Corporate Culture and Analyst Catering

Journal of Accounting and Economics 67 (2019) 120–143
Contents lists available at ScienceDirect
Journal of Accounting and Economics

journal homepage: www.elsevier.com/locate/jae
Corporate culture and analyst catering⁎

T
Joseph Pacelli
Assistant Professor of Accounting, Kelley School of Business, Department of Accounting, 1309 E. Tenth Street, HH 5159 Bloomington, IN 47405
ARTICLE INFO ABSTRACT
Keywords: This study examines the relation between financial institutions’ corporate culture and the quality of
Analysts analysts’ research services. Using data collected from the Financial Industry Regulatory Authority, I
Corporate Culture measure the weakness of financial institutions’ corporate culture based on violations observed in
Conflicts of Interest securities activities unrelated to equity research. I find evidence demonstrating an association be-
Global Settlement
tween weak corporate culture and analysts’ providing research products catered to institutional
clients at the expense of individual investors. Specifically, FINRA violations are associated with both
(i) less accurate forecasts and less informative reports, and (ii) higher institutional commission
revenues and more broker-hosted conferences for select institutional clients.
1. Introduction
The issue before this Committee today is whether the global settlement will reform the culture of Wall Street […] I believe that the Wall
Street culture must change from the top down, and I am not convinced that the global settlement has done enough to change attitudes at the
top of these banks. (U.S. Senate, 2003)
The activities of sell-side equity analysts have attracted significant attention from both scholars and regulators. Consequently, an
extensive body of literature has emerged examining the influence of financial institutions’ characteristics on the quality and ob-
jectivity of analysts’ research.1 One characteristic, underexplored yet long considered important by regulators, is corporate culture.
This issue is particularly relevant given that regulators have openly questioned whether regulations such as the Global Settlement
would “reform the culture of Wall Street” (U.S. Senate, 2003), and have argued that success of the regulations would depend on firms
following “both the letter and the spirit” of the law (NASAA, 2003). Although such reforms appear to have reduced investment
banking conflicts of interest (Kadan et al., 2009; GAO Report, 2012), the mix of services that financial institutions provide has also
changed. New incentives arose for analysts to cater to a select group of institutional clients by providing them with exclusive research
services including organizing conferences and meetings with management as well as preparing proprietary analyses (e.g., Green
E-mail address: jpacelli@indiana.edu.

⁎
This paper is based on my dissertation work at the Samuel Curtis Johnson School of Management, Cornell University. I am grateful to my
dissertation committee: Eric Yeung (co-chair), Sanjeev Bhojraj (co-chair), Rob Bloomfield, and Roni Michaely. I have benefited from comments and
suggestions from Joanna Wu (the editor), Mark Bradshaw (the reviewer), Janet Gao, Bill Mayew (discussant), Ken Merkley, Eugene Soltes (dis-
cussant), and seminar participants at the AAA Annual Meeting 2015, AFA Annual Meeting 2017, Boston College, Columbia University, Duke
University, Emory University, Georgetown University, Harvard Business School, Indiana University, London Business School, Northwestern
University, Stanford University, University of California, Berkeley, University of California, Los Angeles, University of Chicago, University of
Colorado, University of Illinois at Chicago, University of Southern Illinois, Washington University in St. Louis, and Yale University. I thank Stan
Markov and Musa Subasi for sharing the broker-hosted investor conference data. All remaining errors are my own.
1
For example, prior studies have examined characteristics including business type (Cowen et al., 2006; Groysberg et al., 2013), cross-business
affiliation (Firth et al., 2013), and compensation (Groysberg et al., 2011; Brown et al., 2015).
https://doi.org/10.1016/j.jacceco.2018.08.017
Received 26 May 2016; Received in revised form 25 August 2018; Accepted 30 August 2018
Available online 07 September 2018
0165-4101/ © 2018 Elsevier B.V. All rights reserved.
J. Pacelli Journal of Accounting and Economics 67 (2019) 120–143
et al., 2014; Drake et al., 2018). These new conflicts of interest raise the concern that, without regulation adequately addressing
cultural flaws within corporations, analysts’ objectivity can still be compromised.
