ARTICLE IN PRESS

Journal of Accounting and Economics 46 (2008) 279–293

Contents lists available at ScienceDirect

Journal of Accounting and Economics

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

The market reaction to Arthur Andersen’s role in the Enron scandal:

Loss of reputation or confounding effects?$
Karen K. Nelson, Richard A. Price, Brian R. Rountree 
Jones Graduate School of Management, Rice University, Houston, TX 77005, USA

a r t i c l e in fo abstract

Article history: This paper tests the hypothesis that negative client stock returns following the
Received 18 March 2008 revelation that Enron documents had been shredded are attributable to confounding
Received in revised form effects as opposed to a loss of Andersen’s reputation. We find that a sharp decline
15 September 2008
in oil prices along with differences in the industry composition of the Andersen and
Accepted 15 September 2008
Big 4 client portfolios combine to produce significantly more negative returns for
Available online 21 September 2008
Andersen clients relative to Big 4 clients, and for Andersen’s Houston office clients
JEL classification: relative to its clients in other locations. The market reaction to two other Enron-related
G14 events also offers little support for a reputation effect.
M41
& 2008 Elsevier B.V. All rights reserved.
M49

Keywords:
Auditor reputation
Arthur Andersen
Event studies
Earnings response coefficients

1. Introduction

A recent widely cited study by Chaney and Philipich (2002) (hereafter CP) utilizes damaging revelations regarding
Arthur Andersen’s role in the accounting fraud at Enron to assess the client valuation implications of a loss of auditor
reputation. The study finds that Andersen clients experienced significant negative stock market reactions over a 3-day
period beginning January 10, 2002, the date Andersen publicly acknowledged its employees had destroyed documentation
related to the Enron audit. Moreover, clients of Andersen’s Houston office, where the alleged incidents at Enron occurred,
suffered a larger stock price drop than the firm’s non-Houston clients. The evidence suggests the cumulative loss of
Andersen clients was in excess of $10 billion in market capitalization during this 3-day window, which CP interprets as the
client valuation effect attributable to Andersen’s damaged reputation.1 However, CP does not control for concurrent
negative news events that could explain these results.
As with any event study analysis, it is imperative to isolate the effect of the event being studied (e.g., Dyckman et al.,
1984). Our review of several major news sources reveals that negative macroeconomic and industry-related news was

$
We appreciate the helpful comments and suggestions of Jerry Zimmerman (the editor), Michael Willenborg (the referee), Jennifer Blouin, Gustavo
Grullon, Wayne Landsman, James Weston, and workshop participants at Southern Methodist University and Texas Tech University. Srinivasan
Krishnamurthy, Jian Zhou, and Nan Zhou generously provided some of the data for this study.
 Corresponding author.
E-mail address: rountree@rice.edu (B.R. Rountree).
1
We derive this figure by multiplying the mean 3-day loss of $37,115,000 (CP, Table 8) by 284 Andersen clients (CP, Table 5).

0165-4101/$ - see front matter & 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jacceco.2008.09.001
 ARTICLE IN PRESS
280 K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293

released during the event window. In particular, a number of news sources (see Appendix) attributed the decline in equity
prices during this period to concerns about the prospects for economic recovery and upcoming disappointing earnings
reports, as well as a substantial oil price drop in the Energy sector and overvaluation in the Information Technology sector.
We test whether these confounding events, coupled with differences in the industry composition of Andersen’s clients
relative to clients of the other big auditors (which we refer to as the Big 4 auditors), produce the pattern of stock returns
documented by CP.
Consistent with CP, we find Andersen clients experienced an average abnormal return of approximately 2% in the
(0, +2) event window surrounding the shredding announcement, a market response that is significantly more negative than
that of Big 4 clients. However, we show that the results are strongly influenced by the Energy sector, where Andersen
clients experienced an abnormal return of 5.70%, nearly twice as large in absolute terms as the return in any other sector.
Although this return is not significantly different from the 6.55% abnormal return for Big 4 Energy clients, Andersen
audited a disproportionate share of the companies in this sector. We show that these confounding factors help to produce a
more negative abnormal return for the portfolio of Andersen clients as a whole and for the energy-heavy Houston office in
particular.
Furthermore, the abnormal returns of Andersen and Big 4 clients are not significantly different in nine out of ten
industry sectors, inconsistent with the notion of a widespread client market reaction attributable to a decline in Andersen’s
reputation. Only in the Information Technology sector do we find that Andersen clients experienced more negative returns,
on average, than Big 4 clients. On the one hand, this evidence is consistent with a reputation effect; on the other hand, there
is no a priori reason to expect this reputation effect to be isolated to this sector. It is possible that the single statistically
significant difference out of ten industry sectors is simply random. Moreover, when we control for differences in size by
matching each Andersen client to a Big 4 client in the same sector that is closest in terms of market value of equity, there is
no longer a statistically significant difference in abnormal returns for this sector. Nevertheless, we are unable to rule out the
possibility that the results in this sector are related to Andersen’s damaged reputation, although the evidence taken as a
whole is not consistent with the shredding announcement having a significant reputation effect on the stock prices of
Andersen’s clients.
We also examine the market reaction to two related events – the February 2, 2002 release of the Powers Report
suggesting Andersen was complicit in Enron’s accounting maneuverings and the March 15, 2002 indictment of Andersen
for obstruction of justice. Consistent with CP, we find the market reaction surrounding the release of the Power’s Report
was significantly negative for Andersen clients, but not as pronounced as on the shredding date. In contrast to CP, however,
we also examine the stock price reaction of Big 4 clients. The results show that abnormal returns for Andersen clients are
not significantly different from Big 4 clients, inconsistent with the notion the market was reacting to a differential loss of
auditor reputation.
Prior research reports mixed findings regarding the indictment date. CP finds an insignificant market reaction for their
sample of S&P 1500 firms while Krishnamurthy et al. (2006) finds some evidence of a small effect for a broader sample of
Andersen clients. We replicate both of these findings, and conclude there was a marginal reaction, statistically and
economically, around the indictment date for Andersen’s smaller clients not in the S&P 1500. Because the market response
is limited to small clients, however, the indictment date evidence is inconsistent with a loss of auditor reputation. Overall,
the findings reported in this study show there is no consistent evidence of a reputation effect surrounding any of the
information events examined.
Our final set of tests examines earnings response coefficients (ERCs) of Andersen and Big 4 clients. If the shredding
announcement affected the market’s perception of audit quality, we would expect to observe lower ERCs for Andersen clients
following the announcement. However, we find no evidence that the market reaction to earnings surprises of Andersen
clients was lower following the shredding announcement compared to the same period in the preceding year or to Big 4
clients, as would be expected if damage to Andersen’s reputation was impounded in the market’s valuation of client earnings.
Taken together, our results indicate that the significant market decline for Andersen clients in the days following the
shredding announcement is not a reflection of the auditor’s damaged reputation. To the extent a loss of auditor reputation
triggered a market response, it appears to be limited to the Information Technology sector. However, even in this sector,
confounding effects make it difficult to attribute any stock price effect to Andersen’s sullied reputation. Similarly, the
market reaction to other Enron-related events and to earnings surprises offers no systematic evidence of a response
indicative of a reputation effect. Nevertheless, we caution that our study does not suggest that auditor reputation is
unimportant to the capital markets; rather, our analysis highlights the difficulty of identifying and quantifying a reputation
effect in the context of an event study contaminated by confounding factors.
Section 2 provides background and motivation. Sections 3 and 4 describe the research methods and present the
empirical findings. Section 5 summarizes and concludes.

2. Background and motivation

2.1. Research on auditor reputation

Audits are a mechanism to enhance the credibility of financial statements and reduce the client firm’s agency costs. The
assurance value of the audit is thus directly related to the reputation of the auditor (Watts and Zimmerman, 1983).
 ARTICLE IN PRESS
K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293 281

Consistent with this view, prior research demonstrates that the value of an auditor’s reputation is impounded in the stock
prices of its clients. For example, auditor reputation is positively associated with the value of initial public offerings (Titman
and Trueman, 1986; Beatty, 1989; Datar et al., 1991). Conversely, clients experience a decline in equity values surrounding
the announcement of information damaging to their auditor’s reputation (Franz et al., 1998; Firth, 1990; Weber et al., 2008).
Auditor reputation also affects the market’s valuation of client earnings. ERCs are significantly greater for companies with
more reputable auditors (Teoh and Wong, 1993; Francis and Ke, 2006) but decline following a loss of auditor reputation
(Moreland, 1995).
A second component of the value investors attach to an audit is the option value of the claim financial statement users
have on the auditor in the event of an audit failure (Simunic, 1980; Dye, 1993). Some studies suggest this insurance demand
for auditing dominates reputation concerns (Lennox, 1999; Khurana and Raman, 2004), while other studies conclude it is
difficult to distinguish between reputation and insurance effects (Baber et al., 1995; Franz et al., 1998). Consistent with the
latter view, Weber et al. (2008) suggests the negative returns CP documents likely reflect not only the market’s revised
assessment of Andersen’s reputation but also the auditor’s prospects for survival. In this study, we posit that the negative
client returns around the shredding announcement and related events are the result of confounding events, as discussed
further below. In this case, the distinction between reputation and insurance effects is moot as neither explains the market
response.

