An Empirical Analysis of The Efficiency of Online Auction IPO

The current issue and full text archive of this journal is available at www.emeraldinsight.com/1743-9132.
htm
IJMF 5,3
268
An empirical analysis of the efciency of online auction IPO processes and traditional IPO processes
Nayantara Hensel
Graduate School of Business and Public Policy, US Naval Postgraduate School, Monterey, California, USA
Abstract
Purpose The purpose of this paper is to examine whether the online auction mechanism in the USA is more effective at pricing initial public offerings (IPOs) than the traditional book building process. Design/methodology/approach The analysis compares the performance of online auction IPOs with traditional IPOs issued in the same industry area and in the same year to assess the differences in rst day mispricing and its persistence. The paper compares the characteristics of rms choosing the auction process relative to the traditional process. It also uses regression models to examine whether online auction IPOs had a signicantly lower rst day price increase than traditional IPOs. Findings The results indicate that for 60 percent of the auction IPOs, over 40 percent of the traditional IPOs issued in that year and in that three-digit Standard Industry Classication (SIC) area had greater mispricing. The mispricing of online auction IPOs relative to traditional IPOs persist over time for 50-80 percent of online auction IPOs. Regression analyses controlling for industry effects, year effects, size of the issue, and type of traditional underwriter (low, medium, and high volume underwriters) suggest that the auctions rst day price surges are not signicantly lower than those of traditional underwriters. Moreover, high volume traditional underwriters have statistically signicantly higher rst day price surges than low volume traditional underwriters, supporting the theory that they intentionally misprice to benet their preferred clients. Firms choosing the auction process tend to be smaller in terms of the number of shares of their IPO and their annual sales than rms choosing the traditional IPO process. There is some overlap in industry sector and age, although this varies by year. Originality/value This paper suggests that the auction process may not be as efcient in pricing IPOs as was initially intended and that there are opportunities for further innovation and improvement. Keywords Auctions, Flotation of companies, Pricing, Electronic commerce, United States of America Paper type Research paper
International Journal of Managerial Finance Vol. 5 No. 3, 2009 pp. 268-310 q Emerald Group Publishing Limited 1743-9132 DOI 10.1108/17439130910969729
JEL classication G32, G24, G12, M21 The author appreciated discussions with Professor Dale Jorgenson. The author also appreciated the comments on earlier drafts of this paper from participants at the NBEIC Conference at the Federal Reserve Bank of Dallas in 2008, at the San Francisco Chapter meeting of the National Association of Business Economists at the Bureau of Labor Statistics in 2008, at the 2007 European Financial Management Conference, at the 2007 National Association of Business Economists Annual Conference, at the 2006 Western Economics Association Conference, at the 2006 Midwest Economics Association Conference, and at the DePaul University Economics Seminar in 2006. The views and analysis in this paper represent only those of the author, not any institution with which the author is afliated.
I. Introduction and literature review The resurgence in the initial public offering (IPO) market in 2007 in the USA and overseas raises the question of whether the traditional IPO issuance process is more or less efcient in pricing IPOs than the online auction process. Minimizing the rst day price surge of IPOs has been an important topic in the USA since the dot-com era, when various dot-coms experienced substantial price growth from their initial price on the rst day; examples include Enel (966.9 percent), VA Linux (896.7 percent), and Sycamore Networks (612.8 percent). In recent years, the average rst day price surge for an IPO has fallen. This is partially because the exuberance and uncertainty over new technology in the dot-com boom has ended, and partially due to the initiation and conclusion of legal proceedings by the then New York State Attorney General Elliot Spitzer and the Securities and Exchange Commission (SEC) against many of the underwriting investment banks which alleged that the investment banks had manipulated IPO prices during the dot-com era to benet their preferred clients. Despite this, mispricing has continued[1], although on a smaller scale than was seen during the dot-com boom[2]. This paper compares the performance of the traditional IPO process in pricing new issues with the performance of the online auction process in the USA, as well as provides some possible explanations for the results. In the traditional IPO book building process, the underwriting investment banks take the issue on a road show to various possible investors (often large mutual funds or preferred clients of the investment bank) and build a demand curve of possible prices for the IPO based upon the indications of interest that they receive from the investors in terms of the price that they are willing to pay for the IPO and the number of shares that they are willing to buy. Based on demand curve developed through this book building process, the underwriting investment banks determine an offer price for the IPO. The offer price is usually about 7 percent below the price at which the underwriters purchase the shares of the IPO from the issuing company. The underwriters then resell the shares of the IPO at the offer price to the institutional investors and preferred clients of the investment bank who assisted them in pricing the IPO through the book building process. The 7 percent spread between the price at which the underwriters buy the IPO from the issuing company and the offer price at which they resell the issue to their preferred clients represents the prots for the underwriters[3]. When the IPOs price jumps up from the offer price on the rst day of trading, the investors receiving the initial allotments of the IPO benet because they were allocated the shares at the offer price by the underwriting investment bank. This underpricing , i.e. setting the offer price too low such that it jumps up on the rst day suggests that the IPO may not have been accurately priced and represents money left on the table to the issuing rm, because they sold the issue to the underwriters at 7 percent below the offer price and therefore could have obtained greater proceeds if the offer price had been higher and had more accurately reected the value of the company. Indeed, Aggarwal et al. (2002) nd a stronger positive relationship between institutional allocation and underpricing than would be predicted by the premarket demand. Consequently, developing a more accurate offer price, which will minimize mispricing, is important (the Wall Street Journal, 2005a; Ian, 2005). Loughran and Ritter (2004) discuss some of the causes for the underpricing in the traditional IPO process during the dot-com era. They argue that executives of issuing rms were more willing to accept underpricing of shares of their rm because they were
Efciency of IPO processes
269
IJMF 5,3
270
allocated shares of other new IPOs, so that they could benet from the underpricing (and the resulting rst day price increase) in these IPOs a practice known as spinning. Loughran and Ritter (2004) also discuss the greater importance of analyst coverage during the 1990s for IPOs (relative to other time periods), which made issuing rms more likely to go to investment banks with the top analysts[4]. The development of the Dutch auction process, OpenIPO.com (which was developed by Hambrecht, who had previously co-founded the investment bank Hambrecht & Quist (the San Jose Business Journal, 1999)) was the mechanism for the issuance of Googles IPO and represents one of the most recent of the attempts to efciently price IPOs so that the underwriter receives a more accurate reection of the economic value of the rm. The online auction process debuted in February 1999 with the issuance of the IPO for Ravenswood Wineries. As numerous press articles have noted, the Dutch auction method would supposedly minimize or eliminate the rst day price surge in IPOs[5] by developing an offer price which is a more accurate reection of the companys value through an auction rather than through the book building process. Under this method, bidders post the price that they are willing to pay and the number of shares that they wish to purchase. The auction is open for about two weeks, and bidders can change or cancel their bids until the close of the auction. After the auction ends, OpenIPO.com assembles the bids in order from highest to lowest, and then sets the offer price. All bidders who bid above this price have their bids satised at this offer price, while those who bid below it do not have their bids satised. This offer price is usually the highest price at which all of the shares are sold, based on the cumulation of the bids[6]. If the number of shares in the bids above the offer price exceeds the total number of shares available in the offering, then shares are allocated on a pro-rata basis[7]. Unlike the traditional method, in which the individuals participating in the setting of the offer price are institutional investors and/or preferred clients of the underwriters, bidders in the auction method can include smaller investors. The role of the investment bank as the middleman is minimized. Generally, Hambrechts usual fee ranges from 4 to 6 percent of the proceeds and is sometimes even less than that, which is lower than the 7 percent fee charged by traditional underwriters (the Wall Street Journal, 2005a). But, is the online auction process likely to solve the problem of underpricing and of the resulting rst day price surges for IPOs, or will it generate additional problems? This analysis compares the rst day mispricing of the online auction process through OpenIPO.com with that of the IPO process using traditional underwriters and examines whether the rst day mispricing from the auction process persists over time. The paper is organized as follows: Section II describes the data and provides summary statistics indicating differences in the rst day mispricing of online auction IPOs and traditional IPOs over time, as well as compares the characteristics of IPOs debuting using each process in terms of their size, age, and industry sector. Section III presents empirical results comparing the performance of the online auction IPO process and the traditional IPO process by matching IPOs issued in the same year and in the same industry sector by the two processes, as well as by using regression models controlling for the industry sector and the size of the IPOs. Section III also examines whether the rst day mispricing from the online auction process persists over time. Section IV presents some possible explanations for the lack of success of the online auction process in efciently pricing IPOs and provides some possible solutions. Finally, Section V presents the conclusions.
Little work has been done on the empirical performance of IPO auctions; this analysis contributes to the literature in that it empirically examines the performance of the US online auction process and compares it to the results of the traditional IPO process. Kandel et al. (1999) examine the demand schedules and elasticities of 27 Israeli IPOs between 1993 and 1996, which debuted using a uniform price auction and nd underpricing and an average abnormal return on the rst day of trading of 4.5 percent. Biais and Faugeron-Crouzet (2002) examine the Mise en Vente auction-like mechanism used in France using data on 92 IPOs between 1983 and 1996 and nd average underpricing of about 13 percent. They also argue that the Dutch auction format can lead to tacit collusion on the part of bidders and can lead to underpricing. Several papers examine optimal mechanism design, including Biais and Faugeron-Crouzet (2002) and Jagannathan and Sherman (2005). Sherman (2000) models how in a repeated setting, underwriters can lower the excess returns of uninformed investors and thus lower the degree of underpricing, as well as discusses how hybrid book building can lead to greater underpricing than straight book building. Carter et al. (2000) focus on the use of the internet to distribute an IPO to small investors, but not to price it (as the online auction process does). They explore the characteristics of IPOs which are underwritten by traditional underwriters, but which allocate some portion of the initial shares to an online investment bank, such as Wit Capital, which can then distribute them to small investors. Their analysis examined the characteristics and behavior of 27 IPOs debuting between July and December 1998, so both the dataset and the date of publication of the paper largely pre-dated the online IPO auction pricing process, although the paper notes that the online auction process had recently been developed. Their ndings suggested that rms using the internet for distribution were larger, used more reputable underwriters, and had younger CEOs. They had greater volume and volatility than IPOs that chose the traditional distribution mechanism. They also had higher returns, which the authors suggested was due to the underwriter discounting the offer price to reect the riskiness inherent in this new method of distributing the IPO or due to the rms having younger CEOs. Differences in information sets for investors provide explanations for IPO underpricing in the traditional process. Rock (1986) develops a winners curse model in which informed investors only invest in issues that they know are underpriced, thus leading to greater underpricing by underwriters. Beneviste and Spindt (1989) develop an information-gathering model, in which underpricing rewards informed investors for providing information on demand, price, and quantity to the underwriters. Welch (1989) provides a signaling model in which high quality rms tend to underprice. Lee et al. (1999), using data on the Singapore stock exchange, show that large investors with better information request participation in IPOs with higher initial returns, thus being one of the rst papers to document that larger investors have an informational advantage[8]. Other papers, such as Field (1997), suggest that institutional investors may be better informed about IPO value than other investors. Carter and Manaster (1990) discuss how IPOs with higher returns have more informed investor capital. The analysis in this paper nds that the online process did not exhibit signicantly lower underpricing than the traditional process, as had been initially intended by its founders. In Section V, the paper discusses how one reason for the auction underpricing may be that there is greater participation of smaller investors, who are less informed, and so are less able to accurately price an IPO.
