Journal of Banking & Finance 33 (2009) 1255–1265

Contents lists available at ScienceDirect

Journal of Banking & Finance

journal homepage: www.elsevier.com/locate/jbf

Does the market dole out collective punishment? An empirical analysis

of industry, geography, and Arthur Andersen’s reputation
Roger D. Huang a,*, Hang Li b
a
Department of Finance, Mendoza College of Business, University of Notre Dame, Notre Dame, IN 46556, USA
b
Mendoza College of Business, University of Notre Dame, Notre Dame, IN 46556, USA

a r t i c l e i n f o a b s t r a c t

Article history: Arthur Andersen’s reputation was tarnished following news that its Houston office had shredded docu-
Received 6 October 2008 ments related to the auditing of energy giant Enron. Earlier studies documented widespread spillover
Accepted 21 January 2009 of the reputation effect, suggesting a strong commonality in Big 5 audit practices. We examine whether
Available online 31 January 2009
the market is more discriminating in its assessments. We focus on the roles industry specialization of
auditors and the geography of clients’ audit offices play in accounting for the contagion. Our results
JEL classification: are supportive of investors who differentiate audit practices by industry and who account for the location
M400
of the specific office where the audit work is done. We find that losses suffered by energy firms or firms
G100
G140
located close to Houston are equivalent to approximately 90% of the aggregate abnormal losses suffered
G180 by Big 5 clients. Our evidence suggests the possibility of more localized impact of accounting scandals
G280 and supports accounting regulations targeted at individual industries.
Ó 2009 Elsevier B.V. All rights reserved.
Keywords:
Reputation spillover effect
Audit practices and offices
Industry specialization
Geographic location

1. Introduction impacts on Andersen clients and clients of other Big 5 accounting

Document shredding by Arthur Andersen and the subsequent
Justice Department indictment were unprecedented watershed
events for the business world. The year before it filed for bank-
ruptcy on December 2, 2001, Enron was an energy giant with sales
that exceeded $100 billion. Its external auditor was Arthur Ander-
sen, a member of the Big 5 accounting firms.1 After allegations of
accounting violations in Enron surfaced, Andersen came under suspi-
cion concerning its independence from Enron and its audit quality.
Following Andersen’s announcement that it had shredded Enron-re-
lated documents on January 10, 2002, it was indicted and convicted
on a single count of obstruction of justice by US federal prosecutors
and ceased operations on August 31, 2002.
The purpose of our paper is to examine explanations for the
spread in the reputation effect following news that Andersen’s
Houston office has shredded Enron-related documents. Earlier
studies have documented significant negative abnormal valuation
* Corresponding author. Tel.: +1 574 631 1691; fax: +1 574 631 4825.
E-mail addresses: roger.huang.31@nd.edu (R.D. Huang), hli1@nd.edu (H. Li).
1 academicians to assess the significance of an audit firm’s reputa-
The Big 5 accounting firms are Arthur Andersen, Deloitte & Touche, Ernst &Young,
KPMG, and Price Waterhouse Coopers.
firms. Since those clients were located nationwide, this raises the
possibility of a widespread reputation effect. Did investors penalize
all Big 5 clients collectively? Meting out punishment across the
board suggests a strong commonality in audit practices conducted
by all Big 5 auditors. We examine the possibility that the market is
more discriminating by considering industry and geographic
explanations for the spillover.
Our focus acknowledges the fact that audit practices differ by
industry or by location of specific audit offices. We are motivated
by the fact that auditors are industry specialists and auditing work
is primarily accomplished at an individual office level. Our results
show that energy firms or firms located close to Houston experi-
enced losses equal to around 90% of the aggregate loss suffered
by Big 5 clients over a 3-day period following the shredding news.
Energy firms experienced most of the losses. The picture that
emerges is one of a market that is discriminating in penalizing
firms rather than one that doles out collective punishment. It also
suggests the possibility of more localized impact of accounting
scandals and supports regulations aimed at individual industries.
The Andersen–Enron event provides a unique opportunity for
tion. Although professional reputation is of utmost importance to

