Author’s Accepted Manuscript
Are all analysts created equal? Industry expertise
and monitoring effectiveness of financial analysts
Daniel Bradley, Sinan Gokkaya, Xi Liu, Fei Xie
www.elsevier.com/locate/jae
PII:
DOI:
Reference:
S0165-4101(17)30012-5
http://dx.doi.org/10.1016/j.jacceco.2017.01.003
JAE1134
To appear in: Journal of Accounting and Economics
Received date: 10 November 2015
Revised date: 18 January 2017
Accepted date: 20 January 2017
Cite this article as: Daniel Bradley, Sinan Gokkaya, Xi Liu and Fei Xie, Are all
analysts created equal? Industry expertise and monitoring effectiveness of
financial
analysts, Journal
of
Accounting
and
Economics,
http://dx.doi.org/10.1016/j.jacceco.2017.01.003
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
Are all analysts created equal? Industry expertise and monitoring
effectiveness of financial analysts
Daniel Bradley,a Sinan Gokkaya,b Xi Liu,c Fei Xied
a
University of South Florida, danbradley@usf.edu, 813.974.6358
b
Ohio University, gokkaya@ohio.edu, 740.593.0514
c
Miami University and Ohio University, liux4@ohio.edu, 740.593.2040
dUniversity of Delaware, xief@udel.edu, 302.831.3811
January 2017
______________________________________________________________________________
Abstract
We examine whether analysts’ prior industry experience influences their ability to serve as effective
external firm monitors. Our analyses of firms’ financial disclosure quality, executive compensation
and CEO turnover decisions portray a consistent picture that related pre-analyst industry experience
is of critical importance for analysts to play an effective monitoring role. Coverage by analysts with
such experience is associated with reduced earnings management, lower probability of committing
financial misrepresentation, less CEO excess compensation, and higher performance sensitivity of
CEO turnover. We also provide evidence on several plausible mechanisms through which industry
expert analysts exert monitoring efforts and limit managerial opportunism.
______________________________________________________________________________
We thank Lauren Cohen, Russell Jame, Sebastien Michenaud, Thomas Shohfi, Rong Wang, Jian Xue,
Xiaoyun Yu, Min Zhu, and seminar participants at the 2014 Conference on Financial Economics
and Accounting (CFEA), the 2015 Financial Management Association, the 2015 China International
Conference in Finance and Tsinghua University for helpful comments and suggestions. We are
responsible for any errors.
Are all analysts created equal? Industry expertise and monitoring effectiveness of financial analysts
1
“… management has not historically been the most committed to capital returns, but is seemingly willing to adjust senior ranks
going forward. The current CEO's contract is down to under 10 months.” — Jamie Baker, JP Morgan analyst, on
JetBlue1
1. Introduction
The notion of analyst monitoring goes at least as far back as Jensen and Meckling (1976, p.
354-355), who suggest that security analysts possess comparative advantages in monitoring firm
management and thus can play a large role in reducing agency costs. As the quote above illustrates,
analysts can be vocal critics of management policies, which can ultimately influence firm behavior. A
few months after Jamie Baker’s call, Jetblue announced that its CEO would be resigning at the end
of his contract.2 In this paper, we focus on analysts’ industry expertise and examine its effect on their
monitoring effectiveness. Industry expertise is a critical component of analysts’ human capital.
Analysts are assigned to and specialize in few industries to take advantage of economies of scale in
information production (Boni and Womack (2006), Kadan, Madureira, Wang, and Zach (2012)).
Surveys of financial analysts and institutional investors consistently indicate that industry expertise is
the most important attribute for analysts (Brown, Call, Clement, and Sharpe (2015) and the annual
polls of Institutional Investor magazine). The limited academic attention to analysts’ industry experience
has so far been confined to their earnings forecasts and stock recommendations (Boni and Womack
(2006), Kadan, Madureira, Wang, and Zach (2012), Bradley, Gokkaya, and Liu (2015)). Whether
industry expertise affects analysts’ effectiveness in their other functions remains unknown.
We measure analysts’ industry expertise based on whether they have prior work experience
in industries of the firms they cover. To do so, we manually collect analysts’ employment history
from LinkedIn.com, the largest professional networking service. 3 As elaborated below, despite its
positive effect on traditional metrics of analyst performance, the ramifications of industry expertise
for the monitoring and corporate governance role of analysts are more complex and difficult to
determine ex ante. On the one hand, it is possible that analysts with industry experience in the firms
1
Media outlets attributed the pressure to oust Jetblue’s then-CEO Dave Barger to Jamie Baker’s upgrading of the firm
while noting that management turnover was likely. Source: https://www.thestreet.com/story/12885110/1/why-jetblueceo-dave-barger-was-chased-out-by-wall-street.html.
2 http://fortune.com/2014/09/18/jetblue-ceo-david-barger-to-take-off/. For more examples of analyst monitoring
activities, see http://www.wsj.com/articles/noble-promises-more-transparency-1429664102 (questioning accounting
methods and demanding more transparency) and http://www.broadcastingcable.com/news/news-articles/bernsteinresearch-criticizes-media-ceo-pay/110241 (criticizing CEO pay).
3 For instance, Jamie Baker, the analyst mentioned in the opening quote of this paper is an example of an analyst with
related pre-analyst industry experience. According to his LinkedIn profile, he worked for two airline companies prior to
becoming a sell-side analyst. See section 2 for complete details on our data collection procedure.
2
they cover can provide more effective external monitoring because their prior industry experience
may allow them to develop a better understanding of the firm’s industry. Richer and more in-depth
industry knowledge can enhance analysts’ ability to analyze firms’ financial information and evaluate
the strategies and decisions proposed or implemented by firm management. Therefore, analysts with
related industry expertise, i.e., industry expert analysts, are better equipped and thus more likely to
identify and bring attention to firm policies that do not serve shareholders’ best interests. In addition,
these analysts also contribute to a more transparent information environment through their more
efficient information production and more accurate earnings forecasts (Bradley, Gokkaya, and Liu
(2015)). As a result, industry expert analyst coverage can have the dual effects of reducing managers’
latitude and incentives to engage in self-serving behavior as well as providing an impetus for boards
of directors to demand higher accountability of managers for their actions. We term this view the
effective monitor hypothesis.
On the other hand, prior work experience in a firm’s industry may reduce analysts’ incentive
to monitor firm management in subtle, but potentially important ways. For example, having worked
in the firm’s industry increases the likelihood that the analysts and the firm’s managers know each
other if their career paths have crossed or they have met at work functions such as industry trade
shows or conventions. Likewise, related industry work experience also increases the chance of
professional connections being developed between the analysts and firm management through
common friends or acquaintances from within the industry. These professional ties can potentially
cloud analysts’ views, making them more likely to agree with rather than disapprove of the decisions
made by managers. Together, these different channels imply that related industry experience can
impair the incentives of analysts to monitor firm management, allowing corporate insiders to indulge
more in activities that benefit themselves at shareholders’ expense. We term this view the impaired
monitor hypothesis.
We test these hypotheses by examining the effects of industry expert analyst coverage on
several major corporate policies including financial disclosure quality, CEO compensation, and CEO
turnover. A large part of an analyst’s job entails the perusal and analysis of financial information
disclosed by firms to capital markets and the use of such disclosure as a basis to evaluate managerial
decision making and forecast future performance. Therefore, we start our analysis by relating analyst
industry expertise to a number of observable outcomes of the choices made by managers in firms’
financial disclosure. This is also in keeping with earlier studies by Yu (2008) and Irani and Oesch
(2013) that examine the effect of analyst monitoring on corporate financial reporting quality.
3
We first examine the impact of analyst industry expertise on earnings management through
discretionary accruals, which are estimated using the method suggested by Owens, Wu, and
Zimmerman (2015). Earnings management is considered as a manifestation of the agency problems
between managers and shareholders, because managers are able to extract various forms of private
benefits and personal gains by manipulating reported financial results (e.g., Perry and Williams (1994)
and Bergstresser and Philippon (2006)). Shareholders, on the other hand, bear significant costs when
aggressive earnings management leads to financial misreporting that results in earnings restatements,
shareholder lawsuits, and regulatory/legal sanctions against the firm (e.g., Dechow, Sloan, and
Sweeney (1996) and Karpoff, Lee, and Martin (2008)). Analysts have incentives to be vigilant about
aggressive earnings management. Dyck, Morse, and Zingales (2010) find that failure to detect
accounting fraud at covered firms increases an analyst’s probability of being demoted.
As Dichev, Graham, Harvey, and Rajgopal (2013) point out, it is difficult for outsiders to
detect earnings manipulation, but large deviations from industry and peer norms can serve as a red
flag for potential financial misreporting because a substantial portion of a firm’s financial reporting
choices is driven by operations and economic conditions specific to its industry. Therefore, relevant
industry expertise and knowledge are essential for the evaluation of many aspects of corporate
financial reporting. Consistent with the effective monitoring hypothesis, we find that coverage by analysts
with related industry experience is significantly and negatively associated with firms’ earnings
management. In contrast, coverage by other analysts is not related to earnings management behavior.
These results are robust to controlling for a wide range of analyst and firm-specific attributes.
Next, we use the F-score developed by Dechow, Ge, Larson, and Sloan (2011) to capture
the ex-ante probability of firms engaging in financial misreporting and earnings restatements as an ex
post measure of financial misreporting. We find that both measures are significantly lower when
firms are followed by more analysts with related industry expertise. As is the case with our earnings
management results, coverage by other analysts is not related to the probability of financial
misreporting.
In further analysis, we examine firms’ CEO compensation and CEO turnover decisions. We
find that coverage by industry expert analysts is associated with significantly lower levels of CEO
excess compensation and higher responsiveness of firms in replacing poorly performing CEOs.
Closely echoing the results from our financial disclosure analysis, we find that coverage by other
analysts is not related to either CEO compensation or CEO turnover decisions. These results
provide additional support for the effective monitor hypothesis.
4
A potentially serious econometric concern with respect to our analyses is identification.
Because analyst coverage in general, and coverage by industry expert analysts in particular are not
random, baseline OLS regressions may suffer from endogeneity biases. For example, industry expert
analysts could utilize their superior industry knowledge to identify and provide coverage for firms
with fewer agency problems. To mitigate endogeneity concerns, whenever possible we conduct
additional analysis in which we focus on exogenous disappearances of analysts caused by brokerage
house mergers or closures following Kelly and Ljungqvist (2012). Employing a difference-indifferences (DID) approach, we observe a significant increase in firms’ earnings management, ex
ante probability of committing material financial misstatement (F-score), and CEO excess
compensation after they lose analyst coverage due to a brokerage merger or closure. However,
consistent with the baseline OLS results, this pattern only holds for the loss of coverage by analysts
with related industry experience.
Our collective findings indicate that industry expert analysts provide external monitoring of
firm managers and improve corporate governance. However, it is not clear how these analysts play
such a role. While much of analyst monitoring may take place behind the scenes in private
communication with management and hence is not observable, we uncover some evidence on how
they exert monitoring influence. First, using textual analysis, we find that industry expert analysts are
31% more likely to discuss governance-related issues in their reports compared to other analysts. In
addition, coverage by industry expert analysts discussing corporate governance issues in their reports
is associated with even a lesser degree of agency problems at covered firms. Second, we examine
whether issuing cash flow forecasts is a method that industry expert analysts utilize to discourage
earnings management by firms. This idea is based on the insight from McInnis and Collins (2011),
who find that analysts issuing cash flow forecasts alongside earnings forecasts increase firms’
difficulties in engaging in accruals manipulation. Our analysis shows that industry expert analysts
indeed are more likely to issue cash flow forecasts, and firms covered by industry expert analysts
issuing cash flow forecasts display even better financial reporting quality. Finally, we examine analyst
deviations from management guidance when making earnings forecasts, which may represent a less
costly means for analysts to signal their disagreements with management (e.g., Feng and McVay
(2010) and Louis, Sun and Urcan (2013)). We find that industry expert analysts are more likely to
deviate from management guidance and their deviation is associated with improved monitoring
effectiveness. Thus, we highlight several subtle mechanisms whereby industry expert analysts can
improve firms’ corporate governance outcomes.
5
Overall, our evidence provides strong support for the effective monitoring hypothesis. However,
certain situations may arise in which industry expert analysts’ judgement may be impaired. We
examine two such situations—analysts with professional ties to firm management and analysts
whose employers have investment banking relationships with firms. We find that having a
professional tie to firm management emanating from overlapping industry employment experience
does not hinder analysts’ monitoring ability. On the other hand, industry expert analysts with
investment banking relationships with the covered firm do not provide effective monitoring,
suggesting that the benefit of industry experience is offset by the incentives created by investment
banking business. This is consistent with the literature on investment banking conflicts where such
affiliation results in analysts catering to firm management.
Our study makes three primary contributions to the literature. First, our investigation into
multiple major corporate policies conveys a consistent message that industry expertise is a critical
element in enabling financial analysts to perform their external monitoring function in followed
firms. We complement recent studies on analysts’ monitoring role by identifying which analysts are
more effective monitors. In this respect, our evidence also complements the finding by Yu (2008)
that analysts working for top brokers and with longer forecasting experience have a larger effect on
firms’ earnings management. These results together highlight the importance of exploring analyst
heterogeneities. It is worth noting that the effect of analyst industry expertise we document is
incremental to those of an analyst’s general, industry-specific, and firm-specific forecasting
experience and thus represents a new and salient attribute that impacts analysts’ ability to fulfill their
monitoring roles. Furthermore, unlike general, industry-specific, or firm-specific forecasting
experiences, which analysts develop as outsiders looking in, the industry knowledge we focus on is
obtained by analysts as insiders from previously working in certain industries and thus is capable of
providing a deeper and more complete understanding of those sectors.
Second, we add to the emerging literature that investigates the implications of industry
expertise for a variety of economic agents in corporate and financial market settings, such as CEOs,
corporate directors, and investment banks (e.g., Liu and Ritter (2011), Custodio and Mertzer (2013),
Wang, Xie and Zhang (2015)). Our paper is the first to demonstrate the importance of industry
expertise for the monitoring and corporate governance role of financial analysts. As such, our
findings complement the evidence in Bradley, Gokkaya, and Liu (2015) that industry expertise
enables analysts to make more accurate earnings forecasts as well as the evidence in Dass et al.
6
(2013), Masulis et al. (2014) and Wang, Xie, and Zhu (2015) that industry expertise facilitates the
corporate governance functions of corporate boards.
Finally, we advance the literature’s understanding of how sell-side analysts provide external
monitoring of their coverage firms, an issue on which prior research is largely silent. We explore a
large set of potential channels and demonstrate the importance of three analyst monitoring
mechanisms: discussing corporate governance issues in research reports, issuing cash flow forecasts
simultaneously with earnings forecasts, and deviating from management guidance in earnings
forecasts.
2. Data and descriptive statistics
The primary data employed in this study are constructed from a number of sources.
Appendix A gives a detailed description of our sample construction process. We first pull all analysts’
annual EPS forecasts from I/B/E/S over the period of 1988 to 2011.4 This results in 17,775 unique
firms and 16,766 unique analysts. We then merge this sample with CRSP/Compustat to obtain firm
financial statement and stock price and return information. We next identify sell-side analysts who
provided at least one annual earnings forecast and merge this sample with the I/B/E/S
recommendation file to retrieve analyst last names, first name initials and brokerage house
information. We eliminate analysts with missing information on first name initials and last names
and also discard analyst research teams since only the last names of analyst team members can be
obtained from I/B/E/S. This initial screening results in 9,275 unique analysts following 6,713 firms
from 1988 to 2011.
We search Zoominfo.com, an employment background indexing website to capture the
surviving analysts’ full first names. We follow a very conservative approach in our web search and
require that analyst last names, first name initials, and brokerage houses match the information
gathered from I/B/E/S. This leaves us with 4,845 analysts covering 6,391 unique firms. For each
remaining analyst in our sample, we manually collect information on employment backgrounds from
LinkedIn.com, the world’s largest professional network.5 We search for the analyst’s full name along
with the corresponding brokerage house’s name and collect detailed information on the names of
the analyst’s pre-analyst employers and years of employment.
4
Our sample starts in 1988 because we follow Collins and Hribar (2002) in measuring accruals based on information
from the cash flow statement, which is not available until 1988.
5 In section 5.4 we discuss two potential selection issues with using analysts’ LinkedIn profiles.
7
We next break down analysts’ employment experience into “related” and “unrelated” at the
firm-level within their coverage portfolio. An analyst is coded as having “related industry experience”
in a covered firm if the firm and the analyst’s prior employer(s) are from the same industry based on
the 4-digit Global Industry Classification System (GICS) classification codes, which create 24
industry groups. Pre-analyst work experience is defined as “unrelated experience” if the GICSindustry classifications of the followed firm and the analyst’s prior employer(s) do not match.