271
IJMF 5,3
272
Another strand of the literature examines how IPO rms perform following their debut, such as Jain and Kini (1994), Ritter (1991), and Teoh et al. (1998). Krignam et al. (1999) examine underwriters pricing errors and show that rst-day winners continue to be winners and that rst-day dogs continue to be dogs. Section IV of this analysis examines whether the mispricing of online auction IPOs continues to exceed the mispricing of traditional IPOs over time (one week, two weeks, four weeks, 60 days, 90 days, 180 days, one year after debuting) and nds that a greater degree of mispricing relative to traditional IPOs persists for 50-80 percent of online auction IPOs. II. Data and summary statistics This section describes some of the differences between the online auction process and the traditional IPO issuance process, both in terms of their comparative performance over the years, and in terms of their characteristics, such as the size and the industry sector of the rms using each of the processes to issue an IPO. The data, which are from SDC Platinum, consist of all IPOs issued by both the traditional underwriters and the online underwriter in the USA between February 1999 and June 2005. Table I subdivides the IPOs into those issued by the online auction process and those issued by traditional underwriters. Traditional underwriters are then subdivided into low volume underwriters, medium volume underwriters, and high volume underwriters. This paper denes low volume underwriters to be underwriting investment banks which served as primary lead bookrunner on under 14 IPOs between 1999 and 2005, medium volume underwriters to be underwriting investment banks which served as primary lead bookrunner for between 14 IPOs and 40 IPOs, and high volume underwriters to be underwriting investment banks which served as primary lead bookrunner for over 40 IPOs. Table I shows summary statistics, by underwriter category, on the total number of underwriters in that category, the number of IPOs, the percentage of total IPOs underwritten by each category of underwriter, and the average principal amount, size, and rst day price surge of an IPO underwritten by an underwriter in each category. Several interesting observations can be drawn from Table I. First, despite the publicity surrounding the process, the number of IPOs issued by the online auction process is small relative to the number issued by the traditional process. Within the traditional process, there are comparatively few high and medium volume underwriters, but they issue collectively over 80 percent of the IPOs. Second, based on these averages, the rst day price increase of an IPO issued through the auction process is 29-30 percent, which is lower than the average for all of the traditional underwriters at 38 percent. This is consistent with the intention of the auction process to minimize mispricing. Nevertheless, when one segments the traditional underwriters into low, medium, and high volume underwriters, the auction process is more efcient than high volume underwriters (48 percent), as efcient as a medium volume underwriter (29 percent), and less efcient than low volume underwriters (10 percent). One would think that low volume underwriters, since they have less experience, should be less efcient at pricing IPOs than high volume underwriters. Nevertheless, the high volume underwriters were the investment banks who paid nes for allegedly intentionally underpricing IPOs to benet their preferred clients, who received the initial allocations. Third, despite the fact that the degree of rst day mispricing is much greater for the auction process than for low volume underwriters, auction IPOs tended
Number of underwriters 1 1 148 13 11 125 311 17.75 2.5 57.5 281 16.04 25.5 101.6 1,146 65.41 88.2 167.6 1,738 99.2 11.7 136.31 13 0.74 13 49.15 3,851,129 11,700,000 14,600,000 7,806,649 5,841,266 14 0.80 14 164.4 4,976,409
Number Percentage of of IPOs total IPOs
Average number of IPOs
Average principal Average size (number Average rst day amount ($ million) of IPO of shares) of IPO price increase 0.290 0.30 0.383 0.482 0.289 0.101
Online process (with Google) Online process (without Google) All traditional underwriters High volume traditional underwriters Medium volume traditional underwriters Low volume traditional underwriters
Notes: This table provides summary statistics on number of underwriters, number of IPOs, percentage of total IPOs, average number of IPOs per underwriter, average principal amount, average number of shares per IPO, and average rst day percentage price increase; the table compares the entire sample of traditional IPO underwriters with the online auction process, and then segments the traditional IPO underwriters into low volume, medium volume, and high volume traditional underwriters Source: SDC Platinum
273
Table I. Summary statistics on various types of underwriters
IJMF 5,3
274
to be similar to IPOs underwritten by low volume underwriters in terms of the number of shares in the average IPO and in terms of the average principal amount raised in the IPO (if one excludes Google in examining the auction process). Fourth, Google raised an unusually large amount for an auction IPO at almost $2 billion. Table II compares the online auction IPOs and the IPOs issued through the traditional process based on industry sector as measured by the two-digit SIC code. Almost two third of traditional IPOs were in the same industry area as the IPOs issued by the auction process. Both processes had over 20 percent of their IPOs in the software/online services area, due to the growth of the internet at the time, and both had 7-8 percent of their IPOs in the area of manufacturing telecomm equipment and energy components. The auction process, however, had a greater emphasis than the traditional process in the soft drink/coffee industry area, the biotech and pharmaceutical area, the restaurant area, the online retail and catalog area, and the online nancial services and loan services area. Table III shows the industry breakdown for traditional IPOs across all industry sectors (not just the ones overlapping with auction IPOs). Two of the top four industry areas for auction IPOs were also in the top four sectors for traditional IPOs (SIC 7300 and 2800). Six of the top ten industry sectors for traditional IPO issuance were the same industry areas as for auction IPOs. Nevertheless, when one compares IPOs issued in the same industry sector between the traditional process and the auction process (Table II), the size of the issue in terms of number of shares is much smaller. The average number of shares of an auction IPO ranges from 5 percent of the average number of shares of a traditional IPO (SIC 2800) to over 50 percent (SIC 6700). Similarly, the average principal amount raised in an auction IPO ranged from 12 percent of the amount raised by the traditional IPO (SIC 5800) to 66 percent of the amount raised in a traditional IPO (SIC 6200). In many industry areas, the principal amount raised in auction IPOs was 30-50 percent of the amount raised in traditional IPOs (SIC 2800, 3600, 5900, 6100, and 6700). The only sector in which the share size and the average principal amount raised was greater for auction IPOs than traditional IPOs was in SIC 7300 when Google is included. If Google is excluded (see note in Table II), then the average number of shares for auction IPOs in SIC 7300 is about 58 percent of the average number of shares of a traditional IPO in SIC 7300 and the average principal amount for auction IPOs is 66 percent of the amount raised by traditional IPOs. In short, while there was some overlap in industry area between online auction IPOs and traditional IPOs, auction IPOs tended to be smaller in terms of the principal amount raised and the number of shares in the issue. Table IV compares the percentage of technology IPOs issued in each year through the online auction process and through the traditional IPO process. The denition of a technology IPO follows Loughran and Ritter (2004), and includes IPOs in SIC codes 3571, 3572, 3575, 3577, 3578, 3661, 3663, 3669, 3671, 3672, 3674, 3675, 3677, 3678, 3679, 3812, 3823, 3825, 3826, 3827, 3829, 3841, 3845, 4812, 4813, 4899, 7371, 7372, 7373, 7374, 7375, 7378, and 7379. In four of the seven years, the online auction process had no tech IPOs and, consequently, had a smaller percentage of IPOs debuting in this more broadly dened technology sector than in the traditional process. In three of the seven years, the online auction process had a higher percentage of technology IPOs than the traditional process. One explanation for the greater degree of mispricing in the online auction IPO process relative to the traditional process (seen in the averages in Table I) could have been that a greater percentage of technology IPOs were debuting using the
Two-digit SIC code (industry area) 0.63 14.29 35,700,000 (111.18) 2,150,000 (18.45)
Business description for companies going public in this SIC code between 1999 Percentage of traditional and 2005 IPOs in this industry area Percentage of online IPOs in this industry area Average number of shares of Average number of shares of traditional IPOs (principal online IPOs (principal amount in $s million) amount in $s million)
2000
2800 7.88 7.14 12,200,000 (127.85)
6.79
14.29
7,656,754 (88.1)
3,950,000 (33.45) 3,500,000 (42)
3600
5800 5900
0.81 2.13
7.14 14.29
8,539,446 (136.31) 5,337,011 (80.88)
2,000,000 (16) 2,600,000 (34.9)
6100 1.84 18.47 26.70 66.40 21.43 100 7.14 7.14
1.15
7.14
16,500,000 (94.39) 21,400,000 (212.24) 18,900,000 (254.05) 5,551,762 (74.32)
3,052,174 (35.1) 7,612,500 (140.8) 10,000,000 (130) 8,701,684a (588.1)a
6200
6700
7300
Meat packing plants; manufacturing of canned foods, soft drinks, coffee, fruit juices, etc. Manufacturing of pharmaceuticals and biotech drugs Manufacturing of electrical and energy components and telecommunications equipment Own and operate restaurants Own and operate various service businesses: sporting goods, catalog companies, online retail Loan services (payday loans, student loans), mortgage banks, credit card services Provide brokerage and investment banking services; online nancial services Bank holding companies and closed-end investment funds Development of software and online services
Total percentage of IPOs
Notes: aNote that these numbers include Google, if one excludes Google, then the average number of shares in SIC 7300 issued by the online auction process was 3,250,000 shares and the average principal amount raised was $49.15 million Source: The source of the underlying raw data in the table is the SDC Platinum database
275
Table II. Comparison of traditional IPOs and online IPOs by industry, 1999-2005
IJMF 5,3
Two-digit SIC code 7300 6700
Business description for companies going public in this SIC code between 1999 and 2005 Development of software and online services Bank holding companies and closed-end investment funds Manufacturing of electrical and energy components and telecommunications equipment Manufacturing of pharmaceuticals and biotech drugs Provide wireless internet and telecomm service, broadband, VOIP; own and operate radio, cable, and TV stations Manufacturing medical and biotech equipment Drug R&D services, cancer treatment research Health services (including insurance) Own and operate various services: sporting goods, online retail, catalog companies Own and operate various service businesses: sporting goods, catalog companies, and online retail Provide brokerage and investment banking services; online nancial services Provide brokerage and investment banking services; online nancial services Manufacturing surface mining machinery, data storage, ber optics, printing equipment, and ight info computers Oil and gas exploration and production Commercial banks, savings and loans, and associated holding companies Electric and gas utilities; electrical and geothermal power Loan services (payday loans, student loans), mortgage banks, credit card services Loan services (payday loans, student loans); mortgage banks; and credit card services Own and operate hospitals, surgery centers, laboratories Shipping companies; water transportation services Own and operate schools, educational services, educational software, etc. Wholesaler of autos, computers, and roong materials Own and operate restaurants Meat-packing plants; canned foods, soft drinks, coffee, fruit juices Manufacturing aircraft, railroad cars, tanks, snowmobiles, and auto parts Real services Motion picture services (incl. distribution); cable TV services
Percentage of traditional IPOs in this category 26.70 18.47 7.88 6.79 5.98 4.09 2.47 2.42 2.13 2.13 1.84 1.84 1.78 1.61 1.38 1.21 1.15 1.15 1.04 0.92 0.86 0.86 0.81 0.63 0.58 0.46 0.46 (continued)
276
3600 2800 4800
3800 8700 6300 5900 5900 6200 6200 3500
1300 6000 4900 6100 6100 8000 4400 8200 5000 5800 2000 3700 Table III. Industry breakdown for traditional IPOs, 1999-2005 6500 7800
Two-digit SIC code 3300 4600 4500 1200 5100 5600 4200 1500 5700 7000 3900 6400 5400 2700 1000 7900 3000 3200 7200 2500 2300 5200 7500 3100 1600 3400 1400 2200 740 1700 180 2600 5500 5300 2900 750 490
Business description for companies going public in this SIC code between 1999 and 2005 Steel and aluminum production Crude oil pipeline operator Airlines Coal mining services Manufacturing herbal drugs, dietary supplements; pharmacies Clothing retail stores Provided trucking, relocation, and courier services Residential and commercial construction Furniture and appliance stores; video game stores; online lm stores Own and operate casino hotels and hotel casinos Manufacturing collectibles candles, golf equation Insurance agents and brokers Convenience stores; bread stores, doughnut and ice cream stores Publishing Metal, gold, and palladium mining Operate gyms; golf instruction Manufacturing plastic containers and products Manufacturing glass containers, ready-mix concrete, and chemicals Misc coin-operated laundries; tax preparation services, and weight reduction services
Percentage of traditional IPOs in this category 0.40 0.40 0.40 0.40 0.40 0.40 0.40 0.29 0.29 0.29 0.29 0.29 0.23 0.23 0.17 0.17 0.17 0.17 0.17
277
, 0.12
Table III.
process and that technology IPOs were more underpriced because they were in developing industries whose potential was harder to estimate at the time. Nevertheless, the data do not support this hypothesis because there was often a smaller percentage of technology IPOs debuting using the auction process than in the traditional process.
IJMF 5,3
1999 2000 2001 2002 2003 2004 2005
Percentage of technology IPOs through traditional underwriters 53.97 57.25 20.64 10.47 13.18 21.12 15.32
Percentage of technology IPOs through online underwriters 66.6 0 0 0 50 33 0
278
Table IV. Yearly comparison of the percentage of technology IPOs issued through the traditional IPO process and through the online auction process: 1999-2005