0378-4266/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankfin.2009.01.003
1256 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265
financial services firms in general, it is especially true for financial
auditors. An auditor’s primary role is to give credence to its clients’
financial statements. A prerequisite for this role is the presumption
by the market that the integrity of the audit firm itself is of the
highest standing. The credibility of the audited financial state-
ments rests on the reputation of the auditor. In addition to the
assurance role, auditors also play an insurance role whereby they
insure audit-related litigation claims against their clients with
their ‘‘deep pockets” or ability to cover the claims (see Watts and
Zimmerman, 1983). When an auditor’s reputation is sullied, its
ability both to assure its clients’ financial statements and to insure
their litigation risks is diminished. The collapse of Andersen per-
mits researchers to measure the effect of a weakened reputation
by its impact on stock prices.
Regulatory authorities have long been concerned about impair-
ment of auditor credibility due to pervasive material misstate-
ments or misrepresentations of financial statements (SEC, 2000,
2001). They are particularly apprehensive about spillover effects
arising from damage to an auditor’s reputation. If spillover effects
are widespread and severe, more regulations may be called for in
order to protect the integrity and stability of financial markets.
At the other extreme, increased market-wide accounting regula-
tions may be unwarranted if repercussions from an auditor’s loss
of reputation capital is highly localized or industry specific. Specif-
ically, one could argue that the comprehensiveness of Sarbanes–
Oxley Act of 2002 or parts of it may be misguided to the extent that
Andersen effect was localized.
Reputation spillover effects may arise from various sources fol-
lowing damaging revelations about an auditor’s credibility. If
investors regard the audit practices in an accounting firm as having
a strong common component, the market may question the accu-
racy and reliability of audited financial statements of all the clients,
even when bad publicity arises from the audit work of a single of-
fice. It may even be that investors regard the audit practices of dif-
ferent accounting firms as having a strong commonality. If so, the
market may extend its suspicion of audited works beyond the im-
paired auditor to other auditors. A systemic failure of the market
for audit services occurs when the mistrust infects many clients
of many auditors. The extent of the contagion will depend on the
market perception of commonality in audit services across clients
and auditing firms. Spillover effects are limited when investors re-
gard audit services to be highly idiosyncratic, so that any perceived
accounting fraud is primarily confined to the client where the
deception originated.
We consider three hypotheses of market perception of com-
monality in audit services. The collective hypothesis states that
the market considers the audit practices across all auditors to be
highly comparable. The industry hypothesis affirms that the market
regards the audit practices of firms belonging to the same industry
to be highly comparable. The geography hypothesis contends that
the audit practices of auditors in the same audit office are highly
comparable. The collective hypothesis directly competes with the
other two hypotheses. The collective hypothesis predicts market
impact on all Big 5 clients. However, this hypothesis permits the
impact on the clients to differ by auditor firm. In contrast, the other
two hypotheses predict differential market impact on Big 5 clients
that follows industry and geography patterns. These hypotheses
imply that equity investors are much more discriminating and do
not simply penalize firms that are Big 5 clients.
Our focus is motivated by the organizational structure of audit
firms. It is well-known that auditors specialize in certain indus-
tries, and numerous papers have studied the implications of this
specialization for the audit industry.2 Industry specialization sug-
2 6
Gramling and Stone (2001) provide a review of this literature.
gests that the market demands specific industry expertise and
knowledge, presumably because audit practices differ across indus-
tries. Enron was an energy firm. Therefore, it is reasonable to exam-
ine whether any Andersen-induced spillover effect would be related
to energy industry concentrations. A finding of the existence of a
large energy industry effect would suggest that the market is skep-
tical of audit practices in the energy industry. Such a finding would
be in sharp contrast to one where the reputation effect spillover oc-
curs across different industries. It would also suggest that regula-
tions to prevent spillovers may need to be tailor-made for specific
industries.
Audit firms also tend to have a decentralized organizational
structure whereby contracting for and administration and delivery
of audit reports are conducted at the city-based office level (Francis
et al., 1999; Reynolds and Francis, 2001).3,4 This suggests the possi-
bility that stock price reputation effects are largely confined to cli-
ents of individual practicing offices; the market would primarily
penalize the audit office where the fraud occurs. In effect, the spill-
over of reputation effect is influenced by the location of individual
offices.
Our geography hypothesis is related to the literature on local
bias of investors. This literature reveals, for example, that domestic
investors overweight home securities and underweight foreign
securities in their portfolios.5 Recent studies show that the home
bias documented in the earlier literature is present even within
the confines of one country. Several studies have found that the
behavior of asset prices depends on geographic location (Coval and
Moskowitz, 2001; Loughran and Schultz, 2005; Barker and Loughran,
2007).6 This local bias of investors has been attributed to local
informational advantages, asset liquidity, familiarity, and other
behavioral characteristics.
Enron was headquartered in Houston and the shredding event
that ultimately led to the demise of Andersen originated in its
Houston office. Therefore, we consider whether the price effects
are most pronounced for Big 5 clients located in Houston and
whether they become less pronounced with increasing distance
from Houston. Since reasons for the presence of a local bias are less
applicable to big firms and because we examine the biggest audi-
tors and their biggest clients, evidence of a ripple effect emanating
from Houston would provide strong evidence in favor of a Houston
effect. Such evidence also has implications for regulators. A finding
of only a local spillover effect would stand in sharp contrast to a
contagion that broadly contaminates the entire market. It may sig-
nal the need for a market-wide regulation aimed at preventing a
systemic effect. More generally, evidence of a local office bias
would suggest that any reputation effect may be tempered by local
equity investors’ preferences.
By separately examining the industry and geography hypothe-
ses, we investigate alternative avenues through which reputation
spillover effect may occur. However, industry specialization and
office-level analysis may interact with one another. For example,
a study by Ferguson et al. (2003) investigated the effect of industry
specialization on audit fees in Australia and found that the effect
was apparent at the office level in city-specific audit markets and
not at the audit firm level. Moreover, firms in some industries tend
to cluster in the same location (Krugman, 1991; Ellison and Glae-
ser, 1997). This is true of energy firms, and nearly half of Ander-
sen’s Houston clients were in the energy industry. By
simultaneously examining the energy industry and Houston
3
Partner compensation schemes are also weighted disproportionately toward
office-level profits and not firm-wide profits.
4
Although audits are primarily conducted at the office level, some large multi-site
audits are performed by several offices.
5
See Kalev et al. (2008) and Karlsson and Nordén (2007) for recent examples.
Jiménez et al. (2009) relate distance to the use of collateral for business loans.
 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265
effects, we examine whether the reputation effect is influenced by
industry specialization at the (Andersen’s) Houston office level.
A growing literature examines the link between auditor reputa-
tion and equity prices.7 Earlier papers examine this link in the con-
text of IPO pricing, auditor litigation, or actions by regulatory
authorities against auditors. For example, in his study of UK auditors,
Firth (1990) found that the share price of auditor clients reacted neg-
atively when the auditors were criticized by the British Department
of Trade.
More recent papers analyze the valuation effects around Ander-
sen-related events. They show that equity prices are affected by
auditor reputation and provide evidence of spillover effects. Cha-
ney and Philipich (2002) showed that when the shredding of En-
ron-related documents became public, Andersen audit clients
experienced abnormal stock price declines. Chaney and Philipich
also reported that Andersen’s Houston clients suffered more severe
negative abnormal returns than other clients. However, they did
not examine the severity by distance from Houston nor did their
analysis focus on other Big 5 audit clients.8 Autore et al. (2009) find
that the stock price declines are bigger for firms with greater infor-
mation uncertainty. Rezaee et al. (2004) examine the events that oc-
curred between October 16, 2001 to September 3, 2002 and the
effect of auditor’s reputation on Andersen’s clients. Krishnan
(2005) focuses on the timeliness of Andersen’s Houston clients in
recognizing bad news on their earnings. Papers by Asthana et al.
(2003), Callen and Morel (2003), and Doogar et al. (2003) report that
the Andersen affair spilled over not only to Andersen clients but to
other Big 5 clients as well.9 Cahan et al. (2005) have argued that this
spillover effect extended beyond Andersen’s US clients to its non-US
clients. Krishnamurthy et al. (2006) focused on the abnormal returns
around March 14, 2002 when Andersen was criminally indicted. In
short, the earlier literature provides direct evidence of damage to
an auditor’s reputation as having a negative impact on its clients
as well as those of other auditors.
Our empirical analysis first replicates the earlier results. We
found that the market penalized Andersen’s non-Enron clients,
especially after news that Andersen had shredded Enron-related
documents became known. We also found that the market penal-
ized clients of other Big 5 accounting firms as well. However, our
results show that not all clients were equally penalized by the
shredding news. After Enron documents were shredded by Ander-
sen’s Houston office, we observed that the market impact declined
systematically with increasing distance from Houston, the epicen-
ter of the event. This result is inconsistent with the collective
hypothesis but is supportive of the geography hypothesis. We also
found that the market reaction to the shredding news can be differ-
entiated by industry. The negative abnormal returns were most
pronounced among energy clients of Big 5 accounting firms. This
result is again inconsistent with the collective hypothesis but is
supportive of the industry hypothesis.
The energy industry and the Houston office effects are econom-
ically important explanations for the spillover of Andersen’s repu-
tation effect. Based on three-day market model-adjusted
cumulative abnormal return (CAR) following the release of the
shredding news, the aggregate decline in market value of firms
in our sample is over $21.5 billion. Energy firms’ losses amount
7 office. We excluded five firms with headquarters outside the United
We differentiate the literature on auditor reputation and stock prices from the
large literature on auditor reputation and audit quality. The latter examines the
relation between auditor brand name and proxies for audit quality.
8
Chaney and Philipich (2002) provide some evidence on the other Big 5 clients in
their Table 7, but they do not focus on whether the market impact for Andersen may
spill over to other audit firms and conclude that the analysis ‘‘is yet to be determined”
(p. 1244). For example, they do not test the difference between Andersen and other
Big 5.
9 10
Asthana et al. (2003) also find that the share price reaction of clients is due to
Andersen-related events and not Enron-related events.
1257
to 71% of the total and firms within 100 miles of Houston equal
46% of the aggregate amount. Jointly, the total losses suffered by
firms that are energy firms or are within 100 miles of Houston
are equivalent to $19.6 billion or 91% of the aggregate decline.
These results are all the more impressive because out of a total
of 1364 firms in our sample, there are only 60 firms within 100
miles of Houston, and of the 1364 firms, only 56 are energy firms.
Together, firms either in energy industry or within 100 miles of
Houston account for less than 7% of the total sample.
Our evidence suggests that the stock market was much more dis-
cerning in its indictment of Andersen. Stock price effects are much
more pronounced for energy firms. This indicates that the market
delineates between the industry that Enron belongs to and other
industries. The evidence suggests accounting regulations that are
tailor-made for individual industries. Stock price effects are also
influenced by location. Investors chose to penalize the Houston of-
fice’s clients specifically and other clients in close geographic prox-
imity to it. If the affected industry is highly concentrated, the
evidence raises doubts about the need to adopt market-wide or
even firm-wide regulations to prevent systemic reputation effects.
The remainder of the paper is organized as follows: the next
section describes the sample data and the methodology. Section
3 presents the empirical results. Section 4 contains the conclusion.
2. Sample data
2.1. Data sources
Our sample consists of firms in the S&P 1500 companies that are
audited by the Big 5 accounting firms. Therefore, our results may
not generalize to all Big 5 clients or even those in the larger COM-
PUSTAT population. We also restrict our sample to clients who
have the same auditor between November 8, 2001 and February
4, 2002. The start and end dates span important Andersen events
and correspond to two of them. Most notably, this restriction ex-
cludes Enron for changing auditors on January 11, 2002.
We are concerned with spillover arising from the Enron–Ander-
sen debacle. However, the contagion risk inherent in the banking
industry is well recognized, and banks are supervised and regu-
lated by numerous government agencies in order to mitigate this
danger. Therefore, we have excluded banks in the S&P 1500 from
our analyses.
We obtained the S&P 1500 companies by selecting firms with
S&P Primary Index Marker (CPSPIN) coded as 1, 2, or 3 in the
2001 Compustat industrial annual files. Compustat annual files as
of the end of 2001 are used to obtain total assets, market value,
book value of common equity, long-term debt, sales growth from
2000 to 2001, and accruals. Price data and daily returns to compute
volatility come from Center for Research in Security Prices (CRSP).
Finally, Auditor/Auditor’s Opinion codes from Compustat and
firms’ 8-Ks and 10-Ks from EDGAR are used to identify the auditor
of individual companies.
The industry classification and distance from Houston play cru-
cial roles in our analyses. Firms are classified into industry catego-
ries according to the scheme used in Fama and French’s 10 Industry
Portfolios.10 Distance from Houston is measured as the distance be-
tween the location of company headquarters and Andersen Houston
States. Zip codes of corporate headquarters were obtained from the
SEC Compact Disclosure database. We then matched the zip codes to
the Census 2000 US Gazetteer ZIP Code File to acquire their latitude
and longitude coordinates. Using the coordinates, we computed an
The industry classifications are available from http://mba.tuck.dartmouth.edu/
pages/faculty/ken.french/data_library.html.
1258 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265