Analysts without pre-analyst industry work experience are classified as “inexperienced.” We use the
information from Compustat and CRSP to identify the industry of analysts’ prior employers that are
publicly traded. For former employers that are privately held, we manually obtain their business
descriptions from a variety of sources, including LinkedIn, Hoover’s, Thomson Financial’s Security Data
Corporation (SDC) database, the Securities and Exchange Commission’s (SEC) website, business
news websites (e.g., Bloomberg and Business Week) and official corporate websites.
To illustrate our data collection process, consider a sell-side analyst in our sample that
previously worked at Conexant Systems (a private company) as a Process Development Engineer for
3 years and 1 month. We obtain Conexant Systems’s industry classification (semiconductors,
GICS=4520) from the company’s LinkedIn page. In a given year, this analyst covers Ultratech Inc.
and Cabot Microelectronics, both of which share the same primary 4-digit GICS as Conexant
Systems. Therefore, he is defined as having related industry experience for these two firms. The
analyst also covers Polypore International and Dynamic Materials (capital goods, GICS=2010) RTI
International Metals and Titanium Metals (materials, GICS=1510), which do not have the same
GICS as his prior employer. Therefore, the analyst is defined as having unrelated industry experience
for his coverage in these latter cases.
***Insert Table 1 here***
In Panel A of Table 1 we provide the top five 4-digit GICS industries from which analysts
obtain pre-analyst work experience as well as the top five industries with the most industry expert
analysts. 6 Pre-analyst work experience most often comes from Software & Services followed by
Diversified Financials. Not surprisingly, these industries also have the highest percentage of analysts
6
In the left half of the panel, we aim to show the distribution of industries from which analysts in our sample obtained
any (not necessarily related) pre-analyst industry experience. For instance, of the 1,272 analysts in our sample who have
pre-analyst industry experience, 162 or 12.7% (=162/1,272) obtained experience in the Software & Services industry.
In the right half of the panel, we aim to show the distribution of industry expert analysts across industries. In our sample,
there are 4,419 instances where in a given year an analyst covers at least one stock on which he/she has related industry
expertise. We then separate these related analyst-industry-year combinations into 24 4-digit GICS industries. There are
791 related analyst-year-industry combinations in the Software & Services industry. The 17.9% figure shown in the
column for this industry is computed as 791/4,419.
8
with related prior industry experience. However, the third through fifth most populated industries in
which analysts have prior work experience are Commercial Services & Supplies, Banks, and Media,
while Health Care Equipment & Services, Retailing, and Technology Hardware & Equipment
represent the corresponding top industries with industry expert analysts. Overall, both pre-analyst
work experience and industry expert analysts have a fairly wide distribution across all 24 GICS
industries.7
Panel B of Table 1 presents summary statistics of key variables. Appendix B provides a
detailed explanation for the construction of these variables. All variables are winsorized at the 0.5%
and 99.5% tails to reduce the impact of outliers. We note that the number of firm-year observations
associated with each dependent variable is different, because the construction, time period, and size
of samples vary across the different analyses we conduct in the paper. For expositional convenience,
the summary statistics of firm financial and analyst coverage characteristics are based on the sample
used for the accrual-based earnings management analysis, which consists of 24,918 firm-year
observations from 1988 to 2011.
For the dependent variables, the average firm has a ratio of absolute discretionary accruals to
total assets of 0.047, an F-score residual of -0.007, and total CEO compensation of $3.434 million.
About 1.7% of firm-year observations witness intentional financial misreporting that leads to
earnings restatements, and about 2.7% of firm-year observations experience forced CEO turnovers.
In terms of firm financial characteristics, the average firm has a market value of equity
(Market cap) of $4.6 billion, a market-to-book ratio (Market-to-book) of 3.2, a return on assets (ROA)
of 3.4%, a growth rate in the book value of total assets (Total assets growth) of 22.9%, a standard
deviation of cash flows divided by total assets (Cash flow volatility) of 22.2%, and a ratio of external
financing to total assets (External financing) of 2.6%. In addition, we control for equity ownership
held by dedicated and non-dedicated institutional investors in each firm.8 The average ownership by
dedicated institutional investors is 8.3% and the average ownership by non-dedicated institutional
investors is 55.3%. These values are in line to those reported in other studies (e.g., Bushee and Noe
(2000) and Chen, Harford, and Li (2007)).
Regarding analyst coverage variables, the average firm is covered by 3.97 sell-side analysts,
with 0.90 having related pre-analyst industry experience (Related analysts), 1.70 having unrelated pre7
To ensure our results are not driven by the most represented industries, in untabulated analysis we limit our sample to
the non-top 5 industries and continue to find similar results.
8 We follow Chen, Harford, and Li (2007) and identify dedicated institutional investors based on Bushee’s (1998)
classification methodology.
9
analyst industry experience (Unrelated analysts), and 1.38 having no pre-analyst industry experience
(Inexperienced analysts). 9 We separately calculate an analyst’s general and firm-specific forecasting
experience as the number of years since the analyst initially appeared in I/B/E/S (Experience as
analyst) and the number of years in which the analyst provided coverage for a particular firm
(Experience with firm). For the average firm in our sample, the average analyst covering the firm has
6.31 years of general experience, 2.45 years of firm-specific experience, a research portfolio
consisting of 12.56 unique firms (Analyst portfolio size), and 56.5% of analysts work at a top brokerage
house (Analysts from top brokers) while 13.1% attain all-star status (All-star analyst).
3. Empirical results
In this section we present the results from our examination of the effective monitor hypothesis
versus the impaired monitor hypothesis. Sections 3.1 and 3.2 evaluate the quality of a firm’s financial
disclosure policy in relation to coverage by analysts with related industry experience, with Section 3.1
based on accruals-based earnings management and Section 3.2 on measures of financial
misreporting. Section 3.3 investigates the effect of analyst industry experience on CEO
compensation policies. Section 3.4 examines whether the presence of analysts with related industry
experience affects the sensitivity of forced CEO turnovers to firm performance.
3.1. The effect of analyst industry expertise on earnings management
We examine whether the presence of industry expert analysts (Related analysts) is related to
firms’ earnings management, which we measure using the absolute value of discretionary accruals
(Abs_DA). Managers have incentives to manage earnings upward (e.g., to inflate the stock price
prior to their stock sales, or to hit performance targets for bonuses) as well as downward (e.g., prior
to option repricing or management buyouts or as an attempt to smooth earnings). Because our
hypotheses are related to the magnitude rather than the direction of earnings management, we
follow prior studies, such as Bergstresser and Philippon (2006) and Yu (2008), and use the absolute
value of discretionary accruals scaled by total assets as our dependent variable of interest. Dopuch et
al. (2011) and Owen, Wu, and Zimmerman (2015) show that measurement error in discretionary
9
Because we only include analysts with a LinkedIn profile, the number of analysts following a firm in our sample is
naturally lower. An alternative sampling strategy is to include analysts without a LinkedIn profile and simply classify them
as either unrelated or inexperienced analysts. Of course, doing so will misclassify some analysts with related industry
experience as otherwise and therefore bias against our finding any difference in monitoring effectiveness between related
analysts and the other two groups. Nevertheless, all our results continue to hold under this sampling approach.
10
accruals is related to the homogeneity of the underlying industry. Therefore, we estimate
discretionary accruals based on the accrual models developed by Owens, Wu, and Zimmerman
(2015) that explicitly take into account heterogeneities in firms’ economic fundamentals.10 However,
to the extent that empirical proxies for firms’ business model shocks are noisy, model
misspecification remains possible and the discretionary accrual estimates may still be contaminated
by measurement errors (Owens et al. (2015)). We mitigate this concern by supplementing our
discretionary accruals analysis with an examination of two alternative financial misreporting
measures that do not require the modeling of firms’ accrual generating processes (see Section 3.2).
3.1.1. Baseline OLS regressions
We first estimate OLS regressions of earnings management against analyst coverage while
explicitly controlling for a battery of firm and analyst-level characteristics as discussed in Section 2.
We include firm and year-fixed effects in the regressions to mitigate potential concerns about timeinvariant unobservable firm characteristics being correlated with both earnings management and
analyst coverage. Our formal model is specified in equation (1). To mitigate reverse causality
concerns, all firm financial characteristics included as control variables are lagged by one year.
Abs_DAi,t= β0 +β1Total analystsi,t-1 [or β2Related analysts i,t-1 +β3Unrelated analysts i,t-1
+β4Inexperienced analysts i,t-1] +β5Market cap i,t-1 +β6Market-to-book i,t-1 +β7ROA i,t-1
+β8Total assets growth i,t-1 +β9Cash flow volatilityi,t-1 +β10External financing i,t-1
+β11Institutional ownership (dedicated) i,t-1 +β12Institutional ownership (non-dedicated) i,t-1
+β13Experience as analysti,t-1+β14Experience with firmi,t-1+β15Analysts from top brokersi,t-1
+β16All-star analysts i,t-1 +β17Analyst portfolio size i,t-1 +Firm and year fixed effects + ε (1)
***Insert Table 2 here***
Table 2, Panel A reports the regressions results. As in the rest of the tables, robust t-statistics
based on standard errors adjusted for heteroskedasticity and firm-level clustering are in the
parentheses below coefficient estimates. As a starting point of our analysis, we replicate the finding
by Yu (2008) and Irani and Oesch (2013) that analyst coverage is negatively related to firms’ earnings
10
Specifically, we use the accrual model specified in equation (13) in Owens et al. (2015) and the firm-specific stock
return variation measure in Chun, Kim, Morck, and Yeung (2008) to capture the heterogeneity in firms’ underlying
economics and proxy for business model shocks. Our results are also robust to using the models specified in equations
(14) and (15). In addition, we obtain qualitatively similar results with higher statistical significance when we use the
discretionary accruals estimated from the modified Jones (1991) model in Dechow, Sloan, and Sweeney (1996) and
Warfield, Wild, and Wild (1995) and the performance-adjusted discretionary accruals in Kothari, Leone, and Wasley
(2005) and Gong, Louis, and Sun (2008).
11
management. Specifically, in model 1, we regress a firm’s absolute value of discretionary accruals
(Abs_DA) against the total number of analysts following the firm. Consistent with analysts playing
an external monitoring role, we find that total analyst coverage is significantly and negatively
associated with Abs_DA.
Next we decompose the total number of analysts following a firm into analysts with related
industry experience (Related analysts), analysts with unrelated industry experience (Unrelated analysts),
and analysts without pre-analyst industry work experience (Inexperienced analysts). We reestimate the
abnormal accruals regression and present the results in model 2. We find that only coverage by
analysts with related industry experience is significantly and negatively associated with a firm’s
abnormal accruals. While the coefficients on the other two types of analyst coverage are also
negative, they are not statistically significant. These findings suggest that the negative association
between analyst coverage and earnings management documented by prior studies (Yu (2008), Irani
and Oesch (2013), and Chen, Harford, and Lin (2015)) is largely driven by analysts with related
industry experience. As such, they support the effective monitor hypothesis that financial analysts with
related industry experience provide more effective monitoring of firms’ financial disclosure.
An empirical challenge in identifying a causal relation between analyst coverage and
corporate financial disclosure is the possibility that an analyst’s decision to cover a firm is associated
with observable factors that could also affect the firm’s financial reporting (Healy, Hutton, and
Palepu (1999)). Therefore, we follow Yu (2008) and construct residual analyst coverage to mitigate
this potential endogeneity concern. Residual coverage is defined as the component of analyst
coverage uncorrelated with firm-specific control variables and the main determinants of earnings
management as listed in Panel B of Table 1.11 We then use residual coverage as our main proxy for
analyst coverage in the remaining analysis. The regression results reported in models 3 and 4 of
Table 2, Panel A are very similar to those in models 1 and 2; only coverage by analysts with related
industry experience is significantly associated with firms’ earnings management behavior.
3.1.2. Brokerage mergers and closures as quasi-natural experiments
11
We estimate a series of analyst coverage regressions, where the dependent variables are the number of 1) total analysts
2) analysts with related industry experience, 3) analysts with unrelated industry experience, and 4) analysts without prior
industry experience. The residuals from these regressions are then labeled as “residual coverage” for the corresponding
type of analysts. Untabulated results indicate that the determinants of analyst coverage are quantitatively and qualitatively
similar across these three types of analysts. For instance, analyst coverage is positively related to firm size and growth
rate of assets, but negatively related to external financing activities.
12
As an alternative approach to mitigating endogeneity concerns, we exploit a unique setting
where there is an unexpected exogenous shock to a firm’s analyst coverage independent of the firm’s
characteristics or earnings management behavior. Specifically, this identification strategy relies on
two plausibly exogenous quasi-natural experiments, namely, brokerage closures (Kelly and
Ljungqvist (2012)) and brokerage mergers (Hong and Kacpercyzk (2010)). These brokerage-related
events are used by a growing body of research to study the impact of financial analyst coverage on a
wide array of firm policies.
Kelly and Ljungqvist (2012) argue that following a merger, termination of coverage by the
surviving broker could be endogenous because it chose to no longer cover the stock. Thus, we follow
their approach and focus on coverage losses resulting from a merger where a stock was covered by
analysts from both brokers before the merger and by only one of the analysts after the merger. We
identify 17 brokerage house closures and 37 brokerage house mergers over the period of 1988 to
2008 and construct a sample of 701 treatment firms that were covered by the closed or merged
brokers prior to these events.12 We employ a difference-in-differences (DID) approach where we
focus on the change in a firm’s abnormal accruals from year t-1 to year t+1, with year t being the
year in which the firm loses analyst coverage due to a brokerage closure or merger. We compare the
change in abnormal accruals of treatment firms (treatment firm difference) with the change in
abnormal accruals of control firms over the same time period (control firm difference). This allows
us to control for time trends that may be common to both treatment and control firms and mitigates
concerns related to reverse causality or omitted variables. Following Kelly and Ljungqvist (2012), we
construct an initial sample of control firms that are covered by at least one analyst and are in the
same size and book-to-market quintiles as the treatment firms in the month of June preceding the
broker events. We further require that control firms do not experience analyst losses in the year of
or the year before the broker event. We retain the control firm closest to the treatment firm in the
number of analysts providing coverage.13
Column 1 of Panel B in Table 2 presents the results for all analyst losses. Consistent with the
baseline regressions in Panel A, we find that firms losing analyst coverage emanating from brokerage
12
Brokerage closure data is from Kelly and Ljungqvist (2012) and He and Tian (2013) over 1993 and 2008. Brokerage
house merger information comes from a list of M&A activities in the investment banking industry compiled by Wang,
Xie, and Zhang (2015) from various sources. As discussed earlier, the sample of treatment firms is smaller than in
previous studies because we limit our attention to analysts with a LinkedIn profile. However, if we include all nonLinkedIn analysts and classify them as inexperienced we obtain qualitatively similar results.
13 For robustness, we also require that treatment and control firms come from the same 4-digit GICS industry. This
additional matching criterion reduces the number of treatment firms to 637. Results from the smaller sample are very
similar to those reported in the paper.
13
mergers or closures engage in more aggressive earnings management behavior in the post-event year
compared to the pre-event year (average treatment firm difference=0.27%, t-stat=1.88). Focusing on
the difference in differences (DID) between treatment and control firms, our results indicate
significantly more aggressive earnings manipulation for treatment firms (average DID=0.39%, tstat=2.15). More important for our purpose, when we differentiate among the types of analysts lost
by firms in columns 2 and 3, we find that only the loss of coverage by analysts with related industry
expertise precipitates a significant increase in earnings management. Specifically, in the case of
industry expert analyst losses, the average DID in abnormal accruals is 1.21%, which is significant at
the 1% level. It is also economically important given the mean and median values of Abs_DA in our
sample are 4.7% and 2.9%, respectively. In contrast, in the case of other analyst losses, the average
DID in abnormal accruals is 0.23%, insignificantly different from zero. A further comparison of the
DID between firms losing industry expert analysts versus other analysts indicates that the loss of
industry expert analysts leads to more pronounced increases in earnings management (difference in
DID=0.98%, t-stat=-2.21). Overall, these results provide further support for the effective monitor
hypothesis and suggest that related industry expertise is a valuable attribute for financial analysts to
possess in order to effectively perform their monitoring function.14
3.2. Industry expert analyst coverage and financial misreporting by firms
In this section, we examine the relation between industry expert analyst coverage and the
likelihood of firms committing financial misreporting, a more egregious type of behavior than
accrual-based earnings management. Following Wang, Xie, and Zhu (2015), we employ both ex-ante
and ex-post measures of financial misreporting. We use the F-score developed by Dechow, Ge,
Larson, and Sloan (2011) to measure the ex-ante probability of a firm committing financial
misrepresentation. Our ex post indicator of financial misreporting is earnings restatements by firms.