2.2. Enron and the reputation of Arthur Andersen

The revelation of accounting irregularities at Enron in the third quarter of 2001 caused regulators and the media to focus
extensive attention on Andersen. The magnitude of the alleged accounting errors, combined with Andersen’s role as
Enron’s auditor and the widespread media attention, provide a seemingly powerful setting to explore the impact of auditor
reputation on client market prices around an audit failure. CP investigates the share price reaction of Andersen’s clients to
various information events that could lead investors to revise their beliefs regarding Andersen’s reputation.
Perhaps most damaging to Andersen’s reputation was their admission on January 10, 2002 that employees of the firm
had destroyed documents and correspondence related to the Enron engagement. For a sample of S&P 1500 firms, CP reports
that in the 3-day window following the shredding announcement (0, +2), Andersen clients experienced a significant
2.03% market reaction, and this reaction was significantly more negative than for Big 4 clients. Andersen’s Houston office
clients, where Enron was headquartered, experienced an even stronger negative market reaction than Andersen’s non-
Houston clients.2 Overall, CP concludes the shredding announcement had a significant impact on the perceived quality of
Andersen’s audits, and that the resulting loss of reputation had a negative effect on the market values of the firm’s other
clients.
In this study, we report new findings that shed light on whether this event study evidence is consistent with an auditor
reputation effect. In so doing, we do not suggest that auditor reputation does not matter. As discussed above, there is ample
evidence that reputation is important to auditors and their clients. Rather, our purpose is to determine whether client
returns around Andersen’s shredding announcement and related events can be considered evidence of a reputation effect,
or whether the results are confounded by other effects.

2.3. Confounding effects

In conducting any event study, it is imperative to identify other events or factors that could confound the results and
lead to erroneous inferences. There are reasons to suspect that confounding events affect the CP results. Specifically, CP
documents a significant negative market reaction not only for Andersen clients but also for Big 4 clients. Although this
evidence could be explained by Andersen’s shredding announcement causing the market to quickly become suspicious of
the quality of other auditors’ work, as suggested by CP, an alternative explanation is that negative macroeconomic or
industry news occurred during this period.
To identify potentially confounding news events during the shredding announcement window, we obtained from
LexisNexis all articles from the Wall Street Journal, New York Times, and Financial Times for the period January 10–16, 2002.
Assuming a 1-day lag in publication of a news story, this period corresponds to the longest event window, (1, +3),
investigated by CP. This search reveals several contemporaneous negative information events during the shredding
announcement window, which we discuss below. Perhaps not surprising given this negative news, the Dow Jones Industrial
Average fell nearly 1% on each of the two trading days following the shredding announcement, Friday, January 11, 2002

2
The evidence on other event dates is mixed. CP (Table 3) reports no significant market reaction for Andersen clients surrounding Enron’s
announcement that it was restating earnings (November 8, 2001) and Andersen’s admission to Congress that its audit team had made an error in
judgment (December 12, 2001), but there is a significant negative reaction surrounding the release of the Powers Report (February 2, 2002) revealing the
main Chicago office of Andersen was aware of the accounting problems at Enron. Further, CP finds a positive but insignificant market reaction to
Andersen’s indictment for obstruction of justice (March 15, 2002), while for a larger sample of Andersen clients Krishnamurthy et al. (2006) reports a
significant negative reaction. Although our primary event study analyses in Section 3.1 focus on the shredding announcement, we present additional
evidence on the Powers Report and indictment event dates in Section 3.2.
 ARTICLE IN PRESS
282 K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293

(day+1) and Monday, January 14, 2002 (day+2). These negative returns parallel the cumulative loss for Andersen clients
during the same period.
We provide excerpts from a sample of news articles in the Appendix. Several articles indicate the market decline on
January 11, 2002 (day+1) was directly related to cautious remarks provided by then Federal Reserve Chairman Alan
Greenspan concerning the prospects of an economic recovery (New York Times [3,5]). Economic worries tied to Greenspan’s
statements are again mentioned in news articles as contributing to continued market declines on January 14, 2002 (day+2)
(Wall Street Journal [7]; Financial Times [8]). Other news releases suggest the January 14, 2002 market decline was related to
concerns about upcoming corporate earnings reports (New York Times [6]). In contrast, we did not find any news articles
suggesting the market decline was associated with Andersen’s shredding announcement.
Other media coverage during the shredding announcement window focused on industry-specific events. In particular,
the Energy sector was hit by significant negative oil price shocks, which analysts argued were the result of bearish
fundamentals and excess supply (Financial Times [2], New York Times [4]). Moreover, there were reports that stocks in the
Information Technology sector were overvalued (New York Times [3]), leading to profit-taking that drove down market
prices (Wall Street Journal [1]).
The confounding news releases are of further concern if the events differentially affect the client portfolios of Andersen
and the Big 4. It is well known that the large audit firms tend to develop industry-specific expertise, and this specialization
leads to auditor concentration along industry lines (e.g., Hogan and Jeter, 1999; GAO, 2003). If bad news is released for
industries disproportionately represented in the Andersen sample, then we expect a more negative market response for
Andersen clients relative to Big 4 clients even in the absence of any reputation effects associated with the shredding
announcement. Similarly, within the Andersen sample, if bad news is released that disproportionately affects their Houston
office clients, then a more negative market reaction is expected in that location in the absence of any reputation effects. In
the remainder of the paper, we investigate the effects of confounding news events and industry concentration on the
negative market reaction for Andersen’s clients (Section 3). We also extend the event study analysis of CP to consider
whether there is evidence that the damage to Andersen’s reputation is reflected in client ERCs (Section 4).

3. Abnormal returns

3.1. Shredding announcement

Following CP, our sample consists of companies in the S&P 1500. We obtain the listing of the S&P 1500 as of October 31,
2001 from the Standard & Poors website, and from this sample exclude Enron and companies without the necessary
security price data on CRSP. To identify the auditor for each company in our sample, we extract the audit opinion from
10-K filings in 2001 and 2002, and also search 8-K filings to determine if there was a change in auditors prior to the
shredding announcement. Table 1 presents the frequency of sample observations by industry and sector using the Global
Industry Classification Standard (GICS). For comparison purposes, we also present the sample distribution reported in CP
(Table 1).3 The data indicate our sample closely resembles that of CP, and therefore our findings are likely not affected by
differences in sample composition.4

3.1.1. Pooled results

To investigate the market’s reaction to the shredding announcement on January 10, 2002, we compute the abnormal
return on day t (ARit):

ARit ¼ Rit  ða^ þ b^ i Rmt Þ. (1)

Rit is the return for client i on day t, Rmt is the return on the Russell 3000 index, and a^ and b^ are parameter estimates
obtained from estimation of the market model for the period November 1, 2000 to October 31, 2001:
Rit ¼ ai þ bi Rmt þ uit . (2)
We report cumulative abnormal returns for various windows around the shredding announcement date in Table 2.
Consistent with CP (Table 3), we find significant mean abnormal returns for Andersen’s clients that range from 1.10% to
1.98% depending on the length of the event window, with approximately 70% of the firms experiencing negative returns.
Further, consistent with CP (Table 7), mean abnormal returns for Big 4 clients are also significantly negative but
significantly less negative than for Andersen clients in the (0, +1) and (0, +2) event windows, based on a t-statistic for test of

3
CP tabulates the industry distribution of all firms in the S&P 1500 rather than only those with available data. On that basis, our sample consists of
290 (1,185) Andersen (Big 4) firms. Observations in industries with less than three Andersen clients are combined by sector and reported as ‘‘Other.’’
Because we do not know the identities of the 23 companies CP classifies as ‘‘Other,’’ we are not able to show these observations by sector; instead, we
report these observations in one category at the end of the table as in CP (Table 1).
4
In general, we follow the research design procedures of CP as closely as possible. We do not know the source or date used by CP to identify the S&P
1500 listing, but we select October 31, 2001 as this date immediately precedes the first event (November 8, 2001) examined by CP. Because the
composition of the S&P 1500 changes over time, we repeat the analysis of shredding announcement returns using the S&P 1500 listing as of October 1,
2001, November 8, and January 9, 2002, with no change in inferences. We discuss several other robustness tests at the end of this section.
 ARTICLE IN PRESS
K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293 283

Table 1
Industry classification of sample by sector

Chaney and Philipich

Andersen Big 4 Andersen Big 4

N (%) N (%) N N

Energy
Energy equipment and services 9 3.1 23 2.0 9 25
Oil, gas, and consumable fuels 17 5.9 24 2.0 17 24