arc-length distance between Houston and each company i in our
sample as follows:
DFHi ¼ arccosfcosðlat i Þ  cosðloni Þ  cosðlatHouston Þ
 cosðlonHouston Þ þ cosðlati Þ  sinðloni Þ  cosðlatHouston Þ
2pr
 sinðlonHouston Þ þ sinðlat i Þ  sinðlat Houston Þg 
360
where DFH is distance from Houston and lat and lon are latitudes
and longitudes (measured in degrees) of firm headquarters and
Andersen Houston office locations, and r is the radius of the earth
(3963 mi). The zip code for Andersen Houston office would place
it in downtown Houston.
For our empirical estimation, we also needed stock return data.
Individual firm returns and market index returns were taken from
the CRSP Daily database.
2.2. Sample statistics
Tables 1 and 2 present the sample summary statistics. Table 1
contains statistics for the control variables we use in our regression
analyses. They are provided for all Big 5 clients (Panel A) as well as
Andersen (Panel B) and other Big 5 (Panel C) clients, which we refer
to as Big 4 clients. They show that Andersen clients are about 20%
of our sample. On average, Andersen’s clients are significantly
smaller than the Big 4 clients as measured by market value and
book value. However, total assets are statistically insignificantly
different despite their differences in magnitudes. The comparison
also reveals that Andersen’s clients on average are significantly clo-
ser to Houston by over 100 miles and tend to be significantly less
volatile. The two sets of clients have similar average prices, long-
term debts, sales growth, and accruals.
Table 1 also includes statistics for distances of firms’ headquar-
ters from Houston (DFH). An average or median firm’s headquar-
ters is about 1100 miles from Houston. It is also about 100 miles
more for a Big 4 client than for an Andersen client. The lower quar-
tile for DFH is 874 miles for all Big 5 clients, well above the 100-
mile radius we focus on in our analyses below.
Fig. 1 depicts the geography of Big 5 clients’ headquarters. The
map shows that the clients are located all over the US but most
are in the eastern half of the country. The map also highlights
the layout of energy firms from those of other clients and plots a
500 miles radius from Houston. Houston appears to be a favored
location of energy firms, and most of the energy clients seem to
be situated within 500 miles of Houston.
Table 2 presents the industry distribution of our sample based
on Fama and French’s 10 Industry Portfolios. Panel A shows the
number of S&P 1500 clients by Big 5 auditor in each category.
The energy industry has the third smallest number of clients (56)
after the telecommunication and consumer durables industries.
Andersen has by far the highest number of energy clients (23) fol-
lowed by Price Waterhouse Coopers (16). Panel B reports the num-
ber of clients as a percentage of all the Big 5 clients in each
industry. The percentages reveal the specialization of the auditors
in each industry. Not surprisingly, Andersen has the highest per-
centage of energy clients at 41%, but, perhaps more surprisingly,
it is the highest concentration across all industries and auditors.

Table 1
Descriptive statistics for sample firms. The sample consists of S&P 1500 companies, with the exception of banks, that are audited by Big 5 accounting firms. All data for total assets
(TA), market value (MV), book value of common equity (Book), long-term debt (Ltdebt), sales growth (Salegrow) from 2000 to 2001, and accruals are from Compustat annual files
and are as of the end of 2001. We follow Sloan (1996) and measure accruals as the change in non-cash current assets, less the change in current liabilities (exclusive of short-term
debt and taxes payable), less depreciation expense, all scaled by the average of the beginning and end of year total assets. The distance from Houston (DFH) is the arc-length
distance between the headquarters of the companies in our sample and downtown Houston. The DFH variable excludes firms with headquarters outside the United States.
Volatility is the standard deviation of daily stock returns during the estimation window, which includes 250 trading days prior to November 1, 2001. Daily returns and price data
come from CRSP. ** and *** indicate 5% and 1% significance levels, respectively, for t-tests on the difference in means between the Andersen’s clients and Big 4 clients.