3.2.1. Analyst industry expertise and F-score
Dechow et al. (2011) construct a comprehensive database through careful examination of
firms subject to U.S. Securities and Exchange Commission (SEC) enforcement actions as indicated
in Accounting and Auditing Enforcement Releases (AAERs). They then identify a sample of firms
14
As an alternative to the DID approach, we also use the regression approach of Chen, Harford, and Lin (2015) and
Irani and Oesch (2013), in which we regress changes in the absolute value of discretionary accruals against exogenous
losses in analyst coverage along with changes in the control variables. We find that a firm’s abnormal accurals experience
significant increases only after the loss of coverage by analysts with related industry experience.
14
that have misstated at least one of their quarterly or annual financial statements with the intent of
misleading investors. The F-score is created from a model that analyzes the financial characteristics
of these misstating firms. Given that part of the construct for the F-score uses accruals information,
we orthogonalize the F-score against discretionary accruals and use the residuals as the dependent
variable for our analysis.15 This allows us to provide evidence that is truly complementary to the
discretionary accruals-based results reported in Section 3.1. We estimate regressions of F-score
residuals against different types of analyst coverage, along with the same set of control variables as
included in equation (1). Panel A of Table 3 presents the regression results.
***Insert Table 3 here***
The most immediate takeaway from Model 1 of Panel A is that the intensity of analyst
coverage is significantly and negatively related to a firm’s F-score residual, consistent with analysts
playing a monitoring role and reducing firms’ ex-ante likelihood of earnings manipulation. However,
once again, we find that this result is driven by coverage by analysts with related industry experience
(Model 2). Conversely, coverage by analysts with unrelated industry experience or no industry
experience is not significantly associated with the ex ante probability of intentional financial
misrepresentation.
We next reexamine the relation between industry expert analyst coverage and the F-score
residuals using exogenous shocks to a firm’s analyst coverage due to brokerage closures or mergers.
As in the earnings management analysis in Section 3.1.2, we employ a DID approach around a twoyear event window surrounding the brokerage closures or mergers and examine the change in a
firm’s F-score residual from year t-1 to year t+1, with year t representing the year in which the firm
loses analyst coverage due to such brokerage events. Panel B of Table 3 presents these results.
In column 1, we find that firms experience a significant increase in the F-score residuals
when they experience a loss of analyst coverage, consistent with monitoring by analysts reducing
firms’ ex ante probability of committing serious financial misreporting. In column 2, when we
differentiate between the types of analyst coverage lost by firms, we find that only the loss of
coverage by industry expert analysts leads to a significant increase in a firm’s F-scores. This result
reaffirms our finding from Panel A about the importance of related industry experience in
facilitating the monitoring role of analysts, thereby lending further support to the effective monitor
hypothesis.
15
Using the original F-score produces qualitatively similar results.
15
3.2.2. Analyst industry expertise and earnings restatements
We construct our sample of earnings restatements from two data sources. We start with the
U.S. General Accounting Office’s (GAO) Financial Statement Restatement Database as published in
2003 and 2007. The GAO dataset is manually constructed using a Lexis-Nexis text-based search for
the variants of the word “restate,” and contains restatements announced between January 1, 1997
and June 30, 2006. We then extend the GAO dataset to 2011 by using the accounting restatements
obtained from the Audit Analytics restatement database. Hennes, Leone, and Miller (2008) develop a
methodology that classifies a restatement as an irregularity if it satisfies at least one of the three
criteria: (i) variants of the words "irregularity" or "fraud" were explicitly used in restatement
announcements or relevant filings in the four years around the restatement; (ii) the misstatements
came under SEC or DOJ investigations; and (iii) independent investigations were launched by the
boards of directors of restatement firms. In a sample of restatements between 2002 and 2005, they
demonstrate the importance and effectiveness of their classification scheme by showing that
compared to error restatements, irregularity restatements are met with significantly more negative
announcement returns (on average: -14% vs. -2%), are followed by shareholder class action lawsuits
at a significantly higher rate, and lead to significantly more CEO/CFO turnovers. Therefore, we
focus on irregularity restatements because they are more likely to represent instances of deliberate
and material earnings manipulation by managers. For each irregularity restatement, we obtain
information on the specific years for which financial results were restated due to earnings
manipulation.
We merge the restated year information with our comprehensive CRSP-CompustatI/B/E/S sample and obtain a sample of 19,772 firm-year observations from 1997 to 2011. In about
1.7% (336) firm-years, earnings were restated due to accounting irregularities.16 We estimate logistic
regressions to examine whether coverage by industry expert analysts is related to the probability of a
firm committing intentional financial misreporting in a given year. The dependent variable is equal
to one for restated firm-years and zero otherwise. In addition to the covariates in equation (1), we
also control for how prone an industry is to earnings restatements. The rationale for this control is
that industries where firms have a high proclivity to commit accounting fraud and restate earnings
may be considered undesirable and thus less likely to produce employees who would go on to
become financial analysts in the future. This possibility creates a potentially artificial negative relation
16
There are 428 firm-years in which earnings were restated due to accounting errors. Due to their substantially less
serious nature compared to accounting irregularities, we treat the error affected firm-years as non-restated years. Our
results are nearly identical if we remove these observations from our analysis.
16
between a firm’s probability of earnings restatement and coverage by analysts with related industry
experience.17 To mitigate this concern, we construct a measure to capture an industry’s exposure to
restatements. Specifically, we first compute the percentage of firms in each industry that were
associated with an irregularity restatement over the past ten years. We create a binary variable, High
restate exposure, that is equal to one for industry groups with above sample median restatement
frequency and zero otherwise. Table 4 reports the results. The coefficient estimates presented under
Model 1 suggest that the probability of intentional financial misreporting is negatively and
significantly associated with the total number of analysts following. However, as model 2 indicates,
this significant and negative relation is largely attributable to industry expert analysts, and there is no
evidence that coverage by other analysts is related to the probability of financial misreporting.
Economically, the coefficient estimates in model 2 imply that the likelihood of intentionally
misreporting financial restatements is lower by 35% per one standard deviation increase in coverage
by industry expert analysts.18 Finally, it is worth noting that the High restate exposure indicator has a
significantly positive coefficient, suggesting that firms in industries that historically are more fraud
prone indeed have a higher probability of committing financial misrepresentation.
***Insert Table 4 here***
Overall, the evidence in Sections 3.1 and 3.2 lends strong support to the effective monitor
hypothesis that industry expertise improves the effectiveness of analysts’ monitoring and curtailing of
firms’ earnings manipulation activities as reflected in abnormal accruals, F-score, and irregularitydriven financial restatements.19
3.3. Analyst industry expertise and CEO compensation
In the context of firms’ executive compensation policy, Chen, Harford, and Lin (2015) show
that analyst coverage can serve as an additional layer of market discipline to reduce excess
17
We thank the editor and referee for suggesting this possibility. In unreported analyses, we formally evaluate whether
industries that are more stable, homogeneous, and successful are more likely to produce employees that go on to
become analysts of the same industry. We find no evidence that the percentage of industry expert analysts among all
analysts covering a particular industry is related to the homogeneity, volatility, and performance of the industry.
18 We are not able to use brokerage closures/mergers in this setting because only two firms announced irregularity
restatements after losing coverage by industry expert analysts.
19 In untabulated results, we use a firm being the target of a securities class-action lawsuit as an alternative ex post
measure of financial misreporting, to the extent that earnings manipulation with an intention to mislead investors can
lead to class action lawsuits (DuCharme, Malatesta, and Sefcik (2004)). We obtain a comprehensive sample of 3,566
securities class action lawsuits on 2,152 unique firms from 1996 to 2011 from the Securities Class Action Clearinghouse
website (http://securities.stanford.edu/). We find that coverage by industry expert analysts are associated with a
significantly lower probability of a coverage firm being targeted by a class action lawsuit, whereas coverage by other
analysts is not significantly related to the probability of being sued.
17
compensation received by top executives. We argue that analysts’ industry expertise can provide
them with a comparative advantage in understanding industry and firm-specific factors that have
economic impact on the competitive level of executive compensation. As a result, they are more
likely to recognize large discrepancies between a CEO’s compensation and her fair and competitive
level of pay. Their opinions are also likely to be taken seriously by managers and boards because of
analysts’ unique position to influence the stock price.20 As such, the effective monitor hypothesis predicts
that firms followed by more industry expert analysts are less likely to award exorbitant compensation
packages to their top executives. The impaired monitor hypothesis, however, argues that analysts with
related industry experience may be connected to the managers of the firms they follow and thus
have less incentive to question overly generous CEO pay.
To test these predictions, we obtain CEO compensation information including salary, bonus,
stock and option grants, and long-term incentive plans from Standard & Poor’s ExecuComp
database for our sample. We assess the impact of analyst industry expertise on excess CEO
compensation. To estimate excess CEO compensation, we regress the logarithmic transformation
of total CEO compensation against a firm’s market capitalization, buy-and-hold abnormal returns
over CRSP value-weighted returns, stock return volatility, as well as industry (based on two-digit SIC
codes) and year fixed effects. The residuals from this estimation is our measure of excess CEO
compensation (e.g., Hartzell, Ofek, and Yermack (2004), Cai, Garner, and Walkling (2009)). We
estimate OLS regressions of excess CEO compensation against analyst coverage while controlling
for firm financial and governance characteristics as well as firm and year-fixed effects. The
regression model is specified as follows.
CEO excess compensationi,t= β0 +β1Total analysts (Residual) i,t-1 [or β2Related analysts (Residual) i,t-1
+β3Unrelated analysts (Residual)i, t-1+β4Inexperienced analysts (Residual) i, t-1 +β5Market capi, t-1
+β6Tobin’s Q i, t-1 +β7Leverage i, t-1 +β8R&D/sale i, t-1 +β9Capx/sale i, t-1 +β10Xad/sale i, t-1
+β11Abnormal ROA i, t-1 +β12Abnormal stock rets i, t-1 +β13CEO tenure i, t-1 +β14Firm age i, t-1
+β15Institutional ownership (dedicated) i, t-1+β16 Institutional ownership (non-dedicated) i, t-1
+β17Experience as analyst i, t-1 +β18Experience with firm i, t-1 +β19Analysts from top brokers i, t-1
+β20 All-star analysts i, t-1 +β21Analyst portfolio size i, t-1 +Firm and year fixed effects +ε (2)
***Insert Table 5 here***
Panel A of Table 5 presents the regression results. Consistent with the effective monitor
hypothesis, we find that analyst coverage is associated with significantly lower levels of excess CEO
20
According to the CFO survey by Graham, Harvey, and Rajgopal (2005), about 54.6% of respondents rank analysts
among the two most important stock price setters for their companies.
18
compensation (model 1), but the relation is again driven by analysts with related industry expertise in
their followed firms (model 2). In terms of economic significance, the coefficient estimates from
model 2 imply that for a one standard deviation increase in the number of industry expert analysts,
excess CEO compensation is lower by 4.8%, representing roughly a reduction of $165,000 in CEO
compensation given the average CEO pay of $3.43 million in our sample.
In further analysis, we exploit exogenous shocks to a firm’s analyst coverage created by
brokerage house closures and mergers. Specifically, we focus on the change in a firm’s CEO excess
compensation from year t-1 to year t+1, with year t being the year in which the firm loses analyst
coverage due to a brokerage closure or merger. We then compare the changes in CEO compensation
of treatment firms experiencing loss of analyst coverage to those of similar control firms over the
same time period. Panel B of Table 5 presents the results. We find that the excess compensation for
a treatment firm’s CEO experiences a significant increase following the loss of analyst coverage
(column 1) compared to control firms. When we break down the loss of analyst coverage based on
the types of analysts, we find that only the loss of industry expert analysts leads to a significant
increase in CEO compensation (column 2). The difference in DID is also significant between firms
losing industry expert analysts and those losing other analysts (column 4).
3.4. Industry expert analyst coverage and CEO turnover-performance sensitivity
In this section we investigate whether coverage by industry expert analysts is related to the
responsiveness of boards in removing poorly performing CEOs. Analysts with expertise in a firm’s
related industries are likely to have a more thorough understanding of the firm’s operating
environment, and the opportunities and challenges it faces. Therefore, they are likely to be able to
better analyze and evaluate major corporate strategies and decision making. When firm performance
disappoints, they can put more pressure on the firm to change its strategic direction and, if necessary,
change its leadership.
To examine if analyst industry expertise plays a role in forcing out poorly performing CEOs,
we follow Jenter and Kanaan (2015) and construct our primary dataset of CEO forced turnovers
from Standard & Poor’s ExecuComp database. We then use news reports collected from LexisNexis to ascertain that CEO turnover is forced, not due to any of the following reasons such as
19
death, health problems, retirement, or promotion to a higher/comparable position at another firm
(e.g., Parrino (1997), Huson, Malatesta, and Parrino (2004), among others).21
We estimate a logistic model to assess the marginal impact of industry expert analysts on the
CEO turnover-performance sensitivity. The dependent variable equals one for years in which a
forced CEO turnover event occurs, and zero otherwise. The key independent variables include
abnormal firm performance, measured by the firm’s industry-adjusted abnormal return on assets
(ROA) over the previous year, and its interaction terms with the three types of analyst coverage. 22
This model specification allows CEO turnover-performance sensitivity to vary with the coverage of
different types of analysts.
***Insert Table 6 here***
Table 6 reports the results. We find a significantly negative coefficient on abnormal ROA,
affirming the well documented negative relation between the probability of forced CEO turnover
and firm performance (see, e.g., Coughlan and Schmidt (1985), Weisbach, (1988), Fich and
Shivdasani (2006)). More importantly, the interaction term between abnormal performance and
coverage by industry expert analysts has a significantly negative coefficient, while the coefficients on
the other two interaction terms are insignificant. This evidence is consistent with the notion that the
external monitoring by industry expert analysts makes corporate boards more responsive to poor
firm performance in disciplining poorly performing CEOs. Conversely, coverage by analysts without
related industry expertise is not related to the CEO turnover-performance sensitivity.23
4. How do analysts monitor?
So far we have provided evidence consistent with analysts with related industry experience
performing an important external monitoring function for firms they cover. However, it is not
entirely clear how these analysts exert their monitoring efforts. Therefore, in this section we aim to
better understand the mechanisms through which industry expert analysts influence firms’ corporate
21
As described by Jenter and Kannan (2015), “all departures for which the press reports that the CEO is fired, forced
out, or retires or resigns due to policy differences or pressure are classified as forced. All other departures for CEOs
above and including age 60 are classified as voluntary. Departures for CEOs below age 60 are reviewed further and
classified as forced if either the press does not report the reason as death, poor health, or the acceptance of another
position (including the chairmanship of the board), or the press reports that the CEO is retiring, but does not announce
the retirement at least six months before the succession. Finally, the cases classified as forced can be reclassified if the
reports convincingly explain the departure as due to reasons that are unrelated to the firm’s activities.”
22 We obtain similar results when we measure firm performance using abnormal stock returns over the previous year.
23 Only one firm experienced a forced CEO turnover following the loss of industry expert analyst coverage, making it
econometrically unreliable to analyze turnover-performance sensitivity using shocks created by brokerage closures and
mergers.
20
governance and constrain managerial opportunism.24 A challenge for our analysis is that much of
analysts’ monitoring activities may be unobservable to outsiders, and even the observable actions
can be quite subtle because analysts that aggressively question and criticize corporate policies and
strategies risk alienating prized access to firm executives. Therefore, analysts are more likely to
convey their concerns and disapprovals regarding certain managerial decision making to firms
through private communication. Nevertheless, we examine several observable actions of analysts to
shed light on how they play a monitoring role.
4.1. Governance discussions in analyst reports
We first consider discussions of governance-related issues in analyst research reports. We
obtain full-text analyst research reports from Thomson Reuters Investext for our sample of analysts and
extract the header information for each report. We obtain the name of the analyst writing the report,
the broker, the firm and date published. We match the Thomson Reuters Investext header information
with I/B/E/S broker-analyst-firm pairs and manually verify these matches. We then parse analyst
reports for governance-related keywords in the following categories: ‘corporate governance,’
‘financial restatement,’ ‘shareholder proposal,’ ‘shareholder activism,’ ‘CEO compensation,’ or ‘CEO
turnover’. To come up with our list of keywords, we read 1000 randomly selected analyst reports
and identify the ways in which analysts discuss corporate governance.25 We look at every bigram
word combination in compiling the keyword list (see Appendix C).