26 9.0 47 4.0

Material
Chemicals 6 2.1 36 3.1 6 36
Containers and packaging 3 1.0 9 0.8 3 9
Metals and mining 3 1.0 27 2.3 3 27
Paper and forest products 7 2.4 9 0.8 7 9
Other 0 0.0 4 0.3

19 6.6 85 7.2
Industrial
Airlines 3 1.0 7 0.6 3 7
Building products 6 2.1 5 0.4 5 6
Commercial services and supplies 27 9.3 55 4.7 28 57
Construction and engineering 4 1.4 7 0.6 5 7
Electrical equipment 5 1.7 22 1.9 5 22
Machinery 16 5.5 38 3.2 16 37
Road and rail 5 1.7 13 1.1 5 13
Other 6 2.1 38 3.2

72 24.9 185 15.8

Consumer discretionary
Hotels, restaurants, and leisure 7 2.4 39 3.3 7 39
Household durables 8 2.8 33 2.8 8 32
Leisure equipment and products 4 1.4 10 0.9 3 11
Media 6 2.1 26 2.2 6 26
Multiline retail 4 1.4 18 1.5 4 18
Specialty retail 5 1.7 51 4.3 5 51
Textiles, apparel, and luxury goods 5 1.7 15 1.3 5 15
Other 6 2.1 27 2.3

45 15.6 219 18.7

Consumer staples
Food and staples retailing 3 1.0 14 1.2 3 14
Food products 5 1.7 23 2.0 5 23
Other 3 1.0 25 2.1

11 3.8 62 5.3

Health care
Biotechnology 3 1.0 20 1.7 3 20
Health care equipment and supplies 7 2.4 34 2.9 8 34
Health care providers and services 8 2.8 42 3.6 7 44
Pharmaceuticals 4 1.4 15 1.3 4 15

22 7.6 111 9.5

Financial
Banks 12 4.2 74 6.3 12 74
Diversified financial services 7 2.4 30 2.6 7 30
Other 2 0.7 54 4.6

21 7.3 158 13.5

Information technology
Communications equipment 7 2.4 39 3.3 7 43
Electronic equipment and instruments 6 2.1 37 3.2 6 37
 ARTICLE IN PRESS
284 K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293

Table 1 (continued )

Chaney and Philipich

Andersen Big 4 Andersen Big 4

N (%) N (%) N N

Semiconductor equipment and products 5 1.7 60 5.1 5 63

Software 13 4.5 53 4.5 13 53
Other 5 1.7 43 3.7

36 12.4 232 19.8

Telecommunications
Diversified telecommunication services 3 1.0 9 0.8 3 9
Other 2 0.7 5 0.4

5 1.7 14 1.2

Utilities
Electric utilities 16 5.5 38 3.2 16 38
Gas utilities 9 3.1 14 1.2 9 14
Multi-utilities 6 2.1 7 0.6 6 7
Other 1 0.3 2 0.2

32 11.1 61 5.2

Other per CP 23 198

Total firms 289 100.0 1,174 100.0 287 1,187

This table presents industry classifications using the Global Industry Classification Standard (GICS) definitions for the sample of S&P 1500 firms (excluding
Enron) as of October 31, 2001 identified from the Standard and Poors website. Industries with three or more Andersen clients are reported separately;
industries with less than three Andersen clients are combined by sector and reported as ‘‘Other.’’ Excluded from our sample are 1 (11) Andersen (Big 4)
client firms without the necessary returns data to conduct the empirical tests. Auditor information was obtained from the audit opinion in the 2001 and
2002 10-K filings, with any changes in auditors prior to the shredding announcement identified from 8-K filings. Industry frequencies are presented
separately for clients of Andersen and the remaining Big 4 auditors (Deloitte & Touche, Ernst & Young, KPMG, and PricewaterhouseCoopers). The table also
presents the industry classifications reported in Chaney and Philipich (2002), Table 1, panel A, which includes all S&P 1500 firms regardless of data
availability.

differences. We also assess the significance of the difference in returns using the seemingly unrelated regression technique
of Schipper and Thompson (1983) to control for cross-sectional dependencies in the data. These F-statistics indicate the
difference is significant only for the (0, +1) window.5
In summarizing their results, CP notes the negative abnormal returns for both Andersen and Big 4 clients are consistent
with a market-wide effect for all large auditors, suggesting that ‘‘market participants might perceive that if there is a
significant quality problem with Andersen’s audits, perhaps a similar, yet undetected, quality problem exists for all of the
Big 5 audit firms’’ (CP, pp. 1239–1240). However, CP concludes the findings are likely the result of a decline in Andersen’s
reputation attributable to perceived audit quality problems because abnormal returns are more negative for Andersen
clients. In contrast, we propose and test an alternative explanation for these findings. Specifically, we investigate whether
confounding news events coupled with a different industry composition of Andersen and Big 4 clients drive the results
reported in CP and Table 2. Although CP (Table 6) implements industry controls in the analysis of returns for Andersen
clients, the paper does not do so in the comparison of returns for Andersen and Big 4 clients (Table 7). More generally, CP
does not control for potential confounding news events occurring during the event period.

3.1.2. Industry sector results

To investigate the influence of confounding industry effects, Table 3 presents results by GICS sector, the broadest
industry classification provided by GICS, for the (0, +2) event window that is the focus of the analysis in CP. In addition to
the mean abnormal return and fraction of negative returns, we also report the relative frequency of industry sector
observations. As expected given the tendency of auditors to specialize along industry lines, the composition of the
Andersen and Big 4 samples differs, with Andersen auditing a disproportionate share of the Energy, Industrial, and Utilities

5
CP (Table 7) separately tests for differences between the abnormal returns for clients of Andersen and each of the Big 4 in the (0, +1) and (0, +2)
event windows; no tests are reported for the (0, +3) and (1, +3) windows. CP (footnote 5) reports that sensitivity tests using the approach of Schipper and
Thompson (1983) produce findings similar to the tabulated results.
 ARTICLE IN PRESS
K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293 285

Table 2
Comparison of market reaction to January 10, 2002 shredding announcement for Andersen and Big 4 clients

Window Andersen (N ¼ 289) Big 4 (N ¼ 1,174) Difference

CAR (%) Neg. (%) t-Stat. CAR (%) Neg. (%) t-Stat. Mean t-Stat. F-Stat.

(0, +1) 1.10 67.8 5.24** 0.52 56.8 5.17** 0.58 2.47** 2.81*
(0, +2) 1.86 68.9 7.10** 1.24 61.8 8.87** 0.62 2.09** 2.12
(0, +3) 1.98 69.6 6.44** 1.66 60.9 8.91** 0.32 0.90 0.63
(1, +3) 1.92 66.1 5.71** 1.70 59.3 8.24** 0.22 0.57 0.15

This table presents the mean cumulative abnormal return (CAR %) for various windows surrounding January 10, 2002 (day 0), the date Andersen admitted
shredding Enron-related documents. Abnormal returns are calculated as ARit ¼ Rit  ða^ þ b^ i Rmt Þ, where Rit is the return for client i on day t, Rmt is the
return on the Russell 3000 index, and a^ and b^ are parameter estimates obtained from estimation of the market model Rit ¼ ai+biRmt+uit for the period
November 1, 2000 to October 31, 2001. The table also presents the percentage of negative CARs (Neg. %), and the t-statistic for the test of whether the
mean CAR is significantly different from zero. The difference column presents the mean difference in CARs between the Andersen and Big 4 samples, the t-
statistic for the test of whether the difference is significantly different from zero, and the F-statistic calculated using the seemingly unrelated regression
technique of Schipper and Thompson (1983).
*, **Significantly different from zero at the 10% or 5% level, respectively, using two-tail tests.

Table 3
Sector analysis of market reaction to January 10, 2002 shredding announcement for (0, +2) window

Sector Andersen (N ¼ 289) Big 4 (N ¼ 1,174) Difference Matched pairs difference

Freq. (%) CAR (%) Neg. (%) t-Stat. Freq. (%) CAR (%) Neg. (%) t-Stat. Mean t-Stat. Mean t-Stat.