Lower Upper
Variable N Mean Standard deviation Minimum Quartile Median Quartile Maximum
Panel A: Big 5 clients
Price ($) 1382 30.85 22.67 3.04 18.00 26.73 39.02 530.00
TA ($ Mill.) 1382 10,886.21 52,561.02 32.44 517.84 1413.11 4524.62 1,051,450.00
MV ($ Mill.) 1381 7734.14 26,574.04 59.00 615.53 1525.12 4729.91 397,831.59
Book ($ Mill.) 1380 2282.59 7025.45 865.00 256.95 581.40 1677.77 152,071.00
Ltdebt ($ Mill.) 1379 2344.32 15,677.42 0.00 20.40 262.09 996.22 419,975.00
Salegrow (%) 1381 8.01 36.00 96.72 5.48 4.03 16.37 461.95
DFH (miles) 1377 1105 473 0 874 1129 1432 3898
Volatility (%) 1376 3.54 1.57 0.91 2.34 3.13 4.36 9.79
Accruals 1329 0.060 0.077 0.503 0.094 0.056 0.022 0.582
Panel B: Andersen clients
Price ($) 280 29.51 17.24 3.69 17.69 26.25 36.99 136.90
TA ($ Mill.) 280 7659.10 39,340.72 60.58 457.24 1340.12 4016.81 617,340.00
MV ($ Mill.) 279 4777.41*** 12,663.48 69.30 553.79 1241.97 3572.80 133,636.46
Book ($ Mill.) 280 1451.83*** 2960.30 245.55 228.24 500.30 1456.35 36,655.00
Ltdebt ($ Mill.) 280 2644.53 19,129.88 0.00 41.92 318.07 1061.16 314,733.00
Salegrow (%) 280 7.92 28.50 66.43 5.60 3.93 16.23 235.73
DFH (miles) 280 1011*** 482 0 709 1050 1400 1890
Volatility (%) 278 3.33** 1.42 1.12 2.23 3.03 4.06 8.75
Accruals 275 0.057 0.078 0.503 0.093 0.054 0.022 0.233
Panel C: Big 4 clients
Price ($) 1102 31.19 23.85 3.04 18.08 27.08 39.16 530.00
TA ($ Mill.) 1102 11,706.16 55,405.13 32.44 533.37 1440.10 4638.10 1,051,450.00
MV ($ Mill.) 1102 8482.72*** 29,014.86 59.00 645.25 1558.01 5108.31 397,831.59
Book ($ Mill.) 1100 2494.05*** 7712.76 865.00 266.61 593.89 1757.15 152,071.00
Ltdebt ($ Mill.) 1099 2267.83 14,677.92 0.00 19.08 256.12 977.60 419,975.00
Salegrow (%) 1101 8.04 37.68 96.72 5.48 4.03 16.40 461.95
DFH (miles) 1097 1129*** 468 0 894 1162 1446 3898
Volatility (%) 1098 3.59** 1.60 0.91 2.36 3.17 4.46 9.79
Accruals 1054 0.060 0.077 0.435 0.094 0.057 0.022 0.582
 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265 1259

Table 2
Industry distribution of Big 5 S&P 1500 clients. Fama–French 10 Industry Portfolios classification scheme is used for the industry classification. The Big 5 clients in our sample
consist of S&P 1500 companies, with the exception of banks, that are audited by Big 5 accounting firms. The Big 5 firms are Arthur Andersen (AA), Deloitte & Touche (D&T), Ernst &
Young (E&Y), KPMG, and Price Waterhouse Coopers (PWC).

Industries Auditor All

AA D&T E&Y KPMG PWC
Panel A: Number of clients
Consumer non-durables (NoDur) 16 12 26 12 20 86
Consumer durables (Durbl) 3 10 11 5 13 42
Manufacturing (Manuf) 48 36 62 26 71 243
Oil, gas, and coal extraction and prod. (Enrgy) 23 3 8 6 16 56
Business equipment (HiTec) 41 47 56 39 91 274
Telephone and TV transmission (Telcm) 7 3 6 4 6 26
Wholesale, retail, and some services (Shops) 28 32 41 36 35 172
Healthcare, medical equip., and drugs (Hlth) 20 10 28 12 31 101
Utilities (Utils) 31 23 4 4 28 90
Other (Other) 63 50 74 47 58 292
Total 280 226 316 191 369 1382
Panel B: Percentage of clients
Consumer non-durables (NoDur) 19 14 30 14 23 100
Consumer durables (Durbl) 7 24 26 12 31 100
Manufacturing (Manuf) 20 15 26 11 29 100
Oil, gas, and coal extraction and prod. (Enrgy) 41 5 14 11 29 100
Business equipment (HiTec) 15 17 20 14 33 100
Telephone and TV transmission (Telcm) 27 12 23 15 23 100
Wholesale, Retail, and Some services (Shops) 16 19 24 21 20 100
Healthcare, medical equip., and drugs (Hlth) 20 10 28 12 31 100
Utilities (Utils) 34 26 4 4 31 100
Other (Other) 22 17 25 16 20 100
Total 20 16 23 14 27 100

Fig. 1. Locations of Big 5 Clients. The map shows the locations of all Big 5 clients’ headquarters in our sample except for two firms in Hawaii and one firm in Alaska. It also
plots a 500 miles radius from downtown Houston.

3. Empirical results
We begin with the evidence that the energy industry and the
Houston office effects are economically important explanations
for the spillover of Andersen’s reputation effect following the
shredding announcement. This is apparent in the change in market
values over a 3-day window as reported in Table 3.11 Panel A shows
that Andersen clients, excluding Enron, experienced an abnormal de-
cline of close to $4 billion over a 3-day period. The Big 4 clients en-
dured losses that were more than four times those of Andersen. The
Big 5 as a whole suffered a total abnormal drop of over $21.5 billion.
11
The underlying numbers for Table 3 are not presented in the paper, but are
available from the authors upon request.
Panel B shows the changes in market value by energy and non-en-
ergy firms. Losses experienced by the energy firms exceed twice
those of non-energy firms. Energy firms’ losses amounted to 71%
of the total abnormal decline in market value. Panel C presents the
changes in market value by distance from Houston. It shows that
losses are essentially confined to clients who are located less than
500 miles from Houston. Firms within 100 miles of Houston experi-
enced drops equal to 46% of the total abnormal drop. Panel D reports
the changes in market value by energy industry or by firms close to
Houston. On the day of the shredding announcement and 2 days fol-
lowing, energy firms or firms within 100 miles of Houston experi-
enced losses equivalent to more than 90% of the total Big 5
decline, while accounting for less than 7% of the number of firms
in the sample.
1260 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265

Table 3
Estimated impact on market value ($ Mill.) for (0, +2) window period. This table reports the abnormal market value impact of the January 10, 2002 document shredding event. The
cumulative abnormal returns (CARs) are calculated using parameters from the market model. Firms are excluded from the calculation if they do not have at least 120 observations
over the estimation period of 250 trading days. The energy industry is defined according to the Fama–French 10 Industry Portfolios. The distance from Houston (DFH) is the arc-
length distance between the headquarters of the companies in our sample and downtown Houston. The DFH variable excludes firms with headquarters outside the United States.
Panels A and B exclude five firms with headquarters outside the United States to be consistent with Panels C and D.

N Mean 25th%tile Median 75th%tile Sum

Panel A: Estimated impact on market value ($ Mill.) by auditor
Big 5 clients
CAR 1364 1.319 3.694 1.035 1.047
Gain (loss) 1364 15.825 55.365 10.866 13.436 21,584.906
Andersen clients
CAR 277 1.717 4.120 1.718 0.472
Gain (loss) 277 13.822 55.261 17.086 6.701 3828.779
Big 4 clients
CAR 1087 1.217 3.462 0.920 1.194
Gain (loss) 1087 16.335 55.378 10.217 15.141 17,756.127
Panel B: Estimated impact on market value ($ Mill.) by industry
Energy firms
CAR 56 6.242 8.741 6.110 3.561
Gain (loss) 56 272.365 288.692 127.946 36.086 15,252.453
Non-energy firms
CAR 1308 1.108 3.308 0.919 1.151
Gain (loss) 1308 4.841 49.762 9.213 15.051 6,332.454
Panel C: Estimated impact on market value ($ Mill.) by distance from Houston
Firms located within 100 miles from Houston
CAR 60 5.213 8.461 5.171 2.019
Gain (loss) 60 164.978 207.895 41.443 11.892 9898.686
Firms located between 100 and 500 miles from Houston
CAR 107 2.325 4.864 1.914 0.145
Gain (loss) 107 102.833 96.522 28.219 2.461 11,003.120
Firms located more than 500 miles from Houston
CAR 1197 1.034 3.174 0.830 1.213
Gain (loss) 1197 0.571 47.243 8.840 15.764 683.101
Panel D: Estimated impact on market value ($ Mill.) for local or energy firms
Energy firms or firms located within 100 miles from Houston
CAR 93 5.321 8.345 4.928 2.661
Gain (loss) 93 210.695 213.041 83.000 22.138 19,594.653
Energy firms or firms located within 500 miles from Houston
CAR 182 3.509 6.460 3.436 0.563
Gain (loss) 182 137.281 153.992 38.836 5.525 24,985.195