We then merge the parsing results with our main sample. Analyst-year observations are
identified as “governance-related” for the corresponding coverage firm if one of the research reports
issued by the following analyst contains one of our keywords and zero otherwise. We find
(untabulated) that 9.7% of industry expert analysts discuss at least one of these governance issues for
their coverage firms compared to 5.6% for non-expert analysts. These differences are statistically
significant (t-stat=7.3).
As a more formal way to explore the likelihood of governance discussions in research
reports, we estimate a multivariate logistic regression controlling for a host of firm and analyst
characteristics defined previously. The dependent variable is a binary indicator that equals one if the
analyst discusses one of the governance-related issues for their coverage firms in a given year, zero
otherwise. The key independent variables are two indicator variables: Related analyst, which is equal to
24
25
We greatly acknowledge many constructive comments from the referee and the editor on this section.
We thank Tim Loughran for his valuable advice on the keyword identification strategy.
21
one for analysts with related industry experience in the firms they cover, and Unrelated analyst, which
is equal to one for analysts with unrelated industry experience in the firms they cover. Inexperienced
analysts serve as the omitted category. We present the regression results in model 1 of Panel A in
Table 7. We find that Related analyst has a significant and positive coefficient, suggesting that industry
expert analysts are more likely to discuss governance issues in their reports. Economically, the
likelihood of such discussions is 31% higher (t-stat=4.47) for industry expert analysts relative to
inexperienced analysts. On the other hand, unrelated industry work experience is not significantly
associated with discussion of governance issues.
***Insert Table 7 here***
Given that industry expert analysts are more likely to discuss governance issues in their
reports, we next examine if this form of monitoring impacts corporate governance outcomes. We
reestimate the discretionary accruals regression in equation (1) by partitioning Related analysts into
two variables based on whether an industry expert analyst discusses governance issues in research
reports for a coverage firm in a given year.26 Column 1 of Panel B presents the regression results
with the coefficients on control variables suppressed for brevity. We find that the negative relation
between industry expert analyst coverage and discretionary accruals is more pronounced when
industry expert analysts issue research reports discussing governance issues (p-value for the
coefficient difference: 0.029). This suggests that corporate governance discussion in research reports
represents one mechanism through which industry expert analysts monitor firm managers and exert
influence over firm governance.27
4.2. Cash flow forecasts
McInnis and Collins (2011) show that by issuing cash flow forecasts along with earnings
forecasts, analysts may increase firms’ difficulty and costs of engaging in accruals manipulation.
Building on their insight, we examine whether cash flow forecast issuance is a mechanism for
industry expert analysts to improve firms’ financial reporting quality. Toward that end, we first
26
We conduct analogous tests for other outcome variables, i.e., F-score residuals, earnings restatements, CEO
compensation and CEO turnover-performance sensitivity, and find qualitatively similar results (see Table IA.1 in the
Internet Appendix).
27 We perform a similar analysis of analysts’ takeaway reports following analyst investor days. These are firm-hosted
investor relations events where analysts are invited to meet with executives, hear about the strategy, financial position
and provide an opportunity for analysts to ask questions to management. While the sample size is considerably smaller
for these takeaway reports, using the same textual analysis strategy, we continue to find that industry expert analysts are
more likely to discuss governance-related issues in their analyst takeaway reports published following analyst/investor
days compared to non-expert analysts (5.0% versus 2.5%, respectively).
22
examine whether industry expert analysts are more likely to issue cash flow forecasts. In particular,
we estimate a logistic regression where the dependent variable equals one if an analyst
simultaneously provides both earnings and cash flow forecasts for the coverage firm of interest, and
zero otherwise. Results in model 2 of Panel A indicate that industry expert analysts are 37.5% more
likely (t-stat=5.62) to issue cash flow forecasts compared to other analysts.
Next, we investigate the impact of such forecasts on coverage firms’ earnings management
behavior. In particular, we distinguish between industry expert analysts based on issuance of cash
flow forecasts and reestimate equation (1). In model 2 of Panel B, we find that the negative relation
between industry expert analyst coverage and discretional accruals is stronger when industry expert
analysts issue cash flow forecasts (p-value for the coefficient difference: 0.047).28 Thus, our analysis
suggests that cash flow forecasts may represent another mechanism through which industry expert
analysts perform their external monitoring function.
4.3. Deviations from management guidance
In this section, we consider whether industry expert analysts deviate from management
earnings guidance as a means to monitor and improve firm governance, and in particular financial
reporting. Feng and McVay (2010) suggest that analysts deviate less from management earnings
guidance when they have incentives to please management. Similarly, Louis, Sun, and Urcan (2013)
argue that intentionally deviating from management guidance is a way in which analysts can subtly
express their disagreement with management when they believe the earnings guidance contains a
large managed component. Therefore, we first investigate whether industry experts deviate more
from management guidance when issuing earnings forecasts. Similar to Louis, Sun, and Urcan
(2013), the dependent variable of this analysis is the absolute value of the deviation of analyst i’s
earnings forecast from firm j’s earnings guidance for year t scaled by firm j’s stock price on the day
prior to the management guidance (abs DEVAF). To construct this variable, we use the last
management guidance issued for year t and the analyst’s forecast following the guidance but prior to
the actual earnings announcement. The key independent variables are Related analyst and Unrelated
analyst, the binary indicators for analyst types defined in Section 4.1. In addition to the firm financial
and analyst characteristics controls used in equation (1), we follow Louis, Sun, and Urcan (2013) in
adding the following controls: the absolute value of discretionary accruals (Abs_DA); negative
28
In models 1 and 2 of Internet Appendix Table IA.2, we perform similar tests for F-score residuals and restatements.
The results are not sensitive to how financial reporting quality is measured.
23
earnings (Loss); market-adjusted excess returns from two days following the management guidance
to one day prior to the analyst forecast date (Abret); market-adjusted excess returns over the three
months prior to the management guidance’s month (Abretm3); coverage firm earnings and price
adjusted forecast error volatility over the past five years (Evolatility, Fvolatility, respectively); change in
coverage firm’s earnings over year t-1 and year t (Che); management guidance error (Pae), calculated
as the difference between the guidance and the actual earnings for year t; and the average Pae over
the past two years (Appae). We also control for the time (in days) by which analyst forecasts lag the
management guidance (Gap), because a longer delay by analysts in issuing their forecasts may result
in larger deviations from management guidance. The results in Table 7, Panel A, model 3 suggest
that compared to inexperienced analysts, industry expert analysts tend to deviate significantly more
from management guidance in their earnings forecasts. On the other hand, analysts’ unrelated
industry work experience does not factor into such deviations.
We next investigate whether industry expert analysts’ intentional deviation from
management guidance helps discourage firms’ earnings management. To do so, we re-estimate the
abnormal accruals regression specified in equation (1), which we augment by differentiating among
industry expert analysts based on whether the absolute value of their deviations from management
guidance is above/below the sample median. The results presented in model 3 of Panel B show that
firms’ abnormal accruals display a more pronounced negative relation with coverage by industry
expert analysts deviating more from management guidance than with coverage by those deviating
less (p-value for coefficient difference: 0.005). Models 3 and 4 in Internet Appendix Table IA.2 use
F-score residuals and restatements as alternative measures of financial reporting quality and produce
similar results. Therefore, our findings suggest that deviating from management guidance is a
monitoring device that industry expert analysts use to limit managerial opportunism in corporate
financial reporting decisions.
4.4. Earnings conference calls
Another possible channel for analysts to express their opinions on corporate governance
issues is through their interaction with top executives during the Q&A period of earnings
conference calls. Analysts may be vocal about their disapproval of firm policies and directly question
management’s actions during these calls. Thus, we examine if this is a likely conduit. We collect
transcripts of earnings conference calls through Thomson Reuters Street Events and extract the Q&A
section from these transcripts. We then parse the textual content of these sections for the
24
governance keywords described in section 4.1. Of the 24,971 transcripts examined, we find that only
159 conference calls mention governance-related issues. Therefore, it appears that directly
questioning management’s governance practices on conference calls is not a strategy that analysts
engage in. This perhaps is not entirely surprising given that managers have control over key aspects
of their interactions with analysts during conference calls, including who they call on to participate
and how they interact with analysts (Cohen, Lou and Malloy (2014)). Therefore, disapproving
analysts may not able to get the opportunity to voice their concerns. It is, however, possible that
some analysts may discuss governance-related issues during one-on-one “call-backs” immediately
following the public earnings conference call, which would be consistent with the ‘behind-the-scenes’
view that at least some monitoring takes place in private communication with management.
5. Additional analyses and robustness tests
5.1. Income increasing vs. incoming decreasing earnings management
The discretionary accruals analysis in Section 3.1 suggests that industry expert analysts help
restrain firms’ accrual-based earnings management. In this section, we investigate whether these
analysts have stronger incentives to monitor and discourage positive (income-increasing) accruals
than negative (income-decreasing) accruals. We take two approaches to answer this question. In the
first approach, we follow Yu (2008) and create two subsamples based on whether discretionary
accruals for a firm year is positive or negative. We then reestimate the discretionary accruals
regression separately in the two subsamples. The dependent variable continues to be the absolute
value of discretionary accruals (Abs_DA) in each subsample regression. Table 8 presents the
regression results. We find that industry expert analyst coverage has a significant and negative
coefficient in both subsamples, suggesting that industry expert analysts monitor and reduce both
income-increasing and income-decreasing discretionary accruals. Comparing the two coefficients, it
appears that industry expert analyst coverage has a larger effect on reducing income-increasing
discretionary accruals than income-decreasing ones (p-value for the coefficient difference: 0.011).
***Insert Table 8 here***
As an alternative to the subsample regression analysis, we employ a multinomial logistic
regression approach in which we first sort all firm-year observations into quintiles based on the
signed discretionary accruals. We then create a dependent variable that takes the value of 1 for
observations in the bottom quintile (the most negative discretionary accruals), 2 for observations in
the middle three quintiles (moderate discretionary accruals), and 3 for observations in the top
25
quintile (the most positive discretionary accruals). We estimate the multinomial logit regression using
the full sample and therefore avoid any potential econometric concerns related to the subsample
regression approach. Another advantage of this approach is that the most positive and negative
discretionary accruals likely represent the most aggressive earnings management, and by focusing on
these cases, we alleviate concerns about measurement errors in the raw values of discretionary
accruals. Table IA.3 in the internet appendix reports the regression results, with observations in the
middle three quintiles serving as the reference category. We find that the coefficient on industry
expert analyst coverage is significantly negative in both columns, indicating that coverage by industry
expert analysts is significantly and negatively related to the probability of a firm recording either very
negative or very positive discretionary accruals. This suggests that industry expert analysts restrain
both income-increasing and income-decreasing earnings management. In addition, the magnitude of
the coefficient on industry expert analysts is significantly larger (p-value: 0.048) in column (2),
suggesting that coverage by industry expert analysts reduces the likelihood of aggressive upward
earnings management more than aggressive downward earnings management. In terms of economic
significance, all else being equal, a one standard deviation increase in industry expert analyst
coverage reduces the relative probability of a firm recording very negative vs. moderate discretionary
accruals by 6.5%, and it reduces the relative probability of a firm recording very positive vs.
moderate discretionary accruals by almost twice as much, 12.1%. Overall, results from the
multinomial logit analysis reaffirm our findings from the subsample regressions above.
5.2. Analysts’ monitoring costs
Our paper provides ample evidence that industry expertise aids in analysts’ monitoring of
firm management. One question related to the broad literature on analyst monitoring is whether
analysts bear any costs in monitoring firm management and if so, whether such costs can be severe
enough to reduce analysts’ incentive to monitor in the first place. It is entirely conceivable that
certain situations could arise that weaken analysts’ monitoring incentive. We explore two such
possibilities in this section.
First, we examine the impact of investment banking relationships on industry expert analysts’
monitoring incentives. A voluminous literature suggests that analysts affiliated with banks that have
an underwriting relationship with a firm provide overly optimistic recommendations for the firm’s
stock (e.g., Lin and McNichols (1998), Michaely and Womack (1999)). Such a relationship can also
compromise analysts’ willingness to be effective monitors if their monitoring may potentially
26
alienate firm managers and jeopardize existing investment banking relationships. This view is
confirmed by Brown et al. (2015), who survey sell-side analysts about the incentives they face. Citing
one of the respondent’s answers, “Equity analysts… are very, very reluctant—even after the Spitzer
rules—to upset the investment bankers, because the investment bankers bring in so much more
profitability…They certainly realize that the success of their company is tied to the performance of
this much higher-margin business than the business that they’re part of.”
In an attempt to shed some light on this issue, we examine if investment banking pressure
reduces the monitoring incentives of industry expert analysts. We classify industry expert analysts as
affiliated with firm management if their employers provided equity underwriting or M&A advisory
services for the firm within the past 3 years, and unaffiliated otherwise. About 15% of industry
expert analysts are classified as affiliated. We treat affiliated and unaffiliated industry expert analyst
coverage as separate variables and include them simultaneously in the reexamination of financial
reporting, CEO compensation, and CEO turnover policies. For brevity, only the results from the
abnormal accruals regression are reported in model 1 of Table 9. Regression results with the other
dependent variables used throughout the paper are in the Internet Appendix. We find that only the
coverage by unaffiliated industry expert analysts is significantly related to a firm’s abnormal accruals,
likelihood of committing financial fraud, excess CEO compensation, and CEO turnoverperformance sensitivity. In other words, the disciplining effects of industry expert analyst coverage
on these corporate policies are largely driven by unaffiliated industry expert analysts. Our findings
are consistent with the view that analysts whose employers have existing investment banking
relationships with the covered firm face higher costs and pressure in monitoring firm managers and
thus have weaker incentives to do so.29 This evidence furthers our understanding of the impact of
conflicts of interest on sell-side analysts and suggests that investment banking pressure affects more
than analyst research such as earnings forecasts and stock recommendations (e.g., Lin and
McNichols (1998), Michaely and Womack (1999)).
***Insert Table 9 here***
In our second inquiry, we investigate whether potential connections to firm management
through overlapping prior industry work experience reduce analysts’ monitoring effectiveness. To
test this conjecture, we obtain employment history of senior officers and directors in coverage firms
from Boardex of Management Diagnostic Limited (Boardex). Employment connections are defined as
29
We acknowledge that the lack of statistical significance for the coverage by affiliated industry expert analysts could be
due to the smaller number of these analysts. Therefore, readers should exercise caution when interpreting the results
from these tests.
27
those arising from overlapping employment between the industry expert analyst and a member of
management at the same firm prior to the analyst becoming a sell-side analyst. A significant
advantage of pre-existing employment connections is that they are formed ex-ante, ensuring that
their formation is not related to coverage firms and eliminating reverse causality concerns (Cohen et
al. (2010)). An industry expert analyst is defined as professionally connected to management if she
possesses at least one overlapping employment experience with a director or officer at the coverage
firm (32% of industry expert analysts meet this criteria) and zero otherwise. Model 2 of Table 9 reestimates the abnormal accruals regression specified in equation (1), except that we now partition
industry expert analysts based on the analyst’s professional connections to management. We find
that coverage by industry expert analysts is significantly and negatively related to abnormal accruals,
irrespective of professional connections between analysts and firm management. Even though the
coefficient is larger in magnitude for unconnected industry expert analysts than for connected
industry expert analysts, the coefficient difference is not statistically significant. The tenor of the
results is similar when we repeat the analysis for other outcome variables. Therefore, it appears that
professional connections do not have a material impact on industry expert analyst’s monitoring
effectiveness.
5.3. When is industry expert analyst monitoring more important?
In this section, we explore cross-sectional variations in the importance of analysts’ external
monitoring along the dimensions of firms’ information environment and corporate governance. An
important benefit of this analysis is that it can depict a more nuanced picture of the effect of analyst
industry expertise by highlighting the settings in which it is especially valuable. If related industry
experience enhances analysts’ monitoring effectiveness, then its effect should be more pronounced
when monitoring of firm management would be more difficult without a good understanding of the
firm’s industry. To test this conjecture, similar to Lee, Strong, and Zhu (2014), we construct a
composite information asymmetry rank measure that is equal to the sum of quartile ranks for firm
size, age, cash flow and stock return volatilities, accruals quality, analyst coverage, and analyst
earnings forecast dispersion in each year. We partition our sample of firm-years into two subsamples
at the median of this information asymmetry composite measure, and treat industry expert analyst
coverage at high and low information asymmetry firms separately. We reestimate the abnormal
accruals regression and present the results in model 3 of Table 9. We find that even though industry
expert analyst coverage is significantly and negatively related to abnormal accruals at both high and
28
low information asymmetry firms, the coefficient estimate is significantly larger in magnitude in high
information asymmetry firms (p-value for coefficient difference: 0.011). This evidence is consistent
with our conjecture that industry expertise is more important for analysts to perform their
monitoring function in informationally more opaque firms.
