas were higher than their nonbank counterparts. As for expense ratio, there was no significant difference in the ratio

127

Conclusions between bank and non-bank domestic equity funds for the two periods. Table 6.3 below summarizes findings for comparison of bank and non-bank fund characteristics. Table 6.3 Comparison of bank and non-bank fund characteristics
Characteristic Size Beta Expense ratio 1999-2002 No significant difference No significant difference No significant difference 2003-2004 No significant difference Significantly higher beta of bank funds No significant difference

Source: developed for this research. Testing of the last set of hypotheses showed there was no significant difference between bank and non-bank funds in terms of expense ratio and size. In terms of beta, there was evidence that bank funds had more systematic risk than non-bank funds during the later period, as evident by significantly higher beta of bank funds. Overall, these characteristics of bank funds were comparable to their non-bank counterparts. This section drew conclusions for each research question based on results for each set of hypotheses from section 5.4. These conclusions are linked together in the following section to answer the research problem.

6.3 Conclusions about the research problem

Conclusions for research questions from the previous section are used in this section to answer the research problem factors influencing mutual fund performance differences, based on type of financial institution managing the fund. As banks had a reputation for being under-performers relative to non-banks in Australia (Moullakis & Patten 2005) and the USA (Bauman & Miller 1995; McTague 1994) while Frye (2001) reported bond funds from banks in the USA did not under-perform their nonbank counterparts after controlling for differing fiduciary responsibility, this research attempted to detect any performance difference between bank and non-bank equity funds facing the same fiduciary responsibility as well as factors influencing such performance difference, if any.

128

Chapter 6 For this research, the funds had common fiduciary responsibility, as they were Singapores domestic equity unit trusts approved by the CPF Board for managing social security savings. These funds were managed by two types of financial institutionsbanks as well as non-banks (insurance and investment companies). Performance differences between bank and non-bank funds, if any, were attributable to differing fund characteristics, which were identified from finance literature as fund performance determinants. Fund characteristics of bank and non-bank funds considered for this research included (1) asset allocation, (2) systematic risk, (3) expenditures and (4) size, as measured by the funds targeted market index return, beta, expense ratio and value of net assets under management respectively. Five performance measures were used: (1) returns, (2) information ratio, (3) Jensen alpha, (4) Sharpe ratio and (5) Treynor ratio. This research contributes a revised mutual fund performance conceptual model in Figure 6.2. Figure 6.2 Conceptual model of mutual fund performance determinants
Risk-free return Market return Beta

Asset allocation & investment style Expected return Systematic risk Implicit cost Explicit cost

Return

Information ratio Jensen alpha

Size Expense ratio

Transaction costs

Fund performance Sharpe ratio

Bank Financial institution Nonbank Treynor ratio

Source: developed for this research.

129

Conclusions By considering conclusions from this research, this model corrects the initial model presented in Figure 3.7 (page 61). According to this finalized model, fund performance, which can be measured by periodic returns and risk-adjusted measures (information ratio, Jensen alpha as well as Sharpe and Treynor ratios), is influenced strongly by expected returns (indicated by solid line linking expected return to fund performance), which is derived from targeted market asset allocation (measured by market return) as well as the funds level of systematic risk (measured by beta). Fund performance is weakly affected by type of financial institution (bank or non-bank), transaction costs (measured by implicit costs such as size as well as explicit costs such as expense ratio), as indicated by dashed lines connecting financial institution and transaction costs to fund performance. This research found performance of Singapores CPF-approved domestic equity funds generally unimpressive, even though their returns were better than earning guaranteed interest rates. After controlling for differing fiduciary responsibilities, only a minority of performance measures indicated under-performance of band equity funds relative to their non-bank counterparts, supporting Fryes (2001) observation that bond funds from banks did not under-perform their non-bank counterparts. Supporting no significant performance difference between bank and non-bank domestic equity funds, this research found no significant difference in size and expense ratio characteristics of bank and non-bank funds. These characteristics were at most weak determinants of fund performance. Asset allocation, with its corresponding level of systematic risk, was the strongest determinant of fund performance, as it explains close to 90 percent of variability in returns for domestic equity funds. A notable difference among bank and non-bank funds was the significantly higher level of systematic risk in bank funds during the later period. Bank funds with higher levels of systematic risk may expect

130

Chapter 6 higher returns, but their actual returns did not outperform their non-bank counterparts, possibly due to bank-specific non-systematic factors.

6.4 Implications for finance theory

The previous section presented major conclusions from this research. In this section, contributions of these conclusions to finance literature are examined. These contributions are presented in the same order as the four research questions, corresponding to four conference papers based on this thesis.1

6.4.1 Performance of domestic equity funds

Implications for finance literature on overall performance of Singapores CPF-approved domestic equity unit trusts were presented in Tng (2005a). As majority of published research on fund performance were conducted using data from the USA and other large developed markets, leaving many small markets relatively unexplored, this research contributed to finance literature as there was little research published on Singapores fund industry. This research on performance of Singapores domestic equity funds agreed with research using USA data. Specifically, this research agreed with Carlson (1970) and Jensen (1968) in reporting unimpressive fund performance relative to local stock market index. Besides, this research confirmed findings from Brinson, Hood and Beebower (1986), Brinson, Singer and Beebower (1991) Ibbotson and Kaplan (2000) as well as Jensen (1968) in reporting very high correlation between fund and targeted market returns. In terms of research published on Singapores fund performance, this research agreed with Koh (1999) in reporting unimpressive fund performance relative to the STI index and very high correlation between returns of funds and their targeted market, but disagreed with the
1

Page vi provides a listing of research papers.

131

Conclusions earlier study by using data from 1999 to 2004 to show returns from investing in CPFapproved domestic equity funds were better than earning guaranteed interest rates. These contributions are summarized in Table 6.4 below. Table 6.4 Literature contribution on fund performance
Research in USA Unimpressive fund returns relative to local stock market (Carlson 1970; Jensen 1968) Very high correlation between returns of funds and targeted market index (Brinson, Hood & Beebower 1986; Brinson, Singer & Beebower 1991; Ibbotson & Kaplan 2000; Jensen 1968) Research in Singapore Unimpressive fund returns relative to local stock market (Koh 1999; Tng 2005a) Very high correlation between returns of funds and targeted market index (Koh 1999; Tng 2005a) Unit trusts mostly earned lower returns than CPF interest rates (Koh 1999) Unit trusts mostly earned higher returns than CPF interest rates (Tng 2005a)

Source: developed for this research. Besides contributions to finance literature on fund performance, this thesis contributed to research on performance comparisons of funds from banks as well as insurance and investment companies.

6.4.2 Performance comparison of bank and non-bank funds

Implications for finance literature on performance comparison of Singapores CPFapproved domestic equity funds from banks and non-banks were presented in Tng (2005d). This research on Singapores bank and non-bank domestic equity funds agreed with Fryes (2001) study for bond funds in the USA by reporting bank fund performance was not inferior to non-bank funds after controlling for differing fiduciary responsibilities. As a result, this research did not agree with studies and press reports on underperformance of bank funds relative to their non-bank counterparts in Australia (Moullakis & Patten 2005) and the USA (Bauman & Miller 1995; Bogle & Twardowski 1980; McTague 1994). These contributions are summarized in Table 6.5 on the next page. This research provided a unique opportunity to control for differing fiduciary responsibilities by examining domestic equity funds approved by Singapores CPF Board.

132

Chapter 6 After contributing to existing literature on performance of bank and non-bank funds, this research proceeded to explore quantitative mutual fund characteristics as performance determinants.

6.4.3 Expense ratio, size and fund performance

Implications for finance literature on the relation between expense ratio, size and fund performance were presented in Tng (2005c). Table 6.5 Literature contribution on performance of bank and non-bank funds
Bank funds under-perform non-bank funds Australia (Moullakis & Patten 2005) USA (Bauman & Miller 1995; Bogle & Twardowski 1980; McTague 1994) Bank funds do not under-perform non-bank funds Singapore (Tng 2005d) USA (Frye 2001)

Source: developed for this research. Even though there was evidence that large funds outperformed small funds for Singapores domestic equity funds during one holding period, the better performance of large funds was not significant for all holding periods tested, which was consistent with various findings from the USA (Cicotello & Grant 1996; Droms & Walker 1994; Grinblatt & Titman 1994), Australia (Bird, Chin & McCrae 1983; Gallagher 2003; Gallagher & Martin 2005) and Sweden (Dahlquist, Engstrom & Soderlind 2000); while the relation between expense ratio and returns for Singapores domestic equity funds was insignificant, supporting findings in the USA from Ippolito (1989). As for plausible relation between expense ratio and size, even though large funds seemed to have lower expense ratios than small funds, suggesting economies of scale, the difference in expense ratio was insignificant. This result was not reported for research in Australia or the USA. Insignificance of fund size and expense ratio in affecting returns was attributable to dominance of the asset allocation determinant, which was explained in section 6.3. These contributions are summarized in Table 6.6 on the following page.

133

Conclusions Table 6.6 Literature contribution on fund characteristics and performance

Expense ratio No significant relation between fund expense ratio and performance: Singapore (Tng 2005c) USA (Ippolito 1989) No significant relation between expense ratio and fund size: Singapore (Tng 2005c) Size Generally no significant relation between fund size and performance: Australia (Bird, Chin & McCrae 1983; Gallagher 2003; Gallagher & Martin 2005) Singapore (Tng 2005c) Sweden (Dahlquist, Engstrom & Soderlind 2000) USA (Cicotello & Grant 1996; Droms & Walker 1994; Grinblatt & Titman 1994)

Source: developed for this research. Contributions of mutual fund characteristics as performance determinants led to comparison of such characteristics among bank and non-bank funds.

6.4.4 Comparison of bank and non-bank fund characteristics

Implications for literature on comparison of bank and non-bank fund characteristics were presented in Tng (2005b). Existing literature compared performance of bank and non-bank funds, as pointed out in section 6.4.2. However, there was little literature on comparing expenditures and size characteristics of these fund groups. This research contributed to the perceived literature gap by comparing these characteristics for Singapores domestic equity funds managed by banks and non-banks, concluding these characteristics of bank funds were not inferior to their non-bank counterparts. Specifically, there was no significant difference in expense ratio and size for bank and non-bank funds. However, there was evidence bank funds were becoming riskier than their non-bank counterparts, refuting Fryes (2001) finding for bank funds in the USA being less risky than non-bank ones. With four conference papers contributing to finance theory from four aspects of investments, this DBA research made a significant contribution to knowledge in its immediate investments field as well as its parent discipline of finance. The following section presents benefits of this research to financial practices.

134

Chapter 6

6.5 Implications for policy and practice

This section examines implications for policy and practice from three perspectives: (1) financial institutions, (2) individual investors and (3) government policies.

6.5.1 Implications for financial institutions

Evaluating performance of unit trusts managed by local banks was an appropriate research area, as it has important implications for management of local banks. For this study, all CPF-approved bank equity funds were managed by local banks, which face increasing competition from foreign financial institutions with Singapores gradual deregulation of its financial sector. As distinction between banks and non-banks diminished, the mutual fund market is strategically important for financial institutions. According to Gallo, Apilado and Kolari (1996), offering mutual fund products improves profitability of banks. Profitability of banks can be significantly improved if banks shed their image as under-performers in the fund management industry. Banks can offer sustainable mutual funds if they were competitive products that did not under-perform funds from insurance and investment companies. This research has important implications for banks in sustaining their mutual fund products, with the fund management industrys increasing importance for banks profitability. Banks are capable of offering competitive fund products. Sustaining their share of the mutual fund industry requires banks to shed their unjustified reputation as non-aggressive fund managers incapable of generating impressive returns, as this study showed characteristics of domestic bank equity funds were not inferior to their non-bank counterparts. Otherwise, poor reputation and investor apathy can result in small funds being terminated eventually. Switching from competition between banks and non-banks in Singapore to an international phenomenon, it is possible for domestic banks to earn an economic profit with financial

135

Conclusions deregulation, if they depart from perfect competition and gain a competitive advantage over local as well as foreign financial institutions. Domestic banks face the challenge to withstand erosion of investor monies by more competitors. However, there are opportunities for gaining competitive advantage. Key ingredients for competitive advantage, as identified by Porter (1980), are industry characteristics, product differentiation and cost advantages. In terms of the fund industry, financial deregulation can reduce barriers to entry for foreign institutions. With proliferation of fund products from local and foreign institutions, investors need to be informed about the relative performance of bank and non-bank funds. Bank funds were comparable in performance to their non-bank counterparts. However, in terms of investment strategy, there was evidence bank fund managers were more risky than their non-bank counterparts. To gain a competitive advantage, local banks can differentiate their financial products. Even though technology and human resources are relatively uniform across the fund industry, portfolio management costs can be reduced with economies of scale by offering bigger funds. Bank funds were growing steadily. Such growth is sustained when banks develop a quality reputation for fund products. Fund characteristics that determine performance can be differentiated. Determination of fund characteristics that influence performance will open opportunities for further research beneficial to financial institutions. Besides financial institutions, benefits from this research accrue to retail investors.

6.5.2 Implications for individual investors

This research found bank-managed domestic equity funds approved for Singapores CPF Investment Scheme did not under-perform their non-bank counterparts. However, in terms of investment strategy, there was evidence bank fund managers were not more

136

Chapter 6 conservative than their non-bank counterparts. In terms of fund size and expense ratio, there was no significant difference between bank and non-bank funds. Knowledge of these results assist individual investors in building a diversified portfolio of investments, with a portion comprising of mutual funds with different characteristics. These investors take advantage of government policies for investing social security savings to generate more returns than earning guaranteed interest rates.

6.5.3 Implications for government policies

With CPF-approved domestic equity funds failure to outperform their market index and the market determining most of the variability in returns from a typical fund, actively managed funds incurred substantial research and trading costs that failed to translate into additional returns. Instead, index funds, which are passively managed, can be promoted and approved for Singapores CPF Investment Scheme as they are less costly than actively managed funds in terms of research and trading costs, and attain comparable or even better performance than actively managed funds generally. As index funds were not popular in Singapore, the CPF Board can promote advantages of investing in index funds and encourage introduction of index funds by FMCs to be approved for the CPF Investment Scheme. While benefits of this research have practical relevance, there were limitations in conducting this research.

6.6 Research limitations

Even though examining Singapores CPF-approved domestic equity funds offered an opportunity to control for differing fiduciary responsibilities, the countrys fund industry was small and young. This research worked with 19 funds from 1999 to 2004 after liberalization of the CPF Investment Scheme. Plotting histograms of data showed they did

137

Conclusions not follow the normal distribution closely, departing slightly from statistical assumptions. However, it is unlikely for statistical assumptions to be exactly satisfied in empirical research, as highlighted in section 1.6.2 (p. 13). Multivariate analysis was not carried out for this research as such analysis can be limited by insufficient quarterly data (around 20 eligible funds since 1999) with characteristics not very normally distributed from a relatively small and new Singapore market. More data collected for other periods and markets as part of future research may confirm results of this study.

6.7 Implications for further research

This research examined effects of fund characteristics on performance in a small equity market. In such a market, large fund size may not provide significant economies of scale to lower expenses or provide better returns than small funds. Higher expenses may not produce better returns. For further research, turnover ratio can be incorporated as a measure of explicit transaction cost to examine significance of this factor as a fund performance determinant in various equity markets. Other factors to be considered include past performance, measured by best or worst return, as well as a composite gain/loss ratio. As suggested in Chapter 3, testing Singapores stock market efficiency using more recent data at various time intervals can be carried out for further research.

6.8 Implications for research methodology

Additional mutual fund performance determinants were suggested in the previous subsection for further research. However, regression of multivariate data corresponding to dozens of fund characteristics may be complicated by substantial redundancy among these characteristic variables, resulting in high levels of correlation and multicollinearity. Therefore, principal components analysis can be used to re-express data by taking linear

138

Chapter 6 combinations of original variables to produce a small number of components that account for the greatest amount of available information, resulting in a more manageable data analysis process (Lattin, Carroll & Green 2003). In addition, factor analysis identifies underlying sources of variance common to a number of variables when observed variations in each variable are attributable to underlying common factors (Lattin, Carroll & Green 2003). Such multivariate analysis involves fitting secondary data to candidate models before using a partial F test to compare significance of additional terms in each model. Continuing research to find an appropriate mutual fund factor model supported by data becomes an iterative process based on theoretical notions, with data being used to refine each model before testing the refined model with additional data. For more complete understanding from different perspectives, this thesis recommends future research projects incorporating various approaches, including secondary data, case and survey research. For this research, analysis explored a linear functional form based on Sharpes (1964) CAPM model. Multivariate analysis and non-linear regression can be carried out for future research on more established fund markets with sufficient data on multiple fund characteristics over longer periods.

139

References

Anderson, S, Coleman, B, Frohich, C & Steagall, J 2001, 'A multifactor analysis of country fund returns', Journal of Financial Research, vol. 24, no. 3, pp. 331-46. Applications for Admission of Fund Management Companies into CPF Investment Scheme, 2001, CPF Board, Singapore, 7 Feb. Ariff, M 1986, 'Announcement effects in a thin market--a study of the Singapore equity market', paper presented to Proceedings of the Academy of International Business, Southeast Asia Regional Conference, Taipei, 26-28 June. Atkinson, S, Baird, S & Frye, M 2003, 'Do female mutual fund managers manage differently?' Journal of Financial Research, vol. 26, no. 1, pp. 1-18. Avera, W 1994, 'Definition of industry ethics and development of a code', in K Baker (ed.), Good Ethics: the Essential Element of a Firm's Success, Association for Investment Management and Research, Virginia. Bauman, W & Miller, R 1995, 'Portfolio performance rankings in stock market cycles', Financial Analysts Journal, vol. 51, no. 2, pp. 79-87. Berkowitz, M & Qiu, J 2003, 'Ownership, risk and performance of mutual fund management companies', Journal of Economics and Business, vol. 55, pp. 109-34. Bird, R, Chin, H & McCrae, M 1983, 'The performance of Australian superannuation funds', Australian Journal of Management, vol. 8, pp. 49-69. Blake, C, Elton, E & Gruber, M 1993, 'The performance of bond mutual funds', Journal of Business, vol. 66, no. 3, pp. 371-403. Bogle, J 1998, 'The implications of style analysis for mutual fund performance', Journal of Portfolio Management, vol. 24, no. 4, pp. 34-42. Bogle, J & Twardowski, J 1980, 'Institutional investment performance: banks, investment counselors, insurance companies and mutual funds', Financial Analysts Journal, vol. 36, no. 1, pp. 33-41. Brinson, G, Hood, L & Beebower, G 1986, 'Determinants of portfolio performance', Financial Analysts Journal, vol. 42, no. 4, pp. 39-48. Brinson, G, Singer, B & Beebower, G 1991, 'Determinants of portfolio performance II: an update', Financial Analysts Journal, vol. 47, no. 3, pp. 40-8. Brown, S & Goetzmann, W 1995, 'Performance persistence', Journal of Finance, vol. 50, no. 2, pp. 679-98. ---- 1997, 'Mutual fund styles', Journal of Financial Economics, vol. 43, no. 1, pp. 373-99. Brownell, P & Trotman, K 1988, 'Research methods in behavioral accounting', in K Ferris (ed.), Behavioral Accounting Research: a Critical Analysis, Century VII, Ohio. Burnmeister, E, Roll, R & Ross, S 1994, 'A practitioner's guide to arbitrage pricing theory', in J Peavy (ed.), A Practitioner's Guide to Factor Models, Research Foundation of the Institute of Chartered Financial Analysts, Charlottesville. Burns, A & Bush, R 2000, Marketing Research, 3rd edn, Prentice Hall, New Jersey. Cai, J, Chan, K & Yamada, T 1997, 'The performance of Japanese mutual funds', Review of Financial Studies, vol. 10, no. 2, pp. 237-74. Carhart, M 1997, 'On persistence in mutual fund performance', Journal of Finance, vol. 52, no. 1, pp. 57-82. Carhart, M, Carpenter, J, Lynch, A & Musto, D 2002, 'Mutual fund survivorship', Review of Financial Studies, vol. 15, no. 5, pp. 1439-63. Carlson, R 1970, 'Aggregate performance of mutual funds, 1948-1967', Journal of Financial and Quantitative Analysis, vol. 5, no. 1, pp. 1-32. Chan, L, Chen, H & Lakonishok, J 2002, 'On mutual fund investment styles', Review of Financial Studies, vol. 15, no. 5, pp. 1407-37.

140

References
Changes to the CPF Investment Scheme, 2002, CPF Board, Singapore, 29 Aug. Chen, N 1983, 'Some empirical tests of the theory of arbitrage pricing', Journal of Finance, vol. 38, no. 5, pp. 1393-414. Chen, N, Roll, R & Ross, S 1986, 'Economic forces and the stock market', Journal of Business, vol. 59, no. 3, pp. 383-404. Chevalier, J & Ellison, G 1999, 'Are some mutual fund managers better than others? Crosssectional patterns in behavior and performance', Journal of Finance, vol. 54, no. 3, pp. 875-99. Chow, H 2003, 'Impact of falling stock markets: Invesco pulls plug on unit trusts here', The Straits Times, 7 Jun, p. A40. Christopherson, J, Ferson, W & Glassman, D 1998, 'Conditioning manager alphas on economic information: another look at the persistence of performance', Review of Financial Studies, vol. 11, no. 1, pp. 111-42. Cicotello, C & Grant, C 1996, 'Equity fund size and growth: implications for performance and selection', Financial Services Review, vol. 5, pp. 1-12. Code of Consumer Banking Practice, 2002, Association of Banks in Singapore, Singapore. Code of Ethics, 2002, Association for Investment Management and Research, viewed 14 Nov 2002, <http://www.aimr.org/standards/ethics/code/>. Corrado, C & Jordan, B 2005, Fundamentals of investments Valuation and Management, 3 edn, McGraw-Hill Irwin, New York. CPF Handbook: Building Our Future, 2005, CPF Board, viewed 21 Mar 2005, <http://www.cpf.gov.sg>. CPF Investment Scheme, 2005, CPF Board, viewed 2 May 2005, <http://www.cpf.gov.sg>. CPF Investment Scheme Risk Classification System, 2004, CPF Board, viewed 21 Mar 2005, <http://www.cpf.gov.sg>. CPF Investment Scheme: Realized Profit/Loss for the Financial Year Ended 30 September, 19992003, CPF Board, Singapore. CPF Savings can now be Invested in Exchange Traded Funds, 2001, CPF Board, Singapore, 27 Sep. Cumby, R & Glen, J 1990, 'Evaluating the performance of international mutual funds', Journal of Finance, vol. 45, no. 2, pp. 497-522. Dahlquist, M, Engstrom, S & Soderlind, P 2000, 'Performance and characteristics of Swedish mutual funds', Journal of Financial and Quantitative Analysis, vol. 35, no. 3, pp. 409-23. Dibartolomeo, D & Witkowski, E 1997, 'Mutual fund misclassification: evidence based on style analysis', Financial Analysts Journal, vol. 53, no. 5, pp. 32-43. Douglas, P 2003, 'Challenges ahead for unit trusts', The Business Times, 26 Feb. Droms, W & Walker, D 1994, 'Investment performance of international mutual funds', Journal of Financial Research, vol. 17, no. 1, pp. 1-14. Dunn, P & Theinsen, R 1983, 'How consistently do active managers win?' Journal of Portfolio Management, vol. 9, pp. 47-50. Easterby-Smith, M, Thrope, R & Lowe, A 1991, Management Research: an Introduction, Sage, London. EDB 2004, Singapore Facts & Figures, Singapore Economic Development Board, viewed 28 Jun 2005, <http://www.edb.gov.sg>. Elton, E, Gruber, M & Blake, C 1996, 'The persistence of risk-adjusted mutual fund performance', Journal of Business, vol. 69, no. 2, pp. 133-57. ---- 2001, 'A first look at the accuracy of the CRSP mutual fund database and a comparison of the CRSP and Morningstar mutual fund databases', Journal of Finance, vol. 56, no. 6, pp. 2415-30. Elton, E, Gruber, M, Das, S & Hlavka, M 1993, 'Efficiency with costly information: a reinterpretation of evidence from managed portfolios', Review of Financial Studies, vol. 6, no. 1, pp. 1-22. Fama, E 1970, 'Efficient capital markets: a review of theory and empirical work', Journal of Finance, vol. 25, no. 2, pp. 383-417. ---- 1972, 'Components of investment performance', Journal of Finance, vol. 27, no. 3, pp. 551-67.

141

References
---- 1991, 'Efficient capital markets: II', Journal of Finance, vol. 46, no. 5, pp. 1575-617. Fama, E & French, K 1992, 'The cross-section of expected stock returns', Journal of Finance, vol. 47, no. 2, pp. 427-65. ---- 1993, 'Common risk factors in the returns on stocks and bonds', Journal of Financial Economics, vol. 33, no. 1, pp. 3-56. Ferson, W & Schadt, R 1996, 'Measuring fund strategy and performance in changing economic conditions', Journal of Finance, vol. 51, no. 2, pp. 425-62. Friend, I & Blume, M 1970, 'Measurement of portfolio performance under uncertainty', American Economic Review, vol. 60, no. 3, pp. 561-75. Friend, I, Blume, M & Crockett, J 1970, Mutual Funds and Other Institutional Investors, McGrawHill, New York. Frye, M 2001, 'The performance of bank-managed mutual funds', Journal of Financial Research, vol. 24, no. 3, pp. 419-42. Gallagher, D 2003, 'Investment manager characteristics, strategy, top management changes and fund performance', Accounting & Finance, vol. 43, no. 3, pp. 283-309. Gallagher, D & Martin, K 2005, 'Size and investment performance: a research note', Abacus, vol. 41, no. 1, pp. 55-65. Gallo, J, Apilado, V & Kolari, J 1996, 'Commercial bank mutual fund activities: implications for bank risk and profitability', Journal of Banking and Finance, vol. 20, pp. 1775-91. Goetzman, W & Peles, N 1997, 'Cognitive dissonance and mutual fund investors', Journal of Financial Research, vol. 20, no. 2, pp. 145-58. Goodwin, T 1998, 'The information ratio', Financial Analysts Journal, vol. 54, no. 4, pp. 34-43. Grinblatt, M & Titman, S 1992, 'The persistence of mutual fund performance', Journal of Finance, vol. 47, no. 5, pp. 1977-84. ---- 1994, 'A study of monthly fund returns and performance evaluation techniques', Journal of Financial and Quantitative Analysis, vol. 29, no. 3, pp. 419-44. Grinold, R & Kahn, R 1994, 'Multiple-factor models for portfolio risk', in J Peavy (ed.), A Practitioner's Guide to Factor Models, Research Foundation of the Institute of Chartered Financial Analysts, Charlottesville. ---- 1995, Active Portfolio Management, Probus Publishing, Chicago. Grossman, S & Stiglitz, J 1980, 'On the impossibility of informationally efficient markets', American Economic Review, vol. 70, no. 3, pp. 393-408. Gruber, M 1996, 'Another puzzle: the growth in actively managed mutual funds', Journal of Finance, vol. 51, no. 3, pp. 783-810. Guba, E & Lincoln, Y 1994, 'Competing paradigms in qualitative research', in N Denzin & Y Lincoln (eds), Handbook of Qualitative Research, Sage, Thousand Oaks, pp. 105-17. Guercio, D & Tkac, P 2002a, Star tower: the effect of Morningstar ratings on mutual fund flows, Federal Reserve Bank of Atlanta. ---- 2002b, 'The determinants of the flow of funds of managed portfolios: mutual funds vs. pension funds', Journal of Financial and Quantitative Analysis, vol. 37, no. 4, pp. 523-57. Healy, M & Perry, C 2000, 'Comprehensive criteria to judge the validity and reliability of qualitative research within the realism paradigm', Qualitative Market Research: an International Journal, vol. 3, no. 3, pp. 118-26. Hendricks, D, Patel, J & Zeckhauser, R 1993, 'Hot hands in mutual funds: short-run persistence of relative performance, 1974-1988', Journal of Finance, vol. 48, no. 1, pp. 93-130. Holiday, K 1994, 'Mutual funds', Bank Marketing, vol. 26, pp. 23-31. Hunt, S, Chonko, L & Wilcox, J 1984, 'Ethical problems of marketing researcher', Journal of Marketing Research, vol. 21. Ibbotson, R & Brinson, G 1993, Global Investing, McGraw-Hill, New York. Ibbotson, R & Kaplan, P 2000, 'Does asset allocation policy explain 40, 90 or 100 percent of performance?' Financial Analysts Journal, vol. 56, no. 1, pp. 26-33. ICI 2004, Mutual Fund Fact Book, Investment Company Institute, viewed 15 Aug 2005, <http://www.ici.org>. Indro, D, Jiang, C, Hu, M & Lee, W 1999, 'Mutual fund performance: does fund size matter?' Financial Analysts Journal, vol. 55, no. 3, pp. 74-87.