Energy 9.0 5.70 100.0 10.48** 4.0 6.55 95.7 11.68** 0.85 1.08 1.30 1.41
Material 6.6 2.97 73.7 3.20** 7.2 3.26 83.5 6.47** 0.29 0.27 1.50 1.12
Industrial 24.9 2.07 69.4 3.34** 15.8 1.79 68.1 6.26** 0.28 0.41 0.60 0.79
Consumer discretionary 15.6 1.26 75.6 2.48** 18.7 1.53 65.3 4.66** 0.27 0.46 0.08 0.10
Consumer staples 3.8 0.44 27.3 0.65 5.3 0.77 53.2 1.80* 1.21 1.52 1.40 1.06
Health care 7.6 2.44 31.8 1.93* 9.5 1.51 28.8 5.14** 0.83 0.71 0.80 0.53
Financial 7.3 0.62 61.9 1.10 13.5 0.25 60.1 0.83 0.37 0.57 0.46 0.57
Information technology 12.4 3.82 91.7 6.16** 19.8 1.11 65.9 3.21** 2.71 3.81** 1.52 1.20
Telecommunications 1.7 1.63 80.0 1.80 1.2 0.83 42.9 0.74 0.80 0.56 0.65 0.32
Utilities 11.1 0.84 46.9 1.73* 5.2 0.23 36.1 0.26 0.61 0.60 0.40 0.23

Total 100.0 1.86 68.9 7.10** 100.0 1.24 61.8 8.87** 0.62 2.09** 0.30 0.73

This table presents the mean cumulative abnormal return (CAR %) for the (0, +2) event window surrounding January 10, 2002 (day 0), the date Andersen
admitted shredding Enron-related documents. Abnormal returns are calculated as ARit ¼ Rit  ða^ þ b^ i Rmt Þ, where Rit is the return for client i on day t, Rmt is
the return on the Russell 3000 index, and a^ and b^ are parameter estimates obtained from estimation of the market model Rit ¼ ai+biRmt+uit for the period
November 1, 2000 to October 31, 2001. The table also presents the percentage of negative CARs (Neg. %) and the t-statistic for the test of whether the mean
CAR is significantly different from zero. The difference column presents the mean difference in CARs between the Andersen and Big 4 samples, and the t-
statistic for the test of whether the difference is significantly different from zero. The matched pairs difference column presents the mean difference in
CARs and the associated t-statistic for a matched sample of Andersen and Big 4 clients, where each Andersen client is matched to the Big 4 client in the
same sector with the smallest absolute difference in size, as measured by the market value of equity on October 31, 2001. Industry sectors are determined
using the Global Industry Classification Standard (GICS) definitions.
*, **Significantly different from zero at the 10% or 5% level, respectively, using two-tail tests.

sectors and the Big 4 dominating the Financial and Information Technology sectors. Because of differences in industry
composition, we not only test the mean difference in returns for the full sample of Andersen and Big 4 clients (column
labeled ‘‘Difference’’), but also for a matched pairs sample where we match each Andersen client to the Big 4 client in the
same sector with the smallest absolute difference in firm size, as measured by the market value of equity on October 31,
2001 (column labeled ‘‘Matched Pairs Difference’’).
Consistent with negative macroeconomic news being released during this period, as discussed above, Table 3 shows that
abnormal returns for Andersen (Big 4) clients are negative in 8 (9) of the 10 industry sectors, although these negative
returns are statistically significant at the 0.10 level in only 6 sectors for both samples. Abnormal returns of Andersen (Big 4)
clients in the Energy sector are 5.70% (6.55%), nearly twice as large in absolute terms as the returns in any other industry
sector. Thus, it is likely the Energy sector has a significant influence on the overall negative market reaction. The Energy
sector is also of particular interest given the concentration of Andersen clients in this sector and in Houston where the
Enron incidents occurred.
Tests of the mean difference in abnormal returns for the full sample reveal a more negative Andersen abnormal return in
five sectors. However, the difference is only significant for the Information Technology sector, inconsistent with the notion
 ARTICLE IN PRESS
286 K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293

0.06
0.03
0
-5 -4 -3 -2 -1 0 1 2 3 4 5
-0.03

Return
-0.06
-0.09 All other sectors
Energy sector
-0.12
Oil price change
-0.15
Day Relative to January 10, 2002

0.03

0
-5 -4 -3 -2 -1 0 1 2 3 4 5
-0.03
Return

-0.06

-0.09
All other sectors
-0.12 Energy sector
Oil price change
-0.15
Day Relative to January 10, 2002

Fig. 1. Cumulative returns and oil price changes. This figure presents mean compounded cumulative raw (abnormal) returns in panel A (panel B)
separately for companies in the Energy sector and all other sectors. Abnormal returns are calculated as ARit ¼ Rit  ða^ þ b^ i Rmt Þ, where Rit is the return for
client i on day t, Rmt is the return on the Russell 3000 index, and a^ and b^ are parameter estimates obtained from estimation of the market model
Rit ¼ ai+biRmt+uit for the period November 1, 2000 to October 31, 2001. The figure also presents the oil price change, measured as the cumulative
compounded percentage change in the price of West Texas Light Crude Oil from the Department of Energy. Industry sectors are determined using the
Global Industry Classification Standard (GICS) definitions.

of a pervasive market reaction for Andersen clients attributable to a decline in the audit firm’s reputation. Moreover, in the
matched pairs analysis, the mean difference in abnormal returns is more negative in only three sectors and there are no
significant sector differences between Andersen and the Big 4, including the Information Technology sector.6 Thus,
controlling for industry composition and size, the shredding announcement did not have a statistically significant
reputation effect on the stock prices of Andersen’s audit clients relative to those of the Big 4.7 In the remainder of this
section, we provide further evidence on the market reaction in the period surrounding the shredding announcement, with
an emphasis on the Energy and Information Technology sectors.
Because there is no significant difference between the large negative abnormal returns for Andersen and Big 4 clients in
the Energy sector, the returns in this sector appear to be driven by negative industry-related news. Specifically, as noted
above, a search of news stories for the period surrounding the shredding announcement indicates there was a significant
drop in oil prices. Fig. 1 plots the cumulative compounded percentage change in oil prices along with cumulative raw
(panel A) and abnormal (panel B) returns of the Energy sector and all other sectors combined for the (5, +5) event window
(i.e., January 3, 2002 to January 17, 2002).8 The results show a strong correlation between changes in oil prices and stock
price movements in the Energy sector, consistent with evidence in Huang et al. (1996) that oil price shocks lead stock
returns for companies in the oil industry, but not for companies in other industry sectors.
To investigate the market reaction around the shredding announcement controlling for changes in oil prices, we employ
the portfolio approach of Sefcik and Thompson (1986) to estimate the following empirical model for Energy sector clients:

Rpt ¼ ap þ g1 AApt þ g2 SHREDt þ g3 AA  SHREDpt þ g4 DOILPRICEt þ g4 DOILPRICEt1 þ g5 Rmt þ uit . (3)

6
Untabulated results using the Schipper and Thompson (1983) approach are similar to the matched pairs findings as the difference in abnormal
returns is insignificant in all sectors.
7
The distribution of returns for Andersen clients in the Information Technology sector is negatively skewed (91.7% negative returns). However,
Andersen clients in this sector are also smaller, on average, than Big 4 clients, measured using the natural logarithm of market value of equity
(t-statistic ¼ 1.82) and total assets (t-statistic ¼ 1.92).
8
For parsimony, we do not separately plot Andersen and Big 4 returns because there is generally no significant difference between the returns of
Andersen and Big 4 clients (Table 3). Inferences are unchanged when the returns of Andersen and Big 4 clients are considered separately.
 ARTICLE IN PRESS
K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293 287

Table 4
Energy sector returns during the shredding announcement and changes in oil prices

Variable Model 1 Model 2 Model 3

Coeff. est. t-Stat. Coeff. est. t-Stat. Coeff. est. t-Stat.

Intercept 0.000 0.36 0.000 0.10 0.000 0.03

AA 0.000 0.02 0.000 0.02 0.000 0.02
SHRED 0.023 1.82* 0.014 1.22 0.012 1.10
AA  SHRED 0.003 0.16 0.003 0.18 0.003 0.18
DOILPRICEt 0.270 8.99** 0.264 9.17**
DOILPRICEt1 0.143 4.79** 0.146 5.10**
MKTt 0.399 6.41**

Total AASHRED 0.020 1.64* 0.011 0.98 0.012 0.84

Adj. R2 0.006 0.173 0.234

N 510 510 510

This table presents regression summary statistics from estimation of the following model:
Rpt ¼ ap þ g1 AApt þ g2 SHREDt þ g3 AA  SHREDpt þ g4 DOILPRICEt þ g4 DOILPRICEt1 þ g5 Rmt þ uit . Rpt is the equal-weighted daily portfolio return for each
of the 255 trading days in the period January 1, 2001 through January 14, 2002. Separate daily portfolio returns are created for the samples of Andersen
and Big 4 Energy sector clients, and thus there are 510 observations in the regression estimation. AApt is an indicator variable equal to one (zero) if the
portfolio return is for Andersen (Big 4) clients. SHREDt is an indicator variable equal to one for the (0, +2) shredding announcement event window, and
zero otherwise. DOILPRICEt (DOILPRICEt1) is the first difference of the natural log of the daily spot price of West Texas Light Crude Oil gathered from the
Department of Energy. Rmt is the return on the Russell 3000 index. Total AASHRED is the total coefficient for Andersen clients during the shredding event
window, calculated as Intercept+AA+SHRED+AA  SHRED.
*, **Significantly different from zero at the 10% or 5% level, respectively, using two-tail tests.