This study investigates whether a financial institution's corporate culture influences the quality of research services it provides. In
particular, I examine whether corporate cultures that fail to promote market integrity and fair business practices (henceforth, weak
cultures and weak-culture firms) are associated with analysts catering their research services to provide institutional clients a pre-
ferential high-quality product to the detriment of retail clients. Recent evidence suggests that catering has become an important
concern for investors, as analysts increasingly face pressures to favor institutional clients by providing them with preferential
treatment, such as access to information and research results that are unavailable to other clients (Davis, 2004; Fisch, 2007; FINRA,
2013). A prominent example of this behavior is the recent Goldman Sachs case in which the firm was charged with holding research
huddles that provided top clients with exclusive information and trading tips withheld from other clients (Moyer, 2011).2 While such
behavior can be difficult to detect, it nonetheless represents a breach of analysts’ fiduciary duties to their retail and excluded
institutional clients, and is therefore inconsistent with brokerage houses’ claims to provide all clients equal treatment.3 Although
regulatory requirements concerning unequal or exclusive treatment are sometimes unclear (i.e., a “gray” area), such treatment can
potentially violate securities laws requiring analysts to engage in “fair dealing with customers” and not issue contradictory opinions
about a stock (FINRA 2211, FINRA 2241). My central claim is that weak corporate cultures exacerbate conflicts of interest in equity
research departments and increase analysts’ tendency to cater their research products.
Prior studies define corporate culture as the shared assumptions, values, and beliefs that inform employee behavior in the pre-
sence of incomplete contracts (Kreps, 1990; Schein, 1990; Crémer, 1993). For this study, I assess the weakness of financial institu-
tions’ corporate culture in the context of their securities market activities via the number of annual security code violations sanc-
tioned by the Financial Industry Regulatory Authority (FINRA) and disclosed in BrokerCheck reports. FINRA violations provide a
useful proxy for examining culture as these rules are in fact intended to protect and promote fair dealing with clients. Moreover,
violations often arise due to firms’ engaging in dishonest business practices, which contribute to a culture that fails to promote ethical
behavior (Richards, 2007; Egan et al., 2016).
In addition, FINRA violations can facilitate an evaluation of corporate culture distinct in two important ways from measures used
in the prior literature. First, while prior studies examine conflicts of interest at the covered firm level that may be indicative of weak
culture (e.g., Lin and McNichols, 1998; Corwin et al., 2017), FINRA violations are measured at the financial institution level and
reflect concrete episodes of actual misconduct instead of the potential for misconduct. Second, while recent research has assessed
culture based on broad employee-survey data (Garrett et al., 2014; Guiso et al., 2015), FINRA violations provide a more specific
measure of culture relating directly to employees’ actual misconduct in securities business activities. Employee surveys are unlikely to
detect weak cultures if individuals fail to identify and report issues within the firm, due to assortative matching problems or ra-
tionalization of unethical behavior (Becker, 1973; Cohn et al., 2014). By measuring corporate culture using FINRA violations, I
resolve these challenges: I evaluate culture through independent assessments of the regulator.
I test the catering prediction using several proxies gauging the quality of products analysts provide to retail and institutional clients.
First, I examine the quality of analysts’ written research based on the accuracy of analysts’ earnings forecasts and the informativeness of
analysts’ reports derived from return reactions surrounding a report revision date (i.e., forecast, recommendation, or target price revi-
sion). Although written research is an important component of analysts’ services (Maber et al., 2016), it is likely to be most useful to retail
clients as they do not enjoy the same personalized access to analysts granted institutional clients. Weak corporate cultures can com-
promise the quality of written research if they lead analysts to allocate less effort to their reports and/or bias their reports. Second, I assess
the quality of services analysts provide to institutional clients based on the number of broker-hosted investor conferences and total
commission revenues earned by the analyst's firm. Recent studies indicate that broker-hosted investor conferences are exclusive events
that connect select clients with firm management, and also signal an analyst's ability to provide other privileged services to these clients
(Green et al., 2014). Similarly, institutional clients often reward analysts for preferential treatment by routing trades through the analysts’
firm, thereby increasing commissions. The catering hypothesis predicts that weak corporate culture is associated with less accurate
forecasts, less informative reports, and higher levels of broker-hosted conferences and commission revenues.
My sample consists of twenty-nine large diversified financial institutions, all of which are subject to frequent and uniform ex-
aminations by FINRA to assess their compliance with securities regulations. While prior studies demonstrate that large investment
banks generally provide more accurate and less optimistic research, host more conferences, and generate more trading fees (Jacob
et al., 2008; Green et al., 2014), I examine the impact of variation in corporate culture among these firms on the quality of services
they provide. A key feature of my empirical design is the measurement of violations arising outside the firm's research department,
thus ensuring that violations are related to unethical behavior in securities business activities, and not directly related to analyst
activities. This research design further ensures that any associations between FINRA violations and analysts’ research activities are
2
On April 12, 2012, the Securities and Exchange Commission charged Goldman, Sachs & Co. with lacking adequate policies and procedures to
address the risk that weekly huddles would result in the firm's analysts sharing material, nonpublic information about upcoming research changes.