142

References
Ippolito, R 1989, 'Efficiency with costly information: a study of mutual fund performance, 19651984', Quarterly Journal of Economics, vol. 104, no. 1, pp. 1-23. ---- 1992, 'Consumer reaction to measures of poor quality: evidence from the mutual fund industry', Journal of Law and Economics, vol. 35, pp. 45-69. Jain, P & Wu, J 2000, 'Truth in mutual fund advertising: evidence on future performance and fund flows', Journal of Finance, vol. 23, no. 2, pp. 389-416. Jensen, M 1968, 'The performance of mutual funds in the period 1945-1964', Journal of Finance, vol. 23, no. 2, pp. 389-416. ---- 1969, 'Risk, the pricing of capital assets, and the evaluation of investment portfolios', Journal of Business, vol. 42, pp. 167-247. Jones, C 1998, Investments Analysis and Management, 6th edn, John Wiley & Sons, New York. ---- 2003, Your Money, Your Choice... Mutual Funds: Take Control Now and Build Wealth Wisely, Financial Times Prentice Hall, New Jersey. Klemkosky, R 1973, 'The bias in composite performance measures', Journal of Financial and Quantitative Analysis, vol. 8, no. 3, pp. 505-14. Koh, S 1999, 'A survey of unit trusts in Singapore', Singapore Management Review, vol. 21, no. 1, pp. 49-78. Koh, S & Fong, W 2003, Personal Financial Planning, 3 edn, Pearson-Prentice Hall, Singapore. Kon, S 1983, 'The market-timing performance of mutual fund managers', Journal of Business, vol. 56, no. 3, pp. 323-47. Kritzman, M 1990, 'Quantitative methods in performance measurement', in S Brown & M Kritzman (eds), Quantitative Methods for Financial Analysis, Dow Jones-Irwin, Illinois. Latane, H & Tuttle, D 1967, 'Criteria for portfolio building', Journal of Finance, vol. 22, no. 3, pp. 359-73. Lattin, J, Carroll, J & Green, P 2003, Analyzing Multivariate Data, Thomson Learning, California. Leland, H 1999, 'Beyond mean-variance: performance measurement in a nonsymmetrical world', Financial Analysts Journal, vol. 55, no. 1, pp. 27-36. Levine, D, Stephan, D, Krehbiel, T & Berenson, M 2002, Statistics for Managers Using Microsoft Excel, 3 edn, Prentice Hall, New Jersey. Lim, G 1985, 'A Box-Jenkins Approach to Stock Market Analysis on the SES', National University of Singapore. Lintner, J 1965, 'Security prices, risk and maximum gains from diversification', Journal of Finance, vol. 20, no. 4, pp. 587-615. Lynch, A & Musto, D 1997, How Investors Interpret Past Returns, New York University. Malkiel, B 1995, 'Returns from investing in equity mutual funds: 1971-1991', Journal of Finance, vol. 50, no. 2, pp. 549-72. Markowitz, H 1952, 'Portfolio selection', Journal of Finance, vol. 7, no. 1, pp. 77-91. ---- 1976, 'Investment for the long run: new evidence for an old rule', Journal of Finance, vol. 31, no. 5, pp. 1273-86. MAS 1998-2004, Survey of the Singapore Asset Management Industry, Monetary Authority of Singapore, viewed 4 May 2005, <http://www.mas.gov.sg>. McGrath, M & Viney, C 1998, Financial Institutions, Instruments and Markets, 2 edn, McGrawHill, Sydney. McKeon, K 2001, 'Praise at what price', Bloomberg Personal Finance, May, p. 34. McTague, J 1994, 'Laggards no longer', Barron's, vol. 74, pp. F22-F3. Mendenhall, W & Sincich, T 1996, A Second Course in Statistics: Regression Analysis, 5 edn, Prentice Hall, New Jersey. Mercer 1999-2002, Unit Trusts and Investment-linked Insurance Products Included Under CPFIS: Performance and Risk Monitoring Reports, Mercer Investment Consulting, viewed 19 Feb 2005, <http://www.mercerfundwatch.com/>. Meredith, G n.d, Southern Cross University Graduate College of Management DBA Research Design and Thesis Preparation: Seminar Notes, Tweed Gold Coast. Mishkin, F & Eakins, S 2003, Financial Markets + Institutions, 4 edn, Addison-Wesley, New York.

143

References
MOM 1998, Liberalisation of CPF Investment Scheme, Ministry of Manpower, Singapore, 19 May. Morgan, G & Smircich, L 1980, 'The case for qualitative research', Academy of Management Review, vol. 5, pp. 491-500. Mossin, J 1966, 'Equilibrium in a capital asset market', Econometrica, vol. 34, no. 4, pp. 768-83. Moullakis, J & Patten, S 2005, 'Big-spending banks caught with wealth of problems', The Australian Financial Review, 8 Aug, p. 1. Ng, S 2003, 'Most CPF-approved unit trusts lose money over 3 years', The Business Times, 28 Jan. Patton, M 1990, Qualitative Evaluation and Research Methods, Sage, Newbury Park. Perry, C 1998, 'A structured approach for presenting theses', Australasian Marketing Journal, vol. 6, no. 1, pp. 63-85. ---- 2001, 'Case research in marketing', The Marketing Review, vol. 1, pp. 303-23. Perry, C, Riege, A & Brown, L 1999, 'Realism's role among scientific paradigms in marketing research', Irish Marketing Review, vol. 12, no. 2, pp. 16-23. Peterson, J, Pietranico, P, Riepe, M & Xu, F 2002, 'Explaining after-tax mutual fund performance', Financial Analysts Journal, vol. 58, no. 1, pp. 75-86. Porter, M 1980, Competitive Strategy: Techniques of Analyzing Industries and Competitors, The Free Press, New York. Reilly, F & Brown, K 2003, Investment Analysis and Portfolio Management, 7th edn, SouthWestern, Ohio. Roll, R 1978, 'Ambiguity when performance is measured by the security market line', Journal of Finance, vol. 33, no. 4, pp. 1051-69. ---- 1980, 'Performance evaluation and benchmark errors', Journal of Portfolio Management, vol. 6, no. 4, pp. 5-12. ---- 1981, 'Performance evaluation and benchmark errors (II)', Journal of Portfolio Management, vol. 7, no. 2, pp. 17-22. Roll, R & Ross, S 1980, 'An empirical investigation of the arbitrage pricing theory', Journal of Finance, vol. 35, no. 5, pp. 1073-103. Ross, S 1976, 'The arbitrage theory of capital asset pricing', Journal of Economic Theory, vol. 13, no. 2, pp. 341-60. Rubinstein, M 1976, 'The strong case for the generalized logarithmic utility function as the premier model of financial markets', Journal of Finance, vol. 31, no. 2, pp. 551-71. S&P 2003-2004, Performance and Risk Monitoring Report for CPFIS-included Unit Trusts & Investment-linked Insurance Products, S&P Fund Services, viewed 19 Feb 2005, <http://www.imas.org.sg>. Saw, S & Tan, K 1986, 'The Stock Exchange of Singapore all-share price indices: testing of independence', Securities Industry Review, vol. 12, no. 1. Sawicki, J 2000, 'Investors' response to the performance of professional fund managers: evidence from the Australian wholesale funds market', Australian Journal of Management, vol. 25, no. 1, pp. 47-66. ---- 2001, 'Investors' differential response to managed fund performance', Journal of Financial Research, vol. 24, no. 3, pp. 367-84. Sawicki, J & Thomson, K 2000, 'An investigation into the performance of recommended funds: do the managed funds "approved" by research companies outperform the non-gratae?' Advances in Pacific Basin Financial Markets, vol. 6, pp. 101-24. Sawicki, J & Finn, F 2002, 'Smart money and small funds', Journal of Business Finance & Accounting, vol. 29, no. 5 & 6, pp. 825-46. Schramn, S 1998, 'Indexed assets top $1 trillion', Pensions and Investments, 23 Feb, p. 1. Seitz, N & Ellison, M 1999, Capital Budgeting and Long-term Financing Decisions, 3 edn, Dryden, Orlando. Shanken, J 1982, 'The arbitrage pricing theory: is it testable?' Journal of Finance, vol. 37, no. 5, pp. 1129-40. Sharpe, W 1964, 'Capital asset prices: a theory of market equilibrium under condtions of risk', Journal of Finance, vol. 19, no. 3, pp. 424-42. ---- 1966, 'Mutual fund performance', Journal of Business, vol. 39, no. 1, pp. 119-38.

144

References
Singer, B 1996, 'Valuation of portfolio performance: aggregate return and risk analysis', Journal of Performance Measurement, vol. 1, no. 1, pp. 582-603. Sirri, E & Tufano, P 1998, 'Costly search and mutual fund flows', Journal of Finance, vol. 53, no. 5, pp. 1589-622. Stock, J & Watson, M 2003, Introduction to Econometrics, Addison Wesley, Boston. Straits Times 2002a, 12 Wall St giants pressured analysts to promote clients, Nov 7, p. A15. Straits Times 2002b, WorldCom faces more fraud charges, Nov 9, p. A15. Susman, G & Evered, R 1978, 'An assessment of the scientific merits of action research', Administrative Science Quarterly, vol. 23, pp. 582-603. Tan, C 2001, Personal Finance in Singapore, 2 edn, Singapore University Press, Singapore. Teo, C 2003, 'The USFTA', Singapore Business Visitor, Sep, pp. 8-10. Tng, C 2005a, 'Performance of approved equity funds: evidence from Singapore's retail funds', paper presented to 18th Annual Australasian Finance and Banking Conference, Sydney, 14-16 Dec. ---- 2005b, 'Sustainable development of bank-managed mutual funds: evidence from Singapore's retail funds', paper presented to 9th Annual Waikato Management School Student Research Conference, University of Waikato, 25 Oct. ---- 2005c, 'Mutual fund performance determinants: size and expenditures', paper presented to 2nd International Business Research Conference, University of Technology Sydney, 5-8 Dec. ---- 2005d, 'Competitive strategy of international banks: fund management challenges and opportunities', paper presented to 2005 Academy of International Business Southeast Asia Regional Conference, Manila, 24-26 Nov. Treynor, J 1965, 'How to rate management of investment funds', Harvard Business Review, vol. 43, no. 1, pp. 63-75. Veltman, L 2000, Living and Working in Australia, 7 edn, How To Books, Oxford. Williams, T 1996, 'Sandoz indexes all large cap stocks', Pensions & Investments, 8 Jan, p. 1. Wong, K 1988, 'Role of Singapore and Malaysian stock exchanges', in C Tan & K KC (eds), Handbook of Singapore-Malaysian Corporate Finance, Butterworths, Singapore. Wong, S 2004, Funds Vs Common Indices, viewed 21 Jan 2004, <http://www.fundsupermart.com>. Wong, W 2003, 'Big drop in investment funds unlikely: observers point to the vast amounts of untapped CPF funds', The Business Times, 16 Aug. Yin, R 1994, 'Case study research--design and methods', in Applied Social Research Methods Series, 2nd edn, Sage, Newbury Park, vol. 5. Zikmund, W 2000, Business Research Methods, 6 edn, Dryden, Texas.

145

Appendices

Five appendices are attached to this thesis. Appendix A tabulates data collected for all 19 funds as well as computation and regression outputs. Information is arranged in alphabetical order according to fund name. Appendices B to E present two conference papers and two journal articles based on this thesis.

146

Appendix A Data, computation and regression outputs

Aberdeen Singapore Equity Fund (ASE) FMC RFR1 RFR2 Investment 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-99 56.90 56.28 55.90 42.77 42.15 41.77 14.13 2.00 3.00 30-Sep-99 -5.00 -5.63 -6.00 -6.72 -7.35 -7.72 1.72 2.10 2.80 31-Dec-99 14.60 13.98 13.60 22.63 22.01 21.63 -8.03 6.50 2.30 31-Mar-00 -15.80 -16.43 -16.80 -13.99 -14.62 -14.99 -1.81 6.60 2.30 30-Jun-00 -2.50 -3.13 -3.50 -4.44 -5.06 -5.44 1.94 6.90 2.60 30-Sep-00 4.40 3.78 3.40 -2.01 -2.63 -3.01 6.41 7.20 2.50 31-Dec-00 1.50 0.88 0.50 -3.52 -4.14 -4.52 5.02 7.30 2.30 31-Mar-01 -8.20 -8.83 -9.20 -13.11 -13.74 -14.11 4.91 6.90 3.30 30-Jun-01 10.10 9.48 9.10 3.12 2.50 2.12 6.98 9.20 2.80 30-Sep-01 -17.10 -17.73 -18.10 -23.57 -24.20 -24.57 6.47 7.70 2.40 31-Dec-01 18.20 17.58 17.20 23.04 22.42 22.04 -4.84 9.70 2.60 31-Mar-02 13.00 12.38 12.00 11.06 10.44 10.06 1.94 11.50 2.00 Average 5.84 5.22 4.84 2.94 2.31 1.94 2.90 6.97 2.58 Std dev 19.78 19.78 19.78 18.84 18.84 18.84 5.89 2.74 0.36 Beta 1.00 Info ratio 0.49 Jensen 2.90 2 0.91 R Sharpe Treynor Fund STI Fund STI RFR1 0.26 0.12 5.20 2.31 RFR2 0.24 0.10 4.83 1.94 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -2.54 -3.17 -3.54 -5.46 -6.08 -6.46 2.92 11.90 2.24 30-Jun-03 14.38 13.76 13.38 14.20 13.58 13.20 0.18 15.00 2.26 30-Sep-03 9.16 8.54 8.16 12.63 12.01 11.63 -3.47 19.40 2.19 31-Dec-03 4.28 3.66 3.28 8.20 7.57 7.20 -3.92 22.40 2.09 31-Mar-04 6.49 5.87 5.49 5.35 4.72 4.35 1.14 26.40 2.05 30-Jun-04 1.67 1.05 0.67 -1.13 -1.75 -2.13 2.80 24.80 2.02 30-Sep-04 5.69 5.07 4.69 7.98 7.36 6.98 -2.29 27.90 2.02 Average 5.59 4.97 4.59 5.97 5.34 4.97 -0.38 21.11 2.12 Std dev 5.39 5.39 5.39 7.10 7.10 7.10 2.87 5.98 0.10 Beta 0.71 Info ratio -0.13 Jensen 1.19 2 0.87 R Sharpe Treynor Fund STI Fund STI RFR1 0.92 0.75 7.03 5.34 RFR2 0.85 0.70 6.50 4.97

147

Data, computation and regression outputs

Regression: RET-RFR1 & STI-RFR1; f=ASE & t=1999-2002 Regression Statistics Multiple R 0.955 2 R 0.911 2 0.903 Adjusted R Std Err 6.174 Observations 12 ANOVA df Regression Residual Total SS MS F 1 3924.480 3924.480 102.954 10 381.189 38.119 11 4305.669 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 2.897 1.797 1.002 0.099

t Stat P-value Lower 95% Upper 95% 1.612 0.138 -1.107 6.900 10.147 0.000 0.782 1.223

Regression: RET-RFR1 & STI-RFR1; f=ASE & t=2003-2004 Regression Statistics Multiple R 0.931 2 R 0.867 2 0.840 Adjusted R Std Err 2.154 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 150.812 23.196 174.008 MS 150.812 4.639 F 32.508 Sig F 0.002

Intercept STI-RFR1

Coeff Std Err 1.192 1.049 0.706 0.124

t Stat P-value Lower 95% Upper 95% 1.136 0.307 -1.505 3.889 5.702 0.002 0.388 1.024

Regression: RET-RFR2 & STI-RFR2; f=ASE & t=1999-2002 Regression Statistics Multiple R 0.955 2 R 0.911 2 Adjusted R 0.903 Std Err 6.174 Observations 12

148

Appendix A

ANOVA df Regression Residual Total SS MS F 1 3924.480 3924.480 102.954 10 381.189 38.119 11 4305.669 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 2.898 1.793 1.002 0.099

t Stat P-value Lower 95% Upper 95% 1.616 0.137 -1.096 6.892 10.147 0.000 0.782 1.223

Regression: RET-RFR2 & STI-RFR2; f=ASE & t=2003-2004 Regression Statistics Multiple R 0.931 2 R 0.867 2 0.840 Adjusted R Std Err 2.154 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 150.812 23.196 174.008 MS 150.812 4.639 F 32.508 Sig F 0.002

Intercept STI-RFR2

Coeff Std Err 1.082 1.020 0.706 0.124

t Stat P-value Lower 95% Upper 95% 1.060 0.338 -1.542 3.705 5.702 0.002 0.388 1.024

149

Data, computation and regression outputs

AXA Life-Fortress Fund (ALF) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Dec-99 28.10 27.48 27.10 22.63 22.01 21.63 5.47 7.10 1.40 31-Mar-00 6.50 5.88 5.50 -13.99 -14.62 -14.99 20.49 7.90 1.50 30-Jun-00 -12.40 -13.03 -13.40 -4.44 -5.06 -5.44 -7.96 7.10 1.60 30-Sep-00 -5.50 -6.13 -6.50 -2.01 -2.63 -3.01 -3.49 7.10 1.60 31-Dec-00 -9.40 -10.03 -10.40 -3.52 -4.14 -4.52 -5.88 8.50 1.50 31-Mar-01 -8.60 -9.23 -9.60 -13.11 -13.74 -14.11 4.51 12.00 1.50 30-Jun-01 8.70 8.08 7.70 3.12 2.50 2.12 5.58 23.90 1.40 30-Sep-01 -26.40 -27.03 -27.40 -23.57 -24.20 -24.57 -2.83 21.70 1.40 31-Dec-01 26.10 25.48 25.10 23.04 22.42 22.04 3.06 36.60 1.30 31-Mar-02 13.40 12.78 12.40 11.06 10.44 10.06 2.34 45.30 1.30 Average 2.05 1.43 1.05 -0.08 -0.70 -1.08 2.13 17.72 1.45 Std dev 17.54 17.54 17.54 15.40 15.40 15.40 8.09 13.84 0.11 Beta 1.01 Info ratio 0.26 Jensen 2.13 2 0.79 R Sharpe Treynor Fund STI Fund STI RFR1 0.08 -0.05 1.41 -0.70 RFR2 0.06 -0.07 1.04 -1.08 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 3.73 3.11 2.73 -5.46 -6.08 -6.46 9.19 72.60 1.27 30-Jun-03 9.47 8.85 8.47 14.20 13.58 13.20 -4.73 84.20 1.25 30-Sep-03 10.65 10.03 9.65 12.63 12.01 11.63 -1.98 72.65 1.23 31-Dec-03 6.30 5.68 5.30 8.20 7.57 7.20 -1.90 60.46 1.23 31-Mar-04 5.58 4.96 4.58 5.35 4.72 4.35 0.23 52.47 1.24 30-Jun-04 1.16 0.54 0.16 -1.13 -1.75 -2.13 2.29 49.16 1.23 30-Sep-04 5.86 5.24 4.86 7.98 7.36 6.98 -2.12 43.87 1.24 Average 6.11 5.48 5.11 5.97 5.34 4.97 0.14 62.20 1.24 Std dev 3.23 3.23 3.23 7.10 7.10 7.10 4.55 14.75 0.01 Beta 0.40 Info ratio 0.03 Jensen 3.35 2 0.77 R Sharpe Treynor Fund STI Fund STI RFR1 1.70 0.75 13.77 5.34 RFR2 1.58 0.70 12.82 4.97

150

Appendix A

Regression: RET-RFR1 & STI-RFR1; f=ALF & t=1999-2002 Regression Statistics Multiple R 0.887 2 R 0.787 2 0.761 Adjusted R Std Err 8.579 Observations 10 ANOVA df Regression Residual Total SS MS 1 2180.737 2180.737 8 588.848 73.606 9 2769.585 F 29.627 Sig F 0.001

Intercept STI-RFR1

Coeff Std Err 2.135 2.716 1.011 0.186

t Stat P-value Lower 95% Upper 95% 0.786 0.455 -4.129 8.398 5.443 0.001 0.582 1.439

Regression: RET-RFR1 & STI-RFR1; f=ALF & t=2003-2004 Regression Statistics Multiple R 0.877 2 R 0.768 2 0.722 Adjusted R Std Err 1.701 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 47.979 14.468 62.448 MS 47.979 2.894 F 16.581 Sig F 0.010

Intercept STI-RFR1

Coeff Std Err 3.354 0.829 0.398 0.098

t Stat P-value Lower 95% Upper 95% 4.048 0.010 1.224 5.484 4.072 0.010 0.147 0.650

Regression: RET-RFR2 & STI-RFR2; f=ALF & t=1999-2002 Regression Statistics Multiple R 0.887 2 R 0.787 2 0.761 Adjusted R Std Err 8.579 Observations 10

151

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS 1 2180.737 2180.737 8 588.848 73.606 9 2769.585 F 29.627 Sig F 0.001

Intercept STI-RFR2

Coeff Std Err 2.139 2.720 1.011 0.186

t Stat P-value Lower 95% Upper 95% 0.786 0.454 -4.134 8.412 5.443 0.001 0.582 1.439

Regression: RET-RFR2 & STI-RFR2; f=ALF & t=2003-2004 Regression Statistics Multiple R 0.877 2 R 0.768 2 0.722 Adjusted R Std Err 1.701 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 47.979 14.468 62.448 MS 47.979 2.894 F 16.581 Sig F 0.010

Intercept STI-RFR2

Coeff Std Err 3.128 0.806 0.398 0.098

t Stat P-value Lower 95% Upper 95% 3.882 0.012 1.057 5.200 4.072 0.010 0.147 0.650

152

Appendix A

AXA Life-Fortress Fund A (ALFA) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-04 7.14 6.52 6.14 5.35 4.72 4.35 1.79 41.80 1.73 30-Jun-04 1.04 0.42 0.04 -1.13 -1.75 -2.13 2.17 48.77 1.60 30-Sep-04 5.94 5.32 4.94 7.98 7.36 6.98 -2.04 45.01 1.68 Average 4.71 4.08 3.71 4.07 3.44 3.07 0.64 45.19 1.67 Std dev 3.23 3.23 3.23 4.69 4.69 4.69 2.33 3.49 0.07 Beta 0.61 Info ratio 0.27 Jensen 2.13 2 0.79 R Sharpe Treynor Fund STI Fund STI RFR1 1.26 0.73 6.65 3.44 RFR2 1.15 0.65 6.04 3.07 Regression: RET-RFR1 & STI-RFR1; f=ALFA & t=2003-2004 Regression Statistics Multiple R 0.891 2 R 0.794 2 0.587 Adjusted R Std Err 2.076 Observations 3 ANOVA df Regression Residual Total 1 1 2 SS MS 16.577 16.577 4.310 4.310 20.887 F 3.847 Sig F 0.300

Intercept STI-RFR1

Coeff Std Err t Stat P-value Lower 95% Upper 95% 1.966 1.612 1.219 0.437 -18.521 22.454 0.614 0.313 1.961 0.300 -3.365 4.593

Regression: RET-RFR2 & STI-RFR2; f=ALFA & t=2003-2004 Regression Statistics Multiple R 0.891 2 R 0.794 2 0.587 Adjusted R Std Err 2.076 Observations 3

153

Data, computation and regression outputs

ANOVA df Regression Residual Total 1 1 2 SS MS 16.577 16.577 4.310 4.310 20.887 F 3.847 Sig F 0.300

Intercept STI-RFR2

Coeff Std Err t Stat P-value Lower 95% Upper 95% 1.822 1.536 1.186 0.446 -17.700 21.343 0.614 0.313 1.961 0.300 -3.365 4.593

154

Appendix A

CMG First State Singapore Growth Fund (CFSSG) FMC RFR1 RFR2 Investment 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-99 58.50 57.88 57.50 42.77 42.15 41.77 15.73 83.20 2.40 30-Sep-99 -5.60 -6.23 -6.60 -6.72 -7.35 -7.72 1.12 85.40 2.40 31-Dec-99 28.60 27.98 27.60 22.63 22.01 21.63 5.97 105.30 2.30 31-Mar-00 8.60 7.98 7.60 -13.99 -14.62 -14.99 22.59 117.40 2.40 30-Jun-00 -20.00 -20.63 -21.00 -4.44 -5.06 -5.44 -15.56 97.90 2.40 30-Sep-00 -4.50 -5.13 -5.50 -2.01 -2.63 -3.01 -2.49 93.60 2.40 31-Dec-00 -14.60 -15.23 -15.60 -3.52 -4.14 -4.52 -11.08 79.40 2.40 31-Mar-01 -12.60 -13.23 -13.60 -13.11 -13.74 -14.11 0.51 71.90 2.40 30-Jun-01 7.50 6.88 6.50 3.12 2.50 2.12 4.38 79.50 2.30 30-Sep-01 -24.10 -24.73 -25.10 -23.57 -24.20 -24.57 -0.53 59.60 2.30 31-Dec-01 28.20 27.58 27.20 23.04 22.42 22.04 5.16 76.10 2.30 31-Mar-02 14.30 13.68 13.30 11.06 10.44 10.06 3.24 86.60 2.20 Average 5.36 4.73 4.36 2.94 2.31 1.94 2.42 86.33 2.35 Std dev 24.18 24.18 24.18 18.84 18.84 18.84 10.24 15.51 0.07 Beta 1.18 Info ratio 0.24 Jensen 2.01 2 0.84 R Sharpe Treynor Fund STI Fund STI RFR1 0.20 0.12 4.03 2.31 RFR2 0.18 0.10 3.71 1.94 Regression: RET-RFR1 & STI-RFR1; f=CFSSG & t=1999-2002 Regression Statistics Multiple R 0.916 2 R 0.839 2 0.823 Adjusted R Std Err 10.160 Observations 12 ANOVA df Regression Residual Total SS MS 1 5397.780 5397.780 10 1032.169 103.217 11 6429.949 F 52.295 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 2.013 2.957 1.176 0.163

t Stat P-value Lower 95% Upper 95% 0.681 0.512 -4.576 8.601 7.232 0.000 0.813 1.538

155

Data, computation and regression outputs

Regression: RET-RFR2 & STI-RFR2; f=CFSSG & t=1999-2002 Regression Statistics Multiple R 0.916 2 R 0.839 2 0.823 Adjusted R Std Err 10.160 Observations 12 ANOVA df Regression Residual Total SS MS 1 5397.780 5397.780 10 1032.169 103.217 11 6429.949 F 52.295 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 2.078 2.950 1.176 0.163

t Stat P-value Lower 95% Upper 95% 0.705 0.497 -4.494 8.651 7.232 0.000 0.813 1.538

156

Appendix A

DBS Shenton Thrift Fund (DST) FMC RFR1 RFR2 Bank 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-99 51.00 50.38 50.00 42.77 42.15 41.77 8.23 54.70 0.60 30-Sep-99 -10.20 -10.83 -11.20 -6.72 -7.35 -7.72 -3.48 45.00 0.80 31-Dec-99 26.80 26.18 25.80 22.63 22.01 21.63 4.17 53.00 0.90 31-Mar-00 -6.00 -6.63 -7.00 -13.99 -14.62 -14.99 7.99 50.00 1.00 30-Jun-00 -12.10 -12.73 -13.10 -4.44 -5.06 -5.44 -7.66 46.30 1.00 30-Sep-00 -9.50 -10.13 -10.50 -2.01 -2.63 -3.01 -7.49 42.70 1.00 31-Dec-00 -2.60 -3.23 -3.60 -3.52 -4.14 -4.52 0.92 41.80 1.00 31-Mar-01 -15.40 -16.03 -16.40 -13.11 -13.74 -14.11 -2.29 35.80 1.00 30-Jun-01 10.70 10.08 9.70 3.12 2.50 2.12 7.58 40.40 1.00 30-Sep-01 -21.50 -22.13 -22.50 -23.57 -24.20 -24.57 2.07 31.60 1.00 31-Dec-01 23.80 23.18 22.80 23.04 22.42 22.04 0.76 40.00 1.10 31-Mar-02 17.00 16.38 16.00 11.06 10.44 10.06 5.94 46.20 1.10 Average 4.33 3.71 3.33 2.94 2.31 1.94 1.39 43.96 0.96 Std dev 21.63 21.63 21.63 18.84 18.84 18.84 5.69 6.74 0.14 Beta 1.113 Info ratio 0.245 Jensen 1.132 2 0.941 R Sharpe Treynor Fund STI Fund STI RFR1 0.17 0.12 3.33 2.31 RFR2 0.15 0.10 2.99 1.94 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -0.48 -1.11 -1.48 -5.46 -6.08 -6.46 4.98 37.75 1.03 30-Jun-03 16.75 16.13 15.75 14.20 13.58 13.20 2.55 43.56 1.05 30-Sep-03 37.30 36.68 36.30 12.63 12.01 11.63 24.67 58.99 1.07 31-Dec-03 13.13 12.51 12.13 8.20 7.57 7.20 4.93 61.54 1.90 31-Mar-04 8.71 8.09 7.71 5.35 4.72 4.35 3.36 72.29 1.08 30-Jun-04 -2.18 -2.81 -3.18 -1.13 -1.75 -2.13 -1.05 68.17 1.05 30-Sep-04 12.85 12.23 11.85 7.98 7.36 6.98 4.87 74.17 0.96 Average 12.30 11.67 11.30 5.97 5.34 4.97 6.33 59.50 1.16 Std dev 13.11 13.11 13.11 7.10 7.10 7.10 8.37 14.06 0.33 Beta 1.51 Info ratio 0.76 Jensen 3.60 2 0.67 R Sharpe Treynor Fund STI Fund STI RFR1 0.89 0.75 7.72 5.34 RFR2 0.86 0.70 7.48 4.97

157

Data, computation and regression outputs

Regression: RET-RFR1 & STI-RFR1; f=DST & t=1999-2002 Regression Statistics Multiple R 0.970 2 R 0.941 2 0.935 Adjusted R Std Err 5.532 Observations 12 ANOVA df Regression Residual Total SS MS F 1 4840.654 4840.654 158.164 10 306.053 30.605 11 5146.707 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 1.132 1.610 1.113 0.089

t Stat P-value Lower 95% Upper 95% 0.703 0.498 -2.456 4.719 12.576 0.000 0.916 1.310

Regression: RET-RFR1 & STI-RFR1; f=DST & t=2003-2004 Regression Statistics Multiple R 0.818 2 R 0.670 2 0.604 Adjusted R Std Err 8.256 Observations 7 ANOVA df Regression Residual Total SS 1 690.876 5 340.805 6 1031.681 MS 690.876 68.161 F 10.136 Sig F 0.024

Intercept STI-RFR1

Coeff Std Err 3.596 4.021 1.511 0.475

t Stat P-value Lower 95% Upper 95% 0.894 0.412 -6.741 13.934 3.184 0.024 0.291 2.731

Regression: RET-RFR2 & STI-RFR2; f=DST & t=1999-2002 Regression Statistics Multiple R 0.970 2 R 0.941 2 0.935 Adjusted R Std Err 5.532 Observations 12

158

Appendix A

ANOVA df Regression Residual Total SS MS F 1 4840.654 4840.654 158.164 10 306.053 30.605 11 5146.707 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 1.174 1.606 1.113 0.089

t Stat P-value Lower 95% Upper 95% 0.731 0.482 -2.405 4.753 12.576 0.000 0.916 1.310

Regression: RET-RFR2 & STI-RFR2; f=DST & t=2003-2004 Regression Statistics Multiple R 0.818 2 R 0.670 2 0.604 Adjusted R Std Err 8.256 Observations 7 ANOVA df Regression Residual Total SS 1 690.876 5 340.805 6 1031.681 MS 690.876 68.161 F 10.136 Sig F 0.024

Intercept STI-RFR2

Coeff Std Err 3.788 3.912 1.511 0.475

t Stat P-value Lower 95% Upper 95% 0.968 0.377 -6.267 13.843 3.184 0.024 0.291 2.731