Our results provide explanations for the spread of reputation ef-
fect among Big 5 clients. Investors appeared to regard the Big 5
accounting firms as adopting similar auditing practices for energy
firms. They also seemed to attribute clients that were physically
close to the Andersen’s Houston office to be tainted by practices at
the Houston office. It appears to be a case of guilt by physical asso-
ciation. What our results reveal is that the reputation effect did not
spread simply because a firm was a member of the Big 5 club.
3.1. Replication of earlier studies
Chaney and Philipich (2002) examined several Enron–Andersen
event dates and found that January 10, 2002, the day Andersen an-
nounced that its Houston office had shredded Enron-related docu-
ments, had the most pronounced negative impact on Andersen
clients. They also observe that Andersen’s Houston clients suffered
bigger price declines than Andersen’s non-Houston clients. Asthan-
a et al. (2003), Doogar et al. (2003) and others found that Big 4 cli-
ents besides those of Andersen also experienced negative market
impacts.
We replicate the earlier results using the same event dates
examined by Chaney and Philipich. We use the parameters of the
market model with CRSP value-weighted market returns estimated
over 250 days prior to November 1, 2001, which is five trading days
before the first event date we examined. The results are robust
with respect to estimating the market model over 250 days prior
to five trading days before a specific event we examined. In calcu-
lating mean CARs, we used only firms that have at least 120 obser-
vations during the estimation period. To test the significance of the
CARs, we relied on a corrected version of the Z-test based on Mik-
kelson and Partch (1988) and a generalized sign test. Tests of dif-
ferences in CARs are based on the Wilcoxon rank sum tests and
t-tests. We also examine the CARs for 2-day (0, +1), 3-day (0, +2),
4-day (0, +3), and 5-day (1, +3) windows, where 0 denotes the
event day.
Our results for all Andersen clients, Andersen’s Houston clients,
and Andersen’s non-Houston clients clearly highlight the Andersen
shredding date, January 10, 2002, as the event when Andersen cli-
ents endured statistically significant abnormal returns that are also
economically significant.12 Accordingly, we focus on the shredding
event in the rest of our analyses. Additionally, in terms of both mag-
nitudes and significance, the price effect appears to peak 2 days after
the event date, and we only present, in the paper, the results for
(0, +2) window to reduce the number of tables.
Next, we conduct the analyses for Big 4 clients. The results show
that Big 4 clients suffer highly significant abnormal decreases in
market valuations. Further analyses reveal that Andersen clients
experience bigger significant declines than the Big 4 clients only
for 2-day and 3-day windows. These results suggest that the mar-
ket primarily penalize Andersen clients for the shredding incident,
supporting Chaney and Philipich’s focus on Andersen’s reputation
effect. In contrast, the price effects are not significantly different
between the two sets of clients for (0, +3) and (1, +3) windows.
12
Tables of the results are available from the authors upon request.
 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265 1261

Table 4
Cumulative abnormal returns by distance from Houston and by industry for (0, +2) window period. The energy industry is defined according to the Fama–French 10 Industry
Portfolios. The distance from Houston (DFH) is the arc-length distance between the headquarters of the companies in our sample and downtown Houston. The DFH variable
excludes firms with headquarters outside the United States. The document shredding event’s cumulative abnormal returns (CARs) are calculated using parameters from the
market model. Firms are excluded from the calculation if they do not have at least 120 observations over the estimation period of 250 trading days. The test statistics are a
corrected version of the Z-test based on Mikkelson and Partch (1988) and a generalized sign test. *, **, and *** denote significance at 5%, 1%, and 0.1% levels, respectively, based on
one-tailed tests. The Wilcoxon rank sum tests or t-tests are used to test the significance of the differences and +, ++, and +++ denote significance at 10%, 5%, and 1% levels,
respectively, based on one-tailed tests.

Less than 100 miles 100–500 miles Greater than 500 miles Diff. (%)
CAR (%) (1) Z Sign Z CAR (%) (2) Z Sign Z CAR (%) (3) Z Sign Z (1)–(2) (2)–(3) (1)–(3)
Big 5 clients (N = 1364)
Energy 7.28 6.290*** 4.339*** 5.84 4.642*** 4.241*** 5.13 3.648*** 3.307*** 1.44++ 0.71 2.15++
Others 3.93 4.353*** 3.701*** 1.61 3.128*** 3.146*** 0.98 6.539*** 6.204*** 2.32+++ 0.63 2.95+++
Diff. (%) 3.35+++ 4.23+++ 4.15+++
Andersen clients (N = 277)
Energy 6.56 3.516*** 2.968** 5.73 3.374*** 3.145*** 3.07 1.242 1.964* 0.83 2.66++ 3.49++
Others 3.09 1.834* 2.063* 0.82 0.578 0.904 1.32 3.967*** 4.098*** 2.27+ 0.50 1.77+
Diff. (%) 3.47++ 4.91+++ 1.75+
Big 4 clients (N = 1087)
Energy 7.75 5.243*** 3.182*** 5.98 3.191*** 2.845** 5.88 3.511*** 2.678** 1.77+ 0.10 1.87
Others 4.28 3.999*** 3.074** 1.79 3.183*** 3.051** 0.90 5.341*** 4.904*** 2.49++ 0.89+ 3.38+++
Diff. (%) 3.47+++ 4.19+++ 4.98+++

The absence of significant differences indicates that clients do not
enjoy a reduced punishment if Andersen is not their auditing firm.
These results are supportive of the collective hypothesis. Finally,
we find that Andersen’s Houston clients suffer much higher signif-
icant impacts than Big 4 clients. For example, the difference is
more than 300 basis points on average over a 3-day window. This
suggests that Houston may play a special role in the contagion pro-
cess, and we explore this possibility with the industry and geogra-
phy hypotheses.
3.2. Industry and Houston office effects: Preliminary results
Auditors specialize in specific industries, and Enron belongs to
the energy industry. This raises the possibility that the contagion
effect observed for Andersen as well as Big 4 clients may have been
driven by energy clients. To examine the industry hypothesis, we
categorize firms into the 10 Fama and French’s industry portfolios.
Auditing work, from solicitation to final delivery of the report, is
primarily carried out by individual offices. Therefore, the market
may distinguish between various offices in accounting for the rep-
utation effect of the shredding incident. Since shredding occurred
in Andersen’s Houston office, investors may have penalized the
Houston office clients and clients in the Houston neighborhood
more severely. The latter may happen if investors regarded audit
work for clients close to the Houston office as having been tainted
by physical proximity to Andersen’s Houston auditors. Moreover,
the evidence above shows that the CARs are more negative for
Andersen’s Houston clients than for its other clients. Investors
may have also extended this Houston effect to all Big 5 clients. Fi-
nally, academic research has documented the presence of local
biases in many settings. One approach to examine the presence
of a Houston office effect is to categorize Big 5 clients by the dis-
tance of their headquarters from Houston. Specifically, we consider
clients located within 100 miles, between 100 and 500 miles, and
greater than 500 miles from Houston.13
Although not reported in the paper, separate analyses of the en-
ergy industry effect and the Houston effect reveal evidence consis-
tent with the presence of both effects. A natural next step is to
investigate the two effects simultaneously. It is possible that one
of the effects had a stronger reputation impact than the other. For
example, firms in the same industry tend to cluster together. In
the present case, almost half of Andersen’s Houston office clients
are in the energy industry. Therefore, it is important to ensure that
the impact attributed to the energy industry is not erroneously
attributed to the Houston effect and vice versa. To examine these
possibilities, we ask whether the distance effect is still present after
controlling for the industry effect. We also ask whether the energy
effect is still present after controlling for distance from Houston.
Table 4 presents a joint test of the two effects over the 3-day
window. Comparison across rows provides evidence of energy
industry effects after controlling for distance from Houston. The
energy industry effect remains alive and strong. In all cases, energy
firms experience bigger abnormal price declines than non-energy
firms irrespective of distance from Houston. The differences be-
tween energy and non-energy industries are economically impor-
tant and highly statistically significant.
Is the Houston office effect an industry effect? To examine this,
we controlled for industry by comparing the CARs across columns.
Controlling for energy firms, the CARs increase with increasing dis-
tance from Houston for all Big 5 clients. The distance effect is gen-
erally present for clients in non-energy industries. The CARs
generally increase significantly for clients located less than 100
miles to more than 100 miles away.
The results show that for the shredding event, the energy indus-
try effect is more dominant than the Houston effect. The latter ap-
pears to have an independent effect for non-energy firms but
appears to be weak for energy firms. These results suggest that the
earlier widespread spillover impacts observed for Andersen and
Big 4 clients is primarily attributable to the location of energy firms.
3.3. Industry and Houston office effects: Regression results
Are energy and distance effects still present after controlling for
determinants of cross-sectional variability in abnormal returns? To
address this question, we implement a multivariate regression
analysis to control for firm characteristics noted in the literature.
Specifically, we control for auditor, firm size as measured by natu-
ral log of market capitalization, book-to-market ratio, return vari-
ability, leverage as measured by long-term debt to total assets
ratio, sales growth rate, and accruals. The sales growth rate reflects
how aggressively revenues are being recognized. The last variable,
accruals, is a proxy for earnings quality.14