Huddles are an increasingly common practice, and in Goldman's case, their analysts met to provide their best trading ideas to top clients and were
part of the broader program known as the “Asymmetric Service Initiative” (ASI). Goldman agreed to pay a $22 million penalty to settle the charges;
agreed to be censured; accepted a cease-and-desist order; and agreed to review and revise its written policies and procedures to correct deficiencies.
Additional information on this event can be found in SEC Press Release 2012–61.
3
For example, Morgan Stanley's research management policies indicate that research and material views should not be selectively disclosed before
broad dissemination (Morgan Stanley, 2018).
121
attributable to common firm-level forces such as corporate culture. My analyses focus on the sample period 2005–2012, as recent
changes in business models emphasizing specialized exclusive services to top clients have arguably made catering behavior more
prominent in this time period (Green et al., 2014; Drake et al., 2018).
My main results provide evidence consistent with an association between weak corporate cultures and higher degrees of analyst
catering. Specifically, I find that the number of FINRA violations sanctioned against a financial institution in a given year is nega-
tively associated with both the accuracy of forecasts and the informativeness of reports, namely stock market reactions surrounding
forecast, recommendation, or target price revisions produced by a firm's analysts during that year. These findings are consistent with
analysts’ working within a weak corporate culture and providing their retail clients with an inferior written product. Further, I also
find that FINRA violations are associated with more broker-hosted conferences and higher commission revenues, consistent with
analysts in a weak-culture firm providing higher quality products to their institutional clients. Overall, these analyses support the
notion that weak corporate culture exacerbates conflicts of interest associated with analysts catering their products.
I conduct a battery of tests to assess the robustness of my findings. First, I show that corporate culture with respect to securities
activities has incremental effects distinct from other dimensions of corporate culture. Specifically, I find that my results persist after
controlling for overall product quality (Gao et al., 2014), employee satisfaction (Garrett et al., 2014; Guiso et al., 2015), and analysts’
cultural background (Liu, 2016; Du et al., 2017). Second, I show that my findings are unrelated to either rogue employee behavior or
heterogeneity in financial institutions’ business models, such as specialization in a certain type of client. Finally, I validate the catering
measures using a case study based on a specific allegation of catering related to the Goldman Sachs’ research huddle regulatory event.
Ultimately, these robustness analyses strengthen the claim that weak corporate culture increases analysts’ catering tendencies.
In additional analyses, I further explore the relation between corporate culture and analysts’ report bias. While prior studies
provide evidence that analysts often optimistically bias their reports to generate investment banking and/or trading business (e.g.,
Lin and McNichols, 1998; Michaely and Womack, 1999; Jackson, 2005), the Global Settlement has likely reduced many of these
conflicts of interest in recent years.4 Therefore, I expect weak-culture firms to be particularly vigilant in avoiding the appearance of
optimistic bias in the presence of increased regulatory scrutiny. Consistent with this notion, I find that weak-culture firms produce
less optimistically-biased reports in terms of forecasts, recommendations, and target prices. Consistent with the main catering
analyses, my results indicate that regulatory reforms may have altered the nature of analysts’ conflicts of interests without elim-
inating culture as the underlying cause of unethical behavior.
My results contribute to the existing literature across four dimensions. First, I extend the emerging literature on corporate culture
by introducing a new measure of culture and demonstrating its influence on analysts’ research production (e.g., Hoi et al., 2013;
Popadak, 2013; Gao et al., 2014; Garrett et al., 2014; Guiso et al. 2015; Liu, 2016; Du et al., 2017). In contrast to a recent study by
Guiso et al. (2015), I examine the association between culture and employees’, specifically analysts’, interaction with their clients
versus the relation between culture and firm performance. Second, I contribute to the literature on regulations of analysts and
conflicts of interest (e.g., Barber et al., 2006; Ertimur et al., 2007; Barniv et al., 2009; Chen and Chen, 2009; Kadan et al., 2009). My
results indicate that weak culture compromises analysts’ objectivity even after significant regulatory efforts have been made to
improve research quality. Third, my study contributes to the literature examining characteristics of financial institutions that in-
fluence analysts’ research (e.g., Cowen et al., 2006; Groysberg et al., 2011; Green et al., 2014). Specifically, I show that corporate
culture offers a substantive explanation for systematic differences in research quality across financial institutions. Fourth and finally,
my findings have implications for the financial services industry as a whole, as they corroborate recent regulatory concerns regarding
weak culture's being a potential cause of misfeasance in the securities industry.