159

Data, computation and regression outputs

DBS Horizon Singapore Equity Fund (DHSE) FMC RFR1 RFR2 Bank 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Sep-00 -3.70 -4.33 -4.70 -2.01 -2.63 -3.01 -1.69 62.70 1.50 31-Dec-00 -3.50 -4.13 -4.50 -3.52 -4.14 -4.52 0.02 63.80 1.50 31-Mar-01 -15.20 -15.83 -16.20 -13.11 -13.74 -14.11 -2.09 63.50 1.50 30-Jun-01 3.70 3.08 2.70 3.12 2.50 2.12 0.58 75.90 1.50 30-Sep-01 -19.40 -20.03 -20.40 -23.57 -24.20 -24.57 4.17 67.00 1.50 31-Dec-01 22.20 21.58 21.20 23.04 22.42 22.04 -0.84 89.20 1.50 31-Mar-02 12.70 12.08 11.70 11.06 10.44 10.06 1.64 105.10 1.50 Average -0.46 -1.08 -1.46 -0.71 -1.34 -1.71 0.25 75.31 1.50 Std dev 14.72 14.72 14.72 15.29 15.29 15.29 2.16 16.23 0.00 Beta 0.95 Info ratio 0.12 Jensen 0.19 2 0.98 R Sharpe Treynor Fund STI Fund STI RFR1 -0.07 -0.09 -1.14 -1.34 RFR2 -0.10 -0.11 -1.53 -1.71 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -4.17 -4.80 -5.17 -5.46 -6.08 -6.46 1.29 81.01 1.51 30-Jun-03 14.78 14.16 13.78 14.20 13.58 13.20 0.58 94.60 1.52 30-Sep-03 18.94 18.32 17.94 12.63 12.01 11.63 6.31 113.29 1.50 31-Dec-03 8.92 8.30 7.92 8.20 7.57 7.20 0.72 122.39 1.48 31-Mar-04 2.92 2.30 1.92 5.35 4.72 4.35 -2.43 119.53 1.49 30-Jun-04 -2.84 -3.47 -3.84 -1.13 -1.75 -2.13 -1.71 112.81 1.47 30-Sep-04 12.40 11.78 11.40 7.98 7.36 6.98 4.42 121.17 1.45 Average 7.28 6.65 6.28 5.97 5.34 4.97 1.31 109.26 1.49 Std dev 8.88 8.88 8.88 7.10 7.10 7.10 3.13 15.61 0.02 Beta 1.19 Info ratio 0.42 Jensen 0.32 2 0.90 R Sharpe Treynor Fund STI Fund STI RFR1 0.75 0.75 5.61 5.34 RFR2 0.71 0.70 5.29 4.97 Regression: RET-RFR1 & STI-RFR1; f=DHSE & t=1999-2002 Regression Statistics Multiple R 0.990 2 R 0.981 2 0.977 Adjusted R Std Err 2.227 Observations 7

160

Appendix A
ANOVA df Regression Residual Total SS MS F 1 1274.892 1274.892 256.980 5 24.805 4.961 6 1299.697 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 0.191 0.846 0.953 0.059

t Stat P-value Lower 95% Upper 95% 0.226 0.830 -1.983 2.365 16.031 0.000 0.800 1.106

Regression: RET-RFR1 & STI-RFR1; f=DHSE & t=2003-2004 Regression Statistics Multiple R 0.948 2 R 0.898 2 0.878 Adjusted R Std Err 3.105 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 425.440 48.196 473.636 MS 425.440 9.639 F 44.136 Sig F 0.001

Intercept STI-RFR1

Coeff Std Err 0.316 1.512 1.186 0.178

t Stat P-value Lower 95% Upper 95% 0.209 0.843 -3.571 4.204 6.644 0.001 0.727 1.645

Regression: RET-RFR2 & STI-RFR2; f=DHSE & t=1999-2002 Regression Statistics Multiple R 0.990 2 R 0.981 2 0.977 Adjusted R Std Err 2.227 Observations 7 ANOVA df Regression Residual Total SS MS F 1 1274.892 1274.892 256.980 5 24.805 4.961 6 1299.697 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 0.173 0.848 0.953 0.059

t Stat P-value Lower 95% Upper 95% 0.205 0.846 -2.006 2.353 16.031 0.000 0.800 1.106

161

Data, computation and regression outputs

Regression: RET-RFR2 & STI-RFR2; f=DHSE & t=2003-2004 Regression Statistics Multiple R 0.948 2 R 0.898 2 0.878 Adjusted R Std Err 3.105 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 425.440 48.196 473.636 MS 425.440 9.639 F 44.136 Sig F 0.001

Intercept STI-RFR2

Coeff Std Err 0.386 1.471 1.186 0.178

t Stat P-value Lower 95% Upper 95% 0.262 0.803 -3.395 4.167 6.644 0.001 0.727 1.645

162

Appendix A

GE Greatlink Singapore Equities Fund (GGSE) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -5.28 -5.91 -6.28 -5.46 -6.08 -6.46 0.18 15.10 1.45 30-Jun-03 13.02 12.40 12.02 14.20 13.58 13.20 -1.18 18.60 1.34 30-Sep-03 7.97 7.35 6.97 12.63 12.01 11.63 -4.66 22.60 1.29 31-Dec-03 7.27 6.65 6.27 8.20 7.57 7.20 -0.93 31.48 1.20 31-Mar-04 3.72 3.10 2.72 5.35 4.72 4.35 -1.63 36.69 1.05 30-Jun-04 0.11 -0.52 -0.89 -1.13 -1.75 -2.13 1.24 42.23 1.03 30-Sep-04 7.37 6.75 6.37 7.98 7.36 6.98 -0.61 45.45 1.06 Average 4.88 4.26 3.88 5.97 5.34 4.97 -1.09 30.31 1.20 Std dev 5.99 5.99 5.99 7.10 7.10 7.10 1.84 11.84 0.16 Beta 0.82 Info ratio -0.59 Jensen -0.13 2 0.95 R Sharpe Treynor Fund STI Fund STI RFR1 0.71 0.75 5.18 5.34 RFR2 0.65 0.70 4.73 4.97 Regression: RET-RFR1 & STI-RFR1; f=GGSE & t=2003-2004 Regression Statistics Multiple R 0.975 2 R 0.950 2 0.940 Adjusted R Std Err 1.465 Observations 7 ANOVA df Regression Residual Total SS MS 1 204.319 204.319 5 10.725 2.145 6 215.044 F 95.255 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err -0.134 0.713 0.822 0.084

t Stat P-value Lower 95% Upper 95% -0.188 0.859 -1.968 1.700 9.760 0.000 0.605 1.038

Regression: RET-RFR2 & STI-RFR2; f=GGSE & t=2003-2004 Regression Statistics Multiple R 0.975 2 R 0.950 2 0.940 Adjusted R Std Err 1.465 Observations 7

163

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS 1 204.319 204.319 5 10.725 2.145 6 215.044 F 95.255 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err -0.201 0.694 0.822 0.084

t Stat P-value Lower 95% Upper 95% -0.289 0.784 -1.984 1.583 9.760 0.000 0.605 1.038

164

Appendix A

Keppel Managed Fund (KM) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Dec-99 23.60 22.98 22.60 22.63 22.01 21.63 0.97 11.50 1.10 31-Mar-00 -5.30 -5.93 -6.30 -13.99 -14.62 -14.99 8.69 14.00 1.10 30-Jun-00 -13.70 -14.33 -14.70 -4.44 -5.06 -5.44 -9.26 12.20 1.10 30-Sep-00 -9.10 -9.73 -10.10 -2.01 -2.63 -3.01 -7.09 11.10 1.20 31-Dec-00 -5.90 -6.53 -6.90 -3.52 -4.14 -4.52 -2.38 10.60 1.20 31-Mar-01 -13.70 -14.33 -14.70 -13.11 -13.74 -14.11 -0.59 9.00 1.20 30-Jun-01 5.10 4.48 4.10 3.12 2.50 2.12 1.98 9.60 1.20 30-Sep-01 -21.30 -21.93 -22.30 -23.57 -24.20 -24.57 2.27 7.70 1.20 31-Dec-01 28.50 27.88 27.50 23.04 22.42 22.04 5.46 9.40 1.20 31-Mar-02 13.00 12.38 12.00 11.06 10.44 10.06 1.94 10.70 1.10 Average 0.12 -0.51 -0.88 -0.08 -0.70 -1.08 0.20 10.58 1.16 Std dev 16.80 16.80 16.80 15.40 15.40 15.40 5.38 1.78 0.05 Beta 1.03 Info ratio 0.04 Jensen 0.22 2 0.90 R Sharpe Treynor Fund STI Fund STI RFR1 -0.03 -0.05 -0.49 -0.70 RFR2 -0.05 -0.07 -0.85 -1.08 Regression: RET-RFR1 & STI-RFR1; f=KM & t=1999-2002 Regression Statistics Multiple R 0.948 2 R 0.898 2 0.885 Adjusted R Std Err 5.684 Observations 10 ANOVA df Regression Residual Total SS MS 1 2280.418 2280.418 8 258.438 32.305 9 2538.856 F 70.591 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 0.221 1.799 1.034 0.123

t Stat P-value Lower 95% Upper 95% 0.123 0.905 -3.929 4.370 8.402 0.000 0.750 1.317

165

Data, computation and regression outputs

Regression: RET-RFR2 & STI-RFR2; f=KM & t=1999-2002 Regression Statistics Multiple R 0.948 2 R 0.898 2 0.885 Adjusted R Std Err 5.684 Observations 10 ANOVA df Regression Residual Total SS MS 1 2280.418 2280.418 8 258.438 32.305 9 2538.856 F 70.591 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 0.233 1.802 1.034 0.123

t Stat P-value Lower 95% Upper 95% 0.130 0.900 -3.923 4.389 8.402 0.000 0.750 1.317

166

Appendix A

NTUC Income Singapore Equity Fund (NISE) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Sep-03 13.56 12.94 12.56 12.63 12.01 11.63 0.93 1.17 0.30 31-Dec-03 9.45 8.83 8.45 8.20 7.57 7.20 1.25 1.41 0.30 31-Mar-04 4.55 3.93 3.55 5.35 4.72 4.35 -0.80 2.21 0.30 30-Jun-04 -0.58 -1.21 -1.58 -1.13 -1.75 -2.13 0.55 1.94 0.50 30-Sep-04 10.28 9.66 9.28 7.98 7.36 6.98 2.30 1.71 0.77 Average 7.45 6.83 6.45 6.61 5.98 5.61 0.84 1.69 0.43 Std dev 5.53 5.53 5.53 5.05 5.05 5.05 1.13 0.41 0.21 Beta 1.07 Info ratio 0.75 Jensen -0.13 2 0.95 R Sharpe Treynor Fund STI Fund STI RFR1 1.24 1.18 6.36 5.98 RFR2 1.17 1.11 6.01 5.61 Regression: RET-RFR1 & STI-RFR1; f=NISE & t=2003-2004 Regression Statistics Multiple R 0.981 2 R 0.963 2 0.951 Adjusted R Std Err 1.227 Observations 5 ANOVA df Regression Residual Total SS MS 1 117.713 117.713 3 4.519 1.506 4 122.232 F 78.153 Sig F 0.003

Intercept STI-RFR1

Coeff Std Err 0.403 0.911 1.074 0.121

t Stat P-value Lower 95% Upper 95% 0.443 0.688 -2.495 3.301 8.840 0.003 0.687 1.460

Regression: RET-RFR2 & STI-RFR2; f=NISE & t=2003-2004 Regression Statistics Multiple R 0.981 2 R 0.963 2 0.951 Adjusted R Std Err 1.227 Observations 5

167

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS 1 117.713 117.713 3 4.519 1.506 4 122.232 F 78.153 Sig F 0.003

Intercept STI-RFR2

Coeff Std Err 0.431 0.875 1.074 0.121

t Stat P-value Lower 95% Upper 95% 0.492 0.656 -2.353 3.214 8.840 0.003 0.687 1.460

168

Appendix A

OUB Manulife Golden Singapore Growth Fund (OMGSG) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Sep-00 -4.50 -5.13 -5.50 -2.01 -2.63 -3.01 -2.49 93.60 2.40 31-Dec-00 -13.30 -13.93 -14.30 -3.52 -4.14 -4.52 -9.78 3.60 1.70 31-Mar-01 -12.20 -12.83 -13.20 -13.11 -13.74 -14.11 0.91 5.20 1.60 30-Jun-01 7.30 6.68 6.30 3.12 2.50 2.12 4.18 8.00 1.20 30-Sep-01 -24.10 -24.73 -25.10 -23.57 -24.20 -24.57 -0.53 6.80 2.70 31-Dec-01 28.00 27.38 27.00 23.04 22.42 22.04 4.96 9.20 2.40 31-Mar-02 14.20 13.58 13.20 11.06 10.44 10.06 3.14 10.50 2.30 Average -0.66 -1.28 -1.66 -0.71 -1.34 -1.71 0.05 19.56 2.04 Std dev 18.09 18.09 18.09 15.29 15.29 15.29 5.08 32.73 0.54 Beta 1.14 Info ratio 0.01 Jensen 0.25 2 0.94 R Sharpe Treynor Fund STI Fund STI RFR1 -0.07 -0.09 -1.12 -1.34 RFR2 -0.09 -0.11 -1.45 -1.71 Regression: RET-RFR1 & STI-RFR1; f=OMGSG & t=1999-2002 Regression Statistics Multiple R 0.967 2 R 0.936 2 0.923 Adjusted R Std Err 5.015 Observations 7 ANOVA df Regression Residual Total SS MS 1 1836.927 1836.927 5 125.770 25.154 6 1962.697 F 73.027 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 0.246 1.904 1.144 0.134

t Stat P-value Lower 95% Upper 95% 0.129 0.902 -4.648 5.141 8.546 0.000 0.800 1.488

Regression: RET-RFR2 & STI-RFR2; f=OMGSG & t=1999-2002 Regression Statistics Multiple R 0.967 2 R 0.936 2 0.923 Adjusted R Std Err 5.015 Observations 7

169

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS 1 1836.927 1836.927 5 125.770 25.154 6 1962.697 F 73.027 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 0.300 1.909 1.144 0.134

t Stat P-value Lower 95% Upper 95% 0.157 0.881 -4.608 5.208 8.546 0.000 0.800 1.488

170

Appendix A

OCBC Savers Singapore Trust Fund (OSST) FMC RFR1 RFR2 Bank 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-99 49.40 48.78 48.40 42.77 42.15 41.77 6.63 17.90 1.30 30-Sep-99 0.30 -0.33 -0.70 -6.72 -7.35 -7.72 7.02 15.30 1.30 31-Dec-99 24.60 23.98 23.60 22.63 22.01 21.63 1.97 19.20 1.40 31-Mar-00 -12.00 -12.63 -13.00 -13.99 -14.62 -14.99 1.99 18.70 1.40 30-Jun-00 -10.80 -11.43 -11.80 -4.44 -5.06 -5.44 -6.36 18.20 1.50 30-Sep-00 -5.80 -6.43 -6.80 -2.01 -2.63 -3.01 -3.79 17.60 1.50 31-Dec-00 -2.00 -2.63 -3.00 -3.52 -4.14 -4.52 1.52 17.60 1.50 31-Mar-01 -12.10 -12.73 -13.10 -13.11 -13.74 -14.11 1.01 16.70 1.60 30-Jun-01 3.70 3.08 2.70 3.12 2.50 2.12 0.58 16.20 1.60 30-Sep-01 -22.40 -23.03 -23.40 -23.57 -24.20 -24.57 1.17 12.60 1.60 31-Dec-01 24.80 24.18 23.80 23.04 22.42 22.04 1.76 15.90 1.60 31-Mar-02 16.70 16.08 15.70 11.06 10.44 10.06 5.64 18.40 1.60 Average 4.53 3.91 3.53 2.94 2.31 1.94 1.59 17.03 1.49 Std dev 20.57 20.57 20.57 18.84 18.84 18.84 3.88 1.83 0.12 Beta 1.07 Info ratio 0.41 Jensen 1.42 2 0.97 R Sharpe Treynor Fund STI Fund STI RFR1 0.19 0.12 3.64 2.31 RFR2 0.17 0.10 3.29 1.94 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -2.77 -3.40 -3.77 -5.46 -6.08 -6.46 2.69 14.60 1.55 30-Jun-03 13.86 13.24 12.86 14.20 13.58 13.20 -0.34 16.50 1.49 30-Sep-03 23.48 22.86 22.48 12.63 12.01 11.63 10.85 19.80 1.49 31-Dec-03 8.01 7.39 7.01 8.20 7.57 7.20 -0.19 21.00 1.44 31-Mar-04 8.72 8.10 7.72 5.35 4.72 4.35 3.37 21.00 1.47 30-Jun-04 -4.87 -5.50 -5.87 -1.13 -1.75 -2.13 -3.74 20.10 1.56 30-Sep-04 5.99 5.37 4.99 7.98 7.36 6.98 -1.99 20.40 1.65 Average 7.49 6.86 6.49 5.97 5.34 4.97 1.52 19.06 1.52 Std dev 9.65 9.65 9.65 7.10 7.10 7.10 4.80 2.50 0.07 Beta 1.19 Info ratio 0.32 Jensen 0.48 2 0.77 R Sharpe Treynor Fund STI Fund STI RFR1 0.71 0.75 5.75 5.34 RFR2 0.67 0.70 5.43 4.97

171

Data, computation and regression outputs

Regression: RET-RFR1 & STI-RFR1; f=OSST & t=1999-2002 Regression Statistics Multiple R 0.984 2 R 0.969 2 0.966 Adjusted R Std Err 3.791 Observations 12 ANOVA df Regression Residual Total SS MS F 1 4509.344 4509.344 313.753 10 143.723 14.372 11 4653.067 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 1.422 1.103 1.074 0.061

t Stat P-value Lower 95% Upper 95% 1.288 0.227 -1.037 3.880 17.713 0.000 0.939 1.210

Regression: RET-RFR1 & STI-RFR1; f=OSST & t=2003-2004 Regression Statistics Multiple R 0.879 2 R 0.773 2 0.727 Adjusted R Std Err 5.036 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 431.499 126.829 558.327 MS 431.499 25.366 F 17.011 Sig F 0.009

Intercept STI-RFR1

Coeff Std Err 0.481 2.453 1.194 0.290

t Stat P-value Lower 95% Upper 95% 0.196 0.852 -5.825 6.788 4.124 0.009 0.450 1.939

Regression: RET-RFR2 & STI-RFR2; f=OSST & t=1999-2002 Regression Statistics Multiple R 0.984 2 R 0.969 2 0.966 Adjusted R Std Err 3.791 Observations 12

172

Appendix A

ANOVA df Regression Residual Total SS MS F 1 4509.344 4509.344 313.753 10 143.723 14.372 11 4653.067 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 1.449 1.101 1.074 0.061

t Stat P-value Lower 95% Upper 95% 1.317 0.217 -1.003 3.902 17.713 0.000 0.939 1.210

Regression: RET-RFR2 & STI-RFR2; f=OSST & t=2003-2004 Regression Statistics Multiple R 0.879 2 R 0.773 2 0.727 Adjusted R Std Err 5.036 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 431.499 126.829 558.327 MS 431.499 25.366 F 17.011 Sig F 0.009

Intercept STI-RFR2

Coeff Std Err 0.554 2.386 1.194 0.290

t Stat P-value Lower 95% Upper 95% 0.232 0.826 -5.580 6.688 4.124 0.009 0.450 1.939

173

Data, computation and regression outputs

OUB Union Singapore Equity Fund (OUSE) FMC RFR1 RFR2 Bank 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-00 -2.10 -2.73 -3.10 -13.99 -14.62 -14.99 11.89 3.20 3.60 30-Jun-00 -9.70 -10.33 -10.70 -4.44 -5.06 -5.44 -5.26 3.30 3.20 30-Sep-00 -5.60 -6.23 -6.60 -2.01 -2.63 -3.01 -3.59 3.20 3.10 31-Dec-00 -7.10 -7.73 -8.10 -3.52 -4.14 -4.52 -3.58 3.00 3.20 31-Mar-01 -12.90 -13.53 -13.90 -13.11 -13.74 -14.11 0.21 2.60 3.10 30-Jun-01 0.60 -0.03 -0.40 3.12 2.50 2.12 -2.52 2.70 3.70 Average -6.13 -6.76 -7.13 -5.66 -6.28 -6.66 -0.48 3.00 3.32 Std dev 4.93 4.93 4.93 6.66 6.66 6.66 6.32 0.29 0.26 Beta 0.32 Info ratio -0.08 Jensen -4.73 2 0.19 R Sharpe Treynor Fund STI Fund STI RFR1 -1.37 -0.94 -20.95 -6.28 RFR2 -1.45 -1.00 -22.12 -6.66 Regression: RET-RFR1 & STI-RFR1; f=OUSE & t=1999-2002 Regression Statistics Multiple R 0.436 2 R 0.190 2 -0.013 Adjusted R Std Err 4.957 Observations 6 ANOVA df Regression Residual Total SS MS 1 23.053 23.053 4 98.280 24.570 5 121.333 F 0.938 Sig F 0.388

Intercept STI-RFR1

Coeff Std Err t Stat P-value Lower 95% Upper 95% -4.732 2.911 -1.626 0.179 -12.813 3.349 0.323 0.333 0.969 0.388 -0.602 1.247

Regression: RET-RFR2 & STI-RFR2; f=OUSE & t=1999-2002 Regression Statistics Multiple R 0.436 2 R 0.190 2 -0.013 Adjusted R Std Err 4.957 Observations 6

174

Appendix A
ANOVA df Regression Residual Total SS MS 1 23.053 23.053 4 98.280 24.570 5 121.333 F 0.938 Sig F 0.388

Intercept STI-RFR2

Coeff Std Err t Stat P-value Lower 95% Upper 95% -4.986 3.002 -1.661 0.172 -13.320 3.347 0.323 0.333 0.969 0.388 -0.602 1.247

175

Data, computation and regression outputs

Schroder Singapore Trust (SST) FMC RFR1 RFR2 Investment 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-99 60.30 59.68 59.30 42.77 42.15 41.77 17.53 78.00 1.80 30-Sep-99 -4.20 -4.83 -5.20 -6.72 -7.35 -7.72 2.52 70.00 1.80 31-Dec-99 26.60 25.98 25.60 22.63 22.01 21.63 3.97 106.00 1.80 31-Mar-00 -10.30 -10.93 -11.30 -13.99 -14.62 -14.99 3.69 102.00 1.80 30-Jun-00 -2.70 -3.33 -3.70 -4.44 -5.06 -5.44 1.74 105.00 1.80 30-Sep-00 -3.60 -4.23 -4.60 -2.01 -2.63 -3.01 -1.59 104.00 1.70 31-Dec-00 -2.20 -2.83 -3.20 -3.52 -4.14 -4.52 1.32 101.00 1.70 31-Mar-01 -11.40 -12.03 -12.40 -13.11 -13.74 -14.11 1.71 93.00 1.70 30-Jun-01 7.90 7.28 6.90 3.12 2.50 2.12 4.78 106.00 1.60 30-Sep-01 -15.50 -16.13 -16.50 -23.57 -24.20 -24.57 8.07 122.00 1.60 31-Dec-01 26.60 25.98 25.60 23.04 22.42 22.04 3.56 162.00 1.60 31-Mar-02 16.30 15.68 15.30 11.06 10.44 10.06 5.24 212.00 1.60 Average 7.32 6.69 6.32 2.94 2.31 1.94 4.38 113.42 1.71 Std dev 21.86 21.86 21.86 18.84 18.84 18.84 4.78 38.45 0.09 Beta 1.14 Info ratio 0.92 Jensen 4.05 2 0.97 R Sharpe Treynor Fund STI Fund STI RFR1 0.31 0.12 5.87 2.31 RFR2 0.29 0.10 5.54 1.94 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -2.90 -3.53 -3.90 -5.46 -6.08 -6.46 2.56 177.59 1.57 30-Jun-03 15.28 14.66 14.28 14.20 13.58 13.20 1.08 209.98 1.59 30-Sep-03 14.46 13.84 13.46 12.63 12.01 11.63 1.83 250.20 1.54 31-Dec-03 8.42 7.80 7.42 8.20 7.57 7.20 0.22 267.28 1.50 31-Mar-04 3.00 2.38 2.00 5.35 4.72 4.35 -2.35 269.63 1.49 30-Jun-04 -0.97 -1.60 -1.97 -1.13 -1.75 -2.13 0.16 243.86 1.49 30-Sep-04 4.80 4.18 3.80 7.98 7.36 6.98 -3.18 248.57 1.48 Average 6.01 5.39 5.01 5.97 5.34 4.97 0.04 238.16 1.52 Std dev 7.09 7.09 7.09 7.10 7.10 7.10 2.11 33.13 0.04 Beta 0.95 Info ratio 0.02 Jensen 0.29 2 0.91 R Sharpe Treynor Fund STI Fund STI RFR1 0.76 0.75 5.64 5.34 RFR2 0.71 0.70 5.25 4.97

176

Appendix A

Regression: RET-RFR1 & STI-RFR1; f=SST & t=1999-2002 Regression Statistics Multiple R 0.983 2 R 0.967 2 0.964 Adjusted R Std Err 4.172 Observations 12 ANOVA df Regression Residual Total SS MS F 1 5081.869 5081.869 291.948 10 174.067 17.407 11 5255.937 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 4.052 1.214 1.141 0.067

t Stat P-value Lower 95% Upper 95% 3.337 0.008 1.346 6.757 17.086 0.000 0.992 1.289

Regression: RET-RFR1 & STI-RFR1; f=SST & t=2003-2004 Regression Statistics Multiple R 0.956 2 R 0.913 2 0.896 Adjusted R Std Err 2.286 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 275.638 26.138 301.776 MS 275.638 5.228 F 52.728 Sig F 0.001

Intercept STI-RFR1

Coeff Std Err 0.287 1.114 0.954 0.131

t Stat P-value Lower 95% Upper 95% 0.258 0.807 -2.576 3.150 7.261 0.001 0.617 1.292

Regression: RET-RFR2 & STI-RFR2; f=SST & t=1999-2002 Regression Statistics Multiple R 0.983 2 R 0.967 2 0.964 Adjusted R Std Err 4.172 Observations 12

177

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS F 1 5081.869 5081.869 291.948 10 174.067 17.407 11 5255.937 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 4.104 1.211 1.141 0.067

t Stat P-value Lower 95% Upper 95% 3.388 0.007 1.405 6.803 17.086 0.000 0.992 1.289

Regression: RET-RFR2 & STI-RFR2; f=SST & t=2003-2004 Regression Statistics Multiple R 0.956 2 R 0.913 2 0.896 Adjusted R Std Err 2.286 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 275.638 26.138 301.776 MS 275.638 5.228 F 52.728 Sig F 0.001

Intercept STI-RFR2

Coeff Std Err 0.270 1.083 0.954 0.131

t Stat P-value Lower 95% Upper 95% 0.249 0.813 -2.515 3.054 7.261 0.001 0.617 1.292

178

Appendix A

UOB Life FOF-Unifund (ULFU) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-01 6.20 5.58 5.20 3.12 2.50 2.12 3.08 35.60 1.50 30-Sep-01 -12.60 -13.23 -13.60 -23.57 -24.20 -24.57 10.97 30.10 1.50 31-Dec-01 18.80 18.18 17.80 23.04 22.42 22.04 -4.24 35.30 1.60 31-Mar-02 15.60 14.98 14.60 11.06 10.44 10.06 4.54 39.20 1.60 Average 7.00 6.38 6.00 3.41 2.79 2.41 3.59 35.05 1.55 Std dev 14.12 14.12 14.12 19.77 19.77 19.77 6.25 3.75 0.06 Beta 0.71 Info ratio 0.57 Jensen 3.00 2 0.97 R Sharpe Treynor Fund STI Fund STI RFR1 0.45 0.14 9.04 2.79 RFR2 0.42 0.12 8.51 2.41 Regression: RET-RFR1 & STI-RFR1; f=ULFU & t=1999-2002 Regression Statistics Multiple R 0.987 2 R 0.975 2 Adjusted R 0.962 Std Err 2.751 Observations 4 ANOVA df Regression Residual Total SS MS 1 582.869 582.869 2 15.131 7.565 3 598.000 F 77.045 Sig F 0.013

Intercept STI-RFR1

Coeff Std Err 4.408 1.393 0.705 0.080

t Stat P-value Lower 95% Upper 95% 3.163 0.087 -1.588 10.403 8.778 0.013 0.360 1.051

Regression: RET-RFR2 & STI-RFR2; f=ULFU & t=1999-2002 Regression Statistics Multiple R 0.987 2 R 0.975 2 0.962 Adjusted R Std Err 2.751 Observations 4

179

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS 1 582.869 582.869 2 15.131 7.565 3 598.000 F 77.045 Sig F 0.013

Intercept STI-RFR2

Coeff Std Err 4.297 1.389 0.705 0.080

t Stat P-value Lower 95% Upper 95% 3.094 0.091 -1.679 10.273 8.778 0.013 0.360 1.051

180

Appendix A

UOB Life FOF-United Growth Fund (ULFUG) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-01 6.40 5.78 5.40 3.12 2.50 2.12 3.28 95.40 1.30 30-Sep-01 -18.30 -18.93 -19.30 -23.57 -24.20 -24.57 5.27 76.20 1.30 31-Dec-01 20.00 19.38 19.00 23.04 22.42 22.04 -3.04 91.50 1.40 31-Mar-02 16.00 15.38 15.00 11.06 10.44 10.06 4.94 103.50 1.40 Average 6.03 5.40 5.03 3.41 2.79 2.41 2.61 91.65 1.35 Std dev 17.19 17.19 17.19 19.77 19.77 19.77 3.87 11.45 0.06 Beta 0.86 Info ratio 0.67 Jensen 3.00 2 0.98 R Sharpe Treynor Fund STI Fund STI RFR1 0.31 0.14 6.29 2.79 RFR2 0.29 0.12 5.85 2.41 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -1.42 -2.05 -2.42 -5.46 -6.08 -6.46 4.04 83.65 1.18 30-Jun-03 15.05 14.43 14.05 14.20 13.58 13.20 0.85 90.17 1.18 30-Sep-03 13.71 13.09 12.71 12.63 12.01 11.63 1.08 134.45 1.19 31-Dec-03 7.12 6.50 6.12 8.20 7.57 7.20 -1.08 139.78 1.19 31-Mar-04 1.02 0.40 0.02 5.35 4.72 4.35 -4.33 141.43 1.16 30-Jun-04 -1.01 -1.64 -2.01 -1.13 -1.75 -2.13 0.12 0.05 1.18 30-Sep-04 8.32 7.70 7.32 7.98 7.36 6.98 0.34 0.05 1.19 Average 6.11 5.49 5.11 5.97 5.34 4.97 0.14 84.23 1.18 Std dev 6.79 6.79 6.79 7.10 7.10 7.10 2.52 62.03 0.01 Beta 0.89 Info ratio 0.06 Jensen 0.71 2 0.87 R Sharpe Treynor Fund STI Fund STI RFR1 0.81 0.75 6.13 5.34 RFR2 0.75 0.70 5.72 4.97 Regression: RET-RFR1 & STI-RFR1; f=ULFUG & t=1999-2002 Regression Statistics Multiple R 0.988 2 R 0.976 2 0.963 Adjusted R Std Err 3.288 Observations 4

181

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS 1 865.025 865.025 2 21.623 10.811 3 886.648 F 80.010 Sig F 0.012