We next investigate whether the external monitoring of industry expert analysts is more
important in the absence of other strong corporate governance mechanisms limiting managers’
opportunistic behavior. Drawing upon prior literature, we use three variables to capture the strength
of a firm’s corporate governance: the percentage of independent directors (Klein (2002) and Xie,
Wallace, and DaDalt (2003)), the percentage of inside directors with outside directorships (Masulis
and Mobbs (2011)) and the percentage of ownership held by long-term, independent institutional
investors (Chen, Harford and Li (2007)). We construct a composite governance measure that is
equal to the sum of the quartile ranks of each firm-year based on the above three metrics. We
partition our sample of firm-years into two subsamples at the median of this composite governance
measure, and treat industry expert analyst coverage at strong and weak governance firms separately.
We reestimate the abnormal accruals regression and present the results in model 4 of Table 9. We
find that industry expert analyst coverage is significantly and negatively related to abnormal accruals
at both strong and weak governance firms. However, the negative relation is more pronounced
among weak governance firms (p-value for coefficient difference: 0.045). This provides support for
the conjecture that industry expert analyst monitoring is more valuable when other governance
devices are lacking.30
5.4. Duration of related industry work experience
To add more texture to our findings, we examine whether the length of an analyst’s related
industry experience matters. Our conjecture is that longer industry work experience provides
analysts with opportunities to develop more industry knowledge and expertise and thus enhances
their monitoring effectiveness. To test this conjecture, we partition industry expert analysts based on
whether the length of an analyst’s related industry experience is longer or shorter than the sample
median and create two separate industry expert analyst coverage variables, Related analyst long
30
In untabulated analyses we perform analogous regressions for other outcome variables (i.e., the ex ante and ex post
likelihood of earnings management, CEO compensation and CEO turnover-performance sensitivity) and document
similar results for the impact of professional connections, investment banking affiliation, firm information environment
as well as corporate governance on industry expert analysts’ monitoring effectiveness. These results are available from
the authors upon request.
29
experience and Related analyst short experience. We reestimate the discretionary accrual regression and
present the results in model 5 of Table 9. We find that even though both coverage variables have
significant and negative coefficients, the magnitude of the coefficient is significantly larger for Related
analyst long experience (p-value for coefficient difference: 0.065). We also perform analogous tests for
the F-score residuals, restatements, CEO compensation and CEO turnover analyses and find
qualitatively similar results. Overall, the evidence is consistent with our hypothesis and suggests that
analysts with longer related industry experience play an even more effective monitoring role.
5.5. Robustness tests
We conduct an assortment of robustness analyses. We first consider and control for
alternative sources of analyst industry experience. Analysts can develop industry expertise simply
through covering firms in the same industries over an extended period of time. To ensure that our
results are not just an artifact of the effect of analyst experience in covering a particular industry, we
explicitly control for the number of years over which an analyst has been issuing forecasts in the
coverage firm’s industry. Model 1 of Table 10 shows that analyst experience in covering a firm’s
industry has a negative but insignificant effect on earnings management. More importantly, preanalyst related industry experience continues to have a significantly negative and economically
important effect on earnings management.
***Insert Table 10 here***
In a recent paper, Dass et al. (2013) focus on directors from related industries defined as
executives or directors of firms from related upstream and downstream industries (i.e., DRIs) and
provide empirical evidence that DRIs improve the effectiveness of their board’s monitoring.
Similarly, Guan, Wong, and Zhang (2015) show that analysts simultaneously following a covered
firm’s customers issue more accurate earnings forecasts for the suppliers than analysts who do not.
Therefore, to mitigate the potential concern that pre-analyst industry expertise may simply proxy for
analyst experience from covering firms in related industries, we include indicators for analysts
following firms in related industries as well as analysts following the customers or suppliers of
covered firms. The results presented in models 2 and 3 of Table 10 suggest that the effect of preanalyst industry experience on earnings management remains significantly negative. Interestingly, we
find that analyst experience from covering firm’s related industries as well as customers or suppliers
is also associated with less aggressive earnings management.
30
Wang, Xie, and Zhu (2015) find that audit committees with more independent directors with
industry expertise, which they term as industry expert directors (IEDs), reduce firms’ earnings
management behavior and probability to engage in fraud. To ensure that industry expert analysts are
not simply more likely to follow firms with more IEDs on their audit committees, we control for the
percentage of IEDs on a firm’s audit committee (AC_IED_Pct) in the earnings management
regression. Consistent with Wang et al. (2015), we find that a greater presence of IEDs on the audit
committee is significantly related to lower earnings management by firms. More importantly, the
coefficient on industry expert analyst coverage remains negative and highly significant (see model
4).31
Finally, we address two potential selection issues arising from our use of analysts’ LinkedIn
profiles. First, LinkedIn was not founded until 2002. Therefore, for analysts providing coverage for a
firm prior to 2002 to have a LinkedIn profile, he/she must have remained active in the investment
research industry or some other profession long enough (by virtue of either aptitude or age, or both)
for LinkedIn to emerge. Second, having a LinkedIn profile is voluntary, so analysts who have a profile
may be different from analysts that do not. To shed light on these issues, model 5 of Table 10
repeats the earnings management analysis by using firm-year observations in the post-2002 period.
The results are robust. 32 We also examine whether there is any difference in monitoring ability
between analysts with and without a LinkedIn profile. Toward that end, we include a firm’s LinkedIn
analyst coverage and non-LinkedIn analyst coverage together as explanatory variables in the
abnormal accruals regression. Results in model 6 of Table 10 show that both types of coverage have
a significantly negative coefficient, but the coefficients are not statistically different from each other.
In untabulated analyses, we also compare the ability or quality of these two groups of analysts as
proxied by their general analyst experience, firm-specific experience, brokerage firm size, all-star
analyst status and portfolio size. We do not find any significant difference. Thus, selection bias
stemming from our use of LinkedIn data does not appear to be a major concern for our analysis.
6. Conclusion
Jensen and Meckling (1976) first broached the idea of analysts being capable of playing a
monitoring role to reduce agency problems. Consistent with their supposition, recent research has
31
All of our other results are robust to the inclusion of these additional controls as well. We do not report them in the
paper for brevity, but they are available upon request.
32 Results from the F-score residuals, restatements, CEO excess compensation, and CEO turnover analyses are also
robust in the post-2002 subsample.
31
produced a growing body of evidence suggesting that analyst coverage can improve corporate
governance, managerial incentives, and firm decision making. In this paper we examine whether
analysts’ industry experience obtained from pre-analyst employment affects their effectiveness as
external monitors of followed firms. Our analyses of firms’ financial disclosure, executive
compensation, and CEO turnover decisions indicate that only financial analysts with related industry
experience in their followed firms can play an effective monitoring role. More specifically, we find
that coverage by industry expert analysts is associated with reduced earnings management, lower
likelihood of committing financial misrepresentation, less CEO excess compensation, and higher
performance sensitivity of CEO turnover. In contrast, coverage by other analysts is not significantly
related to these corporate policies. We further provide evidence on the channels through which
industry expert analysts exert their monitoring influence, namely, by discussing corporate
governance issues in research reports, issuing cash flow forecasts along with earnings forecasts, and
deviating from management guidance in making earnings forecasts. Overall, our results highlight the
importance of analysts’ pre-analyst industry work experience and resultant industry expertise, which
suggests that not all analysts are equally effective in providing external monitoring of firms.
Our study also points to a few avenues for future research. One implication of our research
is that industry expert analysts are likely to have more influence over the behavior and decision
making of firm management. Parallel to the analyst monitoring literature, Degeorge et al. (2013) and
Derrien and Kecskés (2013) find that analyst coverage and preferences can have a causal effect on
firms’ investment and financial policies. Therefore, a natural question is whether corporate policies
respond more to the coverage and preferences of analysts with pertinent industry experience and to
what extent the previous findings are driven by these industry expert analysts.
Akin to sell-side analysts, auditors are third party agents that also play a monitoring role.
Industry expertise is thought to be important in this context as well (Ferguson, Francis, and Stokes
(2003), Krishnan (2003), Gul, Fung, and Jaggi (2009), McGuire, Omer, and Wang (2012), Francis,
Reichelt, and Wang (2005), and Reichelt and Wang (2010)). Similar to the extant analyst literature,
auditor industry expertise is typically measured by the amount of experience accumulated by an
auditing firm through serving a particular industry. But we know little about the background and
skill set of the individuals who actually perform the auditing and what kinds of qualification or
experience at the micro level can lead to higher audit quality. While prior studies in this literature are
limited to the auditing firm level, a new rule adopted by the Public Company Accounting Oversight
32
Board (PCAOB), which mandates the disclosure of the names of the participants conducting the
audit, would allow these questions to be addressed at the individual auditor level.
33
References
Bergstresser, Daniel, and Thomas Philippon. "CEO incentives and earnings management." Journal of
Financial Economics 80.3 (2006): 511-529.
Boni, Leslie, and Kent L. Womack. "Analysts, industries, and price momentum." Journal of Financial
and Quantitative Analysis 41.01 (2006): 85-109.
Bradley, Daniel, Sinan Gokkaya and Xi Liu, 2015. “Before an Analyst Becomes an Analyst: Does
Industry Experience Matter?” The Journal of Finance, forthcoming.
Brown, Lawrence D., Andrew C. Call, Michael B. Clement, and Nathan Y. Sharp. "Inside the
ĀBlack Boxā of SellϋSide Financial Analysts." Journal of Accounting Research 53.1 (2015): 1-47.
Bushee, Brian. J, “The influence of institutional investors on myopic R&D investment behavior.”
The Accounting Review, (1998): 305-333.
Bushee, Brian, J and Christopher Noe, “Corporate disclosure practices, institutional investors, and
stock return volatility.” Journal of Accounting Research 38 (2000): 171-202.
Cai, Jie, Jacqueline L. Garner, and Ralph A. Walkling "Electing directors." The Journal of Finance 64.5
(2009): 2389-2421.
Chen, Xia, Jarrad Harford, and Kai Li. "Monitoring: Which institutions matter?" Journal of Financial
Economics 86.2 (2007): 279-305.
Chen, Tao, Jarrad Harford, and Chen Lin. "Do analysts matter for governance? Evidence from
natural experiments." Journal of Financial Economics 115.2 (2015): 383-410.
Chun, Hyunbae, Jung-Wook Kim, Randall Morck, and Bernard Yeung. "Creative destruction and
firm-specific performance heterogeneity." Journal of Financial Economics 89.1 (2008): 109-135.
Cohen, Lauren, Andrea Frazzini, and Christopher Malloy. "Sell-side school ties." The Journal of
Finance, 65.4 (2010): 1409-1437.
Cohen, Lauren, Dong Lou, and Christopher Malloy. “Playing favorites: How firms prevent the
revelation of bad news.” (2014), Working paper.
Hribar, Paul, and Daniel W. Collins. "Errors in estimating accruals: Implications for empirical
research." Journal of Accounting Research 40. 1 (2002): 105-134.
Coughlan, Anne T., and Ronald M. Schmidt. "Executive compensation, management turnover, and
firm performance: An empirical investigation." Journal of Accounting and Economics 7.1 (1985): 43-66.
Custódio, Cláudia, and Daniel Metzger. "How do CEOs matter? The effect of industry expertise on
acquisition returns." Review of Financial Studies 26.8 (2013): 2008-2047.
34
Dass, Nishant, Omesh Kini, Vikram Nanda, Bunyamin Onal, and Jun Wang. "Board Expertise: Do
Directors from Related Industries Help Bridge the Information Gap?" Review of Financial Studies 27.5
(2013): 1533-1592.
Dechow, Patricia M., Weili Ge, Chad R. Larson, and Richard G. Sloan. "Predicting Material
Accounting Misstatements." Contemporary Accounting Research 28.1 (2011): 17-82.
Dechow, Patricia M., Richard G. Sloan, and Amy P. Sweeney. "Causes and consequences of earnings
manipulation: An analysis of firms subject to enforcement actions by the sec." Contemporary
Accounting Research 13.1 (1996): 1-36.
Degeorge, Francois, François Derrien, Ambrus Kecskes, and Sebastien Michenaud. "Do Analysts'
Preferences Affect Corporate Policies?" Available at SSRN 2253619 (2013).
Derrien, François, and Ambrus Kecskés. "The real effects of financial shocks: Evidence from
exogenous changes in analyst coverage." The Journal of Finance 68.4 (2013): 1407-1440.
Dichev, Ilia D., John R. Graham, Campbell R. Harvey, and Shiva Rajgopal. "Earnings quality:
Evidence from the field." Journal of Accounting and Economics 56.2 (2013): 1-33.
Dopuch, Nicholas, Raj Mashruwala, Chandra Seethanraju and Tzachi Zach. "The impact of a
heterogeneous accrual-generating process on empirical accrual models." Journal of Accounting, Auditing
& Finance (2011).
DuCharme, Larry L., Paul H. Malatesta, and Stephan E. Sefcik. "Earnings management, stock issues,
and shareholder lawsuits." Journal of Financial Economics 71.1 (2004): 27-49.
Dyck, Alexander, Adair Morse, and Luigi Zingales. "Who blows the whistle on corporate
fraud?" The Journal of Finance 65.6 (2010): 2213-2253.
Feng, Mei, and Sarah McVay. "Analysts' incentives to overweight management guidance when
revising their short-term earnings forecasts." The Accounting Review 85.5 (2010): 1617-1646.
Ferguson, Andrew, Jere R. Francis, and Donald J. Stokes. "The effects of firm-wide and office-level
industry expertise on audit pricing." The Accounting Review 78.2 (2003): 429-448.
Fich, Eliezer M., and Anil Shivdasani. "Are busy boards effective monitors?" The Journal of
Finance 61.2 (2006): 689-724.
Francis, Jere R., Kenneth Reichelt, and Dechun Wang. "The pricing of national and city-specific
reputations for industry expertise in the US audit market." The Accounting Review 80.1 (2005): 113-136.
Gong, Guojin, Henock Louis, and Amy X. Sun. "Earnings management, lawsuits, and stock-forstock acquirers’ market performance." Journal of Accounting and Economics 46.1 (2008): 62-77.
35
Graham, John R., Campbell R. Harvey, and Shiva Rajgopal. "The economic implications of
corporate financial reporting." Journal of Accounting and Economics 40.1 (2005): 3-73.
Guan, Yuyan, MH Franco Wong, and Yue Zhang. "Analyst following along the supply chain."
Review of Accounting Studies 20.1 (2015): 210-241.
Gul, Ferdinand A., Simon Yu Kit Fung, and Bikki Jaggi. "Earnings quality: Some evidence on the
role of auditor tenure and auditors’ industry expertise." Journal of Accounting and Economics 47.3 (2009):
265-287.
Hartzell, Jay C., Eli Ofek, and David Yermack. "What's in it for me? CEOs whose firms are
acquired." Review of Financial Studies 17.1 (2004): 37-61.
He, Jie Jack, and Xuan Tian. "The dark side of analyst coverage: The case of innovation." Journal of
Financial Economics 109.3 (2013): 856-878.
Healy, Paul M., Amy P. Hutton, and Krishna G. Palepu. "Stock performance and intermediation
changes surrounding sustained increases in disclosure." Contemporary Accounting Research 16.3 (1999):
485-520.
Hennes, Karen M., Andrew J. Leone, and Brian P. Miller. "The importance of distinguishing errors
from irregularities in restatement research: The case of restatements and CEO/CFO turnover." The
Accounting Review 83.6 (2008): 1487-1519.
Hong, Harrison, and Marcin Kacperczyk. "Competition and bias." The Quarterly Journal of
Economics 125.4 (2010): 1683-1725.
Huson, Mark R., Paul H. Malatesta, and Robert Parrino. "Managerial succession and firm
performance." Journal of Financial Economics 74.2 (2004): 237-275.
Irani, Rustom M., and David Oesch. "Monitoring and corporate disclosure: Evidence from a natural
experiment." Journal of Financial Economics 109.2 (2013): 398-418.
Jensen, Michael C. and William H. Meckling. “Theory of the Firm: Managerial Behavior, Agency
Costs and Ownership Structure”. Journal of Financial Economics 3.4 (1976): 305-360.
Jenter, Dirk, and Fadi Kanaan. "CEO turnover and relative performance evaluation." The Journal of
Finance 70.5 (2015): 2155-2184.
Jones, Jennifer J. "Earnings management during import relief investigations." Journal of Accounting
Research 29.2 (1991): 193-228.