Rpt is the equal-weighted daily portfolio return for each of the 255 trading days in the period January 1, 2001 through
January 14, 2002. Separate daily portfolio returns are created for the samples of Andersen and Big 4 Energy sector clients,
and thus there are 510 observations in the regression estimation. AApt is an indicator variable equal to one (zero) if the
portfolio return is for Andersen (Big 4) clients. SHREDt is an indicator variable equal to one for the (0, +2) shredding
announcement event window, and zero otherwise. To approximate the percentage change in oil prices, we include both the
contemporaneous (DOILPRICEt) and lagged (DOILPRICEt1) first difference in the natural logarithm of the daily price of West
Texas Light Crude oil.9 Finally, Rmt is the return on the Russell 3000 index.
We report the results of this analysis in Table 4. Model 1 includes only the indicator variables AA and SHRED and their
interaction. The negative and significant coefficient estimate on SHRED indicates an incremental 2.3% average abnormal
return for Big 4 Energy sector clients during each day of the shredding announcement window, consistent with the results
in Table 3. More importantly, the coefficient on the interaction term AA  SHRED is not significantly different from zero. In
other words, consistent with the results reported above, the average abnormal return for Andersen’s Energy sector clients
(total AASHRED ¼ 2.0%) is statistically and economically indistinguishable from the returns for Big 4 Energy clients
(2.3%) during the shredding announcement window.
To control for the effects of oil price movements, we augment Model 1 with DOILPRICEt and DOILPRICEt1. As expected,
the results for Model 2 show that both variables capturing changes in oil prices are positive and significant. Of primary
interest, the coefficient estimate on SHRED is insignificant in this estimation, as is the estimate on the interaction term
AA  SHRED. Thus, controlling for changes in oil prices, the shredding announcement window market reactions of 1.4% for
Big 4 clients and 1.1% for Andersen clients are not statistically different from each other or from zero. Finally, Model 3 in
Table 4 controls for the market return, Rmt. Once again, there is no evidence of a significant difference in the market
reaction of Andersen and Big 4 Energy clients in the shredding announcement window, as returns for the portfolios of both
Andersen (1.2%) and Big 4 clients (1.2%) are insignificantly different from zero and from each other.10
The oil price shock in the Energy sector also likely affects inferences regarding the incremental effects of the shredding
announcement on Andersen’s Houston office clients. CP (Table 5) reports that Andersen’s Houston clients suffered more
negative abnormal returns than their non-Houston clients, and infers that investors were especially concerned by the
quality of audits in this location because of the office’s direct association with Enron. However, the concentration of Energy
sector clients in Houston could confound this result. Of the 26 Andersen clients in the Energy sector, Table 5 shows that 11

9
Inferences are unchanged in analyses including up to six lags of changes in oil prices as in Huang et al. (1996).
10
In untabulated analyses, we estimate panel data regressions, clustering by firm and date to control for dependencies in the data. The results are
consistent with the portfolio regressions reported in Table 4, except the coefficient estimate on SHRED in Models 2 and 3 is significant, as is the total
AASHRED coefficient. These results suggest both Andersen and Big 4 Energy clients experienced significantly negative average abnormal returns during the
shredding announcement window. However, the coefficient on the interaction term AA  SHRED remains insignificant in these estimations, confirming the
finding that the shredding announcement did not have a significant incremental reputation effect on the stock prices of Andersen’s audit clients.
 ARTICLE IN PRESS
288 K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293

Table 5
Market reaction to shredding announcement for Houston and non-Houston Energy sector clients

Location Andersen (N ¼ 26) Big 4 (N ¼ 47) Difference

N CAR (%) Neg. (%) t-Stat. N CAR (%) Neg. (%) t-Stat. Mean t-Stat.

Houston 11 7.07 100.0 8.18** 26 7.62 96.2 9.41** 0.55 0.46

non-Houston 15 4.70 100.0 7.89** 21 5.22 95.2 7.84** 0.52 0.58
Difference 2.37 2.26** 2.40 2.29** 0.03 0.02

This table presents the mean cumulative abnormal return (CAR %) for the (0, +2) event window surrounding January 10, 2002 (day 0), the date Andersen
admitted shredding Enron-related documents, for Energy sector clients audited in Houston or a non-Houston location, identified from the audit opinion.
Abnormal returns are calculated as ARit ¼ Rit  ða^ þ b^ i Rmt Þ, where Rit is the return for client i on day t, Rmt is the return on the Russell 3000 index, and a^
and b^ are parameter estimates obtained from estimation of the market model Rit ¼ ai+biRmt+uit for the period November 1, 2000 to October 31, 2001. The
table also presents the percentage of negative CARs (Neg. %), and the t-statistic for the test of whether the mean CAR is significantly different from zero.
The difference column presents the mean difference in CARs between the Andersen and Big 4 samples, and the t-statistic for the test of whether the
difference is significantly different from zero.
*, **Significantly different from zero at the 10% or 5% level, respectively, using two-tail tests.

are audited by the Houston office, representing approximately one-half of Andersen’s total 20 Houston clients.11 CP does
not compare the returns of Andersen and Big 4 clients in Houston or the returns of Andersen and Big 4 clients in non-
Houston locations to determine whether the results are attributable to a Houston office reputation effect or to industry
concentration.
Findings in Table 5 for the Energy sector show that Andersen’s Houston clients experienced significantly more negative
abnormal returns (7.07%) than their non-Houston clients (4.70%). However, similar findings are observed for Big 4
clients in Houston (7.62%) compared with non-Houston locations (5.22%). It is difficult to reconcile these results with
the reputation hypothesis as there is no reason to expect Houston clients of the Big 4 to suffer more severe stock price
declines as these offices were no more involved in the accounting scandals at Enron than were Big 4 offices in other cities.
Rather, the results likely reflect the concentration of Energy clients sensitive to oil prices in Houston.12 Consistent with this
view, the last column of Table 5 shows the difference in abnormal returns of Andersen and Big 4 Energy clients in Houston
is insignificant (t-statistic ¼ 0.46), as is the difference in abnormal returns for clients in non-Houston locations
(t-statistic ¼ 0.58). Expanding the analysis in Table 5 to include all industry sectors, untabulated results continue to show
no significant difference in abnormal returns of Houston clients of Andersen and the Big 4. There is a significant difference
between non-Houston clients of Andersen and the Big 4; this finding, however, is attributable to the influence of the
Information Technology sector as discussed above and documented in Table 3. Overall, the findings are inconsistent with a
reputation effect concentrated in the Houston office of Andersen.
As shown in Table 3, Information Technology is the only sector displaying some evidence of a pattern of abnormal
returns consistent with the reputation hypothesis, i.e., in the full sample, the mean abnormal return of Andersen clients
(3.82%) is significantly more negative than Big 4 clients (1.11%). There is no a priori reason, however, to expect any
reputation effects of Andersen’s shredding announcement to be isolated in this industry sector.13 It is possible this single
statistically significant difference out of ten industry sectors is a random occurrence. Moreover, as shown above, evidence
of significantly more negative abnormal returns for Andersen clients is not found using a matched pairs control sample of
Big 4 clients. Nevertheless, we cannot rule out the possibility of a differential market reaction in the Information
Technology sector related to the reputation effects of Andersen’s shredding announcement. Thus, with the possible
exception of the Information Technology sector, the evidence indicates the shredding announcement did not have a
significant reputation effect on the stock prices of Andersen’s audit clients.