Intercept STI-RFR1

Coeff Std Err 3.003 1.666 0.859 0.096

t Stat P-value Lower 95% Upper 95% 1.803 0.213 -4.164 10.170 8.945 0.012 0.446 1.272

Regression: RET-RFR1 & STI-RFR1; f=ULFUG & t=2003-2004 Regression Statistics Multiple R 0.935 2 R 0.875 2 0.849 Adjusted R Std Err 2.636 Observations 7 ANOVA df Regression Residual Total SS MS 1 242.162 242.162 5 34.730 6.946 6 276.891 F 34.864 Sig F 0.002

Intercept STI-RFR1

Coeff Std Err 0.707 1.284 0.895 0.152

t Stat P-value Lower 95% Upper 95% 0.551 0.606 -2.593 4.007 5.905 0.002 0.505 1.284

Regression: RET-RFR2 & STI-RFR2; f=ULFUG & t=1999-2002 Regression Statistics Multiple R 0.988 2 R 0.976 2 0.963 Adjusted R Std Err 3.288 Observations 4 ANOVA df Regression Residual Total SS MS 1 865.025 865.025 2 21.623 10.811 3 886.648 F 80.010 Sig F 0.012

Intercept STI-RFR2

Coeff Std Err 2.951 1.660 0.859 0.096

t Stat P-value Lower 95% Upper 95% 1.777 0.218 -4.193 10.094 8.945 0.012 0.446 1.272

182

Appendix A

Regression: RET-RFR2 & STI-RFR2; f=ULFUG & t=2003-2004 Regression Statistics Multiple R 0.935 2 R 0.875 2 0.849 Adjusted R Std Err 2.636 Observations 7 ANOVA df Regression Residual Total SS MS 1 242.162 242.162 5 34.730 6.946 6 276.891 F 34.864 Sig F 0.002

Intercept STI-RFR2

Coeff Std Err 0.667 1.249 0.895 0.152

t Stat P-value Lower 95% Upper 95% 0.534 0.616 -2.543 3.877 5.905 0.002 0.505 1.284

183

Data, computation and regression outputs

UOB LifeLink Growth Fund (ULG) FMC RFR1 RFR2 Insurance 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Dec-99 16.70 16.08 15.70 22.63 22.01 21.63 -5.93 0.40 1.50 31-Mar-00 -17.10 -17.73 -18.10 -13.99 -14.62 -14.99 -3.11 0.30 1.50 30-Jun-00 -1.90 -2.53 -2.90 -4.44 -5.06 -5.44 2.54 0.30 1.50 30-Sep-00 -6.40 -7.03 -7.40 -2.01 -2.63 -3.01 -4.39 0.30 1.60 31-Dec-00 -2.50 -3.13 -3.50 -3.52 -4.14 -4.52 1.02 0.40 1.60 31-Mar-01 -11.10 -11.73 -12.10 -13.11 -13.74 -14.11 2.01 0.40 1.60 30-Jun-01 3.30 2.68 2.30 3.12 2.50 2.12 0.18 0.50 1.50 30-Sep-01 -19.90 -20.53 -20.90 -23.57 -24.20 -24.57 3.67 0.40 1.50 31-Dec-01 18.00 17.38 17.00 23.04 22.42 22.04 -5.04 0.50 1.40 31-Mar-02 14.50 13.88 13.50 11.06 10.44 10.06 3.44 0.60 1.40 Average -0.64 -1.27 -1.64 -0.08 -0.70 -1.08 -0.56 0.41 1.51 Std dev 13.67 13.67 13.67 15.40 15.40 15.40 3.70 0.10 0.07 Beta 0.87 Info ratio -0.15 Jensen -0.66 2 0.95 R Sharpe Treynor Fund STI Fund STI RFR1 -0.09 -0.05 -1.46 -0.70 RFR2 -0.12 -0.07 -1.89 -1.08 Regression: RET-RFR1 & STI-RFR1; f=ULG & t=1999-2002 Regression Statistics Multiple R 0.975 2 R 0.950 2 0.944 Adjusted R Std Err 3.249 Observations 10 ANOVA df Regression Residual Total SS MS F 1 1597.944 1597.944 151.392 8 84.440 10.555 9 1682.384 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err -0.657 1.029 0.865 0.070

t Stat P-value Lower 95% Upper 95% -0.639 0.541 -3.029 1.714 12.304 0.000 0.703 1.027

184

Appendix A

Regression: RET-RFR2 & STI-RFR2; f=ULG & t=1999-2002 Regression Statistics Multiple R 0.975 2 R 0.950 2 0.944 Adjusted R Std Err 3.249 Observations 10 ANOVA df Regression Residual Total SS MS F 1 1597.944 1597.944 151.392 8 84.440 10.555 9 1682.384 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err -0.708 1.030 0.865 0.070

t Stat P-value Lower 95% Upper 95% -0.687 0.511 -3.084 1.668 12.304 0.000 0.703 1.027

185

Data, computation and regression outputs

UOB Optimix Singapore Equity Fund (UOSE) FMC RFR1 RFR2 Bank 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Sep-01 -19.30 -19.93 -20.30 -23.57 -24.20 -24.57 4.27 2.20 3.30 31-Dec-01 15.90 15.28 14.90 23.04 22.42 22.04 -7.14 2.60 4.10 31-Mar-02 11.10 10.48 10.10 11.06 10.44 10.06 0.04 2.90 4.10 Average 2.57 1.94 1.57 3.51 2.89 2.51 -0.94 2.57 3.83 Std dev 19.09 19.09 19.09 24.21 24.21 24.21 5.77 0.35 0.46 Beta 0.78 Info ratio -0.16 Jensen -0.32 2 0.98 R Sharpe Treynor Fund STI Fund STI RFR1 0.10 0.12 2.48 2.89 RFR2 0.08 0.10 2.00 2.51 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -5.62 -6.25 -6.62 -5.46 -6.08 -6.46 -0.16 1.08 3.55 30-Jun-03 10.47 9.85 9.47 14.20 13.58 13.20 -3.73 0.78 4.47 30-Sep-03 13.64 13.02 12.64 12.63 12.01 11.63 1.01 0.78 5.11 31-Dec-03 7.57 6.95 6.57 8.20 7.57 7.20 -0.63 0.82 5.73 31-Mar-04 7.78 7.16 6.78 5.35 4.72 4.35 2.43 0.87 5.17 30-Jun-04 -5.27 -5.90 -6.27 -1.13 -1.75 -2.13 -4.14 0.82 5.35 30-Sep-04 5.96 5.34 4.96 7.98 7.36 6.98 -2.02 0.82 5.79 Average 4.93 4.31 3.93 5.97 5.34 4.97 -1.04 0.85 5.02 Std dev 7.51 7.51 7.51 7.10 7.10 7.10 2.42 0.10 0.79 Beta 1.00 Info ratio -0.43 Jensen -1.04 2 0.90 R Sharpe Treynor Fund STI Fund STI RFR1 0.57 0.75 4.30 5.34 RFR2 0.52 0.70 3.93 4.97 Regression: RET-RFR1 & STI-RFR1; f=UOSE & t=1999-2002 Regression Statistics Multiple R 0.992 2 R 0.985 2 0.969 Adjusted R Std Err 3.339 Observations 3

186

Appendix A

ANOVA df Regression Residual Total SS MS 1 717.601 717.601 1 11.146 11.146 2 728.747 F 64.383 Sig F 0.079

Intercept STI-RFR1

Coeff Std Err -0.317 1.948 0.782 0.098

t Stat P-value Lower 95% Upper 95% -0.163 0.897 -25.068 24.434 8.024 0.079 -0.457 2.022

Regression: RET-RFR1 & STI-RFR1; f=UOSE & t=2003-2004 Regression Statistics Multiple R 0.947 2 R 0.896 2 0.876 Adjusted R Std Err 2.647 Observations 7 ANOVA df Regression Residual Total SS MS 1 303.018 303.018 5 35.033 7.007 6 338.051 F 43.248 Sig F 0.001

Intercept STI-RFR1

Coeff Std Err -1.040 1.289 1.001 0.152

t Stat P-value Lower 95% Upper 95% -0.807 0.456 -4.355 2.274 6.576 0.001 0.610 1.392

Regression: RET-RFR2 & STI-RFR2; f=UOSE & t=1999-2002 Regression Statistics Multiple R 0.992 2 R 0.985 2 0.969 Adjusted R Std Err 3.339 Observations 3 ANOVA df Regression Residual Total SS MS 1 717.601 717.601 1 11.146 11.146 2 728.747 F 64.383 Sig F 0.079

Intercept STI-RFR2

Coeff Std Err -0.399 1.943 0.782 0.098

t Stat P-value Lower 95% Upper 95% -0.205 0.871 -25.087 24.290 8.024 0.079 -0.457 2.022

187

Data, computation and regression outputs

Regression: RET-RFR2 & STI-RFR2; f=UOSE & t=2003-2004 Regression Statistics Multiple R 0.947 2 R 0.896 2 0.876 Adjusted R Std Err 2.647 Observations 7 ANOVA df Regression Residual Total SS MS 1 303.018 303.018 5 35.033 7.007 6 338.051 F 43.248 Sig F 0.001

Intercept STI-RFR2

Coeff Std Err -1.040 1.254 1.001 0.152

t Stat P-value Lower 95% Upper 95% -0.829 0.445 -4.264 2.184 6.576 0.001 0.610 1.392

188

Appendix A

UOB Unifund (UU) FMC RFR1 RFR2 Bank 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Sep-00 -10.50 -11.13 -11.50 -2.01 -2.63 -3.01 -8.49 39.80 1.50 31-Dec-00 -5.50 -6.13 -6.50 -3.52 -4.14 -4.52 -1.98 38.00 1.80 31-Mar-01 -10.40 -11.03 -11.40 -13.11 -13.74 -14.11 2.71 33.70 1.50 30-Jun-01 6.20 5.58 5.20 3.12 2.50 2.12 3.08 35.60 1.50 30-Sep-01 -12.60 -13.23 -13.60 -23.57 -24.20 -24.57 10.97 30.10 1.50 31-Dec-01 18.80 18.18 17.80 23.04 22.42 22.04 -4.24 35.30 1.60 31-Mar-02 15.60 14.98 14.60 11.06 10.44 10.06 4.54 39.20 1.60 Average 0.23 -0.40 -0.77 -0.71 -1.34 -1.71 0.94 35.96 1.57 Std dev 13.18 13.18 13.18 15.29 15.29 15.29 6.40 3.40 0.11 Beta 0.85 Info ratio 0.15 Jensen 0.65 2 0.83 R Sharpe Treynor Fund STI Fund STI RFR1 -0.03 -0.09 -0.47 -1.34 RFR2 -0.06 -0.11 -0.91 -1.71 Regression: RET-RFR1 & STI-RFR1; f=UU & t=1999-2002 Regression Statistics Multiple R 0.910 2 R 0.827 2 0.793 Adjusted R Std Err 6.001 Observations 7 ANOVA df Regression Residual Total SS MS 1 862.257 862.257 5 180.037 36.007 6 1042.294 F 23.947 Sig F 0.004

Intercept STI-RFR1

Coeff Std Err 0.651 2.278 0.784 0.160

t Stat P-value Lower 95% Upper 95% 0.286 0.787 -5.205 6.507 4.894 0.004 0.372 1.196

Regression: RET-RFR2 & STI-RFR2; f=UU & t=1999-2002 Regression Statistics Multiple R 0.910 2 R 0.827 2 0.793 Adjusted R Std Err 6.001 Observations 7

189

Data, computation and regression outputs

ANOVA df Regression Residual Total SS MS 1 862.257 862.257 5 180.037 36.007 6 1042.294 F 23.947 Sig F 0.004

Intercept STI-RFR2

Coeff Std Err 0.570 2.285 0.784 0.160

t Stat P-value Lower 95% Upper 95% 0.249 0.813 -5.303 6.442 4.894 0.004 0.372 1.196

190

Appendix A

UOB United Growth Fund (UUG) FMC RFR1 RFR2 Bank 0.63 1.00

Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 30-Jun-99 50.50 49.88 49.50 42.77 42.15 41.77 7.73 144.10 1.30 30-Sep-99 -6.40 -7.03 -7.40 -6.72 -7.35 -7.72 0.32 110.90 1.30 31-Dec-99 19.30 18.68 18.30 22.63 22.01 21.63 -3.33 124.60 1.40 31-Mar-00 -14.60 -15.23 -15.60 -13.99 -14.62 -14.99 -0.61 103.10 1.40 30-Jun-00 -1.20 -1.83 -2.20 -4.44 -5.06 -5.44 3.24 101.20 1.30 30-Sep-00 -3.60 -4.23 -4.60 -2.01 -2.63 -3.01 -1.59 96.00 1.60 31-Dec-00 -2.90 -3.53 -3.90 -3.52 -4.14 -4.52 0.62 104.40 1.30 31-Mar-01 -11.10 -11.73 -12.10 -13.11 -13.74 -14.11 2.01 89.10 1.30 30-Jun-01 6.40 5.78 5.40 3.12 2.50 2.12 3.28 95.40 1.30 30-Sep-01 -18.30 -18.93 -19.30 -23.57 -24.20 -24.57 5.27 76.20 1.30 31-Dec-01 20.00 19.38 19.00 23.04 22.42 22.04 -3.04 91.50 1.40 31-Mar-02 16.00 15.38 15.00 11.06 10.44 10.06 4.94 103.50 1.40 Average 4.51 3.88 3.51 2.94 2.31 1.94 1.57 103.33 1.36 Std dev 19.33 19.33 19.33 18.84 18.84 18.84 3.46 17.51 0.09 Beta 1.01 Info ratio 0.45 Jensen 1.55 2 0.97 R Sharpe Treynor Fund STI Fund STI RFR1 0.20 0.12 3.85 2.31 RFR2 0.18 0.10 3.48 1.94 Date RET RET-RFR1 RET-RFR2 STI STI-RFR1 RET-RFR2 RET-STI SZE EXP 31-Mar-03 -1.58 -2.21 -2.58 -5.46 -6.08 -6.46 3.88 83.65 1.18 30-Jun-03 15.05 14.43 14.05 14.20 13.58 13.20 0.85 107.59 1.18 30-Sep-03 13.71 13.09 12.71 12.63 12.01 11.63 1.08 134.45 1.19 31-Dec-03 7.12 6.50 6.12 8.20 7.57 7.20 -1.08 139.78 1.19 31-Mar-04 3.10 2.48 2.10 5.35 4.72 4.35 -2.25 141.43 1.16 30-Jun-04 -1.01 -1.64 -2.01 -1.13 -1.75 -2.13 0.12 134.56 1.18 30-Sep-04 8.32 7.70 7.32 7.98 7.36 6.98 0.34 134.97 1.19 Average 6.39 5.76 5.39 5.97 5.34 4.97 0.42 125.20 1.18 Std dev 6.61 6.61 6.61 7.10 7.10 7.10 1.92 21.53 0.01 Beta 0.90 Info ratio 0.22 Jensen 1.04 2 0.93 R Sharpe Treynor Fund STI Fund STI RFR1 0.87 0.75 6.43 5.34 RFR2 0.82 0.70 6.01 4.97

191

Data, computation and regression outputs

Regression: RET-RFR1 & STI-RFR1; f=UUG & t=1999-2002 Regression Statistics Multiple R 0.984 2 R 0.968 2 0.965 Adjusted R Std Err 3.620 Observations 12 ANOVA df Regression Residual Total SS MS F 1 3979.784 3979.784 303.697 10 131.045 13.104 11 4110.829 Sig F 0.000

Intercept STI-RFR1

Coeff Std Err 1.547 1.054 1.009 0.058

t Stat P-value Lower 95% Upper 95% 1.468 0.173 -0.800 3.895 17.427 0.000 0.880 1.138

Regression: RET-RFR1 & STI-RFR1; f=UUG & t=2003-2004 Regression Statistics Multiple R 0.963 2 R 0.928 2 0.914 Adjusted R Std Err 1.940 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 243.129 18.811 261.941 MS 243.129 3.762 F 64.623 Sig F 0.000

Intercept STI

Coeff Std Err 1.036 0.990 0.896 0.112

t Stat P-value Lower 95% Upper 95% 1.046 0.343 -1.509 3.582 8.039 0.000 0.610 1.183

Regression: RET-RFR2 & STI-RFR2; f=UUG & t=1999-2002 Regression Statistics Multiple R 0.984 2 R 0.968 2 0.965 Adjusted R Std Err 3.620 Observations 12

192

Appendix A

ANOVA df Regression Residual Total SS MS F 1 3979.784 3979.784 303.697 10 131.045 13.104 11 4110.829 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 1.551 1.051 1.009 0.058

t Stat P-value Lower 95% Upper 95% 1.475 0.171 -0.791 3.892 17.427 0.000 0.880 1.138

Regression: RET-RFR2 & STI-RFR2; f=UUG & t=2003-2004 Regression Statistics Multiple R 0.963 2 R 0.928 2 0.914 Adjusted R Std Err 1.940 Observations 7 ANOVA df Regression Residual Total 1 5 6 SS 243.129 18.811 261.941 MS 243.129 3.762 F 64.623 Sig F 0.000

Intercept STI-RFR2

Coeff Std Err 0.933 0.919 0.896 0.112

t Stat P-value Lower 95% Upper 95% 1.015 0.357 -1.430 3.295 8.039 0.000 0.610 1.183

193

Appendix B Conference paper 1

Sustainable Development of Bank-managed Mutual Funds: Evidence from Singapores Retail Funds Cheong Sing Tng* University of Tasmania Abstract In order to offer sustainable mutual fund products, banks would need to shed their reputation for being incapable of generating impressive returns, as their funds are not inferior to those from insurance and investment companies in terms of assets under management, expenditures, returns and risk. Previous research reporting the underperformance of bank-managed funds in comparison to their non-bank counterparts may have ignored their differing fiduciary standards. Examining the performance of retail funds approved for the Central Provident Fund Investment Scheme in Singapore is an attempt to evaluate the characteristics of bank and non-bank funds facing the same fiduciary responsibility. Keywords FUND PERFORMANCE; MANAGED FUNDS; SUSTAINABLILITY. 1 Introduction With financial deregulation reducing the distinction between banks, insurance companies and investment firms, sustainable development of mutual funds becomes critical for banks, whose profitability could be improved by offering competitive fund products (Gallo, Apilado & Kolari 1996). However, sustainable development of bank-managed funds requires banks to offer products that do not under-perform their counterparts from insurance and investment companies. Besides offering competitive funds, banks would need to shed their image of under-performers in the asset management industry. In the popular press, the performance of bank-managed mutual funds has been perceived as inferior to their non-bank counterparts. In the United States of America (USA), McTague (1994) claimed that bank funds were known to be not aggressive and incapable of generating impressive returns. As a result, various investors expecting little returns from bank funds were reluctant to invest in them. In Australia, the chief executive of AMP, one of the leading wealth management companies in Australia and New Zealand, mentioned that it was difficult for banks to compete with specialist investment companies. As a result, there were few international banks that were very successful in fund management (Moullakis & Patten 2005). Reviewing the literature, while earlier research indicated the underperformance of bank funds compared to their non-bank counterparts (Bauman & Miller 1995; Bogle & Twardowski 1980), later research did not detect any underperformance (Frye 2001). According to Frye (2001), the earlier research reporting underperformance of bank-managed funds relative to non-bank funds ignored their differing fiduciary standards. However, she focused only on bond funds as banks had more assets under management in bond funds than equity funds in the USA. It is not known how bank and non-bank equity funds compare when they face the same fiduciary standard. Examining equity funds approved by Singapores Central Provident Fund (CPF) Board for its CPF Investment Scheme facilitates a more direct comparison of funds managed by banks and non-banks than previous studies, as CPF-approved funds face the same standard for managing social security savings. This study contributes to understanding the sustainability of bank-managed funds by exploring the relationship between past performance and the type of fund management organization for Singapores retail equity funds.1 The majority of research on fund management is conducted using data from the USA and other large developed markets, leaving many small markets unexplored in the literature. Among the developed equity markets identified by Ibbotson and Brinson (1993, pp. 109-11), little research has been published on Singapores fund industry. Examining equity funds approved for Singapores CPF Investment Scheme offers an opportunity to control for differing fiduciary standards among equity funds. Besides risk and returns, the funds are examined in terms of size and expense ratio characteristics.

The author is an Associate Lecturer at the School of Accounting and Finance, University of Tasmania (Postal address: School of Accounting & Finance, University of Tasmania, Locked Bag 1314, Launceston, Tasmania Australia 7250; Telephone: +61 3 6324 3068; Facsimile: +61 3 6324 3711; Email: cheong.tng@utas.edu.au). He is completing his DBA thesis An Empirical Study of the Factors Influencing the Performance of Unit Trusts and would like to acknowledge the guidance of his supervisor Emeritus Professor Geoffrey Meredith from Southern Cross University. 1 The fund management industry can be divided into two segments: the retail segment accepting investment from the general public and the wholesale segment accepting investment from institutions.

194

Conference paper 1
Whether justified or not, banks reputation for relatively unimpressive fund performance could affect their popularity. Less popular funds would attract less investor cash flows, resulting in smaller amounts of net assets under management. The reputation of a fund could therefore affect its size. Small funds are not sustainable as economies of scale provide cost advantages to large funds. Large funds could lower brokerage commissions without a significant increase in administration and research costs (Indro et al. 1999), as measured by the expense ratio, resulting in better net returns than smaller funds with otherwise similar characteristics. Examples of small funds that were unsustainable in Singapore include ten Invesco technology funds (several being CPF-approved) worth S$80 million, which were shut down in 2003 due to investor apathy from declining stock markets (Chow 2003). This paper proceeds as follows. A conceptual framework for comparing characteristics of bank and non-bank funds is presented in the next section. Section three provides a description of the data and methodology used. Results are interpreted in section four before section five concludes with recommendation. 2 Hypotheses and model Comparing the characteristics of funds managed by banks and non-banks reveals the quality of funds offered by these institutions and verifies the results of previous studies on their performance. The quality of these funds is an indication of their sustainability in a competitive industry. Bank funds have grown rapidly over the years. This is confirmed in the USA by Frye (2001). In Singapore, the average bank fund has grown from around S$40 million during 1999-2002 to more than S$60 million in 2003-2004, as tabulated in Appendix A. If investors chase past performance and bank fund performance were inferior to non-bank funds, bank funds would not experience growth among rational investors. Inferior performance of bank funds coupled with their rapid growth could imply irrational or unsophisticated investors. Still, fund performance should still be important to investors. 2.1 Hypotheses for comparing bank and non-bank funds The first hypothesis under consideration is: H1: There is no significant performance difference between domestic equity funds managed by banks and non-banks. Previous research in the USA controlling for differing fiduciary standards of institutional groups considered only bond funds. Examining Singapores CPF-approved funds allows for comparison of returns for equity funds managed by banks and non-banks facing the same fiduciary standard. According to Berkowitz and Qiu (2003), technology usage in the fund management industry is quite homogenous across companies, which could lead to no overall performance difference between funds from banks and non-banks. Uniform information resources could make it difficult for specialist investment firms to outperform their bank counterparts. The second hypothesis is: H2: There is no significant difference in size between domestic equity funds managed by banks and non-banks. If there is no significant performance difference between bank and non-bank funds, both groups of funds should be equally popular among sophisticated retail investors, who contribute similar amounts of monies into the funds, resulting in no significant difference in size for the two fund groups. However, bank funds may be more popular with unsophisticated investors. As these investors are not knowledgeable in financial planning, they conveniently choose the asset management services of banks as they already have savings accounts with these familiar institutions. This is confirmed by Holiday (1994) who reported that unsophisticated investors go for bank funds. Therefore, non-banks may attract more sophisticated investors than banks. Nevertheless, the net assets under management of both groups of funds should reflect their popularity among investors in general. The third hypothesis is: H3: There is no significant difference in the level of systematic risk between domestic equity funds managed by banks and non-banks. The lack of significant performance difference between bank and non-bank funds could be related to their similar systematic risk level, as measured by beta using Sharpes (1964) capital asset pricing model (CAPM). However, McTague (1994) claimed that portfolio managers in banks have a reputation for being risk-averse. Their investment style could be more conservative than their non-bank counterparts, resulting in less transaction costs, which leads to the forth hypothesis: H4: There is no significant difference in the expense ratio of bank and non-bank funds. If there were no differences between the two groups of funds in expenditures, performance, size and risk, the reputation of bank funds being under-performers compared to their non-bank counterparts

195

Appendix B
would be unjustified. Bank funds would be sustainable when they have shed their image as underperformers, which should lead to improved profitability for banks. For examining the performance of funds for these hypotheses, the model presented in the following sub-section is used. 2.2 Capital asset pricing model for Singapore equity funds To examine the performance of equity funds, the quarterly returns for a unit trust are modelled using Sharpes (1964) CAPM model: Equation 1 Single-index model for returns of Singapores domestic equity funds

RET ft = RFRt + f (STI t RFRt ) + ft

where RETft f = return from fund f at time t; = Covf,STI / 2STI, covariance between the returns for fund f and the Singapore stock index (STI) divided by the variance of returns for the STI; = risk-free rate of return at time t; = return from the STI at time t; and = error term for fund f at time t.

RFRt STIt ft Source: adapted from Sharpe (1964). Equation 1 allows each funds return to depend on three factors: 1. pure time value of money, as measured by the risk-free rate RFRt, which is the reward for not taking any risk. As quarterly returns for CPF-approved funds are used in this model, RFRt corresponds to the guaranteed interest rates RFR1 = 2.5%/4 or 0.625 percent per quarter for the CPF Ordinary account; or RFR2 = 4.0%/4 or 1.0% percent per quarter for the CPF Special account;2 2. reward for bearing systematic risk, as measured by the Singapore stock market risk premium (STIt RFRt), which is the reward for bearing an average amount of systematic risk in the above-mentioned stock market; and 3. amount of systematic risk, as measured by f, which is the amount of systematic risk present in fund f relative to the systematic risk in an average asset from the Singapore stock market. Equation 1 therefore models the behaviour of fund returns according to beta, the market risk premium and risk-free rate of return. While there could be leads or lags in the market returns in relation to fund returns, the use of quarterly data should negate these effects. Lagging quarterly data would imply that a fund takes as long as three months to respond to changes in the market, which would not be consistent with the concept of capital market efficiency (Fama 1970, 1991). Studies on the Singapore stock market showed that the efficiency of the market strengthened as the time interval being considered increased (Wong 1988). Specifically, even though the Singapore market was not efficient using daily or weekly data (Lim 1985; Saw & Tan 1986), Ariff (1986) used monthly data to show that the Singapore market was comparable to the New York, London and Australian stock markets in adjusting prices efficiently to reflect new information. Testing the efficiency of the Singapore stock market using more recent data at various time intervals could be carried out for further research. When using quarterly instead of monthly returns data, this research considers the Singapore stock market to be at least weak-form efficient and comparable to other developed markets in terms of efficiency. Assuming weak-form efficiency implies the past and future rates of returns are independent (Fama 1970, 1991). The following section describes how the hypotheses were tested using the CAPM on secondary data. 3 Data and methodology In this section, the data collection process and its consideration for survivorship bias are described in the first sub-section. In the second sub-section, regression analysis of fund returns is explained before the third sub-section presents the testing performed for each hypothesis.

Information about the CPF Ordinary and Special accounts are presented in Section 3.2.

196

Conference paper 1
3.1 Secondary data collection To carry out this research, five years of quarterly returns from 1999 to 2004 for 19 retail funds approved for Singapores CPF Investment Scheme were examined. These funds were invested in shares from the Singapore Stock Exchange. Table 1 identifies the funds used for this research. For this research, only CPF-approved funds were considered as they followed the same fiduciary standard for managing social security savings, so as to control for differing fiduciary responsibilities. Failure to control for such standards could lead to biased test results, as highlighted by Frye (2001). Among these funds, funds investing only in the local stock market were selected. As each benchmark has a unique market cycle, funds based on benchmarks other than the Singapore Straits Times Index (STI) were excluded. These CPF-approved Singapore equity funds were classified according to the type of organization managing the fund: (1) insurance-linked investment products (ILPs) managed by insurance companies; (2) unit trusts managed by investment firms; or (3) bank-managed unit trusts. Each type of organization differs in terms of operational structure, priorities and benefits for fund managers, which might influence the resulting portfolio returns (Bauman & Miller 1995). Table 1 Research sample: all CPF-approved domestic equity funds from 1999 to 2004 with at least three quarterly periods of data Symbol 1999- 20022002 2004 ! ! DBS Horizon Singapore Equity Fund DHSE ! ! DBS Shenton Thrift Fund DST ! ! OCBC Savers Singapore Trust Fund OSST ! OUB Union Singapore Equity Fund OUSE ! ! UOB Optimix Singapore Equity Fund UOSE ! UOB Unifund UU ! ! UOB United Growth Fund UUG ! ! AXA Life-Fortress Fund ALF Insurance company ! AXA Life-Fortress Fund A ALFA ! GE Greatlink Singapore Equities Fund GGSE ! Keppel Managed Fund KM ! NTUC Income Singapore Equity Fund NISE ! OUB Manulife Golden Singapore Growth Fund OMGSG ! UOB Life FOF-Unifund ULFU ! ! UOB Life FOF-United Growth Fund ULFUG ! UOB LifeLink Growth Fund ULG ! ! Investment Aberdeen Singapore Equity Fund ASE company ! CMG First State Singapore Growth Fund CFSSG ! ! Schroder Singapore Trust SST Note: funds with less than three quarterly periods of complete data were excluded, as they would be insufficient for performing regression analysis. Source: funds identified from Mercer (1999-2002) and S&P (2003-2004). This research incorporates consideration for survivorship bias. As funds that do not survive are usually the worst performing ones, when data for these non-survivors are not considered, the resulting average performance of each fund group would be overstated. To control for survival bias, this study collected data for surviving and non-surviving funds using quarterly reports for all CPF-approved unit trusts. However, for performing regression analysis, only funds with at least three quarters of data were included. 3.2 Regression analysis For each fund, linear regression was performed using its quarterly risk premium (RETft RFRt) as the dependent variable and the STI quarterly risk premium (STIt RFRt) as the independent variable. As there are two risk-free rates: RFR1 = 0.625 percent per quarter and RFR2 = 1 percent per quarter for the CPF Ordinary and Special accounts respectively, as well as two holding periods corresponding to data collected from Mercer (1999-2002) and S&P (2003-2004), four sets of linear regression are performed Organization type Bank Fund

197

Appendix B
for each fund using a combination of risk-free rate and holding period.3 Figure 1 illustrates time-series regression with a single fund from the sample. Figure 1 Illustration of time-series regression for a single fund

Time-series regression of quarterly fund risk premium versus market risk premium for a fund from 1999:Q2 to 2002:Q1
R sq = 0.91 80.00 Investment firm: 60.00 Aberdeen Singapore 40.00 Equity Fund (ASE) 20.00 0.00 Linear (Investment -20.00 firm: Aberdeen -40.00 Singapore Equity -40.00 -20.00 0.00 20.00 40.00 60.00 Fund (ASE)) STIt - RFRt

Source: developed for this research. In this example, 12 quarterly risk premiums from a fund were regressed against the corresponding quarterly risk premiums of the STI index. As most of the data points cluster around the fitted regression line, the value of the coefficient of determination R2 is quite high at around 0.9, which was confirmed by Brinson, Hood and Beebower (1986), Brinson, Singer and Beebower (1991), Ibbotson and Kaplan (2000) and Jensen (1968) for equity funds in the USA, as well as Tng (2005) for domestic equity funds in Singapore. Figure 2 shows a normal probability plot for the same example. Figure 2 Illustration of normal probability plot for a single fund

RETft - RFRt
RETft - RFR1

Normal Probability Plot for ASE

100 50 0 -50 0 20 40 60 80 100 120 Sample Percentile

Source: developed for this research. For residual analysis, the mostly linear trend on the normal probability plot suggests that the normality assumption for linear regression is satisfied (Mendenhall & Sincich 1996). After regression analysis was performed for each fund, the hypotheses developed in Section 2 were tested using sample data.
3

While the Special account is used for old age, contingency purposes and investment in retirementrelated financial products, the Ordinary account is used to pay for education, housing, insurance and investment. The CPF interest rates are revised quarterly based on the rate of interest paid by the local banks on deposit and savings accounts. However, the CPF Act guarantees a minimum annual interest rate of 2.5 and 4.0 percent for the ordinary and special accounts respectively (CPF Handbook: Building Our Future, 2005, CPF Board, viewed 21 Mar 2005, <http://www.cpf.gov.sg>.). Therefore, investors who choose to invest their Ordinary or Special accounts in funds incur the opportunity cost of interest on CPF money.