2. Background on FINRA enforcement process
FINRA is the largest independent regulator of securities firms in the United States; FINRA is the successor to the National
Association of Securities Dealers, Inc. (NASD) and the enforcement operation of the New York Stock Exchange (NYSE). FINRA
operates under supervision of the Securities and Exchange Commission (SEC) and is registered as a self-regulatory organization
(Tuch, 2014). The organization is funded by member firms and associated persons and has the power to discipline member-firms
through sanctioning and expelling members. FINRA's central objective is to create and enforce rules regarding standards of conduct,
the most important of which require firms to “observe high standards of commercial honor and just and equitable practice of trade,”
as stated on FINRA's website (Black, 2013; Tuch, 2014; Edwards, 2016).5
Over the course of a fiscal year, FINRA conducts a series of exams that can lead to enforcement actions. These exams are classified
as cycle exams, cause exams, branch exams, market regulation exams, or sweeps. Over 2,000 cycle exams are conducted each year
and compose the bulk of FINRA's field work (FINRA, 2017a). The purpose of a cycle exam is to determine whether firms are in
compliance with federal securities laws, rules, and regulations. Financial institutions are subject to a cycle exam once every one, two,
or three years depending on the complexity of the firm's business model, size, and risk. The remaining FINRA exams are less frequent
and focus on specific customer complaints, high-risk activity, or general reviews of industry practices. These exams are likely to be
more relevant for smaller firms not subject to the annual cycle exam.
The cycle exam consists of three stages. First, there is a preparation stage in which FINRA assigns a regulatory coordinator and/or
4
Consistently, prior studies also indicate that optimistic bias has declined following Global Settlement (Chen and Chen, 2009; Bradshaw et al.,
2017).
5
http://finra.complinet.com/en/display/display_main.html?rbid=2403&element_id=5504
122
lead examiner who notifies the firm of the upcoming exam. During this stage, FINRA communicates high-level details regarding exam
logistics and begins to review and analyze the firm's information. Next, FINRA begins the on-site component of the cycle exam. At this
point, FINRA generally conducts tours of the firms’ offices, requests meetings with key personnel, and investigates practices at local
offices. For large international firms, examiners commonly stay on-site for three to four months (ABA, 2012). The final stage of the
cycle exam consists of a closeout component, in which the examination team prepares a report for FINRA management to review and
approve. The report identifies any exceptions and potential violations.
Once the cycle exam is complete, any serious violations are escalated to FINRA's Enforcement Department for a decision regarding
formal disciplinary action. If a firm receives a formal disciplinary action, the firm has the option to settle the case or to establish a
defense and fight the case before a hearing panel. In practice, very few cases face a hearing panel (Black, 2013) and most are settled
through a Letter of Acceptance, Waiver, and Consent (AWC) in which the firm neither admits nor denies the allegations but agrees to
the findings and consequent sanctions (Pearce, 2017). The AWC resolves the case and the regulator cannot return with another action
alleging violations based on the same conduct. The findings are then considered final and are published on FINRA's website and made
available through BrokerCheck (Harvard Law, 2016).
Each year FINRA publishes a Regulatory and Examination Priorities Letter to highlight important issues (FINRA, 2017b). This
publication is useful for firms preparing for the cycle exams; the letters themselves indicate that culture is critical for FINRA. The
2016 priority letter, for example, states that “Firm culture, ethics and conflicts of interest also remain a top priority for FINRA.”
Moreover, the 2015 letter specifically identifies weak ethical behavior and a lack of alignment between the firm and its customers as
two key shortcomings. These references alert firms to the possibility that violations can ensue from weak corporate culture.
3. Related literature and hypothesis development
3.1. Corporate culture
Corporate culture represents the shared assumptions, values, and beliefs that inform employee behavior within a firm (Kreps,
1990; Schein, 1990; Crémer, 1993; O’ Reilly and Chatman, 1996). In the presence of incomplete contracts (Grossman and
Hart, 1986), corporate culture helps communicate the appropriate course of action and aligns employee behavior with the firm's
objectives (Hermalin, 1999; Guiso et al., 2015). Corporate culture can be viewed as a form of “social control” that complements
traditional control systems, such as formal incentives, as it helps employees make decisions when they “face choices that cannot be
properly regulated ex ante” (Guiso et al., 2015, 62, 61).