Notes: The yearly percentages of technology IPOs issued through traditional underwriters and the online auction process were derived by the author from underlying raw data on SIC codes from SDC Platinum, the denition of technology IPOs includes IPOs in SIC codes 3571, 3572, 3575, 3577, 3578, 3661, 3663, 3669, 3671, 3672, 3674, 3675, 3677, 3678, 3679, 3812, 3823, 3825, 3826, 3827, 3829, 3841, 3845, 4812, 4813, 4899, 7371, 7372, 7373, 7374, 7375, 7378, and 7379 Source: Loughran and Ritter (2004, Appendix D)
The rms which went public using the online auction process had characteristics similar to rms that went public using the traditional process. The author obtained data on the age of the issuing rm, sales and net income of the rm in the 12 months prior to its IPO, and on the age and salary of its CEO by examining the S-3 that each rm, which debuted using the online auction process, had led with the SEC. IPOs debuting using the online auction process were about nine years old, on average. If one excludes Morningstar and Peets Coffee, which were over 20 years old when they went public, the average age was 6.3 years. Table V compares the age of a rm using the traditional process with the age of a rm using the online process on a year-by-year basis. In four of the seven years, rms using the online auction process were actually slightly older than rms using the traditional process. Examination of the S-3s also indicated that the base salary of the CEOs averaged $185,294 and the age of the CEO averaged around 51 years. The Ernst & Young IPO Study found that the average age for CEOs of rms who immediately qualied for listing in the Russell 2000 Index following their IPO using the traditional process was between 50 and 54 years and that the median age of the rm itself was eight to nine years old. This is similar to the pattern seen in this analysis for rms using the online auction process.
Median age (traditional) 1999 2000 2001 2002 2003 2004 2005 4 6 12 15 11 8 11 Median age (online) 7 7 18 5 7 6 13.5
Table V. Comparison of median age of rms using the online auction process and the traditional process: 1999-2005
Notes: Data for the median age of online issuers comes from the aggregation of information on the age of each online issuer obtained by the author through examining each companys S-3 ling with the SEC; composite median age data on traditional issuers are from the Field-Ritter dataset on founding dates for rms going public in the US during 1975-2006
Table VI compares the distribution of sales for a rm debuting using the traditional IPO process in the year preceding its IPO during the 1980s, the 1990s, the height of the dot-com bubble, and the 2000s, with the distribution of sales for rms debuting using the online auction process. The table suggests that the online auction process had a greater percentage of smaller rms with annual sales of under $10 million than the traditional IPO process at any period, with the exception of the 1999-2000 period (the peak of the dot-com era). The online process had similar percentages to the traditional IPO process in terms of the percentage of rms with sales between $10 and $20 million, between $20 and $50 million, and between $50 and $100 million. The online auction process had a smaller percentage of rms with sales greater than $100 million and greater than $200 million than the traditional process in every period, except during the peak of the dot-com period, when the percentages were similar. The conclusion that the online auction process tended to have a greater fraction of smaller rms than the traditional process, the same fraction of intermediate-size rms, and a smaller fraction of really large rms is consistent with the pattern in the data on the number of shares issued for online auction IPOs relative to the number of shares of IPOs issued by the traditional process (Tables I and II). Table VII compares the average rst day price surge for online auction IPOs with those of low, medium, and high volume traditional underwriters across the entire ve-year period, and on a yearly basis. The rst row reects the ndings previously discussed for Table I: although the average rst day price increase for online issuers is higher than the average for traditional issuers, when traditional issuers are decomposed by volume of IPOs issued, the online process is less efcient than low volume underwriters, as efcient as medium volume underwriters, and more efcient than high volume underwriters. Second, the table compares the rst day price surge by type of underwriter on a yearly basis the degree of mispricing (as measured by absolute value) for online underwriters exceeded that of low volume, medium volume,
1980-1989 (traditional, %) Sales between $0 and $10 million Sales between $10 and $20 million Sales between $20 and $50 million Sales between $50 and $100 million Sales between $100 and $200 million Sales greater than $200 million 19.5 12.4 23.3 17.4 11.8 14.3 1990-1998 (traditional, %) 19.4 11.0 22.2 16.6 12.8 18.0 1999-2000 (traditional, %) 38.9 16.3 17.6 10.4 6.7 10.1 2001-2007 (traditional, %) 15.3 4.7 14.5 15.6 13.5 36.5 1999-2005 (online, %) 36.36 9.09 18.8 18.18 9.09 9.09
279
Notes: The chart compares the sales in the 12 months preceding an IPO for rms using the traditional IPO process and rms using the online auction IPO process; data for the sales of rms using the online auction process comes from the aggregation of information on the sales of each online issuer obtained by the author through examining each companys S-3 ling with the SEC Source: Data on the sales for rms using the traditional IPO process comes from Ritter (2008)
Table VI. Comparison of average annual sales for IPOs debuting using the traditional process and IPOs debuting using the online process
280
IJMF 5,3
Average rst day price surge by year and SIC code 0.383 0.589 0.542 0.890 0.057 0.068 0.079 0.013 0.290 1.098 2 0.216 0.318 0.002 0.0752 0.037 0.042 0.383 0.586 0.544 0.900 0.057 0.067 0.080 0.013 0.101 0.142 0.102 0.231 0.160 0.110 0.0423 20.089
Total During During During During During During During
Notes: This table compares the average rst day price surge between the total sample of traditional underwriters and the online auction process, as well as among three types of traditional underwriters low volume, medium volume, and high volume underwriters; the table examines this for the entire 1999-2005 period, as well as on a yearly basis Source: SDC Platinum
Table VII. Summary statistics of rst day price surges by year and type of underwriter Average rst day price surge (issued through online auction) Average rst day price surge (issued through traditional process) Average rst day price surge (issued by low volume traditional underwriters) Average rst day price surge (issued by medium volume traditional underwriter) 0.289 0.387 0.544 0.137 0.038 0.030 0.079 0.055 Average rst day price surge (issued by high volume traditional underwriters) 0.482 0.780 0.635 1.111 0.042 0.068 0.092 0.0410
Average rst day price surge (total sample)
1999 2000 2001 2002 2003 2004 2005
and high volume underwriters 42.8 percent of the time (in three of the seven years). Consequently, this suggests that the online process may not be as efcient in pricing as was originally intended. Moreover, an analysis of the average rst day price surge per underwriter over the ve-year period across all underwriters indicates that the average rst day price surge for IPOs issued by WR Hambrechts OpenIPO.com online auction mechanism met or exceeded the rst day price surge for 82.4 percent of the primary lead bookrunners (108 out of 131) in the sample. The results in Table VII also suggest that while the rst day surges in an IPOs price are important, the problem that the online auction mechanism was partially developed to solve has lessened. The average rst day price increase has dropped substantially from the 55.4-92.5 percent range during 1999-2001 into the 6.3-8.6 percent range of 2002-2004. This gradual reduction in the magnitude of rst day price surges over time may be a reection of more careful pricing by some underwriters in the wake of substantial litigation concerning their alleged manipulation of IPOs, as discussed in the introduction[9]. Many of the investment banks with the top analysts and which engaged in spinning and laddering included some of the high volume traditional issuers in my analysis, which explains why their underpricing was so substantial in many cases, the executives of the issuing rms were willing to accept underpricing in return for shares in other hot IPOs or in return for favorable coverage from the top analysts in the industry. A second explanation for the high degree of underpricing during the dot-com era and the subsequent reduction in underpricing following the collapse of the bubble may be that many of the technology IPOs during the dot-com boom were in emerging industries, and, since most investors had a supercial understanding of the fundamentals and technology in these industries, they had difculty in accurately pricing the issues and were inuenced by herd behavior. Indeed, at the peak of the dot-com bubble in 1999, over half of the IPOs issued through the online process and the traditional process were technology IPOs, using Loughran and Ritters (2004) criteria, as was evident in Table IV. In short, the summary statistics in this section yield several key ndings. First, the degree of underpricing has fallen over time, possibly due to the role of litigation in eliminating anticompetitive practices, such as spinning, as detailed in Loughran and Ritter (2004), and possibly due to the declining fraction of technology IPOs. Second, IPOs issued through the online auction process had a greater degree of average underpricing than low volume issuers and a lesser degree of average underpricing than high volume issuers. This suggests that the online auction process may not have been successful in its objective to minimize underpricing, although it did have lower average underpricing than traditional issuers when the three types are aggregated. Are there differences in the types of rms choosing the online auction process relative to the traditional process for their IPO? Almost two third of traditional IPOs were in the same industry area as the IPOs issued by the online auction process, although the auction process had a greater emphasis in the soft drink/coffee industry area, the biotech and pharmaceutical area, the restaurant area, the online retail and catalog area, and the online nancial services and loan services area. In four of the seven years, the online auction process had no tech IPOs and, consequently, had a smaller percentage of IPOs debuting in this more broadly dened technology sector than in the traditional process, while in three of the seven years, the online auction
281
IJMF 5,3
282
process had a higher percentage of technology IPOs. Second, the IPOs debuting using the online auction process was smaller than IPOs debuting using the traditional process. This was true even if one matched the industry areas and compared the number of shares in an auction IPO and in a traditional IPO, as in Table II. The average number of shares of an auction IPO ranged from 5 percent of the average number of shares of a traditional IPO to over 50 percent. Similarly, the average principal amount raised in an auction IPO ranged from 12 percent of the amount raised by the traditional IPO to 66 percent. Not surprisingly, when the author examined sales in the 12 months preceding an IPO, the online auction process tended to have a greater fraction of smaller rms (as measured by sales) than the traditional process, the same fraction of intermediate-size rms, and a smaller fraction of larger rms. Age of the debuting rm did not differ substantially between rms choosing the auction process and rms choosing the traditional process: in four of the seven years, the auction IPOs were slightly older, and in three of the seven years, they were somewhat younger than IPOs using the traditional process. III. Empirical results and sensitivity analysis This section analyses whether the rst day price surge exhibited by online auction IPOs is signicantly higher than the rst day price surge exhibited by traditional IPOs, controlling for various factors, such as the year, and several factors capturing the characteristics and risk of these IPOs. These factors include the industry of the issuing rm (as measured by the two-digit SIC code), as well as the size of the rm (as measured by the number of shares in its IPO). Other measures describing the characteristics of the rms going public were highly correlated with the industry and the size of the issue and so were not helpful in describing the differences in underpricing. For example, smaller rms as measured by sales or net income were likely to have smaller IPOs (as measured by the number of shares) than larger rms, so the number of shares variable picked up most of these differences, as was suggested by the summary data in Section II. Also, the number of shares was important to include because IPOs with a smaller number of shares could have greater underpricing if there were alot of pent-up demand relative to the supply of shares offered, which then would manifest itself when the IPO began to trade. The industry variables were important in describing risk because many of the traditional and online IPOs during this period were technology IPOs, and were often smaller, younger companies. Age of the company was not helpful in the regression specications during this period both because, as described in Section II, there were not signicant differences in age between the online IPOs and the traditional IPOs, and also because the industry codes often picked up the same type of information as age variables. The lack of signicance of age was also seen in Loughran and Ritter (2004) during the 1999-2000 period and the 2001-2003 period[10]. Consequently, the regression results reported in the tables focus on the specications using industry xed effects by two-digit SIC code and IPO size in terms of number of shares to capture the rm-specic characteristics, since other iterations using similar metrics (age, etc.) yielded similar results. As noted in the earlier tables, signicantly more traditional IPOs were issued than auction IPOs. Consequently, Table VIII matches each auction IPO with a traditional IPO issued in the same year and in the same industry area as measured by the three-digit SIC code and compares the degree of mispricing[11]. Some auction IPOs had
Issue date 9.3 59.25 2 14.53 7.98 2 83.31 2 6.87 2 4.30 2 89.86 245.4 7375 1
Underwriter
SIC code
First day price increase (%) 0 100 50 77.7 35.3 0 43.3 8.78 97.5 70
Difference between the auction IPO mispricing and the traditional IPO mispricing (%)
Percentage of IPOs issued in the same year and in the same three-digit SIC code with less mispricing than the auction IPO (%)
April 8, 1999 March 30, 1999 January 25, 2001 May 2, 2001 August 5, 2004 2836 3674 5812 6798 7372 7379
2084 2086 2095 2095 2834
October 29, 2003
May 18, 2000
May 2, 2001
March 17, 2004
June 22, 1999 December 8, 1999
August 18, 2004