3.1.3. Additional robustness tests

The inferences discussed above are robust to several sensitivity checks. First, we use all firms with the requisite data on
Compustat and CRSP rather than only those firms in the S&P 1500. Second, following CP, we use security prices from
Datastream rather than CRSP. Abnormal returns from this estimation closely approximate the tabulated results, with
Andersen (Big 4) clients experiencing a significant mean abnormal return of 1.87% (1.18%) in the (0, +2) window.
Moreover, consistent with Table 2, the difference in abnormal returns is significant (t-statistic ¼ 2.31). Third, we use several
other proxies for the market return, including the CRSP equal-weighted return, the CRSP value-weighted return, and the

11
Similarly, the 26 Energy sector clients in the Houston offices of the Big 4 represent approximately one-half of the total 43 clients in the Houston
offices of the Big 4.
12
In untabulated analysis, we augment the market model used to calculate abnormal returns in Table 5 with DOILPRICEt and DOILPRICEt1 to test the
sensitivity to changes in oil prices for Houston and non-Houston Energy sector clients. The mean coefficients on these variables are significantly more
positive for Houston clients relative to non-Houston clients, confirming that Houston Energy sector clients are more sensitive to changes in oil prices.
13
As discussed above, a search of news articles around the shredding announcement reveals the release of negative industry-related news for this
sector. However, we did not locate any news articles suggesting the market decline for Andersen clients in this sector was associated with the shredding
announcement.
 ARTICLE IN PRESS
K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293 289

return on the S&P 500 index. Fourth, because CP does not disclose the window used to estimate market model parameters,
we estimate the market model using 6-, 9-, and 12-month windows over various time periods prior to the shredding date.
Depending on the estimation window, the (0, +2) abnormal returns for Andersen clients range from 1.48% to 2.21%,
which encompasses the abnormal return of 1.86% reported in Table 2 and 2.03% reported by CP. Regardless of the
estimation window, abnormal returns for the Andersen sample are significantly more negative than for the Big 4 sample.
Fifth, we use a Fama-French (1993) three-factor model to calculate abnormal returns. Pooling all sectors, the difference in
abnormal returns of Andersen (0.73%) and Big 4 (0.23%) clients is marginally significant (t-statistic ¼ 1.64). In the sector
analysis, the difference for Information Technology continues to be significant for the full sample comparison, and
insignificant for the matched pairs comparison. Finally, inferences are unchanged if we account for non-synchronous
trading in the market model using the methodology from Scholes and Williams (1977). Abnormal returns in the (0, +2)
window are more negative for Andersen clients (1.39% versus 0.78%), but the significance of the difference is slightly
reduced (t-statistic ¼ 1.93).

3.2. Other Enron-related events

3.2.1. Powers Report

In addition to the shredding announcement, CP finds a significant negative market reaction to the release of the Powers
Report criticizing Andersen for its role in the creation of Enron’s controversial off-balance sheet partnerships, among other
issues. The report was released on Saturday, February 2, 2002. CP (Table 3) finds a significant abnormal return for Andersen
clients of 0.97% for the 3-day window (0, +2) beginning on Monday, February 4, 2002. However, abnormal returns for the
2-day window (0, +1) are not significantly different from zero. From this evidence, CP concludes that Andersen’s reputation
was further tarnished by the information contained in the Powers Report.14
The insignificant return in the 2 days (0, +1) following the weekend release of the Powers Report suggests either a
delayed market reaction or that the reaction was not related to Andersen’s reputation. To investigate this issue further, we
first replicate CP’s results. The findings in Table 6, panel A show a significant mean negative abnormal return for Andersen
clients in the (0, +2) window, but an insignificant return in the (0, +1) window. In contrast to CP, however, we also examine
the abnormal returns of Big 4 clients. The results show that abnormal returns are more negative for Big 4 clients in all event
windows, although the differences are not statistically significant.15
A review of news stories around the release of the Powers Report reveals general anxiety about earnings news along
with concerns about more corporate scandals. For instance, Tyco, a PricewaterhouseCoopers client at the time, lost 19% on
February 4 after Standard and Poors lowered its corporate credit rating from A to BBB and a Wall Street Journal article
questioned the company’s growth without acquisitions (Maremont, 2002). On February 5, the S&P 500 fell 0.4% while the
Nasdaq Composite Index dropped 0.9%. McKay (2002) suggests the market’s lackluster performance was driven by a high
level of uncertainty attributable to concerns about Federal Reserve interest rate policies, a stream of bad earnings news, and
increased scrutiny of firms’ financial statements. Thus, similar to the shredding announcement results discussed above, the
negative stock market performance immediately following the release of the Powers Report is consistent with market-wide
effects as opposed to a loss of Andersen’s reputation.

3.2.2. Andersen’s indictment

CP also investigates the market reaction surrounding Andersen’s indictment by the Justice Department for obstruction
of justice on March 15, 2002. Untabulated results in that paper indicate that abnormal returns in various windows
surrounding this date are positive and statistically insignificant, suggesting the market did not consider the indictment of
Andersen as a significant information event in assessing the damage to Andersen’s reputation. For example, CP reports
abnormal returns for the (0, +2) event window of 0.38%. In Table 6, panel B we report similar positive returns for Andersen
clients in the four event windows used throughout CP. For two of the event windows, (0, +3) and (1, +3), abnormal returns
for Andersen clients are significantly positive, and in all four event windows the market reaction for Andersen clients is
significantly more positive than that of Big 4 clients.
Table 6, panel B also reports results for two other event windows, (1, 0) and (1, +1), examined by Krishnamurthy et al.
(2006). Findings in that paper show a significant mean abnormal return of 0.80% in the (1, 0) event window for a sample
of 874 Andersen clients, which is significantly more negative at the 0.10 level than the 0.42% return for Big 4 clients.
Results for the (1, +1) window are insignificant. For our sample of firms from the S&P 1500, we report in Table 6, panel B a
significant negative return for Andersen clients in the (1, 0) event window, but this return is not significantly different
from that of Big 4 clients (t-statistic ¼ 0.12). To reconcile this latter finding with Krishnamurthy et al. (2006), we repeat the

14
On Sunday, February 3, 2002, Andersen announced the creation of an Independent Oversight Board (IOB) headed by former Federal Reserve Board
Chairman, Paul Volcker. CP suggests this event may offset the effects of the release of Powers Report because the market may view the IOB as a vehicle to
restore Andersen’s quality and reputation. However, based on the negative abnormal returns during the event window, CP concludes the findings reflect
the market response to the Powers Report.
15
There are no industry sectors where Andersen clients had significantly more negative returns than the corresponding portfolio of Big 4 clients.
However, in the Energy sector, Andersen clients experienced a 0.4% return which is significantly less negative than the 2.3% return experienced by Big 4
clients; there are no significant differences in the remaining industry sectors.
 ARTICLE IN PRESS
290 K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293

Table 6
Market reaction to other Enron-related events

Panel A: February 4, 2002 release of Powers Report

Window Andersen (N ¼ 288) Big 4 (N ¼ 1,171) Difference

CAR (%) Neg. (%) t-Stat. CAR (%) Neg. (%) t-Stat. Mean t-Stat.

(0, +1) 0.01 51.4 0.05 0.28 47.3 2.11** 0.27 1.11
(0, +2) 0.60 56.3 1.92* 0.81 54.2 5.15** 0.21 0.57
(0, +3) 0.88 60.4 2.83** 1.33 58.1 7.75** 0.45 1.25
(1, +3) 0.78 57.3 2.33** 0.94 53.8 5.11** 0.16 0.41

Panel B: March 15, 2002 indictment of Andersen

Andersen (N ¼ 288) Big 4 (N ¼ 1,170) Difference

CAR (%) Neg. (%) t-Stat. CAR (%) Neg. (%) t-Stat. Mean t-Stat.

Windows examined by CP
(0, +1) 0.30 44.1 1.49 0.20 51.8 2.14** 0.50 2.26**
(0, +2) 0.41 43.1 1.51 0.28 52.8 2.36** 0.69 2.34**
(0, +3) 0.61 43.8 2.21** 0.08 50.0 0.60 0.69 2.25**
(1, +3) 0.70 39.2 2.33** 0.16 46.2 1.07 0.54 1.64*

Windows examined by Krishnamurthy et al. (2006)

(1, 0) 0.40 55.9 1.88* 0.40 51.5 4.08** 0.00 0.12
(1, +1) 0.40 43.8 1.64 0.03 47.1 0.31 0.37 0.17

Panel A presents mean cumulative abnormal returns (CAR %) surrounding February 4, 2002 (day 0), the first trading day after the release of the Powers
Report on Saturday, February 2, 2002 charging that Andersen’s main office in Chicago was aware of problems at Enron. Panel B presents mean CARs
surrounding March 15, 2002 (day 0), the day Andersen was indicted. In both panels, abnormal returns are calculated as ARit ¼ Rit  ða^ þ b^ i Rmt Þ, where Rit is
the return for client i on day t, Rmt is the return on the Russell 3000 index, and a^ and b^ are parameter estimates obtained from estimation of the market
model Rit ¼ ai+biRmt+uit for the period November 1, 2000 to October 31, 2001. The table also presents the percentage of negative CARs (Neg. %), and the t-
statistic for the test of whether the mean CAR is significantly different from zero. The difference column presents the mean difference in CARs between the
Andersen and Big 4 samples, and the t-statistic for the test of whether the difference is significantly different from zero.
*, **Significantly different from zero at the 10% or 5% level, respectively, two-tail test.

tests using the broader sample of Andersen clients in that paper. In this untabulated analysis, we find a significant
abnormal return of 0.80% for Andersen clients in the (1, 0) event window, which is significantly more negative at the
0.10 level than the return for Big 4 clients. Thus, for this particular event window, Andersen’s indictment had a marginal
effect, statistically and economically, on smaller clients not in the S&P 1500. Why the market response would be limited to
smaller clients is unknown, although it is possible these clients face higher switching costs in the face of Andersen’s
demise. Taken together, however, the indictment date evidence offers little consistent support for the notion of a market
reaction attributable to a loss of reputation.