This study uses FINRA violations to examine weak corporate culture in securities business activities and infer employees’ attitudes
towards misconduct. This study differs from prior research examining culture based on employee surveys (e.g., Garrett et al., 2014;
Guiso et al., 2015) in at least two dimensions. First, I focus specifically on unethical behavior in the securities business industry as this
is directly relevant to analyst research: this industry has previously been criticized for conflicts of interest stemming from alleged
cultural flaws (e.g., Richards, 2006; Tarullo, 2014). Recent experimental evidence also indicates that bankers rationalize dishonest
behavior by considering their professional identity distinct from their personal identity (Cohn et al., 2014).
Second, complementary to prior studies measuring culture based on employee surveys, I measure culture using violations sanctioned
by a third-party regulator to assess firms’ treatment of their clients. This approach enables a focus on the relationship between employees,
in this case analysts, and their clients, as opposed to the relationship between employees and their managers. Existing studies using
surveys assessing employees’ attitudes towards their employers (e.g., Guiso et al., 2015) do not directly address firm-client relationships,
especially the effects of varied clientele on the relationship. Moreover, assortative matching problems may influence survey results as
they lead employees to match to firms with similar values (Becker, 1973; Shimer and Smith 2000).6 These employees may not detect and
report flaws in managers’ behavior, especially if employees compartmentalize their professional identities (Cohn et al., 2014).
3.2. The Global Settlement and Analysts’ conflicts of interest
I concentrate the examination of the relation between corporate culture and analysts’ catering behavior in the post-Global
Settlement period because this regulation led to many important changes to sell-side analyst research business models. Enacted in
December 2002, the Global Settlement was a response to claims that sell-side analyst research was unduly influenced by conflicts of
interest between investment banking and research. Regulators claimed analysts’ research was “not based on principles of fair dealing
and good faith” and contained “exaggerated or unwarranted claims about the covered companies” (FINRA, 2003). One important
concern was that analysts’ reports were often optimistically biased towards investment banking clients (Lin and McNichols, 1998;
Michaely and Womack, 1999; Bradshaw et al., 2006). The Global Settlement introduced reforms to reduce these conflicts, such as
requiring a physical separation of research and investment banking, and prohibiting firms from cross-subsidizing research with
investment banking activities.
Despite the Global Settlement having imposed the highest sanction to date, $1.4 billion, in a civil securities enforcement action,
some regulators expressed concerns whether the Settlement would address underlying cultural problems in the securities industry.
During a hearing held by the U.S. Senate Committee on Banking, Housing, and Urban Affairs, the committee questioned whether the
6
For example, individuals who are not concerned about ethical behavior may choose to work at firms that also have little concern for ethical
behavior, and vice versa.
123
Global Settlement would “reform the culture of Wall Street, restore the integrity of stock analysts, and regenerate investor con-
fidence” (U.S. Senate, 2003). During her tenure as SEC director, Lori Richards presented a series of speeches encouraging financial
institutions to develop cultures that “foster ethical behavior and decision-making” and instill in employees “an obligation to do what's
right” (Richards, 2006). These comments provide strong evidence for corporate culture having been a persistent regulatory concern,
and provoke uncertainty regarding the ability of the Global Settlement to adequately address these problems.
Evidence for the success of the Global Settlement in reducing analysts’ conflicts of interest is mixed. On the one hand, regulatory
reviews indicate that the Global Settlement appears to have been successful in reducing investment banking conflicts (GAO
Report, 2012). Prior academic studies also find that optimistic recommendations became less frequent after the Global Settlement
(Barber et al., 2006; Kadan et al., 2009; Clarke et al., 2011).
On the other hand, evidence exists to suggest that the Global Settlement may have been unsuccessful in addressing analysts’
conflicts of interest due to new business models serving to change the nature of these conflicts.7 These business models have reduced
the degree to which investment banking cross-subsidizes investment research, while increasing the importance of specialized ex-
clusive services for key clients. Fisch (2007, 79) notes that the Global Settlement had “a limited impact on potential conflicts that do
not involve investment banking” and that the regulation has led many research departments to segment their product to produce
disparate access to information for retail and institutional investors. Fisch (2007, 53) also states via Peter Lynch: “If, prior to the
regulatory reforms, the smart money got the information first, now in many cases, retail investors do not get the information at all.”