198

Conference paper 1
3.3 Hypothesis testing To compare the returns, beta, net assets under management and expense ratio of funds managed by banks and non-banks, the usual two-tailed pooled-variance t-test for differences in two means was conducted for bank and non-bank funds during each of the time periods 1999-2002 and 2003-2004. The hypothesis that there is no difference between an average characteristic of bank funds (1) and non-bank funds (2) is tested by the following t value: Equation 2 t-statistic for comparison of a characteristic of bank and non-bank funds

t=

X 1 X 2 ( 1 2 ) 1 2 1 Sp + n n 2 1

where
2 SP =

(n1 1)S12 + (n2 1)S 22 (n1 1) + (n2 1)

and
2 Sp

= pooled variance; = mean characteristic of the sample taken from the population of bank funds; = variance of the characteristic for the sample taken from bank funds; = size of the sample taken from the population of bank funds; = mean characteristic of the sample taken from the population of non-bank funds; = variance of the characteristic for the sample taken from non-bank funds; and

X1 S12
n1

X2 S 22

n2 = size of the sample taken from the population of non-bank funds. Source: adapted from Levine et al.(2002, p. 375) The t-statistic was compared to the critical t value t( = .05, df = n1 + n2 2) to determine whether the difference between the means was significantly not equal to zero. A t value greater than t( = .05, df = n1 + n2 2) would lead to rejection of the hypothesis. Results of the hypothesis testing are interpreted in the next section. 4 Results and interpretation In this section, the four sub-sections report findings for the hypotheses developed in Section 3 in the same order as they were presented. 4.1 Performance of bank and non-bank funds Referring to Appendix A, during 1999-2002, when the STI posted an average quarterly return of 2.94 percent, the average bank equity fund under-performed the market at 1.37 percent, while the average non-bank fund managed to outperform the market at 3.60 percent. For 2003-2004, with an STI average quarterly return of 5.97 percent, the reverse seemed to be true as bank funds outperformed the market at 7.68 percent while non-bank funds under-performed the market at 5.84 percent. However, performing a two-sample t-test assuming unequal variances for returns of bank and non-bank funds in Appendix B showed no significant difference between the performance of bank and non-bank equity funds during each holding period, supporting Fryes (2001) findings for bond funds. This result was also confirmed by Tng (2005) who used ratios developed by Goodwin (1998), Jensen (1968), Sharpe (1966) and Treynor (1965) to evaluate the performance of the these CPF-approved equity funds. 4.2 Size of bank and non-bank funds Referring to Appendix A, during 1999-2002, the average bank equity fund managed about S$40 million of net assets, while its non-bank counterpart managed about $42 million. For 2003-2004, the average bank fund managed about S$63 million, while its non-bank counterpart managed about $69 million. While both groups of funds had experienced significant growth in size, bank equity funds seemed to be smaller than those from non-banks. However, performing a two-sample t-test assuming unequal variances for the size of bank and nonbank funds in Appendix C showed no significant difference in size between bank and non-bank equity

199

Appendix B
funds during each holding period, supporting the hypothesis that both groups of funds were equally popular when there was no significant difference in returns. 4.3 Beta of bank and non-bank funds Referring to Appendix A, during 1999-2002, the average bank equity fund beta was 0.87, while its non-bank counterpart was 0.99. For 2003-2004, the average bank fund beta was 1.16, while its nonbank counterpart was 0.78. With a beta less than one, non-bank equity funds had maintained their systematic risk below that of the market. Even though the bank funds initially had less systematic risk than their non-bank counterparts, the systematic risk of bank equity funds seemed to have exceeded both the market and the non-bank funds during the later period. Performing a two-sample t-test assuming unequal variances for the beta of bank and non-bank funds in Appendix D showed no significant difference in beta between bank and non-bank equity funds during 1999-2002, but during 2003-2004, the average beta for bank funds was significantly higher than its non-bank counterpart, providing evidence that bank equity funds were systematically more risky than their non-bank counterparts during the later period. Besides, the bank-managed funds had higher total risk than their non-bank counterparts during both periods, as measured by the variance of quarterly returns in Appendix B. This differed from Fryes (2001) results, who found that bank-managed bond funds were more conservative than their non-bank counterparts in the USA. 4.4 Expense ratio of bank and non-bank funds Referring to Appendix A, during 1999-2002, the average expense ratio for bank equity funds was 2.0, while its non-bank counterpart was 1.74. For 2003-2004, the average bank fund expense ratio was 2.07, while the expense ratio for its non-bank counterpart was 1.34. The expense ratio for bank equity funds was generally higher than for non-bank funds. While proportionate expenditures were reduced for nonbank funds, they were increased for bank funds. However, performing a two-sample t-test assuming unequal variances for the expense ratio of bank and non-bank funds in Appendix E showed no significant difference in proportionate expenditures between the two groups during each of the two periods. 5 Conclusion and recommendation Banks could offer sustainable mutual funds if they are competitive products that do not under-perform the funds from insurance and investment companies. Most of the literature claimed that bank-managed funds under-performed their non-bank counterparts. However, these studies did not control for differing fiduciary standards. This research found that bank-managed domestic equity funds approved for Singapores CPF Investment Scheme did not under-perform their non-bank counterparts. However, in terms of investment strategy, there was some evidence that bank fund managers were not more conservative than their non-bank counterparts. In terms of fund size and expense ratio, there was no significant difference between bank and non-bank funds. These empirical results have important implications for banks in terms of the sustainability of their mutual fund products, as the fund management industry is important for the profitability of banks. This research has shown that banks were capable of offering competitive fund products. To sustain their share of the mutual fund industry, banks should shed their unjustified reputation as non-aggressive fund managers incapable of generating impressive returns, as this study has shown that the characteristics of domestic bank equity funds were not inferior to their non-bank counterparts. Otherwise, bad reputation leading to investor apathy could result in small funds that have to be terminated eventually. References Ariff, M 1986, 'Announcement effects in a thin market--a study of the Singapore equity market', paper presented to Proceedings of the Academy of International Business, Southeast Asia Regional Conference, Taipei, 26-28 June. Bauman, W & Miller, R 1995, 'Portfolio performance rankings in stock market cycles', Financial Analysts Journal, vol. 51, no. 2, pp. 79-87. Berkowitz, M & Qiu, J 2003, 'Ownership, risk and performance of mutual fund management companies', Journal of Economics and Business, vol. 55, pp. 109-34. Bogle, J & Twardowski, J 1980, 'Institutional investment performance: banks, investment counselors, insurance companies and mutual funds', Financial Analysts Journal, vol. 36, no. 1, pp. 33-41. Brinson, G, Hood, L & Beebower, G 1986, 'Determinants of portfolio performance', Financial Analysts Journal, vol. 42, no. 4, pp. 39-48.

200

Conference paper 1
Brinson, G, Singer, B & Beebower, G 1991, 'Determinants of portfolio performance II: an update', Financial Analysts Journal, vol. 47, no. 3, pp. 40-8. Chow, H 2003, 'Impact of falling stock markets: Invesco pulls plug on unit trusts here', The Straits Times, 7 Jun, p. A40. CPF Handbook: Building Our Future, 2005, CPF Board, viewed 21 Mar 2005, <http://www.cpf.gov.sg>. Fama, E 1970, 'Efficient capital markets: a review of theory and empirical work', Journal of Finance, vol. 25, no. 2, pp. 383-417. ---- 1991, 'Efficient capital markets: II', Journal of Finance, vol. 46, no. 5, pp. 1575-617. Frye, M 2001, 'The performance of bank-managed mutual funds', Journal of Financial Research, vol. 24, no. 3, pp. 419-42. Gallo, J, Apilado, V & Kolari, J 1996, 'Commercial bank mutual fund activities: implications for bank risk and profitability', Journal of Banking and Finance, vol. 20, pp. 1775-91. Goodwin, T 1998, 'The information ratio', Financial Analysts Journal, vol. 54, no. 4, pp. 34-43. Holiday, K 1994, 'Mutual funds', Bank Marketing, vol. 26, pp. 23-31. Ibbotson, R & Brinson, G 1993, Global Investing, McGraw-Hill, New York. Ibbotson, R & Kaplan, P 2000, 'Does asset allocation policy explain 40, 90 or 100 percent of performance?' Financial Analysts Journal, vol. 56, no. 1, pp. 26-33. Indro, D, Jiang, C, Hu, M & Lee, W 1999, 'Mutual fund performance: does fund size matter?' Financial Analysts Journal, vol. 55, no. 3, pp. 74-87. Jensen, M 1968, 'The performance of mutual funds in the period 1945-1964', Journal of Finance, vol. 23, no. 2, pp. 389-416. Levine, D, Stephan, D, Krehbiel, T & Berenson, M 2002, Statistics for Managers Using Microsoft Excel, 3 edn, Prentice Hall, New Jersey. Lim, G 1985, 'A Box-Jenkins Approach to Stock Market Analysis on the SES', National University of Singapore. McTague, J 1994, 'Laggards no longer', Barron's, vol. 74, pp. F22-F3. Mendenhall, W & Sincich, T 1996, A Second Course in Statistics: Regression Analysis, 5 edn, Prentice Hall, New Jersey. Mercer 1999-2002, Unit Trusts and Investment-linked Insurance Products Included Under CPFIS: Performance and Risk Monitoring Reports, Mercer Investment Consulting, viewed 19 Feb 2005, <http://www.mercerfundwatch.com/>. Moullakis, J & Patten, S 2005, 'Big-spending banks caught with wealth of problems', The Australian Financial Review, 8 Aug, p. 1. S&P 2003-2004, Performance and Risk Monitoring Report for CPFIS-included Unit Trusts & Investment-linked Insurance Products, S&P Fund Services, viewed 19 Feb 2005, <http://www.imas.org.sg>. Saw, S & Tan, K 1986, 'The Stock Exchange of Singapore all-share price indices: testing of independence', Securities Industry Review, vol. 12, no. 1. Sharpe, W 1964, 'Capital asset prices: a theory of market equilibrium under condtions of risk', Journal of Finance, vol. 19, no. 3, pp. 424-42. ---- 1966, 'Mutual fund performance', Journal of Business, vol. 39, no. 1, pp. 119-38. Tng, C 2005, 'Performance of Approved Equity Funds: Evidence from Singapore's Retail Funds', paper presented to 18th Annual Australasian Finance and Banking Conference, Sydney, 14-16 Dec. Treynor, J 1965, 'How to rate management of investment funds', Harvard Business Review, vol. 43, no. 1, pp. 63-75. Wong, K 1988, 'Role of Singapore and Malaysian stock exchanges', in C Tan & K KC (eds), Handbook of Singapore-Malaysian Corporate Finance, Butterworths, Singapore.

201

Appendix B

Appendix A Summary characteristics of Singapore's CPF-approved domestic equity funds

Fund Type 1999:Q2 to 2002:Q1 Average Size (S$ Expense quarterly million) ratio return (%) -0.46 75.310 1.500 4.33 43.96 0.96 4.53 17.030 1.490 -6.13 3.000 3.32 2.57 2.570 3.830 0.23 35.960 1.570 4.51 103.330 1.360 2.05 17.720 1.450 2003:Q1 to 2004:Q3 Average Size (S$ Expense quarterly million) ratio return (%) 0.953 7.28 109.260 1.490 1.113 12.30 59.50 1.16 1.074 7.49 19.060 1.520 0.323 0.782 4.93 0.850 5.020 0.851 1.009 6.39 125.200 1.180 1.011 6.11 62.200 1.240 4.71 45.190 1.670 4.88 30.310 1.200 1.033 7.45 1.690 0.430 1.144 0.705 0.859 6.11 84.230 1.180 0.867 1.000 5.59 21.110 2.120 1.176 1.141 6.01 238.160 1.520 Beta 6.604 7.68 5.84 5.97 0.63 1.00 66.396 62.77 68.98

Beta

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST

Bank Bank Bank Bank Bank Bank Bank Insurance Insurance Insurance Insurance Insurance Insurance Insurance Insurance Insurance Investment Investment Investment

1.186 1.511 1.194 1.001 0.896 0.398 0.614 0.822 1.074

0.12 -0.66 7.00 6.03 -0.64 5.84 5.36 7.32 2.625 1.37 3.60 2.94 0.63 1.00

10.580 19.560 35.050 91.650 0.410 6.970 86.330 113.420 41.428 40.17 42.41

1.160 2.040 1.550 1.350 1.510 2.580 2.350 1.710

0.895 0.706 0.954

Average fund Average bank fund Average non-bank fund Straits Times Index Ord a/c interest rate Sp a/c interest rate

1.858 0.940 2.00 0.87 1.74 0.99 1.00

1.644 0.938 2.07 1.16 1.34 0.78 1.00

Source: developed for this research using data extracted from Mercer (1999-2002) and S&P (20032004).

202

Conference paper 1

Appendix B Two-sample t-test assuming unequal variances for returns of bank and non-bank funds Alpha = 0.05

Source: developed for this research using data extracted from Mercer (1999-2002) and S&P (20031999:Q2 to 2002:Q1 Bank Non-bank Mean 1.368571429 3.60222222 Variance 15.15754762 11.2570694 Observations 7 9 Hypothesized Mean Difference 0 df 12 t Stat -1.208500534 P(T<=t) one-tail 0.125065181 t Critical one-tail 1.782286745 P(T<=t) two-tail 0.250130362 t Critical two-tail 2.178812792 2003:Q1 to 2004:Q3 Mean Variance Observations Hypothesized Mean Difference df t Stat P(T<=t) one-tail t Critical one-tail P(T<=t) two-tail t Critical two-tail 2004). Bank Non-bank 7.678 5.837142857 7.69177 0.837957143 5 7 0 5 1.42961693 0.106107504 2.015049176 0.212215008 2.570577635

203

Appendix B

Appendix C Two-sample t-test assuming unequal variances for size of bank and nonbank funds Alpha = 0.05 1999:Q2 to 2002:Q1 Mean Variance Observations Hypothesized Mean Difference df t Stat P(T<=t) one-tail t Critical one-tail P(T<=t) two-tail t Critical two-tail Bank Non-bank 40.16547619 42.41 1431.151521 1826.87435 7 9 0 14 -0.111197335 0.456519161 1.76130925 0.913038322 2.144788596

2003:Q1 to 2004:Q3 Bank Non-bank Mean 62.77314286 68.98428571 Variance 2953.549099 6296.869529 Observations 5 7 Hypothesized Mean Difference 0 df 10 t Stat -0.160894099 P(T<=t) one-tail 0.437690385 t Critical one-tail 1.812461505 P(T<=t) two-tail 0.875380769 t Critical two-tail 2.228139238 Source: developed for this research using data extracted from Mercer (1999-2002) and S&P (20032004).

204

Conference paper 1

Appendix D Two-sample t-test assuming unequal variances for beta of bank and nonbank funds Alpha = 0.05 1999:Q2 to 2002:Q1 Mean Variance Observations Hypothesized Mean Difference df t Stat P(T<=t) one-tail t Critical one-tail P(T<=t) two-tail t Critical two-tail 2003:Q1 to 2004:Q3 Mean Variance Observations Hypothesized Mean Difference df t Stat P(T<=t) one-tail t Critical one-tail P(T<=t) two-tail t Critical two-tail Bank Non-bank 1.1576 0.780428571 0.0549953 0.051775952 5 7 0 9 2.780875316 0.010685085 1.833113856 0.021370171 2.262158887 Bank Non-bank 0.872142857 0.99288889 0.072359476 0.02461786 7 9 0 9 -1.05607664 0.159228053 1.833113856 0.318456107 2.262158887

Source: developed for this research using data extracted from Mercer (1999-2002) and S&P (20032004).

205

Appendix B

Appendix E Two-sample t-test assuming unequal variances for expense ratio of bank and non-bank funds Alpha = 0.05 1999:Q2 to 2002:Q1 Mean Variance Observations Hypothesized Mean Difference df t Stat P(T<=t) one-tail t Critical one-tail P(T<=t) two-tail t Critical two-tail 2003:Q1 to 2004:Q3 Mean Variance Observations Hypothesized Mean Difference df t Stat P(T<=t) one-tail t Critical one-tail P(T<=t) two-tail t Critical two-tail Bank Non-bank 2.074571429 1.337142857 2.739075918 0.272157143 5 7 0 5 0.962750715 0.189944652 2.015049176 0.379889304 2.570577635 Bank Non-bank 2.004047619 1.74444444 1.213475794 0.22875278 7 9 0 8 0.582282662 0.288207173 1.85954832 0.576414346 2.306005626

Source: developed for this research using data extracted from Mercer (1999-2002) and S&P (20032004).

206

Appendix C Conference paper 2

Performance of Approved Equity Funds: Evidence from Singapores Retail Funds Cheong S. Tng* University of Tasmania Abstract Bank-managed equity funds are not inferior to their non-bank counterparts. However, the performance comparison is complicated by their differing fiduciary standards. Previous research reporting the relative underperformance of bank-managed funds may have ignored their differing fiduciary standards. Examining the performance of domestic retail funds approved for the Central Provident Fund Investment Scheme in Singapore is an attempt to evaluate bank and non-bank funds facing the same fiduciary responsibility. The returns from these funds correlated highly with their market returns. Even though returns from these funds were higher than the guaranteed interest rates, they did not outperform their market index. Keywords FUND MANAGEMENT COMPANIES; MUTUAL FUNDS; PERFORMANCE EVALUATION.

1 Introduction When choosing asset management companies to generate returns from fund investments, investors prefer fund managers who have a track record of outperforming the industry average. Past performance was reported by Kritzman (1983) as a common criterion for selecting fund managers; while Gruber (1996) as well as Sirri and Tufano (1998) confirmed that investors chase past performance. Performance rankings of funds compiled by financial periodicals are used to identify best-performing funds for new investments.1 Even though the performance of fund managers has been studied extensively, existing literature do not agree on the performance of funds managed by banks, compared to their counterparts from insurance and investment companies. While earlier research indicated the underperformance of bank funds compared to their non-bank counterparts (Bauman & Miller 1995; Bogle & Twardowski 1980), later research did not detect any underperformance (Frye 2001). According to Frye (2001), the earlier research reporting underperformance of bank-managed funds relative to non-bank funds ignored their differing fiduciary standards. However, she focused only on bond funds as banks had more assets under management in bond funds than equity funds in the United States of America (USA). It is not known how bank and non-bank equity funds compare when they face the same fiduciary standard. Examining retail equity funds approved by Singapores Central Provident Fund (CPF) Board for its CPF Investment Scheme facilitates a more direct comparison of funds managed by banks and non-banks than previous studies, as CPF-approved funds face the same standard for managing social security savings.2 This study contributes to the understanding of relative performance of equity funds managed by banks and nonbanks facing the same fiduciary standard by exploring the relationship between past performance and the type of fund management organization. The majority of research on fund management is conducted using data from the USA and other large developed markets, leaving many small markets unexplored in the literature. Among the developed equity markets identified by Ibbotson and Brinson (1993, pp. 109-11), little research has been published on Singapores fund industry. Examining equity funds approved for Singapores CPF Investment Scheme offers an opportunity to control for differing fiduciary standards among equity funds. The paper proceeds as follows. The next section presents hypotheses and models on performance of funds managed by banks and non-banks. Section three provides a description of the data and methodology used. Results are interpreted in section four before section five concludes with recommendation.

The author is an Associate Lecturer at the School of Accounting and Finance, University of Tasmania (Postal address: School of Accounting & Finance, University of Tasmania, Locked Bag 1314, Launceston, Tasmania Australia 7250; Telephone: +61 3 6324 3068; Facsimile: +61 3 6324 3711; Email: cheong.tng@utas.edu.au) and would like to acknowledge the guidance of his DBA supervisor Emeritus Professor Geoffrey Meredith from Southern Cross University. 1 Examples of periodicals reporting fund performance include the Australian Financial Review, Forbes, Fortune and the Wall Street Journal. 2 The fund management industry can be divided into two segments: the retail segment accepting investment from the general public and the wholesale segment accepting investment from institutions. The CPF is a public defined-contribution pension plan for Singaporean employees, who decide how the funds in their accounts will be invested in CPF-approved securities.

207

Appendix C
2 Hypotheses and models Comparing the performance of funds managed by banks and non-banks reveals the quality of funds offered by these institutions and verifies the results of previous studies on their performance. Bank funds have grown rapidly over the years. This is confirmed in the USA by Frye (2001). In Singapore, the average bank fund has grown from S$40 million during 1999-2002 to more than S$60 million in 2003-2004, as tabulated in Appendix A. If investors chase past performance and bank fund performance were inferior to nonbank funds, bank funds would not experience growth among rational investors. Inferior performance of bank funds coupled with their rapid growth could imply irrational or unsophisticated investors. Unsophisticated investors may not be knowledgeable in financial planning. They conveniently choose the asset management services of banks as they already have savings accounts with these familiar institutions. This is confirmed by Holiday (1994) who reported that unsophisticated investors go for bank funds. Therefore, non-banks may attract more sophisticated investors than banks. Still, fund performance should still be important to investors. 2.1 Hypotheses The first hypothesis under consideration is: H1: CPF-approved domestic equity funds do not outperform the benchmark STI index. Among studies on fund performance in the USA, Jensen (1968) found that on a yearly basis, the funds on average earned about one percent less than their expected returns, given the level of risk. Adding expenses to derive gross risk-adjusted returns showed that slightly more than 40 percent outperformed the market while deriving net returns revealed one-third of the funds outperforming the market. As less than 50 percent of the funds could outperform the market on a risk-adjusted basis, the funds on average could not beat the market. Carlson (1970) reported worse fund performance. Specifically, a few groups of funds had gross returns better than the S&P 500 and the New York Stock Exchange composite indexes, but all the groups had net returns under-performing the market indexes. Besides, no consistency was reported for performance on a risk-adjusted basis as less than one-third of the funds managed to perform above average for two sub-periods. Therefore, the average fund was consistently inferior to the overall market. Extending the range of funds studied to international funds, Cumby and Glen (1990) found that fund performance did not outperform the Morgan Stanley world equity index. As for bond funds, Blake, Elton and Gruber (1993) confirmed these funds under-performed their indexes too. On a country basis, Cai, Chan and Yamada (1997) showed that Japanese mutual funds generally underperformed their benchmarks; while Koh (1999) showed that Singapores unit trusts generally under-performed their market indexes from 1976 to 1994. This study examines the performance of CPF-approved domestic equity funds from 1999 to 2004. The second hypothesis is: H2: There is a positive relation between fund returns and the returns to the market index. Jensen (1968) found that funds correlated with their markets at rates higher than 0.90. As the benchmark market reflects each funds asset allocation policy, this result was also confirmed by Brinson, Hood and Beebower (1986), Brinson, Singer and Beebower (1991), as well as Ibbotson and Kaplan (2000), who reported that about 90 percent of the variability in returns of a typical fund across time is explained by its asset allocation policy. If CPF-approved domestic equity funds could not outperform their market index, as stated in the first hypothesis, and the returns from the benchmark market determines more than 90 percent of the variability in returns of a typical fund across time, the majority of funds should not be managed actively. Instead, they should be managed passively as index funds to match the market performance and minimize the cost of research and trading. Index funds duplicate the composition and performance of their targeted markets and have provided comparable or even better performance than actively managed funds most of the time (Reilly & Brown 2003, p. 202). The third hypothesis is: H3: There is no performance difference between CPF-approved domestic equity funds managed by banks and non-banks. Previous research controlling for differing fiduciary standards of institutional groups considered only bond funds. Examining Singapores CPF-approved funds allows for performance comparison of equity funds managed by banks and non-banks facing the same fiduciary standard. According to Berkowitz and Qiu (2003), technology usage in the fund management industry is quite homogenous across companies. This could lead to no significant performance difference between equity funds from banks and non-banks. H4: There is no difference between the returns from investing CPF savings in domestic equity funds and earning the guaranteed interest rates from the ordinary and special accounts. Assuming fund investors are rational, they will leave their monies to earn guaranteed interest rates if the rate of return from unit trusts does not outperform the guaranteed interest rates. Sirri and Tufano (1998) agreed with Gruber (1996) that past performance of funds could influence the selection decision. The large amount of monies being invested in unit trusts and their growth might not support this hypothesis, as funds should have good performance in order to attract investors. However, Holiday (1994) reasoned that novice investors with

208

Conference paper 2
little financial sophistication relied more on marketing than performance information. These disadvantaged investors fund purchasing decision could be based on advertisement or advice from brokers, which might be the reason why poorly performing funds are still popular, as Gruber (1996) hypothesized. In Singapore, Koh (1999) reported the majority of unit trusts earned returns lower than the CPF interest rates. To examine the performance of funds for these hypotheses, the models presented in the following sub-section are used. 2.2 Models In this sub-section, the most popular asset-pricing model, as well as four related fund performance measures are presented. 2.2.1 Capital asset pricing model To examine the performance of equity funds, the quarterly returns for a unit trust are modelled using Sharpes (1964) capital asset pricing model (CAPM): Equation 1 Single-index model for returns of Singapores domestic equity funds

RET ft = RFRt + f (STI t RFRt ) + ft

= return from fund f at time t; = Covf,STI / 2STI, covariance between the returns for fund f and the Singapore stock index (STI) divided by the variance of returns for the STI; RFRt = risk-free rate of return at time t; STIt = return from the STI at time t; and = error term for fund f at time t. ft Source: adapted from Sharpe (1964). Equation 1 allows each funds return to depend on three factors: 1. pure time value of money, as measured by the risk-free rate RFRt, which is the reward for not taking any risk. As quarterly returns for CPF-approved funds are used in this model, RFRt corresponds to the guaranteed interest rates RFR1 = 2.5%/4 or 0.625 percent per quarter for the CPF Ordinary account; or RFR2 = 4.0%/4 or 1.0% percent per quarter for the CPF Special account; 2. reward for bearing systematic risk, as measured by the Singapore stock market risk premium (STIt RFRt), which is the reward for bearing an average amount of systematic risk in the above-mentioned stock market; and 3. amount of systematic risk, as measured by f, which is the amount of systematic risk present in fund f relative to the systematic risk in an average asset from the Singapore stock market. Equation 1 therefore models the behaviour of fund returns according to beta, the market risk premium and riskfree rate of return in the Singapore stock market. While there could be leads or lags in the market returns in relation to the fund returns, the use of quarterly data should negate these effects. Lagging quarterly data would imply that a fund takes as long as three months to respond to changes in the market, which would not be consistent with the concept of capital market efficiency (Fama 1970, 1991). Studies on the Singapore stock market showed that the efficiency of the market strengthened as the time interval being considered increased (Wong 1988). Specifically, even though the Singapore market was not efficient using daily or weekly data (Lim 1985; Saw & Tan 1986), Ariff (1986) used monthly data to show that the Singapore market was comparable to the New York, London and Australian stock markets in adjusting prices efficiently to reflect new information. Testing the efficiency of the Singapore stock market using more recent data at various time intervals could be carried out for further research. When using quarterly instead of monthly returns data, this research considers the Singapore stock market to be at least weak-form efficient and comparable to other developed markets in terms of efficiency. Assuming weak-form efficiency implies the past and future rates of returns are independent (Fama 1970, 1991). 2.2.2 Fund performance evaluation measures To compare fund performance, four risk-adjusted portfolio performance measures are used, unlike financial periodicals that emphasize raw returns. Among these measures, the information ratio (Goodwin 1998) was developed more recently while the measures developed by Jensen (1968), Sharpe (1966) and Treynor (1965) are based on the CAPM. The information ratio IRf for an equity fund f is computed as the arithmetic average of a funds excess return divided by the standard deviation of the excess return (Goodwin 1998): where RETft f

209

Appendix C
Equation 2 Information ratio for evaluating Singapores domestic equity funds

IR f =
where

ER f ER

= average excess return for fund f during a specified time period, which is equal to the average return for fund f during the period minus the average return for the benchmark STI index during that period; and = standard deviation of the excess return during the period. ER Source: adapted from Goodwin (1998). The IR could be interpreted as the mean excess return per unit of unsystematic risk. As for its relation with other evaluation measures, the IR could be expressed in terms of the Jensen alpha when excess returns are estimated with historical data using a single-factor regression equation, while the Sharpe ratio is a special case of the IR (Goodwin 1998). The Jensen alpha ft for an equity fund f during a time period t is:

ER f

Equation 3 Jensens alpha for evaluating Singapores domestic equity funds

ft = RET ft RFRt f (STI t RFRt )

where = return from fund f for the time period t; RETft = beta coefficient for fund f; f = rate of return on risk-free assets during the time period t; and RFRt = return from the STI during the time period t. STIt Source: adapted from Jensen (1968). As ft is the difference between the actual and expected returns, the measure represents the proportion of the fund return that is attributable to the managers ability to derive above-average returns adjusted for risk. Significantly positive and negative values of ft represent superior and inferior managers respectively in market timing as well as stock selection in the Singapore stock market, while a near-zero value matches the market performance on a risk-adjusted basis. The Sharpe ratio Sf for an equity fund f is defined as: Equation 4 Sharpe ratio for evaluating Singapores domestic equity funds

Sf =
where

RET f RFR f

RET RFR

= average rate of return for fund f during a specified time period; = average rate of return on risk-free assets during the time period; and

= standard deviation of the rate of return for fund f. f Source: adapted from Sharpe (1966). Sf considers the total risk of the portfolio by including the standard deviation of returns in its denominator, resulting in a measure of the risk premium earned per unit of total risk. The Treynor ratio Tf for an equity fund f is computed as the funds risk premium divided by its beta coefficient: Equation 5 Treynor ratio for evaluating Singapores domestic equity funds

Tf =
where

RET fi RFR f

RET RFR

= average rate of return for fund f during a specified time period; = average rate of return on risk-free assets during the time period; and

= beta coefficient for fund f. f Source: adapted from Treynor (1965). As the Jensens alpha and the information ratio compare fund performance with the performance of the market index, they were used for the first and third hypotheses to compare the performance of domestic equity funds

210

Conference paper 2
with the STI as well as the relative performances of funds managed by banks and non-banks in outperforming the market respectively. The Sharpe and Treynor ratios incorporate the risk-free rate of return and were therefore used for the forth hypothesis to determine whether there is excess return over the guaranteed interest rates when investing in CPF-approved funds. Even though these performance measures improve upon mere comparison of returns in many financial periodicals, they are not without problems as some researchers identified the presence of bias in these measures. For example, Friend and Blume (1970) reported the risk-adjusted performance measures of low-risk portfolios were better than their high-risk counterparts; Klemkosky (1973) found a positive relationship between the risk involved and the composite performance measures; while Leland (1999) demonstrated the possibility of the Jensens alpha for a portfolio managers performance being biased downward for portfolios designed to limit downside risk. While the performance measures are not without problems, in the absence of alternative measures, all the four measures were used for this research to minimize errors resulting from relying solely on one measure, as each measure ranked individual fund performance differently (Reilly & Brown 2003, p. 1122) and can yield substantially different performance rankings (Corrado & Jordan 2005, p. 434). These measures use the arithmetic mean return as the measure of central tendency for performance, and a measure of dispersion for risk, which can be the variance (or its square root, the standard deviation) for total risk or beta for relative systematic risk. An alternative performance measure is the geometric mean return, which is also a single measure combining risk and return (Seitz & Ellison 1999, p. 340). If the objective of the investor is to maximize the expected longterm growth rate of wealth and the probability distribution of holding period return is the same for each time period, the fund with the highest geometric mean return also provides the highest long-run growth rate of wealth (Seitz & Ellison 1999, p. 420). However, investors are also concerned with short-term performance as they select fund managers based on recent performance. Therefore, the more popular mean-variance model is used, even though the geometric mean model has a few academic supporters, including Latane and Tuttle (1967), Markowitz (1976) and Rubinstein (1976). The following section describes how to test the hypotheses using the CAPM and the performance measures on secondary data. 3 Data and methodology In this section, the data collection process is described in the first sub-section before the second sub-section considers how survivorship bias is minimized during data collection. In the third sub-section, regression analysis is described to estimate Jensens alpha for each fund before the fourth sub-section presents the testing performed for each hypothesis. 3.1 Secondary data collection To carry out this research, five years of quarterly returns from 1999 to 2004 for 19 retail funds approved for Singapores CPF Investment Scheme were examined. These funds were invested in shares from the Singapore Stock Exchange. Table 1 on the following page identifies the funds used for this research. For this research, only CPF-approved funds were considered as they followed the same fiduciary standard for managing social security savings, so as to control for differing fiduciary standards. Failure to control for such standards could lead to biased test results, as highlighted by Frye (2001). Among these funds, funds investing only in the local stock market were selected. Other funds based on benchmarks other than the STI were excluded, as each of these benchmarks has a unique market cycle. These CPF-approved Singapore equity funds were classified according to the type of organization managing the fund: (1) insurance-linked investment products (ILPs) managed by insurance companies; (2) unit trusts managed by investment firms; or (3) bankmanaged unit trusts. Each type of organization differs in terms of operational structure, priorities and benefits for fund managers, which might influence the resulting portfolio returns (Bauman & Miller 1995). The following section describes how consideration for survivorship bias is incorporated for this study. 3.2Survivorship bias Survival bias affected many mutual fund studies that worked on performance data for funds that were still available. As funds that do not survive are usually the worst performing ones, when evidence from these nonsurvivors are not considered, the resulting average performance of each fund group would be overstated. To control for survival bias, this study collected data for surviving and non-surviving funds using quarterly reports for all CPF-approved unit trusts. However, for performing regression analysis, only funds with at least three quarters of data were included.