Hambrecht IPO: Ravenswood Winery Comparable Traditional IPO: Pepsi Bottling Group Hambrecht IPO: Peets Coffee & Tea Comparable traditional IPO: Coffee Holding Co. Hambrecht IPO: New River Pharmaceuticals Average of traditional IPOs in 2004 for SIC 2830s Hambrecht IPO: Genitope Corp Average of traditional IPOs in 2003 for SIC 2830s Hambrecht IPO: Nogatech Average of traditional IPOs in 2000 for SIC 3670s Hambrecht IPO: Briazz Average of traditional IPOs in 2001 for SIC 5810s Hambrecht IPO: Sunset Financial Resources, Inc. Average of traditional IPOs in 2004 for SIC 6790s Hambrecht IPO: Salon.com Hambrecht IPO: Andover.net Average of traditional IPOs in 1999 for SIC 7370s Hambrecht IPO: Google Average of traditional IPOs in 2004 for SIC 7370s
3.61 2 5.69 63.25 4 2 6.25 8.28 11.1 3.125 2 21.58 61.73 0.375 7.24 2 0.769 4.22 2 4.76 330.5 85.1 18 17
Notes: This table compares IPOs which debuted using the OpenIPO auction method with IPOs which debuted using the traditional method in the same year and in the same three-digit SIC code, note that this table does not include the following OpenIPO IPOs because there were no other comparable traditional IPOs in their three-digit SIC code in the year of issue: RedEnvelope Overstock.com, Morningstar, Bo Holding
283
Table VIII. Comparison of auction IPOs and traditional IPOs in the same three-digit SIC code issued in the same year
IJMF 5,3
284
more than one traditional IPO issued in the same three-digit SIC code and in the same year, so their performance was compared to the average of the traditional IPOs issued in the same three-digit SIC code in the same year. The fth column shows the difference in the percentage mispricing between the auction IPO and the traditional IPO in four tenth of the IPOs (40 percent), the auction IPOs degree of mispricing was more than 7 percent greater than the percentage price surge of the comparable traditional IPO. The last column in Table VIII shows the percentage of traditional IPOs issued in the same three-digit SIC code and in the same year which had greater mispricing than the auction IPO. For 60 percent (six out of ten) of the auction IPOs, over 40 percent of the traditional IPOs issued in that year and in that three-digit SIC area had greater mispricing. This suggests that the online auction mechanism is not as efcient in minimizing the rst day price surge as some have thought. Table IX shows the results of the regression estimating the rst day price surge for an IPO as a function of SIC code dummies (for industry effects), year dummies (for year effects), a dummy variable for whether the IPO was issued through the online auction process, and variable for the number of shares issued in each IPO (to control for the size of the issue and of the rm) during the 1999-2005 period. The results indicate that online auction IPOs do not have statistically signicantly higher or lower rst day price surges than traditional IPOs, when controlling for year/industry effects, and the size of the issue. The industry effects were jointly signicant (F 3.78, p 0.0000), the year effects were jointly signicant (F 4.72, p 0.0001), and the coefcient for the number of shares outstanding was signicant. The models in Tables X-XII include dummy variables for the various types of traditional underwriters low volume traditional underwriters, medium volume traditional underwriters, and high volume traditional underwriters with one of them left out to serve as the base case. The model in Table X includes dummy variables for medium volume underwriters and high volume underwriters, in addition to dummy variables for year effects, industry effects, and whether the IPO debuted using the online auction process. The results indicate that online auction IPOs do not have a statistically signicantly higher or lower rst day price surges than traditional IPOs, when controlling for year and industry effects, relative to the base case. The results also suggested that high volume underwriters had statistically signicantly higher rst day price surges than low volume underwriters (the base case). Although the rst day price surges are not statistically signicantly different between online auction IPOs and low volume underwriters (the base case) or between medium volume underwriters and the base case, the sign on the online auction IPO coefcient is positive and the sign for the coefcient for medium volume underwriters is also positive. The industry effects were jointly signicant (F 4.40, p 0.0000), the year effects were jointly signicant (F 2.43, p 0.0296), and the medium and high volume effects were jointly signicant (F 2.88, p 0.0563). The model in Table XI differs from the model in Table X in that the dummy variables for low volume underwriters and high volume underwriters are included such that the base case is medium volume traditional underwriters, rather than high volume traditional underwriters (Table IV). Again, the results indicate that online auction IPOs do not have a statistically signicantly higher or lower rst day price surge than traditional IPOs, when controlling for year and industry effects, and the type of traditional underwriter. The sign on the coefcient for online auction IPOs
Variable Dummy 1999 Dummy 2000 Dummy 2002 Dummy 2003 Dummy 2004 Dummy 2005 Online auction dummy Number of shares SIC 6300 SIC 1300 SIC 3600 SIC 3800 SIC 5100 SIC 6700 SIC 7300 SIC 2800 SIC 3500 SIC 6000 SIC 1500 SIC 1000 SIC 3700 SIC 3100 SIC 4800 SIC 5900 SIC 5000 SIC 5800 SIC 8000 SIC 8200 SIC 5600 SIC 8700 SIC 1600 SIC 6500 SIC 4200 SIC 4400 SIC 4700 SIC 6200 SIC 7800 SIC 7900 SIC 5700 SIC 3400 SIC 4900 SIC 6100 SIC 2200 SIC 7000 SIC 1200 SIC 3000 SIC 3900 SIC 4500 SIC 740 SIC 2000
Coefcient 0.482 * 0.291 2 0.095 2 0.353 2 0.169 2 0.258 0.526 0.000 * 2 1.672 2 0.570 2 0.378 2 0.422 2 0.452 2 1.043 2 0.154 2 0.519 0.006 2 0.605 2 0.521 2 1.149 2 0.934 2 0.812 2 1.008 2 0.421 2 0.478 2 0.610 2 0.469 2 0.725 2 0.364 2 0.453 2 0.614 2 0.553 2 1.144 2 0.523 2 0.069 2 1.270 2 0.810 2 0.363 2 0.416 2 1.432 2 1.053 2 0.993 2 0.638 2 0.799 2 0.969 2 1.182 2 0.847 2 0.847 2 0.981 2 1.771
SE 0.209 0.212 0.236 0.251 0.211 0.257 0.530 0.000 1.965 1.974 1.950 1.955 2.075 1.946 1.946 1.951 1.974 1.983 2.128 2.244 2.035 2.378 1.954 1.968 2.006 2.003 1.991 2.006 2.077 1.965 2.377 2.061 2.079 2.002 2.077 1.970 2.061 2.242 2.128 2.379 1.986 1.990 2.748 2.128 2.073 2.242 2.129 2.077 2.736 2.015 (continued)
285
Table IX. The impact of the online auction process on an IPOs rst day price surge, controlling for year effects, industry effects, and size of the issue
IJMF 5,3
Variable SIC 3200 SIC 7200 SIC 1400 SIC 5400 SIC 1700 SIC 180 SIC 2300 SIC 2500 SIC 2600 SIC 3300 SIC 4600 SIC 5500 SIC 6400 SIC 5300 SIC 5200 SIC 7500 SIC 2700 SIC 2900 SIC 750 _Cons
Coefcient 2 0.510 2 1.067 2 0.743 0.552 2 0.790 2 1.260 2 0.029 2 0.582 2 2.376 8.606 * 2 0.440 2 0.513 2 0.553 2 0.886 2 0.707 2 1.082 2 1.033 2 0.841 1.384 0.262
SE 2.242 2.239 2.378 2.172 2.744 2.744 2.379 2.379 2.745 2.083 2.075 2.746 2.126 2.744 2.378 2.379 2.172 2.746 2.744 1.935
286
Table IX.
Notes: Signicant at *5 percent; (rst day price surge)i a m (SIC code dummies)i b (year dummies)i C (online auction IPO dummy)I r (number of shares offered in each IPO)i
suggests a higher price surge for online auction IPOs relative to the base case of medium volume underwriters, but, again the coefcient is not signicant. Similarly, neither high volume underwriters nor low volume underwriters have signicantly different rst day price surges from medium volume underwriters, although the sign for high volume underwriters was positive and the sign for low volume underwriters was negative, as is consistent with the averages for low, medium, and high volume underwriters in Tables I and VII. The industry effects were jointly signicant (F 4.40, p 0.0000), the year effects were jointly signicant (F 2.35, p 0.0293), and the medium and high volume effects were jointly signicant (F 2.93, p 0.0539). The model in Table XII differs from the model in Table XI in that the dummy variables for low volume and medium volume underwriters are included such that the base case is high volume traditional underwriters, rather than medium volume traditional underwriters. Again, the results indicate that online auction IPOs do not have a statistically different rst day price surge in magnitude than high volume underwriters (the base case), when controlling for year/industry effects, and the type of traditional underwriter, although the sign on the online auction IPO (relative to the base case) is negative. The industry effects were jointly signicant (F 4.40, p 0.0000), the year effects were jointly signicant (F 2.35, p 0.0290), and the medium and high volume effects were jointly signicant (F 2.97, p 0.0514). The results also suggested that low volume underwriters have a signicantly lower rst day price surge relative to the high volume underwriter base case. Medium volume underwriters did not have a signicantly different rst day price surge (although the sign on the coefcient was negative).
Variable Dummy 1999 Dummy 2000 Dummy 2002 Dummy 2003 Dummy 2004 Dummy 2005 Auction dummy Medium volume traditional underwriter High volume traditional underwriter SIC 6300 SIC 1300 SIC 3600 SIC 3800 SIC 5100 SIC 6700 SIC 7300 SIC 2800 SIC 3500 SIC 6000 SIC 1500 SIC 1000 SIC 3700 SIC 3100 SIC 4800 SIC 5900 SIC 5000 SIC 5800 SIC 8000 SIC 8200 SIC 5600 SIC 8700 SIC 1600 SIC 6500 SIC 4200 SIC 4400 SIC 4700 SIC 6200 SIC 7800 SIC 7900 SIC 5700 SIC 3400 SIC 4900 SIC 6100 SIC 2200 SIC 7000 SIC 1200 SIC 3000 SIC 3900 SIC 4500 SIC 740
Coefcient 2 0.232 2 0.294 2 0.694 * 2 0.680 * 2 0.666 * 2 0.740 * 0.300 0.151 0.354 * 0.075 0.043 0.530 0.184 0.153 0.078 0.517 0.184 0.729 0.085 0.078 2 0.008 0.075 2 0.015 0.092 0.239 0.227 0.113 2 0.009 2 0.174 0.308 0.227 0.032 0.223 0.150 0.217 0.607 2 0.020 0.124 0.278 0.385 2 0.048 2 0.104 0.185 0.167 0.207 2 0.115 2 0.380 0.012 0.072 2 0.639
SE 0.252 0.255 0.284 0.303 0.254 0.310 0.650 0.199 0.156 2.368 2.381 2.352 2.358 2.503 2.346 2.347 2.354 2.380 2.392 2.566 2.707 2.454 2.868 2.356 2.374 2.419 2.418 2.401 2.419 2.505 2.369 2.868 2.486 2.507 2.415 2.505 2.375 2.485 2.704 2.567 2.869 2.395 2.399 3.314 2.567 2.500 2.704 2.567 2.505 3.299 (continued)
287
Table X. The impact of the online auction process on an IPOs rst day price surge, compared with medium and high volume traditional underwriters
IJMF 5,3
Variable SIC 2000 SIC 3200 SIC 7200 SIC 1400 SIC 5400 SIC 1700 SIC 180 SIC 2300 SIC 2500 SIC 2600 SIC 3300 SIC 4600 SIC 5500 SIC 6400 SIC 5300 SIC 5200 SIC 7500 SIC 2700 SIC 2900 SIC 750 _Cons
Coefcient 0.019 0.072 2 0.047 2 0.096 1.183 2 0.279 2 0.337 0.477 0.224 2 0.279 14.290 * 0.014 0.140 0.157 0.013 0.068 0.427 0.223 0.299 2.225 0.218
SE 2.430 2.705 2.699 2.868 2.620 3.309 3.309 2.870 2.868 3.309 2.500 2.503 3.312 2.564 3.312 2.869 2.870 2.619 3.312 3.312 2.338
288
Table X.
Notes: Signicant at *5 percent; (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I t (medium volume underwriter dummy)i w(high volume underwriter dummy)i
In summary, Tables IX-XII suggest that online auction IPOs do not have a statistically signicantly higher or lower price surge than traditional IPOs or any subset of traditional IPOs (low, medium, or high volume underwriters). Nevertheless, within the category of traditional underwriters, low volume underwriters have a statistically signicantly lower rst day price surge than high volume underwriters, which is surprising given that high volume underwriters have had more experience in underwriting IPOs and therefore should exhibit a lesser degree of mispricing. This may be because high volume underwriters were intentionally underpricing IPOs, which is consistent with the allegations against many investment banks by the New York State Attorney Generals Ofce during 2002 and 2003, as previously discussed. Medium volume underwriters did not have statistically signicantly different rst day price surges from high volume underwriters, although the sign on the coefcient was negative. Tables XIII-XX provide a sensitivity analysis of the regression models. They includes several different formulations, some of which include or exclude year effects, industry effects, dummy variables for type of traditional underwriter, and measures of the size of the issue, in various combinations. While the models in Tables IX-XII are the most appropriate, as will be discussed, for measuring whether online auction IPOs have a signicantly lower rst day price surge than traditional underwriters, the models listed in the table also show that, regardless of the specication, the online auction IPOs do not have a statistically signicantly lower price surge than traditional IPOs, as had been originally intended. Tables XIII, XV, XVII, and XIX provide a
Variable Dummy 1999 Dummy 2000 Dummy 2002 Dummy 2003 Dummy 2004 Dummy 2005 Auction dummy Low volume traditional underwriter High volume traditional underwriter SIC 6300 SIC 1300 SIC 3600 SIC 3800 SIC 5100 SIC 6700 SIC 7300 SIC 2800 SIC 3500 SIC 6000 SIC 1500 SIC 1000 SIC 3700 SIC 3100 SIC 4800 SIC 5900 SIC 5000 SIC 5800 SIC 8000 SIC 8200 SIC 5600 SIC 8700 SIC 1600 SIC 6500 SIC 4200 SIC 4400 SIC 4700 SIC 6200 SIC 7800 SIC 7900 SIC 5700 SIC 3400 SIC 4900 SIC 6100 SIC 2200 SIC 7000 SIC 1200 SIC 3000 SIC 3900 SIC 4500 SIC 740
Coefcient 20.232 20.295 20.695 * 20.681 * 20.667 * 20.741 * 0.143 2 0.161 0.197 0.075 0.044 0.530 0.185 0.153 0.079 0.517 0.184 0.730 0.087 0.080 20.004 0.077 20.012 0.093 0.240 0.227 0.112 20.009 20.174 0.309 0.228 0.036 0.224 0.151 0.219 0.609 20.013 0.127 0.281 0.386 20.046 20.102 0.186 0.168 0.208 20.116 20.377 0.015 0.074 20.639
SE 0.252 0.255 0.284 0.303 0.254 0.310 0.651 0.198 0.162 2.368 2.381 2.352 2.358 2.503 2.346 2.346 2.353 2.380 2.392 2.566 2.707 2.454 2.868 2.356 2.374 2.419 2.418 2.401 2.419 2.505 2.369 2.868 2.486 2.507 2.415 2.505 2.375 2.485 2.704 2.566 2.869 2.395 2.399 3.314 2.566 2.500 2.704 2.567 2.505 3.299 (continued)
289
Table XI. The impact of the online auction process on an IPOs rst day price surge with low and high volume traditional underwriters
IJMF 5,3
Coefcient 0.021 0.074 20.047 20.095 1.187 20.278 20.337 0.481 0.225 20.278 14.291 0.014 0.141 0.159 0.018 0.066 0.428 0.225 0.300 2.231 0.375
SE 2.430 2.704 2.699 2.868 2.620 3.309 3.309 2.870 2.868 3.309 2.500 2.502 3.312 2.564 3.311 2.869 2.870 2.619 3.312 3.312 2.339
290
Table XI.
Notes: Signicant at *5 percent; (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i w (high volume underwriter dummy)i