The Wall Street Journal expressed a similar concern in an article issued shortly after the Global Settlement:
When New York Attorney General Eliot Spitzer succeeded last year in separating investment-banking divisions from stock analysts to
eliminate conflicts of interest, the historic settlement was supposed to herald a new era of securities analysis—one benefiting little-guy
stock pickers. But now more than ever, the most pioneering, market-moving research is going exclusively to big mutual funds and the
private investment pools known as hedge funds, not to the small investor for whom regulators waged their campaign. (Davis, 2004)
More recently, Goldman Sachs was charged with holding research huddles that provided top clients with exclusive information
and trading tips that were not distributed to their broader client base (Moyer, 2011).
In general, the Global Settlement represented a milestone in the sell-side analyst industry. By separating investment banking and
research productions, the regulation ostensibly eliminated many conflicts of interest. However, the nature of analysts’ conflicts of
interest has also changed following the Global Settlement. To the extent that the Global Settlement did not address underlying
cultural problems in financial institutions, weak corporate cultures can still jeopardize analysts’ objectivity, thereby creating concerns
for retail investors.
3.3. Hypotheses
My central conjecture is that weak corporate cultures exacerbate conflicts of interest within equity research departments and increase
the likelihood that analysts will cater their research products towards top clients at the expense of retail investors. Culture can play a vital
role, since reputation alone is unlikely to be a sufficient deterrent to unethical behavior; the likelihood of misconduct remains con-
siderable when opportunities are plentiful, short-term rewards are large, business activities are opaque, and reputational damage is slow
to occur (Tuch, 2014; Guiso et al., 2015). Moreover, catering is difficult for regulators to detect since they cannot easily observe the soft
services analysts provide to institutional clients, thus increasing incentives for analysts to engage in problematic behavior.8
The law does not offer precise requirements for the legality of catering, which encourages an active role for culture. On the one hand,
securities laws require firms to engage in “fair dealing with customers” and prohibit analysts from issuing opinions that are at odds with
their true beliefs about a stock (FINRA 2111, FINRA 2241). In this sense, catering is clearly at odds with the spirit of the law in that some
customers are harmed by receiving products of lesser value than others receive. On the other hand, the regulatory ethical requirement of
fair dealing with customers is also sometimes unclear. In the Goldman Sachs case, the firm argued that the tips they provided to select
clients were “market color” and that their published research included disclosures stating that their personnel might take positions
contrary to the opinions expressed in reports (Craig, 2009). In this sense, such behavior might be similar to “tipping” in that the
inappropriateness of the actions depends on whether the firm stated that all clients will be treated equally (Irvine et al., 2006).
I examine four aspects of analysts’ research services to determine the relation between culture and analysts’ catering. The first two
aspects entail the quality of analysts’ written research. Although written research is an important component of analysts’ services
(Maber et al., 2016), it is likely to be most useful to retail clients as they do not enjoy the same personalized access to analysts that
institutional clients can expect and benefit from.9 New business models emphasizing specialized services have also greatly reduced
institutional clients’ reliance on written research, with some institutional clients declaring that they “hate written product and would
rather spend two hours on the phone with the analyst,” as stated by Alpha Magazine (Green et al., 2014, 144). Weak corporate culture
might create incentives for analysts to allocate less effort to reports and more effort to meeting the demands of their institutional
clients by providing specialized services. Weak culture may also lead analysts to introduce intentional bias into the research they
7
Recent academic studies discussing these new business models include Green et al. (2014), Maber et al. (2016), and Drake et al., (2018).
8
The Goldman Sachs case discussed above is the only instance of analyst catering observed in the sample.
9
For example, many online stock brokerage services, such as MerrillEDGE, provide investors with access to research reports (https://www.
stockbrokers.com/guides/online-stock-brokers). Brokers may also use research reports indirectly to generate investment recommendations (http://
www.finra.org/investors/brokers).
124
provide to retail investors, especially if doing so facilitates access to management for their institutional clients. Thus, if culture is
related to analysts’ catering, I expect analysts in weak-culture firms to produce lower quality written research.10
I measure the quality of written research using two analyst outputs. The first is earnings forecast accuracy. While prior studies
indicate that earnings forecasts are crucial in disseminating information about earnings expectations to market participants (Clement
and Tse, 2003; Gleason and Lee, 2003), new business models emphasizing specialized services including private communications and
broker conferences have greatly reduced institutional clients’ demand for these forecasts. Consistent with this claim, earnings esti-
mates have declined in importance in both institutional investor rankings and survey data (Bradshaw, 2011; Brown et al., 2015).