211

Appendix C
3.3Regression analysis For each fund, linear regression was performed using its quarterly risk premium (RETft RFRt) as the dependent variable and the STI quarterly risk premium (STIt RFRt) as the independent variable. As there are two risk-free rates: RFR1 = 0.625 percent per quarter and RFR2 = 1 percent per quarter for the Ordinary and Special accounts respectively; as well as two holding periods corresponding to data collected from Mercer (1999-2002) and S&P (2003-2004), four sets of linear regression are performed for each fund using a combination of risk-free rate and holding period. The intercept value and its significance for each regression equation are reported in the next section as the Jensens alpha for the fund. Table 1 Research sample: all CPF-approved domestic equity funds from 1999 to 2004 with at least three quarterly periods of data Symbol 1999- 20022002 2004 ! ! DBS Horizon Singapore Equity Fund DHSE ! ! DBS Shenton Thrift Fund DST ! ! OCBC Savers Singapore Trust Fund OSST ! OUB Union Singapore Equity Fund OUSE ! ! UOB Optimix Singapore Equity Fund UOSE ! UOB Unifund UU ! ! UOB United Growth Fund UUG ! ! Insurance AXA Life-Fortress Fund ALF company ! AXA Life-Fortress Fund A ALFA ! GE Greatlink Singapore Equities Fund GGSE ! Keppel Managed Fund KM ! NTUC Income Singapore Equity Fund NISE ! OUB Manulife Golden Singapore Growth Fund OMGSG ! UOB Life FOF-Unifund ULFU ! ! UOB Life FOF-United Growth Fund ULFUG ! UOB LifeLink Growth Fund ULG ! ! Investment Aberdeen Singapore Equity Fund ASE company ! CMG First State Singapore Growth Fund CFSSG ! ! Schroder Singapore Trust SST Note: funds with less than three quarterly periods of complete data are excluded, as they are insufficient for performing regression analysis. Source: funds identified from Mercer (1999-2002) and S&P (2003-2004). Figure 1 illustrates time-series regression with a single fund from the sample. Figure 1 Illustration of time-series regression for a single fund Organization type Bank Fund

Time-series regression of quarterly fund risk premium versus market risk premium for a fund from 1999:Q2 to 2002:Q1
80.00 60.00 40.00 20.00 0.00 -20.00 -40.00 -40.00 -20.00 0.00 STIt - RFRt R sq = 0.91 Investment firm: Aberdeen Singapore Equity Fund (ASE) Linear (Investment firm: Aberdeen Singapore Equity Fund (ASE))

RETft - RFRt

20.00 40.00 60.00

Source: developed for this research.

212

Conference paper 2
In this example, 12 quarterly risk premiums from a fund were regressed against the corresponding quarterly risk premiums of the STI index. As most of the data points cluster around the fitted regression line, the value of the coefficient of determination R2 is quite high. Figure 2 shows a normal probability plot for the same example. For residual analysis, the mostly linear trend on the normal probability plot suggests that the normality assumption for linear regression is satisfied (Mendenhall & Sincich 1996). After regression analysis was performed for each fund, the hypotheses developed in Section 3 were tested using sample data. 3.4 Hypothesis testing Starting with a set of returns in excess of the benchmark STI returns, the statistical significance of the excess returns over the market index was determined for the first hypothesis, with the usual statistical assumption that the excess returns were normally distributed with a mean and variance to be estimated. According to Goodwin (1998), the normal distribution of excess returns assumption is unnecessary, as long as the population mean and variance of excess returns exist and the excess returns are independent, which would lead to the average excess returns being normally distributed asymptotically, even though the excess returns may not be normally distributed. Figure 2 Illustration of normal probability plot for a single fund

Normal Probability Plot for ASE

RETft - RFR1 100 50 0 -50 0 20 40 60 80 100 120 Sample Percentile

Source: developed for this research. For this research, the excess returns were assumed to be independent, which is consistent with the concept of capital market efficiency (Fama 1970, 1991).To test the significance of the excess returns, the following tstatistic was computed: Equation 6 t-statistic for determining the statistical significance of excess returns

t = T (IR f

where IRf = information ratio for fund f; = number of time periods; and T = T 1. Df Source: Goodwin (1998). For the second hypothesis, to determine the relationship between fund and market returns, the coefficient of determination R2 for a fund with its market is computed during regression analysis and can serve as a measure of diversification within the marketthe closer the R2 value is to 1.00, the more diversified is the fund (Reilly & Brown 2003, p. 1121), implying a higher correlation with the market. To compare the performance of funds managed by banks and non-banks for the third hypothesis, the usual twotailed pooled-variance t-test for differences in two means was conducted for the information ratios of bank and non-bank funds during each of the time periods 1999-2002 and 2003-2004. The hypothesis that there is no difference between the average information ratios of bank funds (1) and non-bank funds (2) is tested by the t value in Equation 7 on the next page. The t-statistic was compared to the critical t value t( = .05, df = n1 + n2 2) to determine whether the difference between the means was statistically different from zero. A t value greater than t( = .05, df = n1 + n2 2) would lead to a rejection of the hypothesis. For the fourth hypothesis, the Sharpe and Treynor ratios for the average fund were examined. Negative average values would lead to rejection of the hypothesis. Even though the Sharpe ratio is the most commonly used measure for evaluating mutual funds, it is only appropriate when the evaluated fund is actually the entire portfolio of the investor (Corrado & Jordan 2005, p. 434), as the Sharpe ratio uses the standard deviation of return as a measure of total risk. Therefore, the Sharpe ratio can be used to evaluate funds for novice investors investing only in a single fund. For the average investor, each fund to be evaluated should be combined with other funds to compose a well-diversified portfolio. The Treynor ratio would then be more suitable as it uses

213

Appendix C
beta to measure systematic disk (Corrado & Jordan 2005, p. 434). The results of the hypothesis testing are interpreted in the next section. 4 Results and interpretation In this section, the four sub-sections report the findings for the four hypotheses developed in Section 3 in the same order as they were presented. 4.1 Performance of CPF-approved domestic equity funds relative to the stock index Appendix B shows the results of linear regression on each funds quarterly risk premium (RETft RFRt) and the STIs quarterly risk premium (STIt RFRt). The overall results are generally consistent with the findings of earlier studies. Equation 7 t-statistic for comparison of information ratios of bank and non-bank funds

t=

X 1 X 2 ( 1 2 ) 1 2 1 Sp + n n 2 1

where
2 SP =

(n1 1)S12 + (n2 1)S 22 (n1 1) + (n2 1)

and
2 Sp

= pooled variance; = mean information ratio of the sample taken from the population of bank funds; = variance of the information ratios for the sample taken from bank funds; = size of the sample taken from the population of bank funds; = mean information ratio of the sample taken from the population of non-bank funds; = variance of the information ratios for the sample taken from non-bank funds; and

X1 S12
n1

X2 S 22

n2 = size of the sample taken from the population of non-bank funds. Source: adapted from Levine et al.(2002, p. 375) From 1999 to 2002, the average quarterly return for the CPF-approved domestic equity funds was below the STI return (2.63 versus 2.94 percent respectively). However, from 2003 to 2004, the average funds return of 6.60 percent outperformed the 5.97 percent return from the STI index. Still, the significance of the positive Jensens alpha values downplayed any possibility of the average fund beating the market. According to the Jensens alpha criterion, even though the majority of funds registered abnormal returns above expectation for both holding periods, only one of them was statistically significant in each period. As the fund registering significant abnormal return during 1999-2002 became one of the worst performers during 2003-2004, consistency was clearly lacking, supporting the observation made by Carlson (1970). Appendix C reports information ratios for these funds. When evaluating fund performance, reasonable values for the information ratio should range from 0.5 to 1.0 for good to exceptionally good performance (Grinold & Kahn 1995). For a study of various groups of funds in the USA, Goodwin (1998) found that the information ratio of the median fund in each group was positive, but the ratio never exceeded 0.5. Thus, with a positive information ratio, the average fund in each group could add value to their investments, but the performance of these average funds did not qualify as good. Besides, no particular group of funds had more than three percent of its members delivering an excellent information ratio greater than 1.0. To interpret the findings for this study, the funds having positive Jensens alpha values also had positive information ratios. The mean information ratio for the sample was 0.262 during 1999-2002 and deteriorated to 0.141 during 2003-2004, well below 0.5, the standard for good performance according to Grinold and Kahn (1995). Thus, agreeing with Goodwins (1998) findings in the USA, the average fund in the sample could add value to its investments, but its performance did not qualify as good. In fact, none of the funds could deliver an excellent information ratio greater than 1.0, even though there were a few good performers for each holding period. Still, the good performers during the first period could not sustain their performance for the second period.

214

Conference paper 2
4.2 Relation between fund returns and the returns to the market index Returning to Appendix B, the average R2 for the sample was quite high at 0.88 during 1999-2002, with individual R2 values greater than 0.75 except for the bank fund OUSE with an R2 less than 0.20. This indicates that all the funds were well diversified within the local equity maket, except for OUSE. In fact, the majority of funds had R2 greater than 0.90, which agreed with Jensens (1968) finding that funds typically correlated with their markets at rates higher than 0.90. As the benchmark market reflects each funds asset allocation policy, this result also supported Brinson, Hood and Beebower (1986), Brinson, Singer and Beebower (1991), as well as Ibbotson and Kaplan (2000), who reported that about 90 percent of the variability in returns of a typical fund across time could be explained by its asset allocation policy. From 2003 to 2004, the average R2 dropped slightly to 0.86, with less funds registering R2 greater than 0.90. Still, the findings support existing literature in reporting high correlation between fund and market returns. 4.3 Relative performance of funds managed by banks and non-banks Referring to Appendix B, during 1999-2002, the average bank fund seemed to under-perform the market at 1.37 percent, while the average non-bank fund managed to outperform the market at 3.60 percent. From 2003-2004, the reverse seemed to be true as bank funds outperformed the market at 7.68 percent while non-bank funds under-performed the market at 5.84 percent. However, performing a two-sample t-test assuming unequal variances for information ratios of bank and nonbank funds in Appendix D showed there was no significant difference between the performance of bank and non-bank equity funds during each holding period, supporting Fryes (2001) findings for bond funds. 4.4 Excess returns from fund investment over guaranteed interest rates Referring to Appendix E, the average fund had positive Sharpe and Treynor ratios for both risk-free rates during the two holding periods, implying returns exceeding the guaranteed interest rates. However, the average bank fund actually registered negative Sharpe and Treynor ratios during 1999-2002, implying that earning the guaranteed interest rates in the Ordinary and Special accounts were better than investing in bank funds during that holding period. Overall, returns from the CPF-approved equity funds were higher than the guaranteed interest rates of Ordinary and Special accounts for both holding periods, refuting the earlier finding from Koh (1999). Finally, comparing the results obtained from all the performance measures showed that even though the rankings were different, the identification of the best performing funds were similar for the first holding period. For the second holding period, there seemed to be more disagreement among the performance measures. Still, examining the top performers for each measure reveals that fund size may not be a good predictor of performance, while top performing funds are usually associated with below-average expense ratios. Examining the fund size and expense ratios in Appendix A, the average bank equity fund is smaller and has a higher expense ratio than its non-bank counterpart. The smaller bank funds are probably less popular than their nonbank counterparts with CPF investors. Bank funds smaller fund size and higher expense ratio could be indicative of economies of scale in the fund management industry. 5 Conclusion and recommendation As CPF-approved domestic equity funds could not outperform their market index, and the market determines more than 90 percent of the variability in returns of a typical fund across time, the majority of these funds should not be actively managed, which incurred substantial research and trading costs that would not translate to additional returns. Instead, the funds should be passively managed as index funds. Index funds should be promoted and approved for the CPF Investment Scheme as they are less costly than actively managed funds in terms of research and trading costs, and have attained comparable or even better performance than actively managed funds most of the time. As index funds are not popular in Singapore, the CPF Board could promote the advantages of investing in index funds and encourage the introduction of index funds by fund management companies to be approved for the CPF Investment Scheme. Evaluating the performance of unit trusts managed by local banks is an appropriate research area, as it has important implications for the management of local banks. For this study, all CPF-approved bank equity funds were managed by local banks, which face increasing competition from foreign financial institutions, as Singapores financial sector is gradually deregulated. As the distinction between banks and non-banks disappears, the mutual fund market will be strategically important for financial institutions. In fact, Gallo, Apilado & Kolari (1996) had shown that offering mutual fund products can improve the profitability of banks. The profitability of banks will be significantly improved if banks can shed the image of under-performers in the fund management industry.

215

Appendix C
References Ariff, M 1986, 'Announcement effects in a thin market--a study of the Singapore equity market', paper presented to Proceedings of the Academy of International Business, Southeast Asia Regional Conference, Taipei, 26-28 June. Bauman, W & Miller, R 1995, 'Portfolio performance rankings in stock market cycles', Financial Analysts Journal, vol. 51, no. 2, pp. 79-87. Berkowitz, M & Qiu, J 2003, 'Ownership, risk and performance of mutual fund management companies', Journal of Economics and Business, vol. 55, pp. 109-34. Blake, C, Elton, E & Gruber, M 1993, 'The performance of bond mutual funds', Journal of Business, vol. 66, no. 3, pp. 371-403. Bogle, J & Twardowski, J 1980, 'Institutional investment performance: banks, investment counselors, insurance companies and mutual funds', Financial Analysts Journal, vol. 36, no. 1, pp. 33-41. Brinson, G, Hood, L & Beebower, G 1986, 'Determinants of portfolio performance', Financial Analysts Journal, vol. 42, no. 4, pp. 39-48. Brinson, G, Singer, B & Beebower, G 1991, 'Determinants of portfolio performance II: an update', Financial Analysts Journal, vol. 47, no. 3, pp. 40-8. Cai, J, Chan, K & Yamada, T 1997, 'The performance of Japanese mutual funds', Review of Financial Studies, vol. 10, no. 2, pp. 237-74. Carlson, R 1970, 'Aggregate performance of mutual funds, 1948-1967', Journal of Financial and Quantitative Analysis, vol. 5, no. 1, pp. 1-32. Corrado, C & Jordan, B 2005, Fundamentals of investments Valuation and Management, 3 edn, McGraw-Hill Irwin, New York. Cumby, R & Glen, J 1990, 'Evaluating the performance of international mutual funds', Journal of Finance, vol. 45, no. 2, pp. 497-522. Fama, E 1970, 'Efficient capital markets: a review of theory and empirical work', Journal of Finance, vol. 25, no. 2, pp. 383-417. ---- 1991, 'Efficient capital markets: II', Journal of Finance, vol. 46, no. 5, pp. 1575-617. Friend, I & Blume, M 1970, 'Measurement of portfolio performance under uncertainty', American Economic Review, vol. 60, no. 3, pp. 561-75. Frye, M 2001, 'The performance of bank-managed mutual funds', Journal of Financial Research, vol. 24, no. 3, pp. 419-42. Gallo, J, Apilado, V & Kolari, J 1996, 'Commercial bank mutual fund activities: implications for bank risk and profitability', Journal of Banking and Finance, vol. 20, pp. 1775-91. Goodwin, T 1998, 'The information ratio', Financial Analysts Journal, vol. 54, no. 4, pp. 34-43. Grinold, R & Kahn, R 1995, Active Portfolio Management, Probus Publishing, Chicago. Gruber, M 1996, 'Another puzzle: the growth in actively managed mutual funds', Journal of Finance, vol. 51, no. 3, pp. 783-810. Holiday, K 1994, 'Mutual funds', Bank Marketing, vol. 26, pp. 23-31. Ibbotson, R & Brinson, G 1993, Global Investing, McGraw-Hill, New York. Ibbotson, R & Kaplan, P 2000, 'Does asset allocation policy explain 40, 90 or 100 percent of performance?' Financial Analysts Journal, vol. 56, no. 1, pp. 26-33. Jensen, M 1968, 'The performance of mutual funds in the period 1945-1964', Journal of Finance, vol. 23, no. 2, pp. 389-416. Klemkosky, R 1973, 'The bias in composite performance measures', Journal of Financial and Quantitative Analysis, vol. 8, no. 3, pp. 505-14. Koh, S 1999, 'A survey of unit trusts in Singapore', Singapore Management Review, vol. 21, no. 1, pp. 49-78. Kritzman, M 1983, 'Can bond managers perform consistently?' Journal of Portfolio Management, vol. 9, no. 4, pp. 54-6. Latane, H & Tuttle, D 1967, 'Criteria for portfolio building', Journal of Finance, vol. 22, no. 3, pp. 359-73. Leland, H 1999, 'Beyond mean-variance: performance measurement in a nonsymmetrical world', Financial Analysts Journal, vol. 55, no. 1, pp. 27-36. Levine, D, Stephan, D, Krehbiel, T & Berenson, M 2002, Statistics for Managers Using Microsoft Excel, 3 edn, Prentice Hall, New Jersey. Lim, G 1985, 'A Box-Jenkins Approach to Stock Market Analysis on the SES', National University of Singapore. Markowitz, H 1976, 'Investment for the long run: new evidence for an old rule', Journal of Finance, vol. 31, no. 5, pp. 1273-86. Mendenhall, W & Sincich, T 1996, A Second Course in Statistics: Regression Analysis, 5 edn, Prentice Hall, New Jersey.

216

Conference paper 2
Mercer 1999-2002, Unit Trusts and Investment-linked Insurance Products Included Under CPFIS: Performance and Risk Monitoring Reports, Mercer Investment Consulting, viewed 19 Feb 2005, <http://www.mercerfundwatch.com/>. Reilly, F & Brown, K 2003, Investment Analysis and Portfolio Management, 7th edn, South-Western, Ohio. Rubinstein, M 1976, 'The strong case for the generalized logarithmic utility function as the premier model of financial markets', Journal of Finance, vol. 31, no. 2, pp. 551-71. S&P 2003-2004, Performance and Risk Monitoring Report for CPFIS-included Unit Trusts & Investment-linked Insurance Products, S&P Fund Services, viewed 19 Feb 2005, <http://www.imas.org.sg>. Saw, S & Tan, K 1986, 'The Stock Exchange of Singapore all-share price indices: testing of independence', Securities Industry Review, vol. 12, no. 1. Seitz, N & Ellison, M 1999, Capital Budgeting and Long-term Financing Decisions, 3 edn, Dryden, Orlando. Sharpe, W 1964, 'Capital asset prices: a theory of market equilibrium under condtions of risk', Journal of Finance, vol. 19, no. 3, pp. 424-42. ---- 1966, 'Mutual fund performance', Journal of Business, vol. 39, no. 1, pp. 119-38. Sirri, E & Tufano, P 1998, 'Costly search and mutual fund flows', Journal of Finance, vol. 53, no. 5, pp. 1589622. Treynor, J 1965, 'How to rate management of investment funds', Harvard Business Review, vol. 43, no. 1, pp. 63-75. Wong, K 1988, 'Role of Singapore and Malaysian stock exchanges', in C Tan & K KC (eds), Handbook of Singapore-Malaysian Corporate Finance, Butterworths, Singapore.

217

Appendix C

Appendix A Characteristics of domestic equity funds from banks and non-banks

Fund 1999:Q2--2002:Q1 2003:Q1--2004:Q3 Avg qtr ret Size (S$ Expense Beta Avg qtr ret Size (S$ Expense Beta million) ratio (%) million) ratio (%) (%) (%) Bank -0.46 75.31 1.50 0.95 7.28 109.26 1.49 1.19 Bank 4.33 43.96 0.96 1.11 12.30 59.50 1.16 1.51 Bank 4.53 17.03 1.49 1.07 7.49 19.06 1.52 1.19 Bank -6.13 3.00 3.32 0.32 Bank 2.57 2.57 3.83 0.78 4.93 0.85 5.02 1.00 Bank 0.23 35.96 1.57 0.85 Bank 4.51 103.33 1.36 1.01 6.39 125.20 1.18 0.90 Insurance 2.05 17.72 1.45 1.01 6.11 62.20 1.24 0.40 Insurance 4.71 45.19 1.67 0.61 Insurance 4.88 30.31 1.20 0.82 Insurance 0.12 10.58 1.16 1.03 Insurance 7.45 1.69 0.43 1.07 Insurance -0.66 19.56 2.04 1.14 Insurance 7.00 35.05 1.55 0.71 Insurance 6.03 91.65 1.35 0.86 6.11 84.23 1.18 0.90 Insurance -0.64 0.41 1.51 0.87 Investment 5.84 6.97 2.58 1.00 5.59 21.11 2.12 0.71 Investment 5.36 86.33 2.35 1.18 Investment 7.32 113.42 1.71 1.14 6.01 238.16 1.52 0.95 Type 2.63 1.37 3.60 2.94 0.63 1.00 41.43 40.17 42.41 1.86 2.00 1.74 0.94 0.87 0.99 1.00 6.60 7.68 5.84 5.97 0.63 1.00 66.40 62.77 68.98 1.64 2.07 1.34 0.94 1.16 0.78 1.00

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST

Average fund Average bank fund Average non-bank fund Straits Times Index Ord a/c interest rate Sp a/c interest rate

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data.

218

Conference paper 2

Appendix B Regression of equity fund and market index risk premiums

Fund 1999:Q2--2002:Q1 Avg Beta R sq Jensen alpha 0.625% 1% RFR qtr ret RFR (%) -0.46 0.95 0.98 0.19 0.17 4.33 1.11 0.94 1.13 1.17 4.53 1.07 0.97 1.42 1.45 -6.13 0.32 0.19 -4.73 -4.99 2.57 0.78 0.99 -0.32 -0.40 0.23 0.85 0.83 0.65 0.57 4.51 1.01 0.97 1.55 1.55 2.05 1.01 0.79 2.14 2.14 Avg qtr ret (%) 7.28 12.30 7.49 2003:Q1--2004:Q3 Jensen alpha Beta R sq 0.625% 1% RFR RFR 1.19 0.90 0.32 0.39 1.51 0.67 3.60 3.79 1.19 0.77 0.48 0.55 0.90 0.93 0.77 0.79 0.95 0.96 -1.04 1.04 3.35 ** 1.97 -0.13 0.40 -1.04 0.93 3.13 * 1.82 -0.20 0.43

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST Avg fund STI

4.93 1.00 6.39 6.11 4.71 4.88 0.90 0.40 0.61 0.82

0.12 -0.66 7.00 6.03 -0.64 5.84 5.36 7.32 2.63 2.94

1.03 1.14 0.71 0.86 0.87 1.00 1.18 1.14 0.94 1.00

0.90 0.94 0.98 0.98 0.95 0.91 0.84 0.97 0.88 1.00

0.22 0.25 4.41 3.00 -0.66 2.90 2.01 4.05 ** 1.14 0.00

0.23 7.45 1.07 0.30 4.30 2.95 -0.71 2.90 2.08 4.10 ** 1.11 0.00

6.11 0.90 5.59 0.71 6.01 0.95 6.60 0.94 5.97 1.00

0.88 0.87 0.91 0.86 1.00

0.71 1.19 0.29 1.01 0.00

0.67 1.08 0.27 0.99 0.00

*Significant excess return at the 5% level. **Significant excess return at the 1% level.

Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data.

219

Appendix C

Appendix C Information ratios for domestic equity funds Fund 1999:Q2--2002:Q1 Info No. of qtr t-statistic ratio periods 0.12 7 0.31 0.25 12 0.85 0.41 12 1.42 -0.08 6 -0.18 -0.16 3 -0.28 0.15 7 0.39 0.45 12 1.57 0.26 10 0.83 2003:Q1--2004:Q3 Info No. of qtr t-statistic ratio periods 0.42 7 1.11 0.76 7 2.00 0.32 7 0.84 -0.43 0.22 0.03 0.27 -0.59 7 7 7 3 7 5 -1.14 0.58 0.08 0.47 -1.56 1.67

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST Avg fund STI

0.04 0.01 0.57 0.68 -0.15 0.49 0.24 0.92 0.26 0.00

10 7 4 4 10 12 12 12

0.12 0.75 0.03 1.15 1.35 0.06 -0.48 1.71 -0.13 0.82 3.17 * 0.02 0.14 0.00

7 7 7

0.15 -0.35 0.06

*Significant excess return at the 5% level. **Significant excess return at the 1% level. Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. Appendix D Two-sample t-test for information ratios of bank and non-bank funds 1999:Q2--2002:Q1 2003:Q1--2004:Q3 Bank Non-bank Bank Non-bank Mean information ratio 0.162 0.339 0.256 0.058 Variance 0.053 0.123 0.188 0.164 Observations 7 9 5 7 Hypothesized mean difference 0 0 Df 14 8 t Stat -1.215 0.801 P(Tt) one-tail 0.122 0.223 t Critical one-tail 1.761 1.860 P(Tt) two-tail 0.245 0.446 t Critical two-tail 2.145 2.306 Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data. =0.05

220

Conference paper 2

Appendix E Sharpe and Treynor ratios for domestic equity funds Fund 1999:Q2--2002:Q1 Sharpe ratio Treynor ratio 0.625% 1% RFR 0.625% 1% RFR RFR RFR -0.07 -0.10 -1.14 -1.53 0.17 0.15 3.33 2.99 0.19 0.17 3.64 3.29 -1.37 -1.45 -20.95 -22.12 0.10 0.08 2.48 2.00 -0.03 -0.06 -0.47 -0.91 0.20 0.18 3.85 3.48 0.08 0.06 1.41 1.04 2003:Q1 to 2004:Q3 Sharpe ratio Treynor ratio 0.625% 1% RFR 0.625% 1% RFR RFR RFR 0.75 0.71 5.61 5.30 0.89 0.86 7.72 7.48 0.71 0.67 5.75 5.43 0.57 0.87 1.70 1.26 0.71 1.24 -0.07 0.45 0.31 -0.09 0.26 0.20 0.31 -0.09 0.43 0.29 -0.12 0.25 0.18 0.29 -1.12 9.04 6.29 -1.46 5.20 4.02 5.87 -1.45 8.51 5.85 -1.89 4.83 3.71 5.54 0.52 0.82 1.58 1.15 0.65 1.17 4.31 6.43 13.77 6.65 5.18 6.36 3.93 6.01 12.83 6.04 4.73 6.01

DHSE DST OSST OUSE UOSE UU UUG ALF ALFA GGSE KM NISE OMGSG ULFU ULFUG ULG ASE CFSSG SST

-0.03

-0.05

-0.49

-0.85

0.81 0.92 0.76

0.75 0.85 0.71

6.13 7.03 5.65

5.72 6.50 5.25 6.27 4.97

Avg fund 0.04 0.01 1.22 0.78 0.93 0.87 6.71 STI 0.12 0.10 2.31 1.94 0.75 0.70 5.34 Source: developed for this research from Mercer (1999-2002) and S&P (2003-2004) data.