Kadan, Ohad, Leonardo Madureira, Rong Wang, and Tzachi Zach. "Analysts' industry
expertise." Journal of Accounting and Economics 54.2 (2012): 95-120.
36
Karpoff, Jonathan M., D. Scott Lee, and Gerald S. Martin. "The consequences to managers for
financial misrepresentation." Journal of Financial Economics 88.2 (2008): 193-215.
Kelly, Bryan, and Alexander Ljungqvist. "Testing asymmetric-information asset pricing
models." Review of Financial Studies 25.5 (2012): 1366-1413.
Klein, April. "Audit committee, boar*d of director characteristics, and earnings management."
Journal of Accounting and Economics 33.3 (2002): 375-400.
Kothari, Sagar P., Andrew J. Leone, and Charles E. Wasley. "Performance matched discretionary
accrual measures." Journal of Accounting and Economics 39.1 (2005): 163-197.
Krishnan, Gopal V. "Does Big 6 auditor industry expertise constrain earnings management?."
Accounting Horizons 17.1 (2003): 1-16
Lee, Edward, Norman Strong, and Zhenmei Judy Zhu. "Did Regulation Fair Disclosure, SOX, and
Other Analyst Regulations Reduce Security Mispricing?" Journal of Accounting Research 52.3 (2014):
733-774.
Lin, Hsiou-wei, and Maureen F. McNichols. "Underwriting relationships, analysts' earnings forecasts
and investment recommendations." Journal of Accounting and Economics 25.1 (1998): 101-127.
Liu, Xiaoding, and Jay R. Ritter. "Local underwriter oligopolies and IPO underpricing." Journal of
Financial Economics 102.3 (2011): 579-601.
Loughran, T., McDonald, B., 2011. When is a Liability not a Liability? Textual Analysis, Dictionaries,
and 10-Ks. The Journal of Finance 66.1: 35-65.
Louis, Henock, Amy X. Sun, and Oktay Urcan. "Do analysts sacrifice forecast accuracy for
informativeness?." Management Science 59.7 (2013): 1688-1708.
Masulis, Ronald W., Christian Ruzzier, Sheng Xiao, and Shan Zhao. 2014. “Do Independent Expert
Directors Matter?” Working paper.
Masulis, Ronald W., and Shawn Mobbs. "Are all inside directors the same? Evidence from the
external directorship market." The Journal of Finance 66.3 (2011): 823-872.
McGuire, Sean T., Thomas C. Omer, and Dechun Wang. "Tax avoidance: Does tax-specific industry
expertise make a difference?." The Accounting Review 87.3 (2012): 975-1003.
McInnis, John, and Daniel W. Collins. "The effect of cash flow forecasts on accrual quality and
benchmark beating." Journal of Accounting and Economics 51.3 (2011): 219-239.
Michaely, Roni, and Kent L. Womack. "Conflict of interest and the credibility of underwriter analyst
recommendations." Review of Financial Studies 12.4 (1999): 653-686.
37
Owens, Edward L., Joanna Shuang Wu, and Jerold L. Zimmerman. "Business model shocks and
abnormal accrual models." Available at SSRN 2365304 (2015).
Parrino, Robert. "CEO turnover and outside succession a cross-sectional analysis." Journal of Financial
Economics 46.2 (1997): 165-197.
Perry, Susan E., and Thomas H. Williams. "Earnings management preceding management buyout
offers." Journal of Accounting and Economics 18.2 (1994): 157-179.
Petersen, Mitchell A. "Estimating standard errors in finance panel data sets: Comparing
Approaches." Review of Financial Studies 22.1 (2009): 435-480.
Reichelt, Kenneth J., and Dechun Wang. "National and office ϋ specific measures of auditor
industry expertise and effects on audit quality." Journal of Accounting Research 48.3 (2010): 647-686.
Wang, Cong, Fei Xie, and Kuo Zhang. 2015. Expert advice: Industry expertise of financial advisors
in mergers and acquisitions. Working paper.
Wang, Cong, Fei Xie, and Min Zhu. "Industry expertise of independent directors and board
monitoring." Journal of Financial and Quantitative Analysis 50.05 (2015): 929-962.
Warfield, Terry D., John J. Wild, and Kenneth L. Wild. "Managerial ownership, accounting choices,
and informativeness of earnings." Journal of Accounting and Economics 20.1 (1995): 61-91.
Weisbach, Michael S. "Outside directors and CEO turnover." Journal of Financial Economics 20 (1988):
431-460.
White, Halbert. "A heteroskedasticity-consistent covariance matrix estimator and a direct test for
heteroskedasticity." Econometrica (1980): 817-838.
Yu, Fang Frank. "Analyst coverage and earnings management." Journal of Financial Economics 88.2
(2008): 245-271.
Xie, Biao, Wallace N. Davidson, and Peter J. DaDalt. "Earnings management and corporate
governance: the role of the board and the audit committee." Journal of Corporate Finance 9.3 (2003):
295-316.
38
Appendix A: Data screening and collection
Firms
Analysts
All analysts annual EPS forecasts between 1988 to 2011 from
I/B/E/S.
17,775
16,766
Merge with CRSP/COMPUSTAT for stock price data and firm
characteristics.
9,684
13,255
Keep the last annual earnings forecast with a horizon between 1
month and 12 month.
6,730
12,584
Merge sample with I/B/E/S recommendation file to get analyst
last name, first initial and brokerage firm estimid, which is used
to identify the brokerage firm names. Remove analysts without
first initial, last name or brokerage estimid, or analyst teams
(those analyst names recorded as “research department” or
contain two analyst last names). This is the ‘clean’ I/B/E/S
sample.
6,713
9,275
Manually search Zoominfo.com for analysts’ names by matching
their brokerage firm, last name and first initial.
6,391
4,845
Search for analysts’ employment history on LinkedIn.com, get
industry information of pre-analyst public firms from CRSP
and pre-analyst private firms from LinkedIn.com, Hoover’s,
Thomson Financial’s SDC, the Securities and Exchange
Commission’s (SEC) website, business news websites (e.g.,
Bloomberg and BusinessWeek) or company official websites.
5,542
2,450
Compute number of analysts covering subject firm for each
firm year pair. Remove observations with missing value on
control variables.
4,267
2,433
39
Appendix B. Variable definitions
Variable
Total analysts
Definition
Number of analysts covering subject firm j
Related analysts
Number of analysts with related industry experience covering
subject firm j
Unrelated analysts
Number of analysts with unrelated pre-analyst industry work
experience covering subject firm j
Inexperienced analysts
Number of analysts without prior industry work experience
covering subject firm j
Discretionary accruals (DA)
Discretionary accruals are estimated from Owen, Wu, and
Zimmerman’s (2015) accrual model
Market cap
Market capitalization of the covered firm (in $billion) by the end
of the fiscal year
Market-to-book
Current market value of equity divided by the book value of
equity in the fiscal year prior to the earnings forecast
ROA
Return on assets, calculated as earnings divided by total assets
(#18/#6)
Total assets growth
Growth rate of assets is calculated by the change of assets scaled
by lagged total assets
Cash flow volatility
The standard deviations of cash flows of a firm in the entire
sample period, scaled by lagged total assets
External financing
The sum of net cash received from equity and debt issuance
scaled by lagged total assets
Institutional ownership (dedicated)
The percentage of institutional holding by dedicated
institutional investors. Dedicated institutional investors are
classified based on Bushee (1998) and provided on his website.
Institutional ownership (non-dedicated)
The percentage of institutional holding by non-dedicated
institutional investors calculated as total institutional investor
holdings less dedicated institutional investor holdings.
Experience as analyst
Average number of years that analyst’s i appeared in I/B/E/S
Experience with firm
Average number of years since analyst’s i first earnings forecast
for the covered firm j
Analysts from top brokers
Number of analysts working at a top decile brokerage house
40
that cover subject firm j
All-star analysts
Number of All-star analysts covering subject firm j
Analyst portfolio size
Average number of firms followed by analysts that cover the
subject firm
High restate exposure
An indicator for industry groups with above sample median
restatement frequency, which is computed as the percentage of
firms in each industry that were associated with an irregularity
restatement over a 10-year rolling window
Tobin’s Q
Market-to-book ratio of assets, where the market value of assets
is computed as the book value of assets plus the market value of
equity, minus the book value of equity ((#6+#199×#25-#60#74)/#6)
Leverage
Long-term plus short-term debt, scaled by total assets
((#9+#34)/#6)
R&D/sale
R&D expenditures scaled by total sales (#46/#12)
Capx/sale
Capital expenditures scaled by total sales (#128/#12)
Xad/sale
Advertising expenditures scaled by total sales (#45/#12)
Abnormal ROA
ROA-adjusted by industry median ROA
Abnormal stock rets
A firm’s annual stock returns minus its corresponding GICS
industry annual stock returns
CEO tenure
Number of years since the CEO has been employed with the
company
CEO total/excess compensation
CEO total compensation includes salary, bonus, stock and
option grants, and long-term incentive plans from Standard &
Poor’s ExecuComp database. To compute CEO excess
compensation, the logarithmic transformation of total CEO
compensation is regressed against a firm’s market capitalization,
buy-and-hold abnormal returns over CRSP value-weighted
returns, stock return volatility, as well as industry (based on twodigit SIC codes) and year fixed effects. CEO excess
compensation is defined as the residuals from this estimation.
Firm age
Number of years since the firm’s first appearance in CRSP
Past return
CRSP VW-index adjusted buy-and hold abnormal returns
(BHARs) over the prior 250 trading days
41
Restatement
Indicator variable=1 if firm has at least one restatement in the
fiscal year, 0 otherwise
Abnormal ROA
Firm performance by the industry-adjusted return on assets
(ROA)
Sales volatility
Standard deviation of sales of a firm in the entire sample period,
scaled by lagged assets
CEO/Chairman
Indicator for whether CEO is also chairman of the board
Abs (Devaf)
Absolute value of forecasting analyst i’s deviation from firm j’s
preceding preannounced earnings for the forecasted year t
scaled by stock price of firm j on the day before the
preannouncement
Abret
Market-adjusted returns starting two days after the
preannouncement date and ending one day before the analyst
forecast date
Loss
A binary variable taking the value one if previous earnings are
negative and zero otherwise
Evoltil
The standard deviation of earnings (as a percentage of price at
the beginning of the fiscal year) over past 5 years
Fvoltil
Price adjusted forecast error volatility over past 5 years
Che
Change in earnings from year t-1 to year t
Pae
Management guidance minus reported earnings, deflated by
price on the day prior to the earnings preannouncement
Appae
Average percentage management earnings forecast error over
the past two years.
Abretm3
Market-adjusted returns over the three months ending the
month prior to the preannouncement month
Gap
Number of days between management forecasts and
analyst forecasts.
Related analyst with/without prof connections
Number of industry expert analysts with/without professional
connections to the covering subject firm j’s management
Related analyst affiliated/unaffiliated
Number of industry expert analysts working for brokers that
have/haven’t provided equity underwriting or M&A advisory
services for the firm within the past 3 years to the covering
subject firm j’s
42
Related analyst high/low info asymmetry
Number of analysts covering subject firm j with above/below
sample median composite information asymmetry measure. The
composite information asymmetry measure is constructed as the
sum of quartile ranks based on a firm’s size, age, cash flow and
stock return volatilities, accruals quality, analyst coverage, and
earnings forecast dispersion.
Related analyst weak/strong governance
Number of analysts covering subject firm j with above/below
sample median composite corporate governance measure. The
composite governance measure is constructed as the sum of
quartile ranks based on the percentage of independent directors,
the percentage of inside directors with outside directorships,
and the percentage of equity ownership held by long-term
independent institutional investors
Industry covering experience
Average number of years that covering analysts has issued
forecasts in the coverage firm’s 2-digit industry
Covering related industry
Indicator variable=1 if at least one of analysts also covers the
coverage firm’s related industry, 0 otherwise
Covering customer/supplier
Indicator variable=1 if at least one of analysts also covers the
coverage firm’s supplier or customer, 0 otherwise
AC_IED_Pct
The number of independent directors on the audit committee
with industry expertise divided by the total number of
independent directors on the audit committee
Total LinkedIn analysts
Number of LinkedIn analysts covering subject firm j
Total non-LinkedIn analysts
Number of non-LinkedIn analysts covering subject firm j
43
Appendix C. List of keywords used for textual analysis of analyst research reports
We read 1000 randomly selected analyst reports to look for every bigram word combination that analysts use to describe
governance and compiled a robust and extensive keyword listed below:
Financial restatement:
restate results
restating results
restatement of financials
restatement risk
material weaknesses
risk of restatement
accounting restatement
financial misconduct
financial fraud
materially false
materially misleading
restatement announcement
restatement numbers
accounting shenanigans
restatement of financial results
restatement filings
restatement activities
Corporate governance:
governance rules
governance committee
governance risk
governance culture
governance issues
governance overhang
governance record
governance practices
governance policies
governance snapshot
governance concerns
governance standards
governance questions
governance procedures
governance provisions
governance problem
weak governance
strong governance
good governance
bad governance
governance features
governance principles
governance concern
governance changes
governance shortcomings
governance discount
governance quality
governance measures
governance rating
governance fear
CEO compensation:
CEO salary
CEO option
CEO stock
CEO cash
CEO bonus
CEO stock option
CEO restricted stock
CEO other compensation
CEO annual compensation
CEO total compensation
CEO turnover:
CEO change
CEO is fired
CEO is laid off
CEO is forced out
CEO job turnover
CEO has resigned
CEO resigned
CEO left
CEO has left
Shareholder proposal:
activist proposal
shareholder proposal
Shareholder activism:
activist
activist hedge fund
proxy fight
44
Table 1. Descriptive statistics
This table reports summary statistics of our sample. Panel A provides the top five 4-digit GICS industries from which
analysts obtain pre-analyst work experience as well as the top 5 industries with the most industry expert analysts. Panel B
is based on the earnings management sample, which consists of 24,918 firm-year observations from 1988 to 2011. The
number of observations used in different analyses varies. See Appendix B for a detailed description of these variables.
Panel A: Top five four-digit GICS industries (pre-analyst work experience and industry expert analysts)
Industry
Software & Services
Diversified Financials
Commercial Services &
Supplies
Banks
Media
Pre-analyst work experience
12.74%
11.95
9.83
Industry
Software & Services
Diversified Financials
Health Care Equipment
& Services
Retailing
Technology Hardware &
Equipment
8.18
7.63
Industry expert analysts
17.90%
8.08
6.74
6.72
6.31
Panel B: Variable descriptive statistics
Dependent Variables
Absolute value of discretionary accruals (Abs_DA)
F-score (residuals)
Restated due to irregularity
CEO compensation ($ mil)
Forced CEO turnover
N
Mean
1st
quartile
Median
3rd
quartile
Std
Dev
24,918
23,233
19,772
9,428
6,623
0.047
-0.007
0.017
3.434
0.027
0.013
-0.541
0.000
1.211
0.000
0.029
-0.189
0.000
2.373
0.000
0.059
0.315
0.000
4.412
0.000
0.058
0.760
0.129
3.782
0.161
Firm Characteristics
Market cap ($ bil)
Market to book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
24,918
24,918
24,918
24,918
24,918
24,918
24,918
24,918
4.618
3.175
0.034
0.229
0.222
0.026
0.083
0.553
0.355
1.492
0.015
0.009
0.062
-0.047
0.000
0.363
0.944
2.310
0.052
0.097
0.101
0.000
0.043
0.553
2.990
3.753
0.095
0.244
0.213
0.050
0.125
0.754
12.927
4.299
0.137
0.637
0.395
0.158
0.120
0.252
Analyst characteristics
Total analysts
Related analysts
Unrelated analysts
Inexperienced analysts
Experience as analyst
Experience with firm
Analyst portfolio size
Analysts from top brokers
All-star analysts
24,918
24,918
24,918
24,918
24,918
24,918
24,918
24,918
24,918
3.972
0.896
1.698
1.377
6.312
2.454
12.561
0.565
0.131
1.000
0.000
0.000
0.000
4.000
1.000
9.333
0.333
0.000
3.000
0.000
1.000
1.000
6.000
2.000
12.000
0.600
0.000
5.000
1.000
2.000
2.000
8.250
3.500
14.545
0.900
0.200
3.524
1.499
1.987
1.387
3.362
2.137
5.724
0.350
0.244
45
Table 2. Industry expert analyst coverage and earnings management
Panel A presents baseline regressions of earnings management. The sample consists of 24,918 firm-year observations
from 1988 to 2011. The dependent variable is the absolute value of discretionary accruals (Abs_DA) adjusted for
business model shocks using Owens, Wu and Zimmerman (2015) model. See Appendix B for a detailed description of
variables. Firm and year-fixed effects are included. Coefficient values are reported as percentages and robust t-statistics
based on standard errors adjusted for heteroskedasticity (White (1980)) and firm clustering (Petersen (2009)) are in
parentheses. Panel B reports the effect of analyst departures from broker mergers and closures on changes in coverage
firms’ earnings management. The first column provides the cross-sectional means of the mean treatment difference,
mean control difference and differences-in-difference (DID) for the full sample of treatment firms affected by the loss
of analyst coverage (Lost Analyst). Lost Related Experienced Analyst (Lost Other Analyst) provides results for the
sample of coverage firms with terminations of coverage by related industry experienced analysts (other analysts). The last
column reports the differences between these two groups. Control firms are required to be in the same size and bookto-market quintiles in the month of June with the treatment firms preceding the broker events and are covered by at
least one analyst. We retain the control firm with the smallest number of difference in the number of analysts relative to
that of treatment firms. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%, respectively.