13 14
We also consider a continuous distance variable in our regressions below. Variability of returns and accruals are suggested by one of the referees.
1262 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265
We use pooled firm-level regressions. Therefore, to the extent
that the Enron shredding event spilled over to other firms, it af-
fected all firms simultaneously so that the regression error terms
in the regressions are likely to be cross-sectionally correlated. If
so, the standard errors may be understated. This problem has been
studied by, for example, Schipper and Thompson (1983), Sefcik and
Thompson (1986), Thompson (1985). Moreover, the regression er-
ror terms may also be heteroskedastic so that the estimation may
produce biased standard errors. To address these problems, our ap-
proach is to use bootstrapping procedures. This is a distribution-
free approach that has the added advantage of not requiring any
specific parametric assumption.
Following Chou (2004) and Hein and Westfall (2004), we boot-
strap on the regression residuals for the non-event period (T1) and
construct pseudo-returns for both the non-event and event periods
(T1 + T2). The specific steps in our bootstrapping procedure are
1. Run the market model regression for each firm over the estima-
tion window (T1) to obtain the OLS estimates and residuals.
2. Use the estimated coefficients from step 1 to calculate CAR for
each firm over the event window (T2).
3. Regress individual firm CAR on firm characteristics to obtain
cross-sectional regression coefficients.
4. For each firm, randomly draw a bootstrap sample (with replace-
ment) of T1 + T2 residuals from the residuals in step 1.
5. For each firm, use the bootstrap residuals from step 4 and esti-
mated coefficients from step 1 to generate pseudo firm returns
for both the estimation window and the event window.
6. Repeat steps 1, 2, and 3 for the pseudo-return data created in
step 5.
7. Repeat steps 4, 5, and 6 for 1000 times.
8. Use the 1000 bootstrap replications of the coefficients in step 3
to compute the bootstrap standard error and t statistics for
these coefficients.
Table 5 contains the regression evidence. The dependent vari-
ables are 3-day CARs. Regressions (1)–(6) show the results of add-
ing additional control variables. Regressions (5) and (6) include all
the variables and differ only in using discrete versus continuous
distance variables.
The distance indicator for less than 100 miles (Distance 1) is
negative and significant in all cases, even when the full regression
is estimated with all the variables. The Distance 2 indicator vari-
able for firms between 100 and 500 miles away becomes insignif-
icant when the energy classification is introduced in regressions 2–
6. When a continuous distance variable, defined as the logarithm of
the inverse of DFH is used, it is significantly negative, indicating
that market impact declines with distance. These results support
the presence of an independent Houston office effect and are
inconsistent with the collective hypothesis.
With regard to industry indicator variables, Table 5 shows that
the energy indicator variable has very negative coefficients and the
estimates are highly statistically significant. We also estimated the
regressions in Table 5 using the full set of industry variables, but
these are discussed in the next section and are not presented in
the paper. The results confirm the presence of a strong energy
industry effect and are inconsistent with the collective hypothesis.
Of the remaining significant coefficients, those for accruals,
which is an inverse measure of earnings quality, are significantly
negative. They suggest that the lower the earnings quality, the big-
ger the drop in market value after the shredding announcement
date. Accruals may also have influenced another variable that has
implications for aggressive accounting practices, i.e. firm size. Firm
size is insignificant in all regressions except regression (6) where it
is barely positively significant. Ex ante, bigger firms tend to be more
reputable, which may increase their CAR values. However, bigger
firms may also offer more opportunities for aggressive accounting
practices, which may decrease their CAR values. We observe that
net effect is neutral, suggesting that the two effects tend to cancel
each other.
Regression (7) in Table 5 shows the combined effect of being an
energy firm that is located within 100 miles from Houston by
including an interaction term between energy and distance, Enrgy-
AndDist1. For this regression, in addition to the interaction term,
we constructed two new indicator variables. First, the variable
EnrgyNotDist1 is one for energy firms located more than 100 miles
from Houston, and zero otherwise. Second, the variable Dist1NotE-
nrgy is one for non-energy firms within 100 miles from Houston
and zero otherwise. The result shows highly significant negative
effects of all three variables with the interaction term exhibiting
the most impact and significance, revealing the importance of
being close to Houston and being an energy firm.
3.4. Robustness checks
Our results are highly robust. Even though the results based on
value-weighted market returns are reported, they are robust with
respect to equal-weighted market returns, S&P 500 index returns,
and Russell 3000 index returns. The results are also robust with re-
spect to different models for computing abnormal returns. Our ba-
sic model for calculating abnormal returns is the market model.
This is the predominant choice of Chaney and Philipich and other
researchers. We obtain similar results when we employ instead
multiple-factor Fama–French models and a model of raw returns.
In the latter, we use raw returns net of value-weighted CRSP re-
turns. This model has the advantage of not losing observations
due to missing values prior to the events. The robustness of our re-
sults with respect to alternative models of abnormal returns is not
surprising given the short-horizon nature of our tests. Kothari and
Warner (2006) find that unlike long-horizon tests, risk adjustment
in short-horizon tests is unimportant and risk adjustment errors
are likely to be small. Brown and Warner (1985) also conclude that
simple risk adjustment is adequate for short-window event
studies.
We conduct three sets of additional robustness checks of our
regression analyses. First, we estimate the regressions in Table 5
using the full set of industry indicator variables. We find that the
manufacturing clients have significant coefficients but are about
three times less negative than the energy coefficients. Clients in
the health and the utility industries have significant coefficients
as well but they are positive. Therefore, membership in these
industries increased their CARs following the shredding incident.
In short, the results continue to be supportive of a strong energy
industry effect.
Second, we estimate the regressions using the raw-return mod-
el for computing abnormal returns. The evidence is again support-
ive of the Houston effect. Both the discrete and continuous
distance variables are negative and significant, and the results re-
main strongly supportive of the energy industry effect. The coeffi-
cients for the industry portfolios are compatible with those
reported in Table 5 and provide the same inferences.
Third, the evidence presented in the paper so far compared
Andersen to Big 4 clients. There are many more of the latter than
the former. Therefore, we check to examine whether our results
may have been affected by an unbalanced number of observations.
Specifically, we compare Andersen to E&Y. We choose E&Y as the
comparison auditor because E&Y with 316 clients is closest to
Andersen with 280 clients, giving us a more balanced sample.
The results show that the discrete and continuous distance vari-
ables remain negative and significant. As for industry effects, en-
ergy retains its strong negative influence on the CARs and are
highly significant. Therefore, the evidence supportive of distance
 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265 1263