summary of the coefcients and their p-values in the various model formulations, while Tables XIV, XVI, XVIII, and XX provide a summary across the models of joint signicance for year effects, industry effects, and inclusion of dummy variables by type of traditional underwriter. Tables XIII and XIV show the statistical results and the tests for the model specications for Models I-III. All three models indicate that the online auction dummy is statistically insignicant. Model I only includes year effects, which are jointly signicant. Model II includes year and industry effects, both of which are jointly signicant. Model III includes year effects, industry effects, and a variable controlling for the size of the issue. The year effects and industry effects are jointly signicant, and the variable controlling for the size of the issue is statistically signicant. Inclusion of the variable controlling for the size of the issue increased the adjusted R 2 of the model to 0.392, which is consequently an improvement on Model I (0.009) and Model II (0.114). Model III is the basic model in Table IX and is the most effective of the basic models in estimating whether online auction IPOs have a lower rst day surge than traditional IPOs, without including variables on whether the traditional underwriters are low volume, medium volume, or high volume underwriters. Tables XV and XVI show the statistical results and the tests for model specications for Models IV-VI. Again, all three models indicate that the online auction dummy is statistically insignicant. Models IV-VI have year effects, but not industry effects or a variable for the size of the issue, and include dummy variables for whether the traditional underwriters are low volume, medium volume, or high volume.
Variable Dummy 1999 Dummy 2000 Dummy 2002 Dummy 2003 Dummy 2004 Dummy 2005 Auction dummy Low volume traditional underwriter Medium volume traditional underwriter SIC 6300 SIC 1300 SIC 3600 SIC 3800 SIC 5100 SIC 6700 SIC 7300 SIC 2800 SIC 3500 SIC 6000 SIC 1500 SIC 1000 SIC 3700 SIC 3100 SIC 4800 SIC 5900 SIC 5000 SIC 5800 SIC 8000 SIC 8200 SIC 5600 SIC 8700 SIC 1600 SIC 6500 SIC 4200 SIC 4400 SIC 4700 SIC 6200 SIC 7800 SIC 7900 SIC 5700 SIC 3400 SIC 4900 SIC 6100 SIC 2200 SIC 7000 SIC 1200 SIC 3000 SIC 3900 SIC 4500 SIC 740
Coefcient 20.234 20.296 20.697 * 20.682 * 20.668 * 20.742 2 0.056 2 0.360 * 2 0.204 0.077 0.046 0.532 0.187 0.157 0.081 0.520 0.187 0.732 0.090 0.081 20.002 0.079 20.010 0.095 0.243 0.230 0.117 20.007 20.171 0.311 0.230 0.037 0.227 0.152 0.221 0.610 20.006 0.128 0.282 0.389 20.044 20.101 0.189 0.169 0.209 20.114 20.376 0.016 0.075 20.639
SE 0.252 0.255 0.284 0.303 0.254 0.310 0.641 0.155 0.163 2.368 2.381 2.352 2.358 2.503 2.346 2.346 2.353 2.380 2.392 2.566 2.707 2.454 2.868 2.355 2.373 2.419 2.418 2.401 2.419 2.505 2.369 2.868 2.486 2.507 2.415 2.505 2.375 2.485 2.704 2.566 2.869 2.395 2.399 3.314 2.566 2.499 2.704 2.567 2.505 3.299 (continued)
291
Table XII. The impact of the online auction process on an IPOs rst day price surge compared with low and medium volume traditional underwriters
IJMF 5,3
Coefcient 0.023 0.075 20.046 20.094 1.188 20.277 20.335 0.482 0.226 20.277 14.293 * 0.016 0.142 0.161 0.021 0.071 0.433 0.226 0.301 2.233 0.573
SE 2.430 2.704 2.699 2.868 2.620 3.309 3.309 2.870 2.868 3.309 2.500 2.502 3.311 2.564 3.311 2.869 2.870 2.619 3.311 3.312 2.333
292
Table XII.
Notes: Signicant at *5 percent; (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i t(medium volume underwriter dummy)i
Models IV-VI differ based on which one is the base case. The year effects and the traditional volume IPO underwriter dummies are jointly signicant in all of the models. Tables XVII and XVIII show the statistical results and the tests for model specications for Models VII-IX. Again, all three models indicate that the online auction dummy is statistically insignicant. Models VII-IX differ from Models IV-VI in that they have added industry effects, which are jointly signicant. Again, the year effects and the traditional underwriter indicator variables are jointly signicant. Owing to inclusion of year effects, these models are an improvement on Models IV-VI and their results are described in more detail in Tables X-XII, which were discussed earlier in this section. Tables XIX and XX show the statistical results and the tests for model specications for Models X-XII. As in the other models, the online auction dummy is statistically insignicant. These models differ from Models VII-IX in that they include a variable for the size of the IPO, in addition to year effects, industry effects, and dummy variables for the type of underwriter. With the inclusion of a variable controlling for the number of shares in each issue, the volume dummies for traditional underwriters become jointly insignicant, although their inclusion was jointly signicant in the models which did not include a variable controlling for the number of shares in each issue. This is not surprising since high volume underwriters have, on average, a greater number of shares in the IPOs that they issue (the average number of shares in an IPO is 14,600,000) than medium volume underwriters (7,806,649 shares) and low volume underwriters (5,841,266 shares) so the dummy variables for type
Models NA NA NA NA NA NA NA NA NA
Coefcient for online auction IPO dummy ( p-value)
Coefcient for low volume underwriter dummy ( p-value)
Coefcient for medium volume underwriter ( p-value)
Coefcient for high volume underwriter ( p-value)
Coefcient for number of shares offered ( p-value) NA NA 4.73e 2 08 (0.000)
Adjusted R 2 (number of observations) 0.0090 (1,752) 0.1143 (1,752) 0.3924 (1,752)
Model I Model II Model III
20.0348 (0.958) 0.0639 (0.920) 0.5288 (0.321)
Notes: Model I, (rst day price surge)i a b (year dummies)i C(online auction IPO dummy)I; Model II, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I; Model III, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I r (number of shares offered in each IPO)i
Table XIII. Impact of the online auction process on an IPOs rst day price increase
293
294
IJMF 5,3
Models NA NA NA NA NA NA
Model I F 3.80 (0.0009) NA NA Model II F 2.40 (0.0258) F 4.40 (0.0000) NA Model III F 4.72 (0.0000) F 3.78 (0.0000) NA
Notes: Model I, (rst day price surge)i a b (year dummies)i C (online auction IPO dummy)I; Model II, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I; Model III, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I r (number of shares offered in each IPO)i
Table XIV. Model specications F-test for joint signicance of industry effects F-test for joint signicance of low and medium volume underwriters F-test for joint signicance of low and high volume underwriters F-test for joint t-Test for signicance of Adjusted R 2 signicance of medium and high number of shares for (number of volume each IPO underwriters observations) NA NA 27.77 (0.0000) 0.0090 (1,752) 0.1143 (1,752) 0.3924 (1,752)
F-test for joint signicance of year effects
Models 20.3800 (0.017) 20.1981 (0.334) NA 2 0.1890 (0.256) NA 0.1859 (0.366) NA 0.1810 (0.275) 0.3732 (0.019) NA NA NA
Coefcient for number of shares offered ( p-value)
Model IV Model V Model VI
2 0.1341 (0.840) 0.0486 (0.943) 0.2411 (0.721)
Notes: Model IV, (rst day price surge)i a b (year dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i t (medium volume underwriter dummy)I; Model V, (rst day price surge)i a b (year dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i w (high volume underwriter dummy)I; Model VI, (rst day price surge)i a b (year dummies)i C(online auction IPO dummy)I t (medium volume underwriter dummy)i w(high volume underwriter dummy)i
Table XV. Impact of the online auction process on an IPOs rst day price increase
295
296
IJMF 5,3
Models F 3.08 (0.0461) NA NA NA NA NA
Model IV F 3.78 (0.0010) NA Model V F 3.78 (0.0010) NA Model VI F 3.77 (0.0010) NA
Notes: Model IV, (rst day price surge)i a b (year dummies)i C(online auction IPO dummy)I y (low volume underwriter dummy)i t (medium volume underwriter dummy)I; Model V, (rst day price surge)i a b(year dummies)i C(online auction IPO dummy)I y (low volume underwriter dummy)i w (high volume underwriter dummy)I; Model VI, (rst day price surge) i a b (year dummies) i C (online auction IPO dummy)I t (medium volume underwriter dummy)i w (high volume underwriter dummy)i
Table XVI. Model specications F-test for joint signicance of low F-test for joint signicance of and medium volume underwriters industry effects F-test for joint F-test for joint signicance of signicance of low medium and high and high volume volume underwriters underwriters NA F 3.03 (0.0485) F 2.97 (0.0514) t-Test for Adjusted R 2 signicance of number of shares for (number of each IPO observations) NA NA NA 0.0113 (1,752) 0.0113 (1,752) 0.0112 (1,752)
Models 20.3597 (0.021) 20.1615 (0.416) NA 20.2041 (0.209) NA 0.1513 (0.447) NA 0.1975 (0.224) 0.3541 (0.023) NA NA NA
Coefcient for number of shares offered ( p-value)
Model VII Model VIII Model IX
20.0563 (0.930) 0.1432 (0.826) 0.3002 (0.644)
Notes: Model VII, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i t (medium volume underwriter dummy)I; Model VIII, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i w (high volume underwriter dummy)I; Model IX, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C(online auction IPO dummy)I t(medium volume underwriter dummy)i w (high volume underwriter dummy)i
Table XVII. Impact of the online auction process on an IPOs rst day price increase
297
298
IJMF 5,3
Models F 2.97 (0.0514) NA NA NA F 2.93 (0.0539) NA
Model VII F 2.35 (0.0290) F 4.40 (0.0000) Model VIII F 2.35 (0.0296) F 4.40 (0.0000) Model IX F 2.34 (0.0296) F 4.40 (0.0000)
Notes: Model VII, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C(online auction IPO dummy)I y (low volume underwriter dummy)i t (medium volume underwriter dummy)I; Model VIII, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i w (high volume underwriter dummy)I; Model IX, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I t(medium volume underwriter dummy)i w (high volume underwriter dummy)i
Table XVIII. Model specications F-test for joint signicance of industry effects F-test for joint signicance of F-test for joint F-test for joint signicance of low signicance of low medium and high volume and medium volume and high volume underwriters underwriters underwriters t-test for Adjusted R 2 signicance of number of shares for (number of each IPO observations) 0.1164 (1,752) 0.1163 (1,752) 0.1163 (1,752) NA NA NA NA F 2.88 (0.0563) NA
Models 20.0180 (0.890) 20.0933 (0.571) NA 0.0337 (0.803) NA 0.0136 (0.934) NA 20.0871 (0.519) 20.0316 (0.808)
Coefcient for number of shares offered ( p-value) 4.73e 2 08 (0.000) 4.73e 2 08 (0.000) 4.73e 2 08 (0.000)
Model X Model XI Model XII
0.5286 (0.321) 0.4552 (0.400) 0.5109 (0.344)
Notes: Model X, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i t (medium volume underwriter dummy)i r (number of shares offered in IPO)I; Model XI, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i w (high volume underwriter dummy)i r (number of shares offered in each IPO)I; Model XII, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C (online auction IPO dummy)I t (medium volume underwriter dummy)i w(high volume underwriter dummy)i r (number of shares offered in each IPO)i
Table XIX. Impact of the online auction process on an IPOs rst day price increase
299
300
IJMF 5,3
Models F 0.05 (0.9504) NA NA NA F 0.23 (0.7961) NA
Model X F 4.73 (0.0001) F 3.76 (0.0000) Model XI F 4.77 (0.0001) F 3.75 (0.0000) Model XII F 4.70 (0.0001) F 3.76 (0.0000)
Notes: Model X, (rst day price surge)i a b (year dummies)i m (SIC code dummies)i C(online auction IPO dummy)I y (low volume underwriter dummy)i t (medium volume underwriter dummy)i r (number of shares offered in IPO)I; Model XI, (rst day price surge)i a b (year dummies)i m(SIC code dummies)i C (online auction IPO dummy)I y (low volume underwriter dummy)i w (high volume underwriter dummy)i r (number of shares offered in each IPO)I; Model XII, (rst day price surge)i a b (year dummies)i m(SIC code dummies)i C(online auction IPO dummy)I t (medium volume underwriter dummy)i w (high volume underwriter dummy)i r(number of shares offered in each IPO)i
Table XX. Model specications F-test for joint signicance of industry effects F-test for joint t-test for signicance of F-test for joint F-test for joint Adjusted R 2 signicance of signicance of low signicance of low medium and high number of shares for (number of volume and medium volume and high volume each IPO underwriters underwriters underwriters observations) NA NA F 0.07 (0.9317) 27.60 (0.0000) 27.62 (0.0000) 27.61 (0.0000) 0.3917 (1,752) 0.3919 (1,752) 0.3918 (1,752)
of underwriter are picking up some of the same effects that were incorporated into the variable for the number of shares of the issue. Also, as seen in other studies, such as Loughran and Ritter (2004), smaller rms are less likely to go public using a high prestige underwriter[12]. These high prestige underwriters, as dened by Loughran and Ritter (2004), Carter and Manaster (1990), and Carter et al. (1998), are largely the same as the high volume underwriters in this analysis large, well-known investment banks which underwrite a number of IPOs. Since separation of traditional underwriters into low, medium, and high volume underwriters is important to the analysis, the models in Tables XVII and XVIII are more appropriate than the models in Tables XIX and XX for comparing the size of the rst day price surge. These models were described in more detail in Tables X-XII. For models in which all of the traditional underwriters are combined together, the most appropriate model is one which includes a variable controlling for the size of the issue and of the rm, but which excludes dummy variables for the type of traditional underwriter. This was the model in Table IX. In summary, these models suggest that regardless of the specication of the model, IPOs which are issued through the online auction mechanism do not have rst price surges which are statistically signicantly different from the traditional IPOs. The sign on the online auction IPO is positive in nine of the 12 models; the only models in which the sign is negative are in three of the models with among the poorer ts (Models I, IV, and VII). This suggests that the rst day price increase tends to be statistically insignicantly higher for IPOs issued through the online auction process, controlling for year xed effects, industry xed effects, size of the IPO, and whether the traditional underwriter is a high volume underwriter, a low volume underwriter, or a medium volume underwriter. The models also provide information on the size of the rst day price surge between the various types of traditional underwriters. Models IV-VI, which have only year effects, but not industry effects, suggest that low volume underwriters have a statistically signicantly lower price surge than high volume underwriters, and that medium volume underwriters have a lower price surge than high volume underwriters and a higher price surge than low volume underwriters, although the results are not statistically signicant. These ndings are consistent with the average rst day price surges for low volume, medium volume, and high volume underwriters in Tables I and VII. Models VII-IX differ from Models IV-VI in that they add industry xed effects, which are jointly signicant; the conclusions and results, however, are identical. Models X-XII differ from Models VII-IX in that they add a variable controlling for the size of the issue. Since inclusion of this variable makes the volume dummies for traditional IPO underwriters unnecessary (as can be seen in their statistical insignicance) because high volume underwriters, on average, issue larger IPOs, these models are not as useful as the previous sets in evaluating the relative behavior of the three types of traditional underwriters. Nevertheless, although the results are statistically insignicant, the sign of the low volume coefcients tend to support the hypothesis that low volume underwriters have lower rst day price increases than medium or high volume underwriters. Although the online auction IPOs exhibit some degree of mispricing, does it persist over time to a greater degree than the mispricing exhibited by IPOs issued through the traditional process? Table XXI examines the degree of mispricing for IPOs over several