Earnings forecasts also have useful empirical properties as they are continuous, have the common benchmark of actual earnings, and
facilitate comparisons across analysts at different financial institutions. Thus, my first hypothesis is as follows:
H1: Weak corporate culture is negatively associated with analysts’ earnings forecast accuracy.
The second output related to the quality of analysts’ written research is report informativeness, as evidenced by the market
reaction to analyst report output, i.e., forecast, recommendation, or target price revision. Prior studies suggest that analysts’ reports
do not always provide new information to the market, as analysts often repackage or retransmit material that is not necessarily
informative (Frankel et al., 2006; Merkley et al., 2017). If analysts at financial institutions with weak corporate cultures produce
written research that is less useful for individual investors, their reports should elicit weaker market reactions. Thus, my second
hypothesis is as follows:
H2: Weak corporate culture is negatively associated with analysts’ report informativeness.
The second two aspects of analysts’ research that I examine relate to the products analysts provide to their institutional clients. If
analysts employed by weak-culture firms produce a more catered research product, I expect weak culture to be not only negatively
associated with the quality of products provided to retail clients, but also positively associated with the quality of products provided
to institutional clients.
One challenge with measuring the quality of analysts’ institutional products is that the soft services that analysts provide in-
cluding private phone calls and one-on-one meetings, are not easily observable. Thus, I examine both a direct and indirect measure of
institutional product quality. The direct measure is based on broker-hosted investor conferences. These conferences are invitation-
only events that provide a select group of institutional clients with access to management (Green et al., 2014). As noted by
Green et al., (2014), these events also signal analysts’ ability or willingness to provide other types of special services, such as
arranging private meetings with firms’ managers. Thus, my next hypothesis is as follows:
H3: Weak corporate culture is positively associated with analysts hosting investor conferences.
Since investor conferences are only one type of institutional product analysts provide to select investors, I also consider a broader
measure of institutional product quality. This measure is based on the commission revenues a broker generates for routing trades of
an analyst's covered firm. Institutional investors often reward brokers who provide high quality services by allocating a greater
portion of transaction order flow and thus higher commissions to the broker (Irvine, 2000; Goldstein et al., 2009). In turn, a
brokerage house can compensate analysts for providing premium services, since analysts’ bonuses are often tied directly to com-
missions (Goldstein et al., 2009). These observations regarding commission revenues lead to my final hypothesis:
H4: Weak corporate culture is positively associated with brokerage commission revenues.
4. Data and research design
4.1. FINRA data and sample selection
I begin my sample selection by obtaining from the Federal Reserve a list of financial conglomerates with U.S. security subsidiaries.
For each security subsidiary, I collect registration numbers from the SEC to ensure an accurate match to security code data.11 The
primary benefit of examining these large conglomerate banks is that they are likely subject to similar annual risk-based cycle ex-
aminations (as discussed in Section 2) and thus exhibit similar likelihoods of being targeted by regulators.12 My sample is further
10
The above discussion does not imply that written research is unimportant. Written research may be important for analysts who want to build
their client base (Maber et al., 2016). Financial institutions may also want to encourage high quality written research to prevent reputational
damage (Jackson, 2005; Mehran and Stulz, 2007).
11
Several institutions in the sample, such as Wells Fargo & Company, are a result of large mergers and acquisitions. For these institutions, I
exclude their observations prior to the merger date since the FINRA data does not clearly distinguish between the acquirer and target prior to the
merger date.
12
Assuming that these exams follow a similar procedure, this also implies that violations are equally likely to be detected among the firms within
my sample.
125
restricted by the availability of analyst data. For each financial institution in the sample, I collect its name from I/B/E/S and require the
institution to employ at least one analyst covering a firm that is also covered by at least one other institution in the sample. This
facilitates relative comparisons of analysts’ outputs such as forecast accuracy across institutions within my sample.
Table 1 presents the final sample of financial institutions. Panel A lists the sample of financial institutions, which consists of 48
security subsidiaries across 29 financial institutions. Panel B describes the business activities of these financial institutions based on
the BrokerCheck header page. The table indicates that 97% of the institutions engage in brokerage activity, 66% have investment
advisory businesses, and all financial institutions have underwriting or other activity.13 These statistics demonstrate that financial
institutions’ business activities span a wide spectrum of services subject to FINRA enforcement.