Table 5
Cross-sectional regressions for (0, +2) window period.
This table reports cross-sectional regression results for the document shredding event. The dependent variable is the individual firm’s 3-day cumulative abnormal returns
(CARs). The explanatory variables consist of the following indicator variables and firm characteristics:
Distance1 = 1 if the firm headquarter is located within 100 miles of downtown Houston, and 0 otherwise
Dist1NotEnrgy = 1 if the firm headquarter is located within 100 miles of downtown Houston but the firm is not in oil, gas, and coal extraction and products industry, and 0
otherwise
Distance2 = 1 if the firm headquarters are between 100 miles and 500 miles from downtown Houston, and 0 otherwise
Ln(1/DFH) is the natural log of inverse of the arc-length distance between company headquarters and downtown Houston
Enrgy = 1 for firm in oil, gas, and coal extraction and products industry, and 0 otherwise
EnrgyNotDist1 = 1 if the firm is in oil, gas, and coal extraction and products industry but is not within 100 miles of Houston, and 0 otherwise
EnrgyAndDist1 = 1 if the firm is in oil, gas, and coal extraction and products industry and is within 100 miles of Houston, and 0 otherwise
AA = 1 for Arthur Andersen clients, and 0 otherwise
E&Y = 1 for Ernst &Young clients, and 0 otherwise
D&T = 1 for Deloitte & Touche clients, and 0 otherwise
PWC = 1 for PriceWaterhouse Coopers clients, and 0 otherwise
Size is the natural log of market value as of the end of 2001
B/M is the book-to-market ratio as of the end of 2001
Salegrow is the sale growth from 2000 to 2001
Leverage is long-term debt divided by total assets
Volatility is the standard deviation of daily stock returns during the estimation window
Accruals is the change in non-cash current assets, less the change in current liabilities (exclusive of short-term debt and taxes payable), less depreciation expense, all scaled
by the average of the beginning and end of year total assets
The bootstrap t value is in parenthesis. *, **, and *** denote two-tailed significance at the 10%, 5%, and 1% levels, respectively
(1) (2) (3) (4) (5) (6) (7)
Intercept 1.034*** 0.984*** 0.949*** 2.013 1.850 3.420** 1.841
(5.88) (5.59) (4.86) (1.45) (1.29) (2.16) (1.28)
Distance1 4.180*** 2.727*** 2.713*** 2.567*** 2.565***
(5.52) (3.35) (3.33) (3.07) (3.07)
Dist1NotEnrgy 2.813***
(2.84)
Distance2 1.292** 0.682 0.678 0.545 0.556 0.513
(2.27) (1.17) (1.16) (0.92) (0.93) (0.85)
ln(1/DFH) 0.222***
(2.71)
Enrgy 3.919*** 3.885*** 4.171*** 4.161*** 4.698***
(5.06) (5.04) (5.22) (5.20) (6.25)
EnrgyNotDist1 4.447***
(4.67)
EnrgyAndDist1 6.359***
(5.30)
AA 0.186 0.057 0.273 0.297 0.269
(0.47) (0.14) (0.47) (0.52) (0.47)
E&Y 0.210 0.243 0.208
(0.38) (0.44) (0.37)
D&T 0.201 0.216 0.203
(0.33) (0.35) (0.33)
PWC 0.350 0.366 0.357
(0.67) (0.69) (0.68)
Size 0.196 0.201 0.204* 0.201
(1.59) (1.63) (1.66) (1.63)
B/M 0.472 0.466 0.496 0.462
(0.79) (0.78) (0.84) (0.77)
Salegrow 0.091 0.086 0.128 0.084
(0.17) (0.16) (0.24) (0.15)
Leverage 0.880 0.877 0.845 0.896
(0.83) (0.82) (0.79) (0.84)
Volatility 0.109 0.107 0.119 0.108
(0.66) (0.65) (0.72) (0.65)
Accruals 5.476** 5.547** 5.686** 5.538**
(2.07) (2.09) (2.14) (2.09)
Adjusted R2 0.041 0.066 0.066 0.079 0.078 0.074 0.077
Obs. 1364 1364 1364 1309 1309 1309 1309