301
IJMF 5,3
Time period After After After After After After After After one day of trading one week of trading two weeks of trading four weeks of trading 60 days of trading 90 days of trading 180 days of trading one year of trading
Percentage of online auction IPOs whose absolute percentage price change from the offer price exceeds that of comparable traditional IPOs issued in their three-digit SIC industry area in the same year 77.78 (7/9) 66.67 (6/9) 62.5 (5/8) 75 (6/8) 50 (4/8) 37.5 (3/8) 62.5 (5/8) 83.3 (5/6)
302
Table XXI. Percentage of online auction IPOs whose post-issue mispricing exceeds that of comparable traditional IPOs
Notes: This table compares the frequency for which the absolute value of the percentage change in the stock price for IPOs issued through the online auction mechanism exceeds the absolute value of the percentage change of the average of IPOs issued through the traditional process in the same three-digit SIC industry area in the same year, the analysis compares these numbers at various time intervals following issuance of the online auction IPO and the traditional IPOs; the absolute value of the percentage change measures the degree to which mispricing (relative to the offer price) occurs; the ndings in this table are consistent with the hypothesis that the online auction IPO process does not reduce mispricing relative to the traditional process Source: SDC Platinum
time periods: after the rst day of trading, after one week of trading, after two weeks of trading, after four weeks of trading, after 60 days of trading, after 90 days of trading, after 180 days of trading, and after one year of trading. The second column shows the percentage of online auction IPOs whose absolute percentage price increase from the original offer price exceeds that of the comparable traditional IPO, issued in the same year and in the same three-digit industry SIC code. The table indicates that, for most time periods, generally over 60 percent of online IPOs still have a greater degree of mispricing relative to their original offer price, when compared to the degree of mispricing exhibited by their traditional counterpart. This suggests that the offer price set in auctions is less reective of the true value of the company than the offer price set through the traditional process. In summary, the analysis suggests that the auction process tends to be more efcient, on average, than the traditional processes of many medium to high volume underwriters, and less efcient on average than the traditional processes of many low volume underwriters. Although one would have expected medium and high volume underwriters to be more efcient at pricing IPOs because they have had the opportunity to price so many of them, they are also more likely than low volume underwriters to have a large client base and to possibly reward their clients who assist them in the initial book-building process by mispricing the IPO so that these clients can benet from the price increase on the rst day of trading. When analysing the average rst day price surges by underwriter across all underwriters, the average rst day price surge for OpenIPO.com exceeds the average rst day price increase for 82.4 percent of the underwriters. When examining together all IPOs in all SIC codes issued by all underwriters, use of the online auction issuance process had a consistently statistically insignicant impact on the rst day price increase. This was
also conrmed by t-tests of the differences between the mean degree of underpricing for online auction IPOs and low, medium, and high volume underwriters. IV. Possible causes and solutions to potential problems in the online auction process The online process is not as successful as had been hoped in minimizing the rst-day price surges of IPOs. Although, by its structure, it eliminates potentially intentional mispricing by underwriters to benet the larger nancial institutions who assist them in building a book of orders, the online auction process still leads to mispricing. Possible causes of this mispricing include: . a lack of information on the part of the small investors relative to larger nancial institutions; . an adverse selection problem concerning the types of rms which choose to issue online; and . inherent conservatism on the part of investors in using a new process. This section discusses these factors as possible causes of the mispricing problem inherent in auctions, examines applications of these factors in the IPO debut of Google (August 2004) and Morningstar (May 2005), and assesses possible solutions. The mispricing in online auctions may be due to the involvement of small investors, who do not have access to detailed sources of information on the company. The investment banks and institutional investors who price the IPO in the traditional process analyse companies routinely, and also have a greater capability than small investors to meet with and interview representatives of the issuing rm. Firms that do their IPO through the online auction process do not have to provide as much information concerning their uses of funds as they would in the traditional process. This is because the investment banks, under the online auction process, do not engage in the book-building process. Examples of types of information that issuing rms could provide which would be useful to investors include: . the uses for the capital that they are raising; . their strategies for overcoming potential challenges; . the corporate governance mechanisms within the rm (share of outsiders on the board, etc.); . their reasons for using the online auction process, rather than the traditional process; and . their involvement in any current or potential litigation. As investors become more familiar with the auction process, they may be less likely to bid conservatively in the presence of greater information. The degree of mispricing for Google was more substantial than the performance of comparable IPOs in its industry group which debuted in the same year using the traditional method. Googles offer price was $85 and it closed at $100.34, reecting an 18 percent increase, while the average rst day price increase for all IPOs issued in 2004 was 8.6 percent. Indeed, analysis of the data shows that 82 percent of the IPOs issued in 2004 experienced less of a jump from the offer price to the closing price on the
303
IJMF 5,3
304
rst day of trading than Google did. Googles rst day price increase also exceeded that of IPOs in its peer group: about 70 percent of the IPOs in its three-digit SIC code issued in 2004 experienced less of a rst day price increase and about 60 percent of the IPOs in Googles four-digit SIC code, focusing largely on the internet search arena, exhibited less of a rst day price increase. One of the criticisms of Google was that it was secretive in how it would use its funds and conveyed little detailed information (the Wall Street Journal, 2004a). At the time, Google faced several strategic issues, which in the absence of more detailed information on the uses of the capital to be raised, may have been difcult for the smaller investors to evaluate. These included Googles lack of diversication in revenue sources and reliance on online advertising, rather than on the other subscription-based services (unlike Yahoo and Microsoft). Despite the lack of external transparency in Googles strategic processes, it has continued to surpass analysts earnings expectations. Nevertheless, the online process could be used more by companies, which may not have a clear sense of the uses for the funds that they are raising or their uses of funds may not be appropriately assessed by smaller investors if enough information is not provided to them. In the case of Googles auction, many informed investors institutional investors and hedge funds entered in the market after the stock had debuted, contributing to its upward price momentum (The Washington Post, 2004a, b); their earlier entry might have provided greater price support in developing the offer price. Nevertheless, the larger investment banks felt locked out of the process, and lacked nancial incentives to push Googles IPO to their clients. This may also have been the case for other online auction IPOs. The recent SEC proposals to liberalize the quiet period may provide a solution to the possible informational problems inherent in the online auction process. During the quiet period, companies traditionally have been only allowed to give out information orally (in presentations), but not in written form (except for the companys prospectus). Consequently, the quiet period increased the informational disadvantage of small investors relative to larger institutional investors, since smaller investors are less likely to be able to attend company presentations and more likely to be reliant on written documents available to the public. In late October 2004, the SEC voted to liberalize these rules by allowing companies planning an IPO to communicate information to investors verbally or in writing, provided that this information would be led with the SEC (the Wall Street Journal, 2004b). Indeed, the SEC has also proposed to allow the marketing roadshows of IPOs to be broadcast online to all investors, although this is not a requirement such that companies which only want to present materials before the traditional audience can do so (the Wall Street Journal, 2005b). Smaller and less well-known companies, previously handicapped by their inability to use the media and the web during the quiet period to generate interest, are more likely to stimulate investor enthusiasm in their IPO (the Investment Dealers Digest, 2004). This relaxation of the quiet period restrictions will make it easier for companies using the online auction process to provide small investors with information if the company wants to do so. With the relaxation of the quiet period restrictions, companies can minimize the informational problems inherent in auctions. A second potential aw of the online process lies in the possibility of an adverse selection problem inherent in the types of rms choosing to use the online process,
which could lead to a lemons discount being placed on these issues by uninformed investors[13]. Since rms using the online process are not subjected to the rigorous interviews of the investment banks in the traditional book-building process, rms which are well-known, but which would have had difculty in going public for some reason using the traditional IPO process may be more likely to use the online auction mechanism. For example, when Morningstar, which rates mutual funds, decided to go public using the online auction process in January 2005, it did so at a time when it was under legal investigation and when its IPO ling under the traditional process had consequently been dormant since May 2004. In September 2004, Morningstar was the subject of an SEC investigation concerning inaccurate data which had been on its website. In December 2004, Eliot Spitzer, as well as the SEC, began looking into possible conicts of interest from Morningstar Associates recommendations to 401(k) on mutual fund investment options, while Morningstar itself provides fund ratings (the Wall Street Journal, 2005a; Stein, 2005). Indeed, Morgan Stanley executives warned Morningstar that an auction carried a high risk of an adverse outcome, such that when Morningstar continued with the auction, instead of switching back to the traditional process, Morgan Stanley resigned as the lead underwriter (Smith, 2005). Morningstars pricing behavior suggests that the online auction initially underpriced it: the stock opened on May 3, 2005 at $18.66 and closed at $20.05. By the second week in June 2005, it was trading in the high $20s per share, by November 2005, it was trading in the mid $30s per share, and, by December 30, 2005, it was trading around $34-$35 per share. As of April 2008, it was trading around $54 per share. This explanation is similar to the changing risk composition hypothesis, discussed in Ritter (1984) in which underpricing increases in an IPO market as the proportion of IPOs that are perceived as risky increases. Nevertheless, the sources of risk would lie more in the lack of information provided by the issuing rm than in other characteristics. As we showed previously in our comparison of characteristics of issuing rms using the online auction IPO process and the traditional IPO process, the rms using the online auction IPO process have many similarities to rms using the traditional IPO process in terms of company age, and industry. The main difference is that, on average, they tend to issue a smaller number of shares, and that a higher percentage of online auction IPOs are likely to have sales below $10 million than traditional IPOs. Inherent conservatism of investors may be a third explanation for the underpricing. One of the lessons that investors learned in the dot-com boom is that a knowledge of the underlying company is important in valuing it. Consequently, at least initially, investors may be setting low offer prices for IPOs because of this lesson and because of the relative youth of the auction process. Indeed, OpenIPO served as lead manager for only four IPOs prior to 2001, by which time the market was declining and investors were more wary in their bids. The inherent conservatism of investors in auctions may have been a factor in the Google auction: the fund manager of the Legg Mason Value Trust noted in his letter to shareholders that:
[. . .] many investors may have steered clear of the Google auction because the determination of what to bid would require considerable work [. . .] we were delighted when the so-called FUD [fear, uncertainty, doubt] dominated the process, resulting in the shares coming in at the bargain price of $85 (Ian, 2005).
305
IJMF 5,3
306