4. Earnings response coefficients

In this section, we present evidence from an alternative empirical approach to assess whether the reputation effects of
Andersen’s shredding announcement had a negative impact on its other audit clients. Specifically, prior research suggests
that an auditor’s reputation lends credibility to the earnings reports of its clients that is reflected in higher ERCs (Teoh and
Wong, 1993). Conversely, situations that damage an auditor’s reputation result in lower client ERCs (Moreland, 1995;
Francis and Ke, 2006). Following this literature, we compare ERCs of Andersen and Big 4 clients in the period after the
shredding announcement. We also compare ERCs of Andersen clients before and after the shredding announcement. If the
market perceived systematic audit quality concerns at Andersen as a result of the shredding announcement, then we expect
lower ERCs for Andersen clients relative to Big 4 clients in the post-shredding announcement period, and a decline in ERCs
for Andersen clients relative to the pre-shredding announcement period.
We estimate the following empirical model to test these predictions:

CARit ¼ l0 þ l1 UEit þ l2 AAPOST it þ l3 UE  AAPOST it þ l4 MBit þ l5 bit þ l6 LNMVEit þ it . (4)

CAR is the cumulative abnormal return in the (1, +1) window surrounding the earnings announcement, calculated using
the market model parameters from Table 2. To measure unexpected earnings (UE), we subtract the most recent IBES
median consensus quarterly forecast prior to the earnings announcement from IBES actual earnings, scaled by price at the
end of the fiscal quarter. AAPOST is an indicator variable equal to one if the observation is for an Andersen client in the post-
shredding announcement period. Because the majority of clients did not switch from Andersen until after its indictment on
March 15, 2002 (Blouin et al., 2007), we define the post-shredding announcement period as January 11, 2002 through
 ARTICLE IN PRESS
K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293 291

Table 7
Earnings response coefficients of Andersen and Big 4 clients in the period surrounding the shredding announcement

Variable Post-shredding announcement: Andersen vs. Big 4 Pre- vs. post-shredding announcement: Andersen only

Coeff. est. t-Statistic Coeff. est. t-Statistic

Intercept 0.009 5.49** 0.007 1.63

UE 1.097 3.92** 1.892 2.82**
AAPOST 0.006 1.62 0.004 0.57
UE  AAPOST 0.522 0.74 0.675 0.59
MB 0.001 1.96** 0.001 0.90
b 0.009 4.19** 0.001 0.22
LNMVE 0.000 0.44 0.000 0.07

Adj. R2 0.035 0.025

N 1,083 430
Andersen clients 215 215

This table presents regression summary statistics from the estimation of a model of cumulative abnormal returns in the (1, +1) window surrounding
earnings announcement dates on unexpected earnings and control variables: CARit ¼ l0 þ l1 UEit þ l2 AAPOST it þ l3 UE  AAPOST it þ l4 MBit þ l5 bit þ
l6 LNMVEit þ it . CAR is the cumulative abnormal return in the (1, +1) window surrounding the earnings announcement, calculated using the market
model parameters from Table 2. UE is unexpected earnings, defined as actual earnings less the most recent median consensus forecast of earnings from
IBES, scaled by the fiscal quarter end stock price (Compustat quarterly data14). AAPOST is an indicator variable equal to 1 if the observation is from an
Andersen client in the post-shredding announcement period, and 0 otherwise. MB is the market value of equity (data14  data61) divided by book value of
common equity (data59) as of the fiscal quarter end. b is the market model beta estimated for the period November 1, 2000 (1999) to October 31, 2001
(2000) for earnings announcements after (before) the shredding announcement. LNMVE is the natural logarithm of market value of equity as of the fiscal
quarter end. Observations with Studentized residuals greater than 2 in absolute value are eliminated. The post-shredding announcement estimation
includes Andersen and Big 4 clients’ earnings announcements occurring between January 11, 2002 and March 14, 2002. The pre- vs. post-shredding
announcement estimation includes Andersen clients’ earnings announcements occurring between January 11, 2002 and March 14, 2002 and in the same
period during the previous year.
*, **Significantly different from zero at the 10% or 5% level, respectively, using two-tail tests.

March 14, 2002. We define the pre-shredding announcement period as January 11, 2001 through March 14, 2001, i.e., the
same interval in the prior year as the post-shredding announcement period. Following Easton and Zmijewski (1989), we
control for growth (market value of equity divided by book value of common equity, MB), risk (b from Table 2), and size
(natural logarithm of market value of equity, LNMVE). Inferences are unchanged if we interact each control variable with
UE, or control for industry fixed effects. Finally, we delete observations with Studentized residuals greater than 2 in
absolute value. Imposing these additional data requirements reduces the sample to 215 (1,083) Andersen (Big 4) clients.
We present the results comparing ERCs of Andersen and Big 4 clients in the post-shredding announcement period on
the left side of Table 7. Consistent with prior research (see Kothari (2001) for a summary), the coefficient estimate on UE is
positive and significant.16 As discussed above, if Andersen suffered a loss of reputation that caused the market to discount
earnings-related information of its clients, then the coefficient estimate on the interaction of UE and AAPOST will be
negative and significant. In contrast, we find the estimated coefficient on the interaction variable is positive and
insignificant. Finally, the control variables MB and b are negative and significant consistent with the prior literature, while
LNMVE is insignificant.
We obtain similar inferences from the comparison of ERCs for Andersen’s clients in the pre- and post-shredding
announcement periods on the right side of Table 7. UE is again positive and significant as expected. As in the
previous analysis, the interaction of UE and AAPOST is positive and insignificant. The control variables are
insignificant in this estimation, which is not surprising given the reduced sample size and the fact the same set of S&P
1500 Andersen clients are in both periods which limits the cross-sectional variation related to these variables. Overall, the
inferences from the ERC tests presented in Table 7 are consistent with the abnormal returns analysis above, and suggest
the shredding announcement did not have a significant effect on the market’s valuation of Andersen clients’ equity
and earnings.

5. Discussion of results and conclusion

Prior research finds that in the 3 days following Andersen’s admission they shredded Enron-related documents,
Andersen’s other audit clients, particularly those in the Houston office, experienced a statistically significant negative stock
market reaction (Chaney and Philipich, 2002). This finding has been widely interpreted as evidence that the damage to

16
The magnitude of the UE coefficient estimates in Table 7 are also consistent with empirical estimates in the range from 1 to 3 documented in prior
research (Kothari, 2001).
 ARTICLE IN PRESS
292 K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293

Andersen’s reputation caused by this admission was impounded in the equity prices of other Andersen clients. In this
study, we provide new evidence that indicates the negative abnormal returns are not attributable to Andersen’s damaged
reputation, but rather to confounding events.
We document that negative macroeconomic and industry-related news was concurrently released during the shredding
announcement window. We posit that these confounding events, along with differences in the composition of the
Andersen and Big 4 samples attributable to auditor industry specialization, combine to produce the pattern of stock market
returns documented by CP. Andersen’s client portfolio contains a disproportionate share of companies in the Energy sector
which was rocked by a sharp decline in oil prices during the shredding announcement window. The key findings of our
paper show that the overall more negative market reaction for Andersen clients, particularly those in the energy-heavy
Houston office, is largely driven by these confounding effects. There is no evidence the shredding announcement had a
significant reputation effect on the stock prices of Andersen’s audit clients controlling either for industry concentration or
for oil price movements within the Energy sector.
There is a single industry sector, Information Technology, for which the market reaction of Andersen clients is
significantly more negative than Big 4 clients. We document that a substantial amount of negative industry-related news
was released during the shredding announcement window, including suggestions that most firms in this sector were
overvalued. There is no apparent reason why the bad news in this sector disproportionately affected Andersen clients.
Moreover, it is possible that this one statistically significant difference out of ten industry sectors is a random occurrence.
As a result, we cannot rule out the possibility of a reputation effect in the Information Technology sector, although it is
difficult to rationalize why this sector would be the only one affected by Andersen’s damaged reputation.
Our results are robust to a variety of specification checks. We also present additional evidence showing no systematic
market response for two other Enron-related events and that after the shredding announcement Andersen clients do not
have lower ERCs compared to the same period in the prior year or to Big 4 clients as would be expected if damage to
Andersen’s reputation was impounded in the market’s valuation of client earnings.
We caution that our results cannot be interpreted as evidence that Andersen’s reputation did not suffer as a result of the
Enron events. Given the amount of negative media coverage of Andersen’s role in the fraud, and research showing that
auditor reputation was important in client switching decisions in the aftermath of the scandal (Barton, 2005; Blouin et al.,
2007), it seems clear that Andersen’s reputation was damaged. However, these studies show that most clients remained
with Andersen following the shredding announcement, switching only after the indictment of Andersen for obstruction of
justice on March 15, 2002. Thus, the evidence suggests the damage done to Andersen’s reputation as a result of the
shredding announcement was not so costly to its clients as to cause widespread client defections. In other words, even
though the shredding announcement occurred during the annual audit for clients with a December fiscal year-end,
concerns about the effects of releasing financial statements audited by Andersen were outweighed by the costs of
switching auditors. Our evidence showing no significant abnormal market reaction for Andersen clients, controlling for
confounding effects, is consistent with this view. Moreover, our study highlights the difficulty of detecting and quantifying
the cost of Andersen’s damaged reputation using an event study methodology, particularly given the confounding effects
that contaminate the analysis.