and industry effects does not appear to be driven by an unbalance
number of observations.
3.5. Discussion
When Andersen shredded Enron-related documents on January
10, 2002, the auditor’s reputation was irreparably damaged. Inves-
tors not only questioned the quality of Andersen’s audit for its non-
Enron clients but also cast doubt on the quality of Big 4 clients as
well. Big 4 clients experienced sizable drops in market value in the
immediate aftermath of the shredding news. Since the clients of
the biggest auditors were located nationwide, this gave the appear-
ance of a widely-dispersed reputation effect. The market seemed to
have imposed a collective punishment on the Big 5 clients.
The results from our analyses suggest a more discriminating
market. We find a strong energy industry effect. Clients in the en-
ergy industry suffered the biggest negative abnormal returns. The
evidence also supports the presence of a Houston office effect.
1264 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265
Audit work is categorized by industry, and the work is carried out
by individual offices. Investors appear accordingly to differentiate
auditing practices by industry and by individual offices where
the work is done. The evidence is inconsistent with a market that
indiscriminately penalizes all Big 5 clients.
Our evidence also suggests different policy implications. The
regulatory authorities are charged with the task of preventing sys-
temic market-wide effects following a reputation loss by a big
auditing firm. Results from earlier academic research on the En-
ron–Andersen saga appear to support the validity of such concerns
and for market-wide regulations. Our results provide support for
more targeted rules. In particular, our evidence calls for regulations
that are tailor-made to individual industries. The geographic prop-
agation of a reputation crisis associated with a client in a particular
industry appears to depend on the locality of the firms in the
industry. For example, to the extent the location of a specific indus-
try is dispersed widely, it may give the appearance of a widely dis-
persed effect. On the other hand, if all the firms in an industry are
concentrated in one location, the reputation effect would be con-
fined to the place where the crisis originated.
Our implication for policy aimed at industry-focused auditing
regulation is very different from the prevailing focus. To date, nei-
ther the Public Company Accounting Oversight Board nor the
Auditing Standards Board have proposed or approved by-industry
standards. They have suggested procedures for individual indus-
tries, but not mandated standards.
4. Conclusion
The demise of Enron and Arthur Andersen are momentous
financial events in recent memory. When Andersen admitted to
shredding Enron-related documents, Andersen clients as well as
those of other Big 5 auditing firms realized significant decline in
their market valuations. Stock price declines due to Andersen’s
reputation impairment arising from its audit work for Enron ap-
pear to spill over far and wide. The market gives the impression
of having imposed a collective punishment on Big 5 clients.
Our empirical analyses of the shredding announcement event
yield several findings. We confirm the earlier reputation effect
for both Andersen and other Big 5 clients. More importantly, we
show that the valuation impacts are not homogeneous across Big
5 clients. Rather, the market impacts are related to Enron being
an energy firm and to its auditor being Andersen’s Houston office.
Specifically, we find that the strongest and most robust abnormal
price declines are reserved for energy clients. The market appears
to fault the audit practices used for the energy industry by the
Big 5 auditors. We also find bigger price drops for firms closest
to Andersen’s Houston office. The market appears to blame the
auditing work done by Andersen’s Houston office and regard cli-
ents located close to the Houston office as having been tainted
by their proximity to Houston. The energy industry effect and
the Houston office effect jointly account for about 91% of the total
abnormal decline in market value of Big 5 clients over the 3 days
following the shredding announcement.
Our evidence is inconsistent with a collective market punish-
ment when Andersen lost its reputation over the shredding inci-
dent. Investors do not appear to regard the quality of audit work
done by all Big 5 auditors to be of similar quality. Instead, the evi-
dence is consistent with a much more discriminating market. It
suggests that Andersen’s reputation effect spills over because of
common audit practices in the energy industry and because of
audit practices unique to Andersen’s Houston office. These expla-
nations of the events following the shredding announcement are
compatible with industry specialization by auditors and with audit
work primarily accomplished at the individual office level. The
shredding spillover is also supportive of accounting regulations
customized for specific industries rather than market-wide rules
and polices that fit all industries.
Acknowledgements
This paper was originally submitted to Professor Giorgio Szego
on November 12, 2007, and was revised twice prior to submission
through EES. We thank Tim Loughran, Rick Mendenhall, Dave Ric-
chiute, and the two referees for their comments and suggestions.
References
Asthana, S.C., Balsam, S., Krishnan, J., 2003. Audit firm reputation and client stock
price reactions: Evidence from the Enron experience. Working Paper, Temple
University.
Autore, D.M., Billingsley, R.S., Schneller, M.I., 2009. Information uncertainty and
auditor reputation. Journal of Banking and Finance 33, 183–192.
Barker, D., Loughran, T., 2007. The geography of S&P 500 stock returns. Journal of
Behavioral Finance 8, 177–190.
Brown, S.J., Warner, J.B., 1985. Using daily stock returns: The case of event studies.
Journal of Financial Economics 14, 3–31.
Cahan, S.F., Emanuel, D., Sun, J., 2005. Shredded reputation: Did Andersen’s non-US
clients suffer and why? Working Paper, University of Auckland.
Callen, J.L., Morel, M., 2003. The Enron–Andersen debacle: Do equity markets react
to auditor reputation? Finance Letters 1, 1–5.
Chaney, P.K., Philipich, K.L., 2002. Shredded reputation: The cost of audit failure.
Journal of Accounting Research 40, 1221–1245.
Chou, P.H., 2004. Bootstrap tests for multivariate event studies. Review of
Quantitative Finance and Accounting 23, 275–290.
Coval, J.D., Moskowitz, T.J., 2001. The geography of investment: Informed trading
and asset prices. Journal of Political Economy 109, 811–841.
Doogar, R., Sougiannis, T., Xie, H., 2003. The impairment of auditor credibility: Stock
market evidence from the Enron–Andersen saga. Working Paper, University of
Illinois.
Ellison, G., Glaeser, E.L., 1997. Geographic concentration in US manufacturing
industries: A dartboard approach. Journal of Political Economy 105, 889–927.
Firth, M., 1990. Auditor reputation: The impact of critical reports issued by
government inspectors. Rand Journal of Economics 21, 374–387.
Francis, J.R., Stokes, D.J., Anderson, D., 1999. City markets as a unit of analysis in
audit research and the re-examination of Big 6 market shares. Abacus 35, 185–
206.
Ferguson, A., Francis, J.R., Stokes, D.J., 2003. The effects of firm-wide and office-level
industry expertise on audit pricing. Accounting Review 78, 429–448.
Gramling, A.A., Stone, D.N., 2001. Audit firm industry expertise: A review and
synthesis of the archival literature. Journal of Accounting Literature 20, 1–20.
Hein, S., Westfall, P., 2004. Improving tests of abnormal returns by bootstrapping
the multivariate regression model with event parameters. Journal of Financial
Econometrics 2, 451–471.
Jiménez, G., Salas, V., Saurina, J., 2009. Organizational distance and use of collateral
for business loans. Journal of Banking and Finance 33, 234–243.
Kalev, P.S., Nguyen, A.H., Oh, N.Y., 2008. Foreign versus local investors: Who knows
more? Who makes more? Journal of Banking and Finance 32, 2376–2389.
Karlsson, A., Nordén, L., 2007. Home sweet home: Home bias and international
diversification among individual investors. Journal of Banking and Finance 31,
317–333.
Kothari, S.P., Warner, J.B., 2006. Econometrics of Event Studies. In: Eckbo Espen, B.
(Ed.), Handbook of Corporate Finance: Empirical Corporate Finance, vol. 1.
Elsevier, North-Holland.
Krishnamurthy, S., Zhou, J., Zhou, N., 2006. Auditor reputation, auditor
independence and the stock market reaction to Andersen’s clients.
Contemporary Accounting Research 23, 465–490.
Krishnan, G.V., 2005. Did Houston clients of Arthur Andersen recognize publicly
available bad news in a timely fashion? Contemporary Accounting Research 22,
165–193.
Krugman, P., 1991. Increasing returns and economic geography. Journal of Political
Economy 99, 483–499.
Loughran, T., Schultz, P., 2005. Liquidity: Urban versus rural firms. Journal of
Financial Economics 78, 341–374.
Mikkelson, W.H., Partch, M.M., 1988. Withdrawn security offerings. Journal of
Financial and Quantitative Analysis 23, 119–133.
Rezaee, Z., Hunt, A.K., Lukawitz, J.M., 2004. Capital market reactions to auditor’s
reputation: Evidence from Andersen’s former clients. Journal of Forensic
Accounting 5 (December), 337–350.
Reynolds, J.K., Francis, J.R., 2001. Does size matter? The influence of large clients on
office-level auditor reporting decisions. Journal of Accounting and Economics
30, 375–400.
SEC (United States Securities and Exchange Commission), 2000. Proposed Rule:
Revision of the Commission’s Auditor Independence Requirements. 17 CFR
Parts 210 and 240, File No. S7-13-00, <http://www.sec.gov/rules/proposed/34-
42994.htm>.
 R.D. Huang, H. Li / Journal of Banking & Finance 33 (2009) 1255–1265 1265

SEC (United States Securities and Exchange Commission), 2001. Final Rule: Revision
of the Commission’s Auditor Independence Requirements. 17 CFR Parts 210 and
240, File No. S7-13-00, <http://www.sec.gov/rules/final/33-7919.htm>.
Sefcik, S.E., Thompson, R., 1986. An approach to statistical inference in cross-
sectional models with security abnormal returns as dependent variable. Journal
of Accounting Research 24, 316–334.
Schipper, K., Thompson, R., 1983. Evidence on the capitalized value of merger
activity for acquiring firms. Journal of Financial Economics 11, 85–119.
Sloan, R.G., 1996. Do stock prices fully reflect information in accruals and cash flows
about future earnings? Accounting Review 71, 289–315.
Thompson, R., 1985. Conditioning the return-generating process on firm-specific
events: A discussion of event study methods. Journal of Financial and
Quantitative Analysis 20, 151–168.
Watts, R.L., Zimmerman, J.L., 1983. Agency problems, auditing, and the theory of the
firm: Some evidence. Journal of Law and Economics 26, 613–633.