V. Conclusion The online auction process was developed in response to the substantial increases between IPOs offer prices and their open prices at the peak of the dot-com era. Table VII suggests that while the rst day surges in an IPOs price are important, the problem that the online auction mechanism was partially developed to solve has lessened. The average rst day price increase has dropped substantially from the 55.4-92.5 percent range during 1999-2001 into the 6.3-8.6 percent range of 2002-2004. This gradual reduction in the magnitude of rst day price surges over time may be a reection of more careful pricing by some underwriters in the wake of substantial litigation concerning their alleged manipulation of IPOs, as well as that many of the IPOs during the dot-com era were in emerging industries and therefore were difcult to price. The ndings of this analysis suggest that the online auction process does not minimize the rst day price surge of an IPO to the degree that was initially intended when the process was developed. The types of rms choosing the auction IPO process relative to the traditional IPO process tended to be smaller rms, as measured by sales and number of shares in their IPO issue. Even controlling for the two-digit SIC industry area, auction IPOs tended to be smaller than traditional IPOs. The percentage of technology rms and the age distribution of the rms were similar between the online and traditional process, although they varied from year to year in four of the seven years, the auction process hosted slightly older rms or had fewer technology IPOs than the traditional process, while in three of the seven years, the opposite effects were seen. Industry sector and number of shares in the IPO reected many of the characteristics of online and traditional rms, which also could be seen in age, sales, etc. Regression analyses controlling for year and industry effects, as well as the size of the IPO, suggest that the rst day price surge of the online auction IPO process did not statistically signicantly differ from the rst day price surge of the traditional process, even when controlling for type of traditional underwriter (low volume, medium volume, and high volume). A comparison of auction IPOs with traditional IPOs issued in the same year and in the same three-digit SIC code suggests that for 60 percent (six out of ten) of the auction IPOs, over 40 percent of the traditional IPOs issued in that year and in that three-digit SIC area had greater mispricing. High volume traditional underwriters have a statistically signicantly higher rst day price surge than low volume traditional underwriters, which is surprising because high volume underwriters, due to the number of IPOs that they price, should be better able to price an issue than low volume underwriters. This nding provides evidence supporting the legal allegations against larger investment banks in 2002-2003 that they may have been manipulating IPO prices during the dot-com era in an effort to reward their preferred clients who assisted them in pricing the IPO. This analysis also nds that the mispricing of online IPOs relative to their traditional counterparts issued in the same three-digit SIC code and in the same year persists over time (one week, two weeks, four weeks, 60 days, 90 days, 180 days, and one year after debuting) for 50-80 percent of online auction IPOs. Although the online auction process increases the ability of small investors to participate in the IPO process, the process may suffer from several problems. First, small investors may lack the ability to efciently price an IPO due to informational asymmetries either because small investors lack access to the sources of information that institutional investors have, or because companies are not required to provide
detailed information in the online process since they do not undergo the rigorous scrutiny of investment banks in the traditional bookbuilding process. Second, since the informational scrutiny is reduced, the online process could be used more by companies which may not have a clear sense of the uses for the funds that they are raising or by well-known companies which may not have been successful in the traditional issuance process. This could lead to an adverse selection problem, as investors could have difculty in distinguishing successful companies from lemons, and consequently could end up discounting the price of all online IPOs. Nevertheless, with interest in IPOs rebounding and the growing belief of small investors that they can become involved early with new issues, it is likely that some of the weaknesses of the online process may be improved. Some of the solutions include: . the SEC reforms on the quiet period, which will minimize the informational asymmetry between small investors and larger institutional investors; and . greater provision of information by issuing companies concerning uses of capital to be raised, etc. so that small investors can better distinguish good companies from lemons. Developing new methods of IPO issuance, increasing available information, and involving more parties in the process are likely to lead to a more egalitarian and transparent process for providing new companies with capital.
Notes 1. Indeed, in the rst quarter of 2008, several IPOs exhibited substantial underpricing on their rst day, despite the declining nancial markets. Asia Time Corporation (debuted February 11, 2008) exhibited a rst day price increase of 142.9 percent, the RiskMetrics Group (debuted January 24, 2008) exhibited a rst day price increase of 35.7 percent, IPC (debuting on January 24, 2008) and Visa (debuting on March 18, 2008) increased about 28-29 percent on their rst day, and Heritage Crystal Clean (debuted March 11, 2008) increased 23.4 percent on its rst day of trading (Renaissance Capital, 2008). 2. The decline in mispricing since the collapse of the dot-com boom is discussed in Loughran and Ritter (2004). 3. Chen and Ritter (2000) examine various explanations for the large spreads in the USA relative to other countries, and as well as why 7 percent is such a common spread. 4. Clarke et al. (2003) discussed how investment banks signicantly boosted their share of IPOs when they gained an all-star analyst who was highly ranked in the Institutional Investor annual survey, and that this effect was greater in the mid to late 1990s than in the preceding periods. Ljungqvist and Wilhelm (2003) noted that the presence of an all-star analyst raised the probability that the underwriting investment bank employing the analyst would be chosen as the lead underwriter. Both of these studies were also discussed in Loughran and Ritter (2004). 5. For example, the Wall Street Journal (2005a) noted that the auction method can sap the rst day price surges that IPOs typically enjoy, while Ian (2005) notes With an auction, the price is usually set after aggregating bids and deciding the highest price at which the company can sell its shares. This often saps any rst-day pops in price that IPOs typically enjoy. 6. Available at: www.wrhambrecht.com (accessed January 21, 2008). 7. For example, if the offer price is set at $11 share, and there are 1 million shares available, but the cumulation of bids above $11 indicates that 1.25 million shares were requested, then
307
IJMF 5,3
8. 9.
308
10.
11.
12.
13.
bidders (who had initially bid above $11) pay $11 per share and receive 80 percent of the shares for which they had bid because 1 million shares divided by 1.25 million shares is 80 percent. Available at: www.wrhambrecht.com (accessed January 21, 2008). Several papers discuss the phenomenon of IPO underpricing in various countries, including Loughran et al. (1994) and Chowdhry and Sherman (1996). Loughran and Ritter (2004, pp. 8-9) discuss some of the legal issues and settlements linked to the behavior of underwriting investment banks during the dot-com era. For example, they note that Robertston Stephens, in its settlement with the SEC and NASD in 2003, admitted to allocating IPOs to clients on the basis of past and future commission business, and that CSFB received commissions which were 65 percent of investor prots from IPOs, such as VA Linux. Many of the studies which use other variables to capture the characteristics of the rms, such as Loughran and Ritter (2004) and Carter et al. (2000), do not use industry variables based on two-digit SIC code or number of shares in the IPO because the effects captured by these variables are often reected in the regressors used in their specications. Table V includes only ten out of the 14 IPOs which debuted using the online auction process, since the other four did not have comparable IPOs issued in the same year in the same three-digit SIC industry area. More prestigious underwriters are listed before the other underwriters in the non-managing underwriting section of an IPOs prospectus in brackets, and underwriters which always appear in the highest bracket receive a 9 on the nine-point scale. These high prestige underwriters, are underwriters with a ranking of 8 or higher on a nine-point scale. Despite the fact that an individual uninformed investor may have less purchasing power than a given informed investor, the aggregation across individual uninformed investors of lemons discounts in the bidding process could lead to substantial underpricing.
References Aggarwal, R., Nagpurnanand, P. and Manju, P. (2002), Institutional allocation in initial public offerings: empirical evidence, Journal of Finance, Vol. 57 No. 3, pp. 1421-42. Beneviste, L.M. and Spindt, P.A. (1989), How investment bankers determine the offer price and allocation of initial public offerings, Journal of Financial Economics, Vol. 24, pp. 343-61. Biais, B. and Faugeron-Crouzet, A.-M. (2002), IPO auctions: English, Dutch, French, and Internet, Journal of Financial Intermediation, Vol. 11, pp. 9-36. Carter, R.B. and Manaster, S. (1990), Initial public offerings and underwriter reputation, Journal of Finance, Vol. 45, pp. 1045-68. Carter, R.B., Dark, F.H. and Singh, A.K. (1998), Underwriter reputation, initial returns, and the long-run performance of IPO stocks, Journal of Finance, Vol. 53, pp. 285-311. Carter, R.B., Strader, T. and Nilakanta, S. (2000), Online investment banking phase I: distribution via the internet and its impact on IPO performance, Journal of the Association for Information Systems, Vol. 1, pp. 1-26. Chen, H.-C. and Ritter, J. (2000), The seven percent solution, Journal of Finance, Vol. 55 No. 3, pp. 1105-31. Chowdhry, B. and Sherman, A. (1996), International differences in over-subscription and underpricing of IPOs, Journal of Corporate Finance, Vol. 2, pp. 359-81. Clarke, J.E., Dunbar, C. and Kahle, K.M. (2003), All-star analyst turnover, investment bank market share, and the performance of initial public offerings, Georgia Tech working paper, University of Pittsburgh, Pittsburgh, PA.
Field, L.C. (1997), Is institutional investment in initial public offerings related to the long-run performance of rms?, working paper, University of California, Los Angeles, CA. Ian, M. (2005), Morningstar bets on bidders for IPO, Wall Street Journal, January 10. Investment Dealers Digest (2004), The quiet period getting overhaul, Investment Dealers Digest, October 18. Jagannathan, R. and Sherman, A. (2005), Reforming the bookbuilding process for IPOs, Journal of Applied Corporate Finance, Vol. 17 No. 1, pp. 67-72. Jain, B.A. and Kini, O. (1994), The post-issue operating performance of IPO rms, Journal of Finance, Vol. 49 No. 5, pp. 1699-726. Kandel, S., Sarig, O. and Wohl, A. (1999), The demand for stocks: an analysis of IPO auctions, The Review of Financial Studies, Vol. 12 No. 2, pp. 227-47. Krignam, L., Shaw, W.H. and Womack, K.J. (1999), The persistence of IPO mispricing and the predictive power of ipping, Journal of Finance, Vol. 54 No. 3, pp. 1015-44. Lee, P.J., Taylor, S.L. and Walter, T.S. (1999), IPO underpricing explanations: implications from investor application and allocation schedules, Journal of Financial and Quantitative Analysis, Vol. 34 No. 4, pp. 425-43. Ljungqvist, A.P. and Wilhelm, W.J. (2003), IPO pricing in the dot-com bubble, Journal of Finance, Vol. 58, pp. 723-52. Loughran, T. and Ritter, J.R. (2004), Why has IPO underpricing changed over time?, Financial Management, Autumn, pp. 5-37. Loughran, T., Ritter, J.R. and Rydqvist, K. (1994), Initial public offerings: international insights, Pacic Basin Finance Journal, Vol. 2, pp. 165-99. Renaissance Capital (2008), Renaissance Capitals rst quarter 2008 IPO review, available at: www.ipohome.com (accessed April 19, 2008). Ritter, J.R. (1984), The Hot Issue market of 1980, Journal of Business, Vol. 57, pp. 215-40. Ritter, J.R. (1991), The long-run performance of initial public offerings, Journal of Finance, Vol. 46 No. 1, pp. 3-27. Ritter, J.R. (2008), Some factoids about the 2007 IPO market, unpublished manuscript, available at: http://bear.cba.u.edu/ritter/ipodata.htm (accessed March 20, 2008). Rock, K. (1986), Why new underwriters are underpriced, Journal of Financial Economics, Vol. 15, pp. 187-212. San Jose Business Journal (1999), Bidding begins on the rst-ever internet-based IPO auction, San Jose Business Journal, February 12. Sherman, A. (2000), IPOs and long-term relationships: an advantage of book building, Review of Financial Studies, Vol. 13 No. 3, pp. 697-714. Smith, R. (2005), Why IPOs still use the old way, Wall Street Journal, July 6. Stein, N. (2005), Morningstars bright future turns cloudy, Fortune, January 10. Teoh, S.H., Welch, I. and Wong, T.J. (1998), Earnings management and the long run market performance of initial public offerings, Journal of Finance, Vol. 53 No. 6, pp. 1935-74. Wall Street Journal (2004a), Engine trouble: how miscalculations and hubris hobbled celebrated Google IPO, Wall Street Journal, August 19. Wall Street Journal (2004b), IPO quiet period could be getting a lot louder, Wall Street Journal, October 27. Wall Street Journal (2005a), Auction IPOs: rst Google, now Morningstar, Wall Street Journal, January 11.
309
IJMF 5,3
310
Wall Street Journal (2005b), IPO outlook: SEC proposes increasing role of web in IPOs, Wall Street Journal, January 3. (The) Washington Post (2004a), Because of, or despite, news, bids placed in IPO, The Washington Post, August 14. (The) Washington Post (2004b), Google ends auction for IPO shares, The Washington Post, August 19. Welch, I. (1989), Seasoned offerings, imitation costs, and the underpricing of initial public offerings, Journal of Finance, Vol. 44 No. 2, pp. 421-49. Further reading Delaney, K.J. (2005), Google revisited: insiders got choice other sellers didnt, Wall Street Journal, September 16. Ernst & Young LLP (2007), Quantitative economics and statistics group on behalf of the strategic growth markets practice, The Ernst & Young IPO Study, available at: www.ey.com (accessed April 15, 2008). Field, L.C. and Jonathan, K. (2002), Takeover defenses of IPO rms, Journal of Finance, Vol. 57 No. 5, pp. 1857-89. Forno, R. (2005), Google to offer instant messaging and voice communications on web, The New York Times, August 24. Karmin, C., Cowan, L. and Fowler, G.A. (2005), Baidu.com sees IPO pop, but some fear a burst bubble, Wall Street Journal, August 14. Pagano, M., Panetta, F. and Zingales, L. (1998), Why do companies go public: an empirical analysis, Journal of Finance, Vol. 53 No. 1, pp. 27-64. Ritter, J.R. and Welch, I. (2002), A review of IPO activity, pricing, and allocations, Journal of Finance, Vol. 57 No. 4, pp. 1795-828. Weisman, R. (2004), Google IPO fails to nd results it sought, Knight Ridder Tribune Business News, August 19. Zingales, L. (1995), Insider ownership and the decision to go public, The Review of Economic Studies, Vol. 62 No. 3, pp. 425-48. Corresponding author Nayantara Hensel can be contacted at: ndhensel@nps.edu
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