Appendix. Sample of press articles during the Andersen shredding announcement window

[1] Wall Street Journal 01/10/02 (Browning, 2002a): ‘‘Traders cited a variety of surprises, from the crash of a Marine aircraft
in Pakistan and an unsubstantiated rumor about bombing in Iraq to comments on the economy from a Federal Reserve
official. But the fundamental problem, y, was simply that technology stocks in particular reached such high levels
during the day that some professional investors started selling to take their recent profits.’’
[2] Financial Times 01/10/02 (Roberts and Singh, 2002): ‘‘Crude oil futures suffered yesterday after weekly US inventory
data showed a steep build-up in distillates, which include heating oil and diesel fuel.’’
[3] New York Times 01/12/02 (Bloomberg News, 2002): ‘‘Stocks fell yesterday, y, after the Federal Reserve chairman, Alan
Greenspan, said there were ‘significant risks’ that an economic recovery would fail to take hold.’’ ‘‘About 70 percent of
50 technology stocks studied are too expensive given their earnings prospects, an analyst for Merrill Lynch, Steve
Milunovich, wrote in a report.’’
[4] New York Times 01/12/02 (2002): ‘‘New doubts over Russia’s commitment to OPEC’s planned supply cutbacks pushed
the price of crude oil down.’’
[5] New York Times 01/12/02 (Stevenson and Leonhardt, 2002): ‘‘It is still premature, [Alan Greenspan] said, to conclude
that the forces restraining economic activity here and abroad have abated enough to allow a steady recovery to take
holdy Mr. Greenspan’s cautionary remarks also seemed to give investors reason to worry that corporate profits were
not on the verge of a recovery.’’
[6] New York Times 01/15/02 (The Associated Press, 2002): ‘‘Investors succumbed to worries about earnings and sent stocks
sharply lower yesterday as they awaited earnings reportsy’’
[7] Wall Street Journal 01/15/02 (Browning, 2002b): ‘‘The Dow Jones Industrial Average finished below 9900 and the
Nasdaq Composite Index below 2000, both for the first time this year, as worries spread that the economy will prove
too weak to sustain the stock market’s big recent gains.’’
 ARTICLE IN PRESS
K.K. Nelson et al. / Journal of Accounting and Economics 46 (2008) 279–293 293

[8] Financial Times 01/15/02 (Morgan, 2002): ‘‘Equities in Europe and Asia showed clear signs of flagging as investors assess
the cautious outlook for the US economy delivered by Federal Reserve chairman, Alan Greenspany. ‘There is a thin line
between a liquidity-driven market that anticipates improving fundamentals and a bubble,’ said Richard Bernstein, chief
US investment strategist at Merrill Lynch. ‘We think the equity market may have stepped over that line.’’’

References

Baber, W.R., Kumar, K.R., Verghese, T., 1995. Client security price reactions to the Laventhol and Horwath bankruptcy. Journal of Accounting Research 33,
385–395.
Barton, J., 2005. Who cares about auditor reputation? Contemporary Accounting Research 22, 549–586.
Beatty, R., 1989. Auditor reputation and the pricing of initial public offerings. The Accounting Review 64, 693–709.
Bloomberg News, 2002. With losing week, this year’s gains are largely gone. The New York Times, January 12, p. C6.
Blouin, J., Grein, B.M., Rountree, B.R., 2007. An analysis of forced auditor change: the case of former Arthur Andersen clients. The Accounting Review 82,
621–650.
Browning, E.S., 2002a. Stocks fall as some lock in tech profits after morning rise. Wall Street Journal, January 10, p. C1.
Browning, E.S., 2002b. Dow industrials fall below 9900, in first for 2002. Wall Street Journal, January 15, p. C1.
Chaney, P.K., Philipich, K.L., 2002. Shredded reputation: the cost of audit failure. The Journal of Accounting Research 40, 1221–1245.
Datar, S., Feltham, G., Hughes, J., 1991. The role of audits and audit quality in valuing new issues. Journal of Accounting and Economics 14, 3–49.
Dyckman, T., Philbrick, D., Stephan, J., 1984. Comparison of event study methodologies using daily stock returns: a simulation approach. Journal of
Accounting Research 22, 1–30.
Dye, R., 1993. Auditing standards, legal liability, and auditor wealth. Journal of Political Economy 101, 887–914.
Easton, P., Zmijewski, M., 1989. Cross-sectional variation in the stock market response to accounting earnings announcements. Journal of Accounting and
Economics 11, 117–141.
Fama, E., French, K., 1993. Common risk factors in returns on stocks and bonds. Journal of Financial Economics 33, 3–56.
Firth, M., 1990. Auditor reputation: the impact of critical reports issued by government inspectors. RAND Journal of Economics 21, 374–387.
Francis, J., Ke, B., 2006. Disclosure of fees paid to auditors and the market valuation of earnings surprises. Review of Accounting Studies 11, 495–523.
Franz, D.R., Crawford, D., Johnson, E.N., 1998. The impact of litigation against an audit firm on the market value of nonlitigating clients. Journal of
Accounting, Auditing, and Finance 13, 117–134.
General Accounting Office (GAO), 2003. Public Accounting Firms: Mandated Study on Consolidation and Competition. United States General Accounting
Office, Washington, DC.
Hogan, C.E., Jeter, D.C., 1999. Industry specialization by auditors. Auditing: A Journal of Practice and Theory 18, 1–17.
Huang, R.D., Masulis, R.W., Stoll, H.R., 1996. Energy shocks and financial markets. Journal of Futures Markets 16, 1–27.
Kothari, S.P., 2001. Capital markets research in accounting. Journal of Accounting and Economics 31, 105–231.
Khurana, I.K., Raman, K.K., 2004. Litigation risk and the financial reporting credibility of Big 4 versus non-Big 4 audits: evidence from Anglo-American
countries. The Accounting Review 79, 473–495.
Krishnamurthy, S., Zhou, J., Zhou, N., 2006. Auditor reputation, auditor independence, and the stock-market impact of Andersen’s indictment on its client
firms. Contemporary Accounting Research 23, 465–490.
Lennox, C.S., 1999. Audit quality and auditor size: an evaluation of reputation and deep pockets hypotheses. Journal of Business Finance & Accounting 26,
779–805.
Maremont, M., 2002. Tyco made $8 billion of acquisitions over 3 years but didn’t disclose them. The Wall Street Journal, February 4, p. A3.
McKay, P.A., 2002. Investors’ mood is blue, leaves indexes in red. The Wall Street Journal, February 6, p. C1.
Moreland, K.A., 1995. Criticisms of auditors and the association between earnings and returns of client firms. Auditing: A Journal of Practice & Theory 14,
94–104.
Morgan, M., 2002. Equities lower on cautious outlook. Financial Times, January 15, p. 42.
New York Times, 2002. The markets: commodities. The New York Times, January 12, p. C6.
Roberts, A., Singh, R., 2002. US inventory data hit crude oil futures prices. Financial Times, January 10, p. 28.
Schipper, K., Thompson, R., 1983. The impact of merger-related regulations on the shareholders of acquiring firms. Journal of Accounting Research 21,
184–221.
Scholes, M., Williams, J., 1977. Estimating betas from nonsynchronous data. Journal of Financial Economics 5, 309–327.
Sefcik, S., Thompson, R., 1986. An approach to statistical inference in cross-sectional models with security abnormal returns as dependent variables.
Journal of Accounting Research 24, 316–334.
Simunic, D., 1980. The pricing of audit services: theory and evidence. Journal of Accounting Research 18, 161–190.
Stevenson, R.W., Leonhardt, D., 2002. Greenspan sounds note of caution on economy. New York Times, January 12, p. C1.
Teoh, S., Wong, T., 1993. Perceived auditor quality and the earnings response coefficient. The Accounting Review 68, 346–366.
The Associated Press, 2002. Shares tumble as Wall Street braces for earnings season. The New York Times, January 15, p. C10.
Titman, S., Trueman, B., 1986. Information quality and the valuation of new issues. Journal of Accounting and Economics 8, 159–172.
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.
Weber, J., Willenborg, M., Zhang, J., 2008. Does auditor reputation matter? The case of KPMG Germany and ComROAD AG. Journal of Accounting Research
46, 941–972.