221

Appendix D Journal article 1

Mutual Fund Performance Determinants: Size and Expenditures Tng Cheong Sing Southern Cross University, Australia Postal address: 171 Lor 1 Toa Payoh #13-1132 S(310171) Singapore Telephone: +65 9871 7218 Facsimile: +65 6259 3558 Email: tcs@pacific.net.sg This article is an original work from its author, accepted for publication with no amendment in the first 2007 issue (vol. 29, no. 1) of Singapore Management Review/Asia-Pacific Journal of Management Theory and Practice. An initial draft of this paper was presented in December 2005 at the Second International Business Research Conference, University of Technology Sydney. Tng Cheong Sing is a Doctor of Business Administration (DBA) candidate at Southern Cross University Graduate College of Management. He has submitted his Doctorate thesis Factors Influencing Unit Trust Performance for examination and is completing a thesis defence. Previously, he was a faculty staff at University of Tasmania School of Accounting and Finance as well as Singapore Polytechnic School of Info-communication Technology. He would like to acknowledge the guidance of his DBA supervisor Emeritus Professor Geoffrey Meredith AM from Southern Cross University. Abstract Mutual fund performance determinants could affect expected returns or transaction costs. Factors affecting expected returns include asset allocation benchmark return and systematic risk, while transaction costs include explicit and implicit costs, which could be measured by expense ratio and fund size respectively. Insignificance of transaction cost determinants could be attributable to dominance of asset allocation in determining returns. Examining expense ratio, performance and size characteristics of domestic equity funds approved for Singapores Central Provident Fund Investment Scheme controls for differing fiduciary standards, as these funds face the same standard of managing social security savings. Keywords ASSETS UNDER MANAGEMENT; EXPENSE RATIO; FINANCIAL INSTITUTION.

1 Introduction Determinants of mutual fund performance could be classified into factors affecting pre-tax or post-tax returns. Factors affecting pre-tax fund performance include (1) expenses, (2) investment style, (3) past pre-tax performance, (4) risk and (5) turnover (Peterson et al. 2001). Post-tax factors are also important as after-tax returns could be much less than before-tax returns for investors with high tax brackets. These post-tax determinants include (1) past pre-tax performance, (2) expenses, (3) risk, (4) investment style, (4) past tax efficiency and (5) recent occurrence of large net redemption (Peterson et al. 2002). Comparison of these two groups of factors shows risk, style, past pre-tax performance and expenses affect performance before and after consideration of taxes. Still, studies on fund performance focused mainly on pre-tax factors. While many funds are taxable on their profits and interests earned, studies on pre-tax factors are appropriate for funds whose profits and interests earned are non-taxable. For example, unit trusts approved for Singapores Central Provident Fund (CPF) Investment Scheme are funds whose profits and interests earned are generally not taxable (CPF Investment Scheme 2005). By studying these funds, this research focused on pre-tax determinants and eliminated complication from differing tax rates. Besides, taxation laws pertaining mutual funds differ from country to country. For example, in the United States of America (USA), while dividends and interests paid by a mutual fund were taxed as ordinary income for the shareholder, short and long-term capital gains were taxed at investors marginal tax rate and capital gains rate of 20 percent respectively (Jones 2003, pp. 136-7). In Australia, income and capital gains derived by several types of superannuation funds were taxable at 15 percent (Veltman 2000, p. 220). As for Singapores CPF-approved unit trusts, even though investment profits were not taxable, dividends received were taxable at investors marginal tax rate (CPF Investment Scheme 2005). Therefore, the following discussion focuses on pre-tax factors. Research on mutual fund performance did not agree on factors affecting fund returns. For example, Peterson et al. (2002) did not consider assets under management an important factor, while Indro et al. (1999) reported fund size affecting performance. While Sharpe (1966) observed funds with higher reward-to-variability ratios incurred lower expenses, Ippolito (1989) found risk-adjusted returns not related with expenses. This research focused on plausible effects of size and expense ratio on pre-tax fund performance. The outline of this paper is as follows. The following section reviews conflicting literature on fund size and expense ratio characteristics to formulate hypotheses for their effect on fund performance. Section three

222

Journal article 1
describes data and methodology used. Results are interpreted in section four before section five concludes with recommendation. 2 Hypotheses and Models This research builds on quantitative characteristics tested by Peterson et al. (2001) as performance determinants for domestic equity funds in the USA. In particular, fund size and expense ratio characteristics are reviewed in this order in the following sub-sections. This is followed by explanation of a plausible relationship between fund size and expense ratio. 2.1 Fund Size as a Measure of Implicit Transaction Costs According to Indro et al. (1999), fund size reflects the following implicit transaction costs of active investment strategieseffect of huge transactions on market prices, opportunity cost of not trading, costs of being under scrutiny by market participants, and administrative stress arising from investment style deviation for large funds. Fund size results from flow of monies into and out of the fund. Studies examining relationship between past performance and flow of funds revealed investors searching for quality funds based on performance record placed their monies in funds with superior recent performance (Guercio & Tkac 2002; Ippolito 1992; Sawicki 2000; Sirri & Tufano 1998). While Ippolito (1992) as well as Sirri and Tufano (1998) found recent poor performance not leading to outflows from retail mutual funds in the USA, Sawicki (2000) reported Australian investors moving monies out of poorly performing wholesale funds. Agreeing with Sawicki (2000), Guercio and Tkac (2002) found change in fund ratings reflecting past performance influencing flow of money to and from the funds. Sawicki (2001) suggested small funds that were relatively young being more likely to abandon unsuccessful strategies for more successful ones, so that investors would not perceive the need to withdraw from such funds. In a later study, Sawicki and Finn (2002) reported small funds being represented disproportionately among the top, but underrepresented among worst performers, suggesting fund size to be a performance determinant. Indeed, net assets under management could affect performance, as funds need to attain a minimum size in order to achieve returns net of research expenses and other costs. However, a fund that is too big could incur excessive costs, resulting in diminishing or even negative marginal returns. Initially, growth in fund size could provide cost advantages, as brokerage costs for larger transactions are lower while research expenses would rise less than proportionately with fund size. After the initial growth, a fund that has grown too large might cause its managers to deviate from its original objectives by investing in lower quality assets (which otherwise would not be considered when the fund is smaller) and increase administrative costs by additional coordination and hiring of staff to manage sub-funds (Indro et al. 1999). Clearly, flow of monies into funds with recent good performance is an attempt by investors to seek maximum risk-adjusted returns in an asymmetric information environment. However, after being larger with injection of additional monies, funds with good recent returns could not sustain their performance, as demonstrated by Carlson (1970), Dunn and Theinsen (1983) as well as Jensen (1969) in the USA and Tng (2005) in Singapore. But as fund managers are compensated proportional to the amount of assets under management, they are rewarded or penalized by clients based on recent performance. Even though past performance and flow of funds based on past performance might not be useful determinants of future performance, amount of assets under management could affect fund performance. Reviewing literature on relation between size and performance of funds produced mixed findings. Cicotello and Grant (1996), Droms and Walker (1994) as well as Grinblatt and Titman (1994) reported absence of such relation for funds in the USA. The relation was also absent in Australia (Bird, Chin & McCrae 1983; Gallagher 2003; Gallagher & Martin 2005) and Sweden (Dahlquist, Engstrom & Soderlind 2000). However, fund size was a performance determinant in the USA (Indro et al. 1999). In Singapore, fund size may be a performance determinant as larger funds could achieve economies of scale compared to smaller funds. As Singapore is relatively small among developed equity markets, domestic equity funds may not suffer from diminishing marginal returns due to excessive fund size. This is tested in the first hypothesis: H1: There is no significant performance difference between small and large funds. As funds need to attain minimum size to achieve returns net of research and other costs, the following section examines such costs as plausible performance determinant. 2.2 Expense Ratio as a Measure of Explicit Transaction Costs Passively managed index funds have lower costs and generally outperform actively managed funds (Bogle 1998). Actively managed funds incur various costs. Examples of such costs are operating and research expenses, which are represented by the funds expense ratio. There are various definitions of expense ratio. According to the Investment Company Institute (ICI), the national association of investment companies in the USA, a funds

223

Appendix D
expense ratio is its cost of doing business, as disclosed in its prospectus and expressed as a percentage of its assets (ICI 2004). However, to view expense ratio as cost of conducting business for a fund is too general. Indro et al. (1999) identified the ratio as proportion of fund assets paid for operating expenses and management fees, including 12b-1 fees (the USA Securities and Exchange Commission Rule 12b-1 permitting funds to use as much as one percent of fund assets per year to pay for distribution and marketing costs), administration fees and other costs incurred by the fund, but excluding brokerage costs. Even though various costs are included in the expense ratio, most of the expenses could be associated with cost of financial market research, as Indro et al. (1999) considered expense ratio to reflect explicit cost of research incurred by the fund manager. As Indro et al. (1999) characterized most retail fund investors as passive and not informed, expense ratio was considered the price paid by investors of a fund to its manager to inform them about the financial market. In order for active management incurring research expenses to be worthwhile, incentives in the form of economic gains from trading based on information from useful research would compensate fund managers for incurring such costs (Grossman & Stiglitz 1980). Therefore, fund managers efficiently incurring research expenses can earn positive risk-adjusted returns net of expenses. Otherwise, inefficient expenses may lead to their income (proportionate to amount of assets under management) being penalized as investors withdraw monies from under-performing funds with excessive expenses. Research on relationship between risk-adjusted fund returns and expenses reported conflicting results in the USA. While Sharpe (1966) observed funds with higher reward-to-variability ratios incurring lower expenses, Friend et al. (1970) reported insignificant negative correlation between risk-adjusted fund returns and expenses. Furthermore, Ippolito (1989) found risk-adjusted returns not related with expenses, while Berkowitz and Qiu (2003) confirmed importance of expenses as determinant of fund performance. For large equity markets, high research expenses could be justified for better performance with more useful information on many investment choices available. For small markets, high research expenses might be wasteful with limited investment choices. As Singapore is relatively small among developed equity markets, a passive investment strategy may be justified when funds with higher research expenses cannot outperform their counterparts with lower expenses. This is tested in the second hypothesis: H2: There is no significant performance difference between funds with high expense ratios and those with low expense ratios. The following sub-section explores plausible relationship between expense ratio and fund size. 2.3 Relation Between Expense Ratio and Fund Size The previous two sub-sections demonstrated performance of mutual funds could be related to implicit and explicit costs, which are measured by expense ratio and fund size respectively. The third and final hypothesis tests relation between fund size and expense ratio. If initial growth in size could provide cost advantages in terms of brokerage costs and research expenses, as explained in section 2.1, large funds could have lower expense ratios than small funds: H3: There is no significant difference in expense ratios of big and small funds. To summarize transaction cost determinants of mutual fund performance explained in this section, Figure 1 provides a conceptual model. [Insert Figure 1: Conceptual Model for Determinants of Mutual Fund Performance near here] As shown in the figure, fund performance determinants can be classified into two categories: expected return and transaction cost determinants. Major determinants of expected return include benchmark index return and systematic risk while transaction cost determinants include expense ratio and fund size measures of explicit and implicit costs respectively. A plausible determinant of explicit costs that is missing from this model is the funds turnover ratio, measuring amount of trading in fund assets and computed by dividing amount of purchases or sales by average assets (Indro et al. 1999). Passive buy-and-hold or active stock selection with market timing can be reflected by low or high turnover ratio respectively. With most variability in fund returns explained by asset allocation benchmark return (Brinson, Hood & Beebower 1986; Brinson, Singer & Beebower 1991; Ibbotson & Kaplan 2000), a remaining significant determinant could be transaction costs, which is considered one of the most important criteria for fund evaluation (Bodie, Kane & Marcus 2005, p. 116; Corrado & Jordan 2005, p. 112). The following section describes how hypotheses were tested using secondary data. 3 Data and Methodology To carry out this research, five years of quarterly returns, net assets under management and expense ratios from 1999 to 2004 for 19 retail funds approved for Singapores CPF Investment Scheme were examined. These funds were invested in shares from the Singapore Stock Exchange. Table 1 identifies funds used for this research. [Insert Table 1: Research Sample near here] For this research, only CPF-approved funds were considered as they followed the same fiduciary standard for managing social security savings, so as to control for differing fiduciary standards. Failure to control for such

224

Journal article 1
standards may lead to biased test results, as highlighted by Frye (2001). Examining CPF-approved equity funds therefore facilitates direct comparison of similar funds. Among these funds, those investing only in the domestic stock market were selected. Excluded from the research sample were funds based on benchmarks other than the local Straits Times Index, as each benchmark has a unique market cycle. As funds investing only in the local stock market have the same risk classification, this approach controls for differing systematic risk. The following sub-sections discuss expense ratio and fund size measurements, consideration of survivorship bias and hypothesis testing. 3.1 Measurement of Expense Ratio and Fund Size For this research, quarterly expense ratio secondary data were calculated according to guidelines developed by the Investment Management Association of Singapore (IMAS). According to IMAS guidelines, the ratio is considered as operating costs (including but not limited to administration fee, amortized expenses, audit fees, custodian and depository fees, goods and services tax on expenses, management fee, printing and distribution costs, registrar fees and trustee fee) expressed as a percentage of funds average net assets for given time period. IMAS guidelines require the ratio to be calculated by taking average of annualised expense ratios for two previous six-month periods. For feeder funds, the ratio is to be calculated as sum of annualised expense ratios of Singapore feeder fund and parent fund to facilitate comparability with direct investment funds (S&P 20032004). For fund size, Indro et al. (1999) measured it using monthly net assets under management. As quarterly data were used for this research, fund size was measured using net assets under management at the end of each quarter. 3.2 Survivorship Bias When collecting expense ratio and fund size data, consideration for survivorship bias was incorporated. As nonsurviving funds are usually the worst performing, when data for non-survivors are not considered, resulting average performance of each fund group can be overstated. To control for survival bias, this study collected data for surviving and non-surviving funds using quarterly reports for all CPF-approved unit trusts. 3.3 Hypothesis Testing To test hypothesis for no significant performance difference between two groups of funds, the two-tailed pooled-variance t-test for difference in two means was conducted for returns of the fund groups during the two holding periods 1999-2002 and 2003-2004 at an alpha level of 0.05. To categorize each fund as big or small for the first hypothesis, the funds were ranked by their average assets under management for each time period in descending order, with top and bottom half being big and small funds respectively. Similarly, for the second hypothesis, funds were categorized into high or low expense ratio groups by ranking their expense ratios for each holding period in descending order, with top and bottom half being high and low expense ratio funds respectively. Results of hypothesis testing are discussed in the next section. 4 Results and Interpretation Summary characteristics of Singapores CPF-approved domestic equity funds were tabulated in Table 2. [Insert Table 2: Summary Characteristics of Singapore's CPF-approved Domestic Equity Funds near here] In this section, the three sub-sections report findings for hypotheses in the same order as they were presented in section two. 4.1 Performance of Large and Small Funds Referring to Table 3, during 1999-2002, average return of big funds was 4.29 percent, versus 0.96 percent for small funds. For 2003-2004, big funds return of 7.37 percent continued to outperform small funds return at 5.84 percent. [Insert Table 3: Two-sample t-test Assuming Unequal Variances for Returns of Big and Small Funds near here] Performing a two-sample t-test assuming unequal variances for returns of big and small funds showed significant out-performance of small funds by large funds during 1999-2002, but no significant performance difference during 2003-2004.

225

Appendix D
4.2 Performance of High and Low Expense Ratio Funds As presented in Table 4, during 1999-2002, average return of funds with low expense ratios was 2.56 percent, which was lower than 2.69 percent average return of high expense ratio counterparts. For 2003-2004, the reverse happened as average returns of low and high expense ratio funds were 7.21 and 6.00 percent respectively. [Insert Table 4: Two-sample t-test Assuming Unequal Variances for Returns of High and Low Expense Ratio Funds near here] However, performing a two-sample t-test assuming unequal variances for returns of high and low expense ratio funds showed no significant performance difference during the two periods. 4.3 Expense Ratio of Large and Small Funds From Table 5, during 1999-2002, average expense ratio of big funds at 1.54 was lower than small funds at 2.17. For 2003-2004, average expense ratio of big funds at 1.30 continued to be lower than small funds at 1.99. [Insert Table 5: Two-sample t-test Assuming Unequal Variances for Expense Ratios of Big and Small Funds near here] Performing a two-sample t-test assuming unequal variances for expense ratios of big and small funds showed no significant expense ratio differences for big and small funds during the two periods. 5 Conclusion and Recommendation Even though there is evidence large funds outperformed small funds, better performance of large funds was not significant for all holding periods tested, which is agreeable with findings from the USA (Cicotello & Grant 1996; Droms & Walker 1994; Grinblatt & Titman 1994), Australia (Bird, Chin & McCrae 1983; Gallagher 2003; Gallagher & Martin 2005) and Sweden (Dahlquist, Engstrom & Soderlind 2000). As for expense ratio, results seemed contradictory for different holding periods, and relation between expense ratio and returns was insignificant, supporting findings from Ippolito (1989). Even though big funds seemed to have lower expense ratios than small funds, suggesting economies of scale, difference in expense ratio was insignificant. To explain these results, insignificance of fund size and expense ratios in affecting returns could be attributable to dominance of asset allocation determinant, which was demonstrated by Brinson, Hood and Beebower (1986), Brinson, Singer and Beebower (1991), as well as Ibbotson and Kaplan (2000) in the USA, and in Singapore by Tng (2005). According to Tng (2005), close to 90 percent of variability in returns of CPF-approved domestic equity funds can be explained by asset allocation policy. This research examined effects of expense ratio and size on fund performance in a small and developed equity market. In such a market, large funds cannot provide significant economies of scale to lower expenses or provide better returns than small funds. Besides, higher expenses cannot produce better returns. For further research, turnover ratio may be incorporated as a measure of explicit transaction cost to examine significance of this factor as a fund performance determinant in various equity markets. References Berkowitz, M & Qiu, J 2003, 'Ownership, risk and performance of mutual fund management companies', Journal of Economics and Business, vol. 55, pp. 109-34. Bird, R, Chin, H & McCrae, M 1983, 'The performance of Australian superannuation funds', Australian Journal of Management, vol. 8, pp. 49-69. Bodie, Z, Kane, A & Marcus, A 2005, Investments, 6 edn, McGraw-Hill, New York. Bogle, J 1998, 'The implications of style analysis for mutual fund performance', Journal of Portfolio Management, vol. 24, no. 4, pp. 34-42. Brinson, G, Hood, L & Beebower, G 1986, 'Determinants of portfolio performance', Financial Analysts Journal, vol. 42, no. 4, pp. 39-48. Brinson, G, Singer, B & Beebower, G 1991, 'Determinants of portfolio performance II: an update', Financial Analysts Journal, vol. 47, no. 3, pp. 40-8. Carlson, R 1970, 'Aggregate performance of mutual funds, 1948-1967', Journal of Financial and Quantitative Analysis, vol. 5, no. 1, pp. 1-32. Cicotello, C & Grant, C 1996, 'Equity fund size and growth: implications for performance and selection', Financial Services Review, vol. 5, pp. 1-12. Corrado, C & Jordan, B 2005, Fundamentals of investments Valuation and Management, 3 edn, McGraw-Hill Irwin, New York. CPF Investment Scheme, 2005, CPF Board, viewed 2 May 2005, <http://www.cpf.gov.sg>. Dahlquist, M, Engstrom, S & Soderlind, P 2000, 'Performance and characteristics of Swedish mutual funds', Journal of Financial and Quantitative Analysis, vol. 35, no. 3, pp. 409-23. Droms, W & Walker, D 1994, 'Investment performance of international mutual funds', Journal of Financial Research, vol. 17, no. 1, pp. 1-14.

226

Journal article 1
Dunn, P & Theinsen, R 1983, 'How consistently do active managers win?' Journal of Portfolio Management, vol. 9, pp. 47-50. Friend, I, Blume, M & Crockett, J 1970, Mutual Funds and Other Institutional Investors, McGraw-Hill, New York. Frye, M 2001, 'The performance of bank-managed mutual funds', Journal of Financial Research, vol. 24, no. 3, pp. 419-42. Gallagher, D 2003, 'Investment manager characteristics, strategy, top management changes and fund performance', Accounting & Finance, vol. 43, no. 3, pp. 283-309. Gallagher, D & Martin, K 2005, 'Size and investment performance: a research note', Abacus, vol. 41, no. 1, pp. 55-65. Grinblatt, M & Titman, S 1994, 'A study of monthly fund returns and performance evaluation techniques', Journal of Financial and Quantitative Analysis, vol. 29, no. 3, pp. 419-44. Grossman, S & Stiglitz, J 1980, 'On the impossibility of informationally efficient markets', American Economic Review, vol. 70, no. 3, pp. 393-408. Guercio, D & Tkac, P 2002, Star tower: the effect of Morningstar ratings on mutual fund flows, Federal Reserve Bank of Atlanta. Ibbotson, R & Kaplan, P 2000, 'Does asset allocation policy explain 40, 90 or 100 percent of performance?' Financial Analysts Journal, vol. 56, no. 1, pp. 26-33. ICI 2004, Mutual Fund Fact Book, Investment Company Institute, viewed 15 Aug 2005, <http://www.ici.org>. Indro, D, Jiang, C, Hu, M & Lee, W 1999, 'Mutual fund performance: does fund size matter?' Financial Analysts Journal, vol. 55, no. 3, pp. 74-87. Ippolito, R 1989, 'Efficiency with costly information: a study of mutual fund performance, 1965-1984', Quarterly Journal of Economics, vol. 104, no. 1, pp. 1-23. ---- 1992, 'Consumer reaction to measures of poor quality: evidence from the mutual fund industry', Journal of Law and Economics, vol. 35, pp. 45-69. Jensen, M 1969, 'Risk, the pricing of capital assets, and the evaluation of investment portfolios', Journal of Business, vol. 42, pp. 167-247. Jones, C 2003, Your Money, Your Choice... Mutual Funds: Take Control Now and Build Wealth Wisely, Financial Times Prentice Hall, New Jersey. Mercer 1999-2002, Unit Trusts and Investment-linked Insurance Products Included Under CPFIS: Performance and Risk Monitoring Reports, Mercer Investment Consulting, viewed 19 Feb 2005, <http://www.mercerfundwatch.com/>. Peterson, J, Pietranico, P, Riepe, M & Xu, F 2001, 'Explaining the performance of domestic equity funds', Journal of Investing, vol. 10, no. 3, pp. 81-91. ---- 2002, 'Explaining after-tax mutual fund performance', Financial Analysts Journal, vol. 58, no. 1, pp. 75-86. S&P 2003-2004, Performance and Risk Monitoring Report for CPFIS-included Unit Trusts & Investment-linked Insurance Products, S&P Fund Services, viewed 19 Feb 2005, <http://www.imas.org.sg>. Sawicki, J 2000, 'Investors' response to the performance of professional fund managers: evidence from the Australian wholesale funds market', Australian Journal of Management, vol. 25, no. 1, pp. 47-66. ---- 2001, 'Investors' differential response to managed fund performance', Journal of Financial Research, vol. 24, no. 3, pp. 367-84. Sawicki, J & Finn, F 2002, 'Smart money and small funds', Journal of Business Finance & Accounting, vol. 29, no. 5 & 6, pp. 825-46. Sharpe, W 1966, 'Mutual fund performance', Journal of Business, vol. 39, no. 1, pp. 119-38. Sirri, E & Tufano, P 1998, 'Costly search and mutual fund flows', Journal of Finance, vol. 53, no. 5, pp. 1589622. Tng, C 2005, 'Performance of approved equity funds: evidence from Singapore's retail funds', paper presented to 18th Annual Australasian Finance and Banking Conference, Sydney, 14-16 Dec. Veltman, L 2000, Living and Working in Australia, 7 edn, How To Books, Oxford.

227

Appendix D

Figure 1: Conceptual Model for Determinants of Mutual Fund Performance

Expected return Size Expense ratio Implicit costs Transaction costs Explicit costs Fund performance

Source: developed for this research. Table 1: Research Sample Organization Fund Symbol type DBS Horizon Singapore Equity Fund DHSE Bank DBS Shenton Thrift Fund DST OCBC Savers Singapore Trust Fund OSST OUB Union Singapore Equity Fund OUSE UOB Optimix Singapore Equity Fund UOSE UOB Unifund UU UOB United Growth Fund UUG AXA Life-Fortress Fund ALF Insurance company AXA Life-Fortress Fund A ALFA GE Greatlink Singapore Equities Fund GGSE Keppel Managed Fund KM NTUC Income Singapore Equity Fund NISE OUB Manulife Golden Singapore Growth Fund OMGSG UOB Life FOF-Unifund ULFU UOB Life FOF-United Growth Fund ULFUG UOB LifeLink Growth Fund ULG Investment Aberdeen Singapore Equity Fund ASE company CMG First State Singapore Growth Fund CFSSG Schroder Singapore Trust SST Source: funds identified from Mercer (1999-2002) and S&P (2003-2004) data. 19992002 ! ! ! ! ! ! ! ! 20022004 ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

! ! !

228

Journal article 1

Table 2: Summary Characteristics of Singapore's CPF-approved Domestic Equity Funds Fund Type 1999:Q2 to 2002:Q1 Average Size (S$ quarterly million) return (%) Expense ratio 2003:Q1 to 2004:Q3 Average Size (S$ quarterly million) return (%) Expense ratio

SST Investment 7.32 113.42 1.71 6.01 238.16 1.52 UUG Bank 4.51 103.33 1.36 6.39 125.20 1.18 ULFUG Insurance 6.03 91.65 1.35 6.11 84.23 1.18 CFSSG Investment 5.36 86.33 2.35 DHSE Bank -0.46 75.31 1.50 7.28 109.26 1.49 DST Bank 4.33 43.96 0.96 12.30 59.50 1.16 UU Bank 0.23 35.96 1.57 ULFU Insurance 7.00 35.05 1.55 OMGSG Insurance -0.66 19.56 2.04 ALF Insurance 2.05 17.72 1.45 6.11 62.20 1.24 OSST Bank 4.53 17.03 1.49 7.49 19.06 1.52 KM Insurance 0.12 10.58 1.16 ASE Investment 5.84 6.97 2.58 5.59 21.11 2.12 OUSE Bank -6.13 3.00 3.32 UOSE Bank 2.57 2.57 3.83 4.93 0.85 5.02 ULG Insurance -0.64 0.41 1.51 ALFA Insurance 4.71 45.19 1.67 GGSE Insurance 4.88 30.31 1.20 NISE Insurance 7.45 1.69 0.43 Note: type refers to the financial institution managing the fundbank, insurance or investment companies. Source: developed using Mercer (1999-2002) and S&P (2003-2004) data.

229

Appendix D

Table 3: Two-sample t-test Assuming Unequal Variances for Returns of Big and Small Funds =0.05 1999:Q2 to 2002:Q1 Mean return (%) Variance Observations Hypothesized Mean Difference Df t Stat P(T t) one-tail t Critical one-tail P(T t) two-tail t Critical two-tail 2003:Q1 to 2004:Q3 Big fund Small fund Mean return (%) 7.366666667 5.841666667 Variance 6.059626667 1.680736667 Observations 6 6 Hypothesized Mean Difference 0 Df 8 t Stat 1.342656119 P(T t) one-tail 0.108116209 t Critical one-tail 1.85954832 P(T t) two-tail 0.216232417 t Critical two-tail 2.306005626 Source: developed using Mercer (1999-2002) and S&P (2003-2004) data. Big fund 4.29 8.541942857 8 0 13 1.993292277 0.033823825 1.770931704 0.067647651 2.16036824 Small fund 0.96 13.78537143 8

230

Journal article 1

Table 4: Two-sample t-test Assuming Unequal Variances for Returns of High and Low Expense Ratio Funds =0.05 1999:Q2 to 2002:Q1 Mean return (%) Variance Observations Hypothesized Mean Difference Df t Stat P(T t) one-tail t Critical one-tail P(T t) two-tail t Critical two-tail 2003:Q1 to 2004:Q3 Low expense ratio High expense ratio Mean return (%) 7.206666667 6.001666667 Variance 6.897386667 1.367136667 Observations 6 6 Hypothesized Mean Difference 0 Df 7 t Stat 1.026724126 P(T t) one-tail 0.169357764 t Critical one-tail 1.894577508 P(T t) two-tail 0.338715529 t Critical two-tail 2.36462256 Source: developed using Mercer (1999-2002) and S&P (2003-2004) data. Low expense ratio 2.55875 6.918755357 8 0 11 -0.07001157 0.472720517 1.795883691 0.945441034 2.200986273 High expense ratio 2.69125 21.73504107 8

231

Appendix D

Table 5: Two-sample t-test Assuming Unequal Variances for Expense Ratios of Big and Small Funds =0.05 Mean expense ratio Variance Observations Hypothesized Mean Difference Df T Stat P(T t) one-tail T Critical one-tail P(T t) two-tail T Critical two-tail Big fund 1.543541667 0.156333681 8 0 9 -1.68592741 0.063044455 1.833113856 0.126088909 2.262158887 Small fund 2.1725 0.957078571 8

Big fund Small fund Mean expense ratio 1.29547619 1.993333333 Variance 0.027117075 2.515746667 Observations 6 6 Hypothesized Mean Difference 0 Df 5 T Stat -1.07196501 P(T t) one-tail 0.166363365 T Critical one-tail 2.015049176 P(T t) two-tail 0.332726729 T Critical two-tail 2.570577635 Source: developed using Mercer (1999-2002) and S&P (2003-2004) data. Total word count: 4,904 words

232

Appendix E Journal article 2

Bank Fund Management Challenges and Opportunities TNG Cheong Sing Southern Cross University Postal address: 171 Lor 1 Toa Payoh #13-1132 S(310171) Singapore Telephone: +65 9871 7218 Facsimile: +65 6259 3558 Email: tcs@pacific.net.sg This article is an original and unpublished work from its author, currently being reviewed by the Singapore Management Review/Asia-Pacific Journal of Management Theory and Practice. The first draft of this paper was presented in November 2005 at the Academy of International Business (AIB) Southeast Asia Regional Conference. Tng Cheong Sing is a Doctor of Business Administration (DBA) candidate at Southern Cross University Graduate College of Management and an AIB member. He has recently submitted his Doctorate thesis Factors Influencing Unit Trust Performance for examination. Previously, he was a faculty staff at University of Tasmania School of Accounting and Finance as well as Singapore Polytechnic School of Infocommunication Technology. He would like to acknowledge the guidance of his DBA supervisor Emeritus Professor Geoffrey Meredith from Southern Cross University. Abstract With financial deregulation in Asia Pacific countries, local banks in small economies need to withstand erosion of business from foreign competitors. Banks, in order to increase profits, compete with local as well as foreign insurance and investment companies by offering mutual fund products. To remain competitive, banks should shed their reputation as incapable of generating impressive fund returns, as customers may not be informed that bank funds are comparable in performance to non-bank counterparts. Banks need to differentiate their fund characteristics and reduce portfolio management costs to gain a competitive advantage. Keywords BANK-MANAGED FUNDS; FUND PERFORMANCE; WEALTH MANAGEMENT.