Panel A. Baseline OLS regressions
Total analysts
Model 1
-0.078***
(-3.761)
Model 2
Related analysts
Model 3
-0.210***
(-5.074)
-0.025
(-0.719)
-0.037
(-0.875)
Unrelated analysts
Inexperienced analysts
Total analysts (Residual)
-0.125***
(-5.760)
Related analysts (Residual)
Unrelated analysts (Residual)
Inexperienced analysts (Residual)
Market cap
Market-to-book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
Experience as analyst
Experience with firm
Model 4
-0.006
(-1.113)
0.067***
(5.505)
0.026
(0.044)
-0.015
(-0.123)
-0.004
(-0.019)
0.845*
(1.746)
-2.522***
(-4.228)
-2.576***
(-7.879)
0.022
(1.138)
-0.083***
-0.005
(-0.998)
0.066***
(5.453)
0.053
(0.088)
-0.017
(-0.139)
-0.014
(-0.062)
0.844*
(1.746)
-2.489***
(-4.167)
-2.507***
(-7.579)
0.017
(0.895)
-0.078***
46
-0.016***
(-2.987)
0.063***
(5.151)
0.020
(0.034)
-0.037
(-0.304)
-0.044
(-0.194)
0.944*
(1.954)
-2.713***
(-4.512)
-2.958***
(-9.221)
0.020
(1.072)
-0.094***
-0.183***
(-4.258)
-0.055
(-1.505)
-0.038
(-0.941)
-0.015***
(-2.742)
0.065***
(5.347)
0.078
(0.130)
-0.043
(-0.354)
0.008
(0.034)
0.946*
(1.964)
-2.698***
(-4.482)
-2.953***
(-9.213)
0.015
(0.768)
-0.087***
Analysts from top brokers
All-star analysts
Analyst portfolio size
Adj-R2
N
Fixed Effects
(-3.078)
-0.356**
(-2.395)
-0.056
(1.102)
0.010
(1.102)
28.36%
24,918
Y
(-2.884)
-0.355**
(-2.388)
-0.065
(1.176)
0.011
(1.176)
28.40%
24,918
Y
(-3.466)
-0.365**
(-2.452)
-0.050
(1.298)
0.012
(1.298)
28.49%
24,918
Y
(-3.197)
-0.355**
(-2.389)
-0.027
(1.347)
0.012
(1.347)
28.42%
24,918
Y
Panel B: Exogenous shocks from brokerage mergers/closures
Mean ΔAbs_DA
(treatment firms)
Mean ΔAbs_DA
(control firms)
Mean DID
(treatment – control)
(1)
Lost Analyst
0.269*
(1.880)
-0.126
(-1.050)
0.395**
(2.150)
(2)
Lost Related Analyst
0.935***
(2.890)
-0.274
(-0.940)
1.208***
(3.080)
47
(3)
Lost Other Analyst
0.133
(0.840)
-0.095
(-0.720)
0.229
(1.110)
(4)
Difference: (2)-(3)
0.802**
(2.220)
-0.178
(-0.560)
0.980**
(2.210)
Table 3. Industry expert analyst coverage and F-score residuals
Panel A presents OLS regression results for F-score residuals from 1988 to 2011 and the change in F-score residuals
around the loss of analyst coverage stemming from brokerage mergers/closures in panel B. See Appendix B for a
detailed description of variables. Firm and year-fixed effects are included. Coefficient values are reported as percentages
and robust t-statistics based on standard errors adjusted for heteroskedasticity (White (1980)) and firm clustering
(Petersen (2009)) are in parentheses. The first column provides the cross-sectional means of the mean treatment
difference, mean control difference and differences-in-difference (DID) for the full sample of treatment firms affected
by the loss of analyst coverage (Lost Analyst). Lost Related Analyst (Lost Other Analyst) provides results for sample of
coverage firms with terminations of coverage by related industry experienced analysts (other analysts). The last column
reports the differences between these two groups. Control firms are required to be in the same size and book-to-market
quintiles in the month of June with the treatment firms preceding the broker events and are covered by at least one
analyst. We retain the control firm with the smallest number of difference in the number of analysts relative to that of
treatment firms. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%, respectively.
Panel. A. Baseline OLS regressions
Model 1
-0.513**
(-2.137)
Total Analysts (Residual)
Related analysts (Residual)
Unrelated analysts (Residual)
Inexperienced analysts (Residual)
Market cap
0.139*
(1.651)
0.079
(0.703)
-4.907
(-0.463)
36.190***
(5.090)
-1.670
(-0.631)
0.784
(0.051)
0.985
(0.180)
4.820
(1.383)
0.409**
(2.078)
-1.309***
(-4.568)
0.259
(0.177)
1.614
(0.806)
0.225**
(2.219)
61.85%
23,233
Y
Market-to-book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
Experience as analyst
Experience with firm
Analysts from top brokers
All-star analysts
Analyst portfolio size
Adj-R2
N
Fixed Effects
48
Model 2
-1.800***
(-3.659)
-0.601
(-1.548)
-0.531
(-1.155)
0.128
(1.512)
0.055
(0.491)
-4.635
(-0.438)
36.048***
(5.076)
-2.240
(-0.846)
1.169
(0.076)
0.695
(0.127)
4.846
(1.391)
0.375*
(1.850)
-1.334***
(-4.651)
0.242
(0.164)
1.503
(0.750)
0.233**
(2.304)
61.89%
23,233
Y
Panel. B. Exogenous shocks from brokerage mergers/closures
(1)
(2)
Lost Analyst
Lost Related Analyst
Mean ΔF-score residual
0.030*
0.097**
(treatment firms)
(1.700)
(2.540)
Mean ΔF-score residual
-0.013
-0.058*
(control firms)
(-0.950)
(-1.670)
Mean DID
0.043*
0.155***
(treatment – control)
(1.890)
(2.750)
49
(3)
Lost Other Analyst
0.019
(0.970)
-0.006
(-0.390)
0.025
(1.000)
(4)
Difference (2)-(3)
0.078*
(1.820)
-0.052
(-1.370)
0.130**
(2.110)
Table 4. Industry expert analyst coverage and restatements
This table presents the results from logit regressions of the probability of firms committing intentional financial
misreporting that leads to earnings restatements from 1997 to 2011. The dependent variable is equal to one for firmyears whose earnings are restated due to accounting irregularities and zero otherwise. See Appendix B for a detailed
description of variables. Coefficient values are reported as percentages and robust t-statistics based on standard errors
adjusted for heteroskedasticity (White (1980)) and firm clustering (Petersen (2009)) are in parentheses. *, **, and ***
indicate statistical significance at the 10%, 5%, and 1%, respectively.
Model 1
-6.040*
(-1.787)
Total analysts (Residual)
Related analysts (Residual)
Unrelated analysts (Residual)
Inexperienced analysts (Residual)
Market cap
0.758
(1.289)
-1.910
(-1.130)
-118.620**
(-2.455)
17.420***
(2.801)
-106.190*
(-1.900)
68.330*
(1.666)
27.500
(0.319)
243.430***
(4.709)
-1.490
(-0.487)
0.899
(0.195)
1.810
(0.077)
28.230
(0.930)
1.440
(0.906)
33.020**
(2.032)
2.04%
19,772
Y
Market-to-book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
Experience as analyst
Experience with firm
Analysts from top brokers
All-star analysts
Analyst portfolio size
High restate exposure
Pseudo-R2
N
Fixed Effects
50
Model 2
-25.310***
(-3.335)
3.400
(0.615)
-9.030
(-1.265)
0.540
(0.912)
-1.940
(-1.128)
-119.880**
(-2.507)
17.430***
(2.762)
-107.370*
(-1.888)
70.120*
(1.711)
25.790
(0.300)
240.470***
(4.642)
-1.360
(-0.453)
0.900
(0.195)
2.530
(0.109)
25.880
(0.843)
1.610
(1.045)
26.730*
(1.647)
2.13%
19,772
Y
Table 5. Industry expert analyst coverage and excess CEO compensation
This table presents regression results for CEO excess compensation (Panel A) and the change in CEO excess
compensation around the loss of analyst coverage due to shocks stemming from brokerage mergers/closures (Panel B).
The dependent variable is the logarithmic transformation of excess CEO compensation. See Appendix B for a detailed
description of variables. Firm and year-fixed effects are included. Coefficient values are reported as percentages and
robust t-statistics based on standard errors adjusted for heteroskedasticity (White (1980)) and firm clustering (Petersen
(2009)) are in parentheses*, **, and *** indicate statistical significance at the 10%, 5%, and 1%, respectively.
Panel A. Baseline OLS regressions
Model 1
Model 2
Total analysts (Residual)
-0.968**
(-2.033)
Related analysts (Residual)
-3.190***
(-3.278)
Unrelated analysts (Residual)
-0.240
(-0.329)
Inexperienced analysts (Residual)
-0.439
(-0.431)
Market cap
14.672***
14.598***
(6.700)
(6.714)
Tobin’s Q
1.718
1.617
(1.392)
(1.334)
Leverage
-0.851
-0.834
(-0.061)
(-0.059)
R&D/sale
-1.597***
-1.595**
(-2.632)
(-2.593)
Capx/sale
15.050***
15.133***
(3.114)
(3.092)
Xad/sale
84.984
87.571
(1.165)
(1.204)
Abnormal ROA
78.000***
78.897***
(5.138)
(5.169)
Abnormal stock rets
-1.553
-1.588
(-0.841)
(-0.861)
CEO tenure
-3.369***
-3.387***
(-7.939)
(-8.004)
Firm age
4.455***
4.483***
(9.584)
(9.785)
Institutional ownership (dedicated)
5.807
5.232
(0.428)
(0.385)
Institutional ownership (non-dedicated)
4.991
5.135
(0.573)
(0.591)
Experience as analyst
0.189
0.116
(0.480)
(0.289)
Experience with firm
0.331
0.355
(0.493)
(0.526)
Analysts from top brokers
-1.877
-2.096
(-0.567)
(-0.633)
All-star analysts
-1.038
-0.892
(-0.211)
(-0.181)
Analyst portfolio size
-0.040
-0.031
(-0.165)
(-0.130)
Adj-R2
61.52%
61.57%
N
9,428
9,428
Fixed Effects
Y
Y
51
Panel B. Exogenous shocks from brokerage mergers/closures
(1)
(2)
Lost Analyst
Lost Related Analyst
Mean ΔCEO excess comp
0.056*
0.178**
(treatment firms)
(1.720)
(2.320)
Mean ΔCEO excess comp
-0.021
-0.050
(control firms)
(-0.740)
(-0.750)
Mean DID
0.077*
0.228**
(treatment – control)
(1.960)
(2.650)
52
(3)
Lost Other Analyst
0.034
(0.970)
-0.016
(-0.510)
0.051
(1.160)
(4)
Difference (2)-(3)
0.143*
(1.690)
-0.034
(-0.450)
0.177*
(1.840)
Table 6. Industry expert analyst coverage and CEO turnover-performance sensitivity
This table presents logit regression results for forced CEO turnovers. The dependent variable is an indicator that equals
1 if there is a forced CEO turnover in year t and 0 otherwise. See Appendix B for a detailed description of variables.
Coefficient values are reported as percentages and robust t-statistics based on standard errors adjusted for
heteroskedasticity (White (1980)) and firm clustering (Petersen (2009)) are in parentheses. *, **, and *** indicate
statistical significance at the 10%, 5%, and 1%, respectively.
Model 1
Model 2
Model 3
Abnormal ROA * Total analysts (Residual)
-12.210
(-0.625)
Abnormal ROA * Related analysts (Residual)
-165.930***
-165.390***
(-2.997)
(-2.999)
Abnormal ROA * Unrelated analysts (Residual)
34.300
31.350
(1.096)
(0.997)
Abnormal ROA * Inexperienced analysts (Residual)
28.700
26.920
(0.511)
(0.475)
Abnormal ROA
-155.960**
-219.390***
-223.910***
(-2.174)
(-3.148)
(-3.161)
Total analysts (Residual)
0.774
(0.269)
Related analysts (Residual)
7.610
7.720
(1.037)
(1.047)
Unrelated analysts (Residual)
-3.420
-1.950
(-0.597)
(-0.336)
Inexperienced analysts (Residual)
1.370
0.082
(0.225)
(0.013)
Market cap
0.008
-0.081
-0.187
(0.014)
(-0.145)
(-0.321)
Firm age
-0.024
-0.048
-0.209
(-0.049)
(-0.097)
(-0.398)
Leverage
13.070
13.050
13.790
(0.199)
(0.200)
(0.207)
Tobin’s Q
-5.910
-4.700
-3.740
(-0.682)
(-0.550)
(-0.438)
Sales Volatility
34.500***
34.790***
34.130***
(4.040)
(4.017)
(3.826)
CEO/Chairman
-11.650
-12.040
-13.210
(-0.718)
(-0.737)
(-0.807)
Institutional ownership (dedicated)
-25.910
-37.310
-49.660
(-0.234)
(-0.334)
(-0.443)
Institutional ownership (non-dedicated)
-48.790
-48.190
-52.070
(-0.953)
(-0.936)
(-0.995)
Abnormal stock rets
-122.290***
-121.290***
-123.290***
(-5.333)
(-5.283)
(-5.260)
Experience as analyst
2.630
(0.807)
Experience with firm
5.070
(1.155)
Analysts from top brokers
-1.640
(-0.055)
All-star analysts
-8.080
(-0.200)
Analyst portfolio size
1.380
(0.738)
Pseudo-R2
1.98%
2.09%
2.18%
N
6,623
6,623
6,623
Fixed Effects
Y
Y
Y
53
Table 7. Industry expert analyst monitoring mechanisms
This table presents the results for monitoring mechanisms by sell-side analysts. Panel A reports logit regressions of the
probability of analysts discussing corporate governance related issues on coverage firms in their research reports (Model
1), issuing cash flow forecasts on coverage firms (Model 2), and OLS regression of the absolute value of analyst earnings
forecast deviation from management guidance (Model 3). Panel B reports the impact of these mechanism on the
earnings management of coverage firms. See Appendix B for a detailed description of variables. Coefficient values are
reported as percentages and robust t-statistics based on standard errors adjusted for heteroskedasticity (White (1980))
and firm clustering (Petersen (2009)) are in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%,
and 1%, respectively.