An Empirical Analysis of The Efficiency of Online Auction IPO

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

An Empirical Analysis of The Efficiency of Online Auction IPO

Uploaded by

Copyright:

Available Formats

The current issue and full text archive of this journal is available at www.emeraldinsight.com/1743-9132.

Efciency of IPO processes

Efciency of IPO processes

Number Percentage of of IPOs total IPOs

Average number of IPOs

Efciency of IPO processes

Table I. Summary statistics on various types of underwriters

2800 7.88 7.14 12,200,000 (127.85)

3,950,000 (33.45) 3,500,000 (42)

8,539,446 (136.31) 5,337,011 (80.88)

2,000,000 (16) 2,600,000 (34.9)

6100 1.84 18.47 26.70 66.40 21.43 100 7.14 7.14

16,500,000 (94.39) 21,400,000 (212.24) 18,900,000 (254.05) 5,551,762 (74.32)

3,052,174 (35.1) 7,612,500 (140.8) 10,000,000 (130) 8,701,684a (588.1)a

Total percentage of IPOs

Efciency of IPO processes

3600 2800 4800

3800 8700 6300 5900 5900 6200 6200 3500

Efciency of IPO processes

Percentage of technology IPOs through online underwriters 66.6 0 0 0 50 33 0

Efciency of IPO processes

Total During During During During During During During

Average rst day price surge (total sample)

1999 2000 2001 2002 2003 2004 2005

Efciency of IPO processes

2084 2086 2095 2095 2834

October 29, 2003

May 18, 2000

March 17, 2004

June 22, 1999 December 8, 1999

August 18, 2004

Efciency of IPO processes

Efciency of IPO processes

Efciency of IPO processes

Efciency of IPO processes

Efciency of IPO processes

Coefcient for online auction IPO dummy ( p-value)

Coefcient for low volume underwriter dummy ( p-value)

Coefcient for medium volume underwriter ( p-value)

Coefcient for high volume underwriter ( p-value)

Coefcient for number of shares offered ( p-value) NA NA 4.73e 2 08 (0.000)

Adjusted R 2 (number of observations) 0.0090 (1,752) 0.1143 (1,752) 0.3924 (1,752)

Model I Model II Model III

20.0348 (0.958) 0.0639 (0.920) 0.5288 (0.321)

Efciency of IPO processes

F-test for joint signicance of year effects

Coefcient for online auction IPO dummy ( p-value)

Coefcient for low volume underwriter dummy ( p-value)

Coefcient for medium volume underwriter ( p-value)

Coefcient for high volume underwriter ( p-value)

Coefcient for number of shares offered ( p-value)

Adjusted R 2 (number of observations) 0.0113 (1,752) 0.0113 (1,752) 0.0112 (1,752)

Model IV Model V Model VI

2 0.1341 (0.840) 0.0486 (0.943) 0.2411 (0.721)

Efciency of IPO processes

Models F 3.08 (0.0461) NA NA NA NA NA

Model IV F 3.78 (0.0010) NA Model V F 3.78 (0.0010) NA Model VI F 3.77 (0.0010) NA

F-test for joint signicance of year effects

Coefcient for online auction IPO dummy ( p-value)

Coefcient for low volume underwriter dummy ( p-value)

Coefcient for medium volume underwriter ( p-value)

Coefcient for high volume underwriter ( p-value)