1 Introduction With financial deregulation increasing competition between local and foreign banks, insurance companies as well as investment firms, the fund management industry has become critical for banks, whose profitability can be improved by offering competitive mutual fund products (Gallo, Apilado & Kolari 1996). Banks, in order to compete in the fund management industry, have to offer products that do not under-perform their counterparts from insurance and investment companies. With financial deregulation in Asia Pacific countries easing entry of foreign competitors, local financial institutions in small economies have to withstand erosion of investor monies by foreign competitors. Besides offering competitive funds, banks need to shed their image of under-performers in the asset management industry. In the popular press, performance of bank-managed funds has been perceived as inferior to their non-bank counterparts, even in the most developed economy. In the United States of America (USA), McTague (1994) claimed bank funds were non-aggressive and incapable of generating impressive returns. As a result, various investors expecting little returns from bank funds were reluctant to invest in them. However, banks in the USA were relatively new to fund management, previously being prohibited by the Glass-Steagall Act before the Federal Reserve Board allowed them to manage funds (Gallo, Apilado & Kolari 1996). From 8 percent of total mutual funds in 1991 to 14 percent in 1999 worth more than US$255 billion, bank funds rapid growth questioned their reputation as under-performers compared to non-bank counterparts (Frye 2001). Besides the USA, banks under-performing image was also perceived in other developed countries. The chief executive of AMP, a popular wealth management company in Australia and New Zealand, commented it was difficult for banks to compete with specialist investment companies. As a result, there were few international banks that were very successful in fund management. The comment was made concerning Australian banks purchases of wealth management companies since 2000 totalling A$19 billion (including Commonwealth Bank of Australias A$9 billion acquisition of Colonial First State, National Australia Banks A$5 billion purchase of MLC, Australia and New Zealand Banking Groups A$4 billion joint venture with Dutch ING Group, and Westpacs A$1 billion acquisition of BT and Rothschild) which had mostly delivered unimpressive returns (Moullakis & Patten 2005). Reviewing the literature, while earlier research indicated underperformance of bank funds compared to nonbank counterparts (Bauman & Miller 1995; Bogle & Twardowski 1980), later research did not detect underperformance (Frye 2001). According to Frye (2001), earlier research reporting relative underperformance

233

Appendix E
of bank-managed funds ignored their differing fiduciary standards. However, she focused only on bond funds as banks in the USA had more assets under management in bond funds rather than equity funds. It is not known how bank and non-bank equity funds compare when faced with the same fiduciary standard. Examining domestic equity funds approved by Singapores Central Provident Fund (CPF) Board for its CPF Investment Scheme facilitates a more direct comparison of funds managed by banks and non-banks than previous studies, as CPF-approved funds face the same standard for managing social security savings. This study contributes to understanding competitiveness of bank-managed funds by exploring relationship between type of fund management organization and past performance for Singapores retail equity funds. Majority of fund management research were conducted using data from the USA and other large developed markets, leaving many small markets unexplored in the literature. Among developed equity markets identified by Ibbotson and Brinson (1993, pp. 109-11), little research was published on the fund industry in Singapore, one of the smallest economies in the world. Examining Singapores CPF-approved equity funds offers an opportunity to control for differing fiduciary standards. As all bank-managed CPF-approved equity funds are from domestic banks, this research reveals performance of these banks in Singapores fund management industry when competing with local as well as foreign insurance and investment companies. Whether justified or not, banks reputation for relatively unimpressive fund performance can affect their popularity. Less popular funds attract less investor cash flows, resulting in smaller amounts of net assets under management. Reputation of a fund can therefore affect its size. Small funds are not competitive as economies of scale provide cost advantages to large funds. Large funds can lower brokerage commissions without significant increase in administration and research costs (Indro et al. 1999), resulting in better net returns than small funds with otherwise similar characteristics. This paper proceeds as follows. Conceptual framework for comparing performance of bank and non-bank funds is presented in the next section. Section three describes data and methodology used. Results are interpreted in section four before section five concludes with recommendation. 2 Comparison of Bank and Non-bank Fund Performance Comparing performance of funds managed by banks and non-banks reveals the quality of funds offered by these institutions and verifies results of previous studies. Quality of funds is an indication of their competitiveness. Bank funds have grown rapidly over the years. This was confirmed in the USA (Frye 2001). In Singapore, the average bank-managed CPF-approved domestic equity fund grew from around S$40 million during 1999-2002 to more than S$60 million in 2003-2004 (Tng 2005). If investors chase past performance and bank fund performance were inferior to their non-bank counterparts, bank funds will not experience growth among rational investors. Inferior performance of bank funds coupled with rapid growth in size can imply irrational or unsophisticated investors. 2.1 Capital Asset Pricing Model for Singapores Equity Funds To examine performance of equity funds, their quarterly returns were modelled in Equation 1 using Sharpes (1964) capital asset pricing model (CAPM). [Insert Equation 1: Single-index Model for Returns of Singapore's Domestic Equity Funds near here] Equation 1 models behaviour of fund returns according to beta, market risk premium and risk-free return. While there may be leads or lags in market returns in relation to fund returns, use of quarterly data should negate these effects. Lagging quarterly data implies funds taking as long as three months to respond to changes in the market, which is not consistent with capital market efficiency (Fama 1970, 1991). Studies on Singapores stock market showed its efficiency strengthened as time interval being considered increased (Wong 1988). Specifically, even though daily or weekly data revealed market inefficiency (Lim 1985; Saw & Tan 1986), Ariff (1986) used monthly data to show Singapores market was comparable to New York, London and Australian stock markets in adjusting prices efficiently to reflect new information. Testing efficiency of Singapores stock market using more recent data at various time intervals can be carried out for further research. By using quarterly instead of monthly returns data, this research considers Singapores market to be at least weak form efficient and comparable in terms of efficiency to other developed markets. Assuming weak-form efficiency implies past and future fund performances are independent (Fama 1970, 1991). 2.2 Fund Performance Evaluation Measures To compare fund performance, four risk-adjusted portfolio performance measures were used, unlike financial periodicals emphasizing raw returns. Among these measures, Goodwins (1998) information ratio (IR) was developed more recently while measures developed by Jensen (1968), Sharpe (1966) and Treynor (1965) were based on CAPM. IR of an equity fund is computed as arithmetic average of funds excess return divided by standard deviation of excess return (Goodwin 1998). This ratio measures mean excess return per unit of unsystematic risk. As for its

234

Journal article 2
relation with other evaluation measures, IR can be expressed in terms of Jensen alpha (raw fund return less return predicted by CAPM) when excess returns are estimated with historical data using the single-factor regression equation mentioned in the previous section, while Sharpe ratio (fund risk premium divided by standard deviation of fund return) is a special case of IR (Goodwin 1998). While Sharpe ratio uses standard deviation measure of total risk (comprising of systematic and non-systematic components), Treynor ratio uses CAPMs beta as relative measure of systematic risk to divide fund risk premium. Even though these performance measures improve upon comparison of raw returns, some researchers identified bias in these measures. For example, Friend and Blume (1970) reported risk-adjusted performance measures of low-risk portfolios better than high-risk counterparts. While these performance measures are not without problems, in the absence of alternative measures, all four measures were used for this research to minimize errors from relying solely on one measure, as each measure ranks individual fund performance differently (Reilly & Brown 2003, p. 1122) and can yield substantially different performance rankings (Corrado & Jordan 2005, p. 434). The following section describes how hypothesis testing was carried out on performance of bank and non-bank funds using CAPM and evaluation measures on secondary data. 3 Data and Methodology In this section, data collection process and its consideration for survivorship bias are described in the first subsection. In the second sub-section, regression analysis of fund returns is explained before the third sub-section describes hypothesis testing of fund performance. 3.1 Secondary Data Collection To carry out this research, five years of quarterly returns from 1999 to 2004 for 19 retail funds approved for Singapores CPF Investment Scheme were examined. These funds were invested in shares from the Singapore Stock Exchange. Table 1 identifies funds used for this research. [Insert Table 1: Research Sample near here] For this research, only CPF-approved funds were considered to control for differing fiduciary responsibilities, as these funds followed the same fiduciary standard for managing social security savings. Failure to control for such standards may lead to biased test results (Frye 2001). Among these funds, only those investing in the local stock market were selected. As each benchmark index has a unique market cycle, funds based on benchmarks other than the Singapore Straits Times Index (STI) were excluded. These CPF-approved domestic equity funds were classified according to type of organization managing the fund: (1) insurance-linked investment products managed by insurance companies; (2) unit trusts managed by investment firms; or (3) bank-managed funds. Organization types differ in terms of operational structure, priorities and benefits for fund managers, which may influence resulting portfolio returns (Bauman & Miller 1995). This research incorporates consideration for survivorship bias. As funds that did not survive are usually the worst performing ones, when data for non-survivors are not considered, resulting average performance of each fund group can be overstated. To control for survival bias, this study collected data for surviving and nonsurviving funds using quarterly reports for all CPF-approved unit trusts. However, for performing regression analysis, only funds with at least three quarters of data were included. 3.2 Regression Analysis For each fund, linear regression was performed using its quarterly risk premium (RETft RFRt) as dependent variable and the STI quarterly risk premium (STIt RFRt) as independent variable. As there are two risk-free rates: RFRo = 0.625 percent per quarter and RFRs = 1 percent per quarter for guaranteed interest rates of CPF Ordinary and Special accounts respectively, as well as two holding periods corresponding to data collected from Mercer (1999-2002) and S&P (2003-2004), four sets of linear regression were performed for each fund using a combination of risk-free rate and holding period. For residual analysis, linear trends on normal probability plots were obtained to confirm normality assumption for linear regression was satisfied (Mendenhall & Sincich 1996). Hypothesis testing was carried out after regression analysis. 3.3 Hypothesis Testing To test hypotheses for no significant performance differences between domestic equity funds managed by banks and non-banks, two-tailed pooled-variance t-test for difference in two means was conducted for returns as well as evaluation measures for bank and non-bank funds during both time periods 1999-2002 and 2003-2004 at an alpha level of 0.05. According to Berkowitz and Qiu (2003), technology usage in the fund management industry was quite homogenous across companies, which may lead to no overall performance difference between bank

235

Appendix E
and non-bank funds. Relatively uniform human and information resources in developed markets can make it difficult for specialist investment firms to outperform their bank counterparts. Results of hypothesis testing are interpreted in the next section. 4 Results and Interpretation Table 2 tabulates summary characteristics computed for funds in the research sample. [Insert Table 2: Summary Characteristics of CPF-approved Domestic Equity Funds near here] The following four sub-sections report findings for comparison of quarterly returns, information ratios, Jensen alphas as well as Sharpe and Treynor ratios for funds managed by banks and non-banks. 4.1 Bank and Non-bank Fund Returns Referring to Table 2 above, during 1999-2002, when the STI posted an average quarterly return of 2.94 percent, the average bank equity fund under-performed the market at 1.37 percent, while the average non-bank fund outperformed the market at 3.60 percent. For 2003-2004, with STI average quarterly return of 5.97 percent, the reverse seemed to be true as bank funds outperformed the market at 7.68 percent while non-bank funds underperformed the market at 5.84 percent. However, performing a two-sample t-test assuming unequal variances for returns of bank and non-bank funds in Table 3 showed no significant difference between returns of bank and non-bank domestic equity funds for both holding periods. [Insert Table 3: Two-sample t-test for Returns of Bank and Non-bank Funds near here] This result supported Fryes (2001) finding for bond funds. 4.2 Information Ratio of Bank and Non-bank Funds Table 2 at the beginning of this section reported IRs for bank and non-bank funds. When evaluating fund performance, reasonable values for the ratio should range from 0.5 to 1.0 for good to exceptionally good performance (Grinold & Kahn 1995). For Singapores CPF-approved domestic equity funds, mean sample IR was 0.262 during 1999-2002 and deteriorated to 0.141 in 2003-2004, well below the 0.5 standard for good performance according to Grinold and Kahn (1995). Agreeing with Goodwins (1998) findings in the USA, average fund in the sample can add value to its investments, but performance did not qualify as good. In fact, none of the funds can deliver an excellent IR greater than 1.0, even though there were a few good performers from banks and non-banks. Still, good performers during the first period cannot sustain their performance for the second period, confirming lack of performance consistency reported in the literature for funds in the USA (Dunn & Theinsen 1983; Jensen 1969). Performing a two-sample t-test assuming unequal variances for IRs of bank and non-bank funds in Table 4 showed no significant difference between IRs of bank and non-bank equity funds during each holding period. [Insert Table 4: Two-sample t-test for Information Ratios of Bank and Non-bank Funds near here] This result again supported Fryes (2001) finding for bond funds. 4.3 Jensen Alpha of Bank and Non-bank Funds Insignificance of positive Jensen alpha values in Table 2 at the beginning of this section downplayed possibility of average fund beating the market, confirming previous research in the USA reporting mutual funds unable to beat the market (Carlson 1970; Jensen 1968). According to the Jensen alpha criterion, even though majority of funds registered abnormal returns above expectation for both holding periods, only one of them was statistically significant in each period. As the fund registering significant abnormal return during 1999-2002 became one of the worst performers in 2003-2004, consistency was clearly lacking, confirming the previous sub-sections observation. Table 5s two-sample t-test assuming unequal variances for Jensen alphas of bank and non-bank funds showed bank funds under-performing non-bank funds significantly during 1999-2002, but for 2003-2004, no significant under-performance of bank funds was detected. [Insert Table 5: Two-sample t-test for Jensen Alphas of Bank and Non-bank Funds near here] Thus, bank funds may have improved their performance to be comparable to non-bank counterparts. 4.4 Sharpe and Treynor Ratios of Bank and Non-bank Funds Table 2 at the beginning of this section showed average fund having positive Sharpe and Treynor ratios during both holding periods, implying returns exceeding guaranteed interest rates. However, the average bank fund actually registered negative Sharpe and Treynor ratios during 1999-2002, implying earning guaranteed interest rates in Ordinary and Special accounts were better than investing in bank funds for that period. Overall, returns

236

Journal article 2
from CPF-approved equity funds were higher than guaranteed interest rates of Ordinary and Special accounts for both periods, refuting an earlier finding from Koh (1999). Performing two-sample t-test assuming unequal variances for Sharpe ratios of bank and non-bank funds in Table 6 showed no significant performance difference during 1999-2002, but significant underperformance of bank funds during 2003-2004. [Insert Table 6: Two-sample t-test for Sharpe Ratios of Bank and Non-bank Funds near here] Table 7s two-sample t-test assuming unequal variances for Treynor ratios of bank and non-bank funds showed no significant performance difference during both periods. [Insert Table 7: Two-sample t-test for Treynor Ratios of Bank and Non-bank Funds near here] These results indicated higher level of non-systematic risk in bank funds than non-bank ones during 2003-2004. 5 Conclusion and Recommendation Financial deregulation does not make it impossible for domestic banks to earn economic profits, if they depart from perfect competition and gain competitive advantage over local as well as foreign financial institutions. Domestic banks face the challenge of withstanding erosion of investor monies by more competitors. However, there are opportunities for gaining competitive advantage. The key ingredients for competitive advantage identified by Porter (1980, 1985) are industry characteristics, product differentiation and cost advantages. In terms of the fund industry, financial deregulation reduced barriers to entry for foreign institutions. With proliferation of fund products from local and foreign institutions, it is important for investors to be informed about relative performance of bank and non-bank funds. This research, by controlling for differing fiduciary standards, showed domestic banks in Singapore can compete with local as well as foreign specialist wealth management companies, as bank funds were comparable in performance to non-bank counterparts. However, in terms of investment strategy, there was evidence bank fund managers were more risky than non-bank counterparts. To gain a competitive advantage, local banks should work towards differentiating their financial products by first shedding unjustified reputation as underperformers. Even though technology and human resources are relatively uniform across the fund industry, portfolio management costs can be reduced with economies of scale when banks offer bigger funds. This research showed bank funds were indeed growing steadily. Such growth can only be sustained when banks develop good reputation for fund products. In fact, expenditures and size are characteristics that should be differentiated, as they can be determinants of fund performance. Determination of characteristics that are significant fund performance determinants will open opportunities for further research. References Ariff, M 1986, 'Announcement effects in a thin market--a study of the Singapore equity market', paper presented to Proceedings of the Academy of International Business, Southeast Asia Regional Conference, Taipei, 26-28 June. Bauman, W & Miller, R 1995, 'Portfolio performance rankings in stock market cycles', Financial Analysts Journal, vol. 51, no. 2, pp. 79-87. Berkowitz, M & Qiu, J 2003, 'Ownership, risk and performance of mutual fund management companies', Journal of Economics and Business, vol. 55, pp. 109-34. Bogle, J & Twardowski, J 1980, 'Institutional investment performance: banks, investment counselors, insurance companies and mutual funds', Financial Analysts Journal, vol. 36, no. 1, pp. 33-41. Carlson, R 1970, 'Aggregate performance of mutual funds, 1948-1967', Journal of Financial and Quantitative Analysis, vol. 5, no. 1, pp. 1-32. Corrado, C & Jordan, B 2005, Fundamentals of investments Valuation and Management, 3 edn, McGraw-Hill Irwin, New York. Dunn, P & Theinsen, R 1983, 'How consistently do active managers win?' Journal of Portfolio Management, vol. 9, pp. 47-50. Fama, E 1970, 'Efficient capital markets: a review of theory and empirical work', Journal of Finance, vol. 25, no. 2, pp. 383-417. ---- 1991, 'Efficient capital markets: II', Journal of Finance, vol. 46, no. 5, pp. 1575-617. Friend, I & Blume, M 1970, 'Measurement of portfolio performance under uncertainty', American Economic Review, vol. 60, no. 3, pp. 561-75. Frye, M 2001, 'The performance of bank-managed mutual funds', Journal of Financial Research, vol. 24, no. 3, pp. 419-42. Gallo, J, Apilado, V & Kolari, J 1996, 'Commercial bank mutual fund activities: implications for bank risk and profitability', Journal of Banking and Finance, vol. 20, pp. 1775-91. Goodwin, T 1998, 'The information ratio', Financial Analysts Journal, vol. 54, no. 4, pp. 34-43. Grinold, R & Kahn, R 1995, Active Portfolio Management, Probus Publishing, Chicago. Ibbotson, R & Brinson, G 1993, Global Investing, McGraw-Hill, New York.

237

Appendix E
Indro, D, Jiang, C, Hu, M & Lee, W 1999, 'Mutual fund performance: does fund size matter?' Financial Analysts Journal, vol. 55, no. 3, pp. 74-87. Jensen, M 1968, 'The performance of mutual funds in the period 1945-1964', Journal of Finance, vol. 23, no. 2, pp. 389-416. ---- 1969, 'Risk, the pricing of capital assets, and the evaluation of investment portfolios', Journal of Business, vol. 42, pp. 167-247. Koh, S 1999, 'A survey of unit trusts in Singapore', Singapore Management Review, vol. 21, no. 1, pp. 49-78. Lim, G 1985, 'A Box-Jenkins Approach to Stock Market Analysis on the SES', National University of Singapore. McTague, J 1994, 'Laggards no longer', Barron's, vol. 74, pp. F22-F3. Mendenhall, W & Sincich, T 1996, A Second Course in Statistics: Regression Analysis, 5 edn, Prentice Hall, New Jersey. Mercer 1999-2002, Unit Trusts and Investment-linked Insurance Products Included Under CPFIS: Performance and Risk Monitoring Reports, Mercer Investment Consulting, viewed 19 Feb 2005, <http://www.mercerfundwatch.com/>. Moullakis, J & Patten, S 2005, 'Big-spending banks caught with wealth of problems', The Australian Financial Review, 8 Aug, p. 1. Porter, M 1980, Competitive Strategy: Techniques of Analyzing Industries and Competitors, The Free Press, New York. ---- 1985, Competitive Advantage: Creating and Sustaining Superior Performance, The Free Press, New York. Reilly, F & Brown, K 2003, Investment Analysis and Portfolio Management, 7th edn, South-Western, Ohio. S&P 2003-2004, Performance and Risk Monitoring Report for CPFIS-included Unit Trusts & Investment-linked Insurance Products, S&P Fund Services, viewed 19 Feb 2005, <http://www.imas.org.sg>. Saw, S & Tan, K 1986, 'The Stock Exchange of Singapore all-share price indices: testing of independence', Securities Industry Review, vol. 12, no. 1. Sharpe, W 1964, 'Capital asset prices: a theory of market equilibrium under condtions of risk', Journal of Finance, vol. 19, no. 3, pp. 424-42. ---- 1966, 'Mutual fund performance', Journal of Business, vol. 39, no. 1, pp. 119-38. Tng, C 2005, 'Performance of approved equity funds: evidence from Singapore's retail funds', paper presented to 18th Annual Australasian Finance and Banking Conference, Sydney, 14-16 Dec. Treynor, J 1965, 'How to rate management of investment funds', Harvard Business Review, vol. 43, no. 1, pp. 63-75. Wong, K 1988, 'Role of Singapore and Malaysian stock exchanges', in C Tan & K KC (eds), Handbook of Singapore-Malaysian Corporate Finance, Butterworths, Singapore.

238

Journal article 2

Equation 1: Single-index Model for Returns of Singapores Domestic Equity Funds

RET ft = RFRt + f (STI t RFRt ) + ft

where RETft f = return from fund f at time t; = Covf,STI / 2STI, covariance between returns for fund f and local stock index (STI) divided by variance of returns for the STI; RFRt = risk-free rate of return at time t; STIt = return from the STI at time t; and = error term for fund f at time t. ft Source: adapted from Sharpe (1964). Table 1: Research Sample Organization Fund Symbol 1999type 2002 ! DBS Horizon Singapore Equity Fund DHSE Bank ! DBS Shenton Thrift Fund DST ! OCBC Savers Singapore Trust Fund OSST ! OUB Union Singapore Equity Fund OUSE ! UOB Optimix Singapore Equity Fund UOSE ! UOB Unifund UU ! UOB United Growth Fund UUG ! AXA Life-Fortress Fund ALF Insurance company (Ins) AXA Life-Fortress Fund A ALFA GE Greatlink Singapore Equities Fund GGSE ! Keppel Managed Fund KM NTUC Income Singapore Equity Fund NISE ! OUB Manulife Golden Singapore OMGSG Growth Fund ! UOB Life FOF-Unifund ULFU ! UOB Life FOF-United Growth Fund ULFUG ! UOB LifeLink Growth Fund ULG ! Investment Aberdeen Singapore Equity Fund ASE company (Inv) CMG First State Singapore Growth ! CFSSG Fund ! Schroder Singapore Trust SST Source: funds identified from Mercer (1999-2002) and S&P (2003-2004) data. 20022004 ! ! ! ! ! ! ! ! !

! ! !

239

Appendix E

Table 2: Summary Characteristics of CPF-approved Domestic Equity Funds Fund Type 1999:Q2--2002:Q1 Ret Info Jensen (%) ratio alpha -0.46 0.12 0.19 4.33 0.25 1.13 4.53 0.41 1.42 -6.13 -0.08 -4.73 2.57 -0.16 -0.32 0.23 0.15 0.65 4.51 0.45 1.55 2.05 0.26 2.14 0.12 0.04 0.22 -0.66 0.01 0.25 7.00 0.57 4.41 6.03 0.68 3.00 -0.64 -0.15 -0.66 5.84 0.49 2.90 5.36 0.24 2.01 7.32 0.92 4.05 2.63 0.26 1.14 1.37 0.16 -0.02 3.60 0.34 2.04 2.94 0.00 0.00 0.63 1.00 Sharpe ratio -0.07 0.17 0.19 -1.37 0.10 -0.03 0.20 0.08 -0.03 -0.07 0.45 0.31 -0.09 0.26 0.20 0.31 0.04 -0.12 0.16 0.12 Treynor ratio -1.14 3.33 3.64 -20.95 2.48 -0.47 3.85 1.41 -0.49 -1.12 9.04 6.29 -1.46 5.20 4.02 5.87 1.22 -1.32 3.20 2.31

DHSE Bank DST Bank OSST Bank OUSE Bank UOSE Bank UU Bank UUG Bank ALF Ins KM Ins OMGSG Ins ULFU Ins ULFUG Ins ULG Ins ASE Inv CFSSG Inv SST Inv Average Average bank Average non-bank STI CPF Ordinary a/c CPF Special a/c

2003:Q12004:Q3 Ret Info Jensen Sharpe Treynor (%) ratio alpha ratio ratio DHSE Bank 7.28 0.42 0.32 0.75 5.61 DST Bank 12.30 0.76 3.60 0.89 7.72 OSST Bank 7.49 0.32 0.48 0.71 5.75 UOSE Bank 4.93 -0.43 -1.04 0.57 4.31 UUG Bank 6.39 0.22 1.04 0.87 6.43 ALF Ins 6.11 0.03 3.35 1.70 13.77 ALFA Ins 4.71 0.27 1.97 1.26 6.65 GGSE Ins 4.88 -0.59 -0.13 0.71 5.18 NISE Ins 7.45 0.75 0.40 1.24 6.36 ULFUG Ins 6.11 0.06 0.71 0.81 6.13 ASE Inv 5.59 -0.13 1.19 0.92 7.03 SST Inv 6.01 0.02 0.29 0.76 5.65 Average 6.60 0.14 1.01 0.93 6.71 Average bank 7.68 0.2 0.88 0.76 5.96 6 Average non-bank 5.84 0.0 1.11 1.06 7.25 6 STI 5.97 0.0 0.00 0.75 5.34 0 CPF Ordinary a/c 0.63 CPF Special a/c 1.00 Note: Jensen alphas, Sharpe and Treynor ratios computed using CPF Ordinary account risk-free interest rate. Source: developed from Mercer (1999-2002) and S&P (2003-2004) data. Fund Type

240

Journal article 2

Table 3: Two-sample t-test for Returns of Bank and Non-bank Funds = 0.05 1999:Q2--2002:Q1 Mean return (%) Variance Observations Hypothesized mean difference Df t statistic P(Tt) one-tail t critical one-tail P(Tt) two-tail t critical two-tail Bank 1.37 15.158 7 0 12 -1.209 0.125 1.782 0.250 2.179 Non-bank 3.60 11.257 9

Bank Non-bank 2003:Q1--2004:Q3 Mean return (%) 7.68 5.84 Variance 7.692 0.838 Observations 5 7 Hypothesized mean difference 0 Df 5 t statistic 1.430 0.106 P(Tt) one-tail t critical one-tail 2.015 0.212 P(Tt) two-tail t critical two-tail 2.571 Source: developed from Mercer (1999-2002) and S&P (2003-2004) data. Table 4: Two-sample t-test for Information Ratios of Bank and Non-bank Funds = 0.05 1999:Q2--2002:Q1 Mean information ratio Variance Observations Hypothesized mean difference Df t statistic P(Tt) one-tail t critical one-tail P(Tt) two-tail t critical two-tail Bank 0.162 0.053 7 0 14 -1.215 0.122 1.761 0.245 2.145 Non-bank 0.339 0.123 9

Bank Non-bank 2003:Q1--2004:Q3 Mean information ratio 0.256 0.058 Variance 0.188 0.164 Observations 5 7 Hypothesized mean difference 0 Df 8 t statistic 0.801 0.223 P(Tt) one-tail t critical one-tail 1.860 0.446 P(Tt) two-tail t critical two-tail 2.306 Source: developed from Mercer (1999-2002) and S&P (2003-2004) data.

241

Appendix E

Table 5: Two-sample t-test for Jensen Alphas of Bank and Non-bank Funds = 0.05 1999:Q2--2002:Q1 Mean Jensen alpha Variance Observations Hypothesized mean difference Df t statistic P(Tt) one-tail t Critical one-tail P(Tt) two-tail t Critical two-tail Bank -0.015 4.775 7 0 11 -2.019 0.034 1.796 0.068 2.201 Non-bank 2.035 3.141 9

Bank Non-bank 2003:Q1--2004:Q3 Mean Jensen alpha 0.878 1.111 Variance 2.891 1.444 Observations 5 7 Hypothesized mean difference 0 df 7 t statistic -0.263 0.400 P(Tt) one-tail t Critical one-tail 1.895 0.800 P(Tt) two-tail t critical two-tail 2.365 Source: developed from Mercer (1999-2002) and S&P (2003-2004) data. Table 6: Two-sample t-test for Sharpe Ratios of Bank and Non-bank Funds = 0.05 1999:Q2--2002:Q1 Mean Sharpe ratio Variance Observations Hypothesized mean difference Df t statistic P(Tt) one-tail t critical one-tail P(Tt) two-tail t critical two-tail Bank -0.116 0.318 7 0 7 -1.228 0.1296 1.895 0.259 2.365 Non-bank 0.158 0.038 9

Bank Non-bank 2003:Q1--2004:Q3 Mean Sharpe ratio 0.759 1.057 Variance 0.017 0.129 Observations 5 7 Hypothesized mean difference 0 Df 8 t statistic -2.015 0.039 P(Tt) one-tail t critical one-tail 1.860 0.079 P(Tt) two-tail t critical two-tail 2.306 Source: developed from Mercer (1999-2002) and S&P (2003-2004) data.

242

Journal article 2

Table 7: Two-sample t-test for Treynor Ratios of Bank and Non-bank Funds = 0.05 1999:Q2--2002:Q1 Mean Treynor ratio Variance Observations Hypothesized mean difference Df t statistic P(Tt) one-tail t critical one-tail P(Tt) two-tail t critical two-tail Bank -1.323 78.928 7 0 8 -1.261 0.121 1.860 0.243 2.306 Non-bank 3.196 14.077 9

Bank Non-bank 2003:Q1--2004:Q3 Mean Treynor ratio 5.963 7.252 Variance 1.559 8.630 Observations 5 7 Hypothesized mean difference 0 Df 9 t statistic -1.037 0.163 P(Tt) one-tail t Critical one-tail 1.833 0.327 P(Tt) two-tail t critical two-tail 2.262 Source: developed from Mercer (1999-2002) and S&P (2003-2004) data. Total word count: 5,042 words

243