Model 1
Model 2
Model 3
Panel A: Monitoring mechanisms
Gov Report (Logit) CF Forecast (Logit) Deviation (OLS)
Related analyst (A)
27.200***
31.850***
0.165**
(4.474)
(5.617)
(2.244)
Unrelated analyst (A)
2.830
-4.270
-0.085
(0.528)
(-0.800)
(-1.377)
Abs_DA
1.775***
(2.655)
Market cap
12.030**
0.257**
-0.051
(2.486)
(2.520)
(-0.664)
Market-to-book
-0.098
-0.080
0.002
(-0.817)
(-0.213)
(0.186)
ROA
0.160
64.470***
-2.174***
(0.404)
(4.007)
(-3.192)
Total assets growth
7.790
-5.510*
-0.151
(0.402)
(-1.813)
(-1.029)
Cash flow volatility
-2.140
1.360*
0.003
(-0.389)
(1.735)
(0.007)
External financing
-9.040
31.660**
0.365
(-0.594)
(2.160)
(1.021)
Institutional ownership (dedicated)
-51.680***
39.250
-0.266
(-2.862)
(1.229)
(-0.548)
Institutional ownership (non-dedicated)
30.290
42.280***
-0.904***
(1.641)
(3.260)
(-2.810)
Experience as analyst (A)
-8.830
-1.680***
-0.009
(-0.837)
(-3.225)
(-1.540)
Experience with firm (A)
1.770***
-1.210
0.008
(3.620)
(-1.131)
(0.824)
Analysts from top brokers (A)
1.070
67.890***
0.213***
(1.285)
(17.015)
(3.861)
All-star analysts (A)
34.710***
23.480***
0.197**
(3.852)
(4.083)
(2.546)
Analyst portfolio size (A)
5.110***
-2.090***
0.000
(12.649)
(-6.764)
(0.031)
Abret
-1.081***
(-7.762)
Loss
0.403***
(3.163)
Evoltil
0.022***
(3.561)
Fvoltil
77.991***
(4.714)
Che
-0.090
(-0.157)
Pae
6.102***
(17.776)
Appae
4.920***
(6.805)
54
Abretm3
Gap
13.43%
106,287
Pseudo-R2/Adj-R2
N
Fixed Effect
Y
Panel B: Impact on earnings management (Abs_DA)
Related Analysts with Gov Report (Residual)
Model 1
-0.360***
(-3.885)
-0.153***
(-3.466)
Related Analysts without Gov Report (Residual)
Related Analysts with CF Forecast (Residual)
15.68%
94,647
Y
Model 2
-0.477***
(-2.872)
0.002***
(9.458)
19.61%
23,675
Y
Model 3
-0.340***
(-3.755)
-0.144***
(-3.087)
Related Analysts without CF Forecast (Residual)
High Deviation Related Analysts (Residual)
Low Deviation Related Analysts (Residual)
Adj-R2
N
Fixed Effect
28.41%
24,918
Y
55
28.44%
24,918
Y
-0.327***
(-5.210)
-0.156***
(-3.462)
28.43%
24,918
Y
Table 8: Income-increasing vs income-decreasing earnings management
This table presents OLS regression results of earnings management for the subsamples of income-increasing
and income-decreasing discretionary accruals. The full sample consists of 24,918 firm-year observations from
1988 to 2011. The dependent variable is the absolute value of discretionary accruals (Abs_DA) in each
subsample regression. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%, respectively.
Income-increasing DA
-0.238***
(-3.582)
-0.071
(-1.210)
-0.043
(-0.672)
-0.012
(-1.419)
0.045**
(2.562)
1.822*
(1.828)
-0.243
(-1.236)
-0.262
(-0.862)
1.379
(1.519)
-1.962*
(-1.823)
-2.832***
(-5.693)
-0.015
(-0.455)
-0.085*
(-1.962)
-0.392*
(-1.650)
-0.197
(-0.668)
0.033**
(2.196)
26.21%
12,015
Y
Related analysts (Residual)
Unrelated analysts (Residual)
Inexperienced analysts (Residual)
Market cap
Market-to-book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
Experience as analyst
Experience with firm
Analysts from top brokers
All-star analysts
Analyst Portfolio size
Adj-R2
N
Fixed Effects
56
Income-decreasing DA
-0.132**
(-2.511)
-0.039
(-0.823)
-0.007
(-0.142)
-0.018**
(-2.387)
0.072***
(4.561)
-1.749**
(-2.244)
0.224
(1.006)
0.189
(0.605)
1.592**
(2.342)
-3.111***
(-4.570)
-3.008***
(-7.598)
0.029
(1.140)
-0.070**
(-2.020)
-0.260
(-1.332)
0.098
(0.424)
0.003
(0.273)
39.50%
12,903
Y
Table 9. Cross-sectional variation of analyst industry expertise on earnings management
This table reports results from cross-sectional regressions of earnings management. The full sample consists of 24,918
firm-year observations from 1988 to 2011. The dependent variable is the absolute value of discretionary accruals
(Abs_DA) adjusted for business model shocks using Owens, Wu and Zimmerman (2015) model. See Appendix B for a
detailed description of variables. Firm and year-fixed effects are included. Coefficient values are reported as percentages
and robust t-statistics based on standard errors adjusted for heteroskedasticity (White (1980)) and firm clustering
(Petersen (2009)) are in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%, respectively.
Related analyst affiliated (residual)
Related analyst unaffiliated (residual)
Model 1
-0.106
(-1.376)
-0.177***
(-4.129)
Model 2
Related analyst with prof connections(residual)
Model 3
Model 4
-0.148**
(-2.172)
-0.204***
(-3.900)
Related analyst without prof connections(residual)
Related analyst high info asymmetry (residual)
-0.269***
(-4.717)
-0.122**
(-2.578)
Related analyst low info asymmetry (residual)
Related analyst weak governance(residual)
-0.313***
(-3.947)
-0.139***
(-2.945)
Related analyst strong governance(residual)
Related analyst long experience (residual)
Related analyst short experience (residual)
Unrelated analysts (Residual)
Inexperienced analysts (Residual)
Market cap
Market-to-book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
Experience as analyst
Experience with firm
Analysts from top brokers
All-star analysts
Analyst portfolio size
Adj-R2
N
Fixed Effects
Model 5
-0.056
(-1.530)
-0.038
(-0.960)
-0.015***
(-2.728)
0.065***
(5.350)
0.075
(0.126)
-0.041
(-0.335)
0.014
(0.063)
0.941*
(1.951)
-2.731***
(-4.534)
-2.972***
(-9.265)
0.015
(0.794)
-0.086***
++(-3.175)
-0.355**
(-2.389)
-0.029
(-0.154)
0.012
(1.343)
28.41%
24,918
Y
-0.056
(-1.529)
-0.037
(-0.914)
-0.015***
(-2.733)
0.065***
(5.353)
0.081
(0.135)
-0.044
(-0.361)
0.009
(0.041)
0.946*
(1.963)
-2.692***
(-4.473)
-2.955***
(-9.220)
0.014
(0.749)
-0.087***
(-3.183)
-0.356**
(-2.394)
-0.026
(-0.137)
0.012
(1.347)
28.42%
24,918
Y
57
-0.054
(-1.461)
-0.036
(-0.901)
-0.015***
(-2.704)
0.065***
(5.344)
0.079
(0.132)
-0.048
(-0.391)
0.014
(0.062)
0.961**
(1.987)
-2.708***
(-4.504)
-2.963***
(-9.256)
0.014
(0.746)
-0.087***
(-3.194)
-0.357**
(-2.402)
-0.023
(-0.121)
0.012
(1.347)
28.45%
24,918
Y
-0.054
(-1.479)
-0.036
(-0.911)
-0.015***
(-2.711)
0.065***
(5.347)
0.091
(0.152)
-0.051
(-0.416)
0.013
(0.056)
0.978**
(2.021)
-2.725***
(-4.518)
-2.992***
(-9.340)
0.014
(0.712)
-0.087***
(-3.196)
-0.351**
(-2.362)
-0.027
(-0.146)
0.012
(1.368)
28.47%
24,918
Y
-0.213***
(-4.305)
-0.124**
(-2.478)
-0.056
(-1.538)
-0.041
(-1.033)
-0.016***
(-2.895)
0.065***
(5.371)
0.118
(0.198)
-0.042
(-0.342)
-0.002
(-0.010)
0.962**
(1.993)
-2.611***
(-4.346)
-2.901***
(-9.048)
0.019
(0.984)
-0.091***
(-3.323)
-0.388***
(-2.607)
-0.023
(-0.123)
0.012
(1.378)
28.42%
24,918
Y
Table 10. Robustness analysis
This table presents earnings management regressions with additional controls (Model 1-4), post-2002 subsample period
(model 5), and for LinkedIn vs. non-LinkedIn analyst residual counts (model 6). The baseline sample consists of 24,918
firm-year observations from 1988 to 2011. The dependent variable is the absolute value of discretionary accruals
(Abs_DA) adjusted for business model shocks using Owens, Wu and Zimmerman (2015) model. See Appendix B for a
detailed description of variables. Firm and year-fixed effects are included. Coefficient values are reported as percentages
and robust t-statistics based on standard errors adjusted for heteroskedasticity (White (1980)) and firm clustering
(Petersen (2009)) are in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%, respectively.
Related analysts (Residual)
Unrelated analysts (Residual)
Inexperienced analysts (Residual)
Model 1
-0.183***
(-4.260)
-0.055
(-1.506)
-0.038
(-0.941)
Model 2
Model 3
-0.181*** -0.183***
(-4.193)
(-4.258)
-0.052
-0.054
(-1.402)
(-1.473)
-0.033
-0.036
(-0.829)
(-0.890)
Model 4
-0.180***
(-4.187)
-0.053
(-1.445)
-0.035
(-0.882)
-0.015***
(-2.740)
0.065***
(5.346)
0.077
(0.129)
-0.043
(-0.352)
0.007
(0.031)
0.946*
(1.964)
-2.699***
(-4.480)
-2.950***
(-9.215)
0.018
(0.722)
-0.085***
(-3.050)
-0.355**
(-2.379)
-0.028
(-0.148)
0.012
(1.350)
-0.006
(-0.224)
-0.015***
(-2.723)
0.065***
(5.385)
0.085
(0.142)
-0.043
(-0.353)
0.006
(0.028)
0.950**
(1.971)
-2.678***
(-4.448)
-2.928***
(-9.118)
0.015
(0.782)
-0.087***
(-3.204)
-0.357**
(-2.398)
-0.028
(-0.152)
0.014
(1.515)
-0.014**
(-2.568)
0.065***
(5.342)
0.111
(0.186)
-0.044
(-0.357)
0.011
(0.048)
0.936*
(1.939)
-2.671***
(-4.439)
-2.890***
(-8.999)
0.014
(0.745)
-0.083***
(-3.030)
-0.354**
(-2.381)
-0.033
(-0.174)
0.012
(1.359)
Model 5
Model 6
-0.220***
(-3.974)
0.006
(0.158)
-0.071
(-1.479)
-0.103***
(-4.852)
-0.093***
(-8.016)
-0.008
-0.038***
(-1.041)
(-5.761)
0.042***
0.061***
(2.843)
(4.970)
-0.347
0.021
(-0.406)
(0.035)
-0.251
0.004
(-0.940)
(0.033)
-0.889
-0.265
(-1.433)
(-1.139)
0.849
1.010**
(1.202)
(2.085)
-0.644
-2.725***
(-0.876)
(-4.552)
-0.747*
-3.012***
(-1.708)
(-9.425)
0.025
0.018
(0.989)
(0.958)
-0.145*** -0.081***
(-3.728)
(-2.990)
0.266
-0.359**
(1.238)
(-2.423)
-0.084
-0.082
(-0.257)
(-0.443)
0.015
0.010
(0.922)
(1.081)
-1.254***
(-4.123)
28.45%
24,918
Y
27.61%
13,170
Y
Total LinkedIn analysts (Residual)
Total non-LinkedIn analysts (Residual)
Market cap
Market-to-book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
Experience as analyst
Experience with firm
Analysts from top brokers
All-star analysts
Analyst portfolio size
Industry covering experience
Cover related industry
-0.014***
(-2.615)
0.065***
(5.346)
0.075
(0.126)
-0.042
(-0.343)
0.003
(0.013)
0.941*
(1.953)
-2.700***
(-4.488)
-2.955***
(-9.228)
0.015
(0.756)
-0.086***
(-3.180)
-0.356**
(-2.392)
-0.026
(-0.138)
0.012
(1.340)
-0.191*
(-1.840)
Cover customer/supplier
-0.690**
(-2.469)
AC_IED_Pct
Adj-R2
N
Fixed Effects
28.41%
24,918
Y
28.43%
24,918
Y
58
28.43%
24,918
Y
28.80%
24,918
Y
Internet Appendix
This internet appendix accompanies, “Are all analysts created equal? Industry expertise and monitoring effectiveness of
financial analysts,” submitted to the Journal of Accounting and Economics.
Table IA.1 presents regression results for the analysis of whether industry expert analysts who discuss corporate
governance issues in their research reports provide more effective monitoring.
Table IA.2 presents regression results for the analysis of whether industry expert analysts who issue cash flow forecasts
and who deviate more from management guidance provide more effective monitoring.
Table IA.3 presents the multinomial logit regression results of the probability of firms’ signed discretionary accruals
falling into the top quintile (the most positive discretionary accruals) or the bottom quintile (the most negative
discretionary accruals).
59
Fixed Effects
N
Adj-R2 /Pseudo
R2
Abnormal ROA * Related analysts without Gov Report (Residual)
Abnormal ROA * Related analysts with Gov Report (Residual)
Related analysts without Gov Report (Residual)
Related analysts with Gov Report (Residual)
60
Y
23,233
Y
19,772
2.14%
(-2.957)
(-3.053)
61.89%
-23.450***
(-1.894)
-104.470*
Restatements
(Logit)
-1.536***
(-3.659)
-4.316***
F-score residual
(OLS)
Y
9,428
61.55%
(-2.304)
-2.357**
(-2.238)
-5.782**
CEO Comp
(OLS)
Y
6,623
2.20%
(-2.924)
-176.440***
(-1.533)
-232.330
CEO Turnover (Logit)
This table presents excerpts of the regression results for the analysis of whether industry expert analysts who discuss corporate governance issues in their research
reports provide more effective monitoring. Discussions of corporate governance-related issues on coverage firms in analyst reports are obtained from Thomson
Reuters Investext. See Appendix B for a detailed description of variables. Coefficient values are reported as percentages. In parentheses below coefficient estimates are
robust t-statistics based on standard errors adjusted for heteroskedasticity (White (1980)) and firm clustering (Petersen (2009)). *, **, and *** indicate statistical
significance at the 10%, 5%, and 1%, respectively.
Internet Appendix Table IA.1. Industry Expert Analyst Monitoring Mechanisms: Governance Discussions
Fixed Effects
N
Adj-R2 /Pseudo
R2
Low Deviation Related analysts (Residual)
High Deviation Related analysts (Residual)
Related analysts without CF Forecast (Residual)
Related analysts with CF Forecast (Residual)
Y
23,233
61
Y
19,772
2.18%
(-2.099)
(-2.000)
61.82%
-15.320**
(-2.420)
-106.810**
Restatements
(Logit)
-1.026**
(-3.022)
-2.692***
F-score residual
(OLS)
Y
23,233
61.88%
Y
19,772
2.12%
(-2.945)
(-2.684)
(-2.215)
-22.150***
(-2.020)
-1.298***
-38.280**
Restatements
(Logit)
-1.879**
F-score residual
(OLS)
This table presents excerpts of the regression results for the analysis of whether industry expert analysts who issue cash flow forecasts and who deviate more from
management guidance provide more effective monitoring. See Appendix B for a detailed description of variables. Coefficient values are reported as percentages. In
parentheses below coefficient estimates are robust t-statistics based on standard errors adjusted for heteroskedasticity (White (1980)) and firm clustering (Petersen
(2009)). *, **, and *** indicate statistical significance at the 10%, 5%, and 1%, respectively.
Internet Appendix Table IA.2. Industry Expert Analyst Monitoring Mechanisms: Cash Flow Forecasts and Divergence from Management Guidance
Internet Appendix Table IA.3. Income-increasing vs income-decreasing earnings management: Multinomial
logit approach
This table presents the results from a multinomial logit regression. The dependent variable is a nominal variable that
takes the value of one for observations with signed discretionary accruals in the bottom quintile (the most negative), 2
for observations with signed discretionary accruals in the middle three quintiles (moderate), and 3 for observations with
signed discretionary accruals in the top quintile (the most positive). The sample consists of 24,918 firm-year observations
from 1988 to 2011. The middle three quintiles serve as the base category. *, **, and *** indicate statistical significance at
the 10%, 5%, and 1%, respectively.
Related analysts (Residual)
Unrelated analysts (Residual)
Inexperienced analysts (Residual)
Market cap
Market-to-book
ROA
Total assets growth
Cash flow volatility
External financing
Institutional ownership (dedicated)
Institutional ownership (non-dedicated)
Experience as analyst
Experience with firm
Analysts from top brokers
All-star analysts
Analyst Portfolio size
Pseudo R2
N
Fixed Effects
Bottom Quintile
Top Quintile
-5.000***
-9.530***
(-2.591)
(-4.990)
-1.650
-0.898
(-1.012)
(-0.736)
1.830
-0.146
(0.938)
(-0.077)
-1.810***
-1.380***
(-4.986)
(-5.018)
0.717
1.670***
(1.469)
(3.450)
-73.670***
125.580***
(-4.573)
(6.673)
17.520***
4.530
(3.784)
(0.972)
-22.320***
4.830
(-2.773)
(0.972)
-6.440
51.240***
(-0.376)
(2.869)
-89.570***
-65.150***
(-4.384)
(-3.314)
-29.770**
-29.880***
(-2.512)
(-2.623)
2.920***
1.510**
(3.957)
(2.163)
-4.710***
-8.810***
(-3.925)
(-7.342)
-15.570***
-15.040***
(-2.621)
(-2.662)
-19.870**
-6.820
(-2.238)
(-0.824)
-1.550***
-0.948**
(-3.944)
(-2.548)
4.98%
24,918
Y