4.2. Description of FINRA data
I download and collect BrokerCheck reports from FINRA's website for each security subsidiary in the sample. As discussed in
Section 2, FINRA frequently conducts cycle exams to determine whether firms are in compliance with securities laws and regulations.
These violations are published online in BrokerCheck reports allowing investors to observe the regulatory histories of all financial
institutions. Security code violations are presented in the Disclosure Events section of the BrokerCheck report.
Table 2 describes the sample of FINRA violations. Panel A presents the sample selection procedure used on the Disclosure Events
data. I retain all completed (i.e., not pending) events with non-missing case numbers issued between 2005 and 2012. I begin my
sample selection in 2005 due to sparse data in the early 2000s and to ensure that the violations are not directly related to the Global
Settlement.14 I delete events with duplicate case numbers, as well as those with no fines indicated. I retain only disclosure events
issued by major regulatory agencies, i.e., FINRA and its predecessors, the National Association of Securities Dealers (NASD) and the
New York Stock Exchange (NYSE). To avoid double counting events I exclude events issued by state agencies, as many of these events
are redundant. Finally, I exclude events directly related to research department activities by deleting 20 observations in which the
allegations mention the word research or contain violations of NASD Code 1050 or NASD Code 2711, which regulate equity research.
Doing so allows me to cleanly attribute the associations between activities in two unrelated divisions of the financial institution to a
common firm-level force such as corporate culture. I measure the weakness of financial institutions’ corporate culture based on the
annual number of disclosure events. The final sample consists of 472 disclosure events issued between 2005 and 2012.
Table 2 Panel B and Appendix A describe the security code violations. Appendix A provides examples of several disclosure events.
In the first example, the firm was sanctioned for violating short-sale regulations around five IPOs. In the second example, the firm was
sanctioned for selling collateralized mortgage obligation securities to unsophisticated small clients.15 Other common examples
(unreported) indicate instances in which financial institutions recommend unsuitable investment products to their clients, mislead
clients, use manipulative sales tactics, and trade ahead of their clients’ orders.
Table 2 Panel B lists the five most frequently occurring security code violations in the sample and indicates several important
statistics. First, the three NASD events listed in this table are also among the top occurring violations as per FINRA's website
(FINRA, 2017c). More importantly, the data indicates that the most frequently occurring violation in the sample involves NASD Rule
2110, which relates to Standards of Commercial Honor and Principles of Trade.16 This rule is commonly referred to as the Business/Bad
Faith Misconduct Rule and it is part of the Duties and Conflicts section of the FINRA manual. This violation is mentioned in 198 (56%) of
the 351 disclosure events that I am able to classify.17 Observing a high frequency of this violation helps validate my claim that violations
are indicative of weak culture, as this violation captures any behavior that FINRA deems “unethical or reflecting bad faith.” 18
Table 2 Panels C through E further describe the security code violations. Panel C presents the frequency of violations by year:
violations peak in 2007 and 2010, indicating that they trend cyclically. Panel D presents the frequency of security code violations by
financial institution, and shows that 86% (25 of 29) of the institutions within the sample receive a violation during the sample period.
Panel E examines the persistence of violations and presents the correlation of violations for up to three lags of violations.19 The
13
In untabulated analyses, I also follow Cowen et al. (2006) and hand-collect business types for each of the financial institutions in my sample
using the 2007 Nelson's Directory of Investment Research. Using their classification system, I find that 24 of the 29 financial institutions in the
sample are classified as full-service investment banks.
14
This is also consistent with Egan et al. (2016). FINRA also appears to have improved the quality of BrokerCheck over time following a review of
the program occurring between 2002 and 2004 (IM-8310-2; FINRA Rule 8312).
15
The allegations section of this report is truncated to preserve space.
16
FINRA Rule 2010 and NASD Rule 2110 both relate to “Standards of Commercial Honor and Principles of Trade.” NASD Rule 2110 has been
superseded by FINRA Rule 2010 after FINRA succeeded NASD.
17
Importantly, Business/Bad Faith Misconduct violations are more prevalent than the above statistic reveals. One challenge with classifying
violations is that self-regulatory organizations have merged over my sample period (e.g., NASD and NYSE merged to form FINRA), leading to
changes in the rulebooks. In untabulated analyses, I create an alternative classification that merges the various rulebooks and find that 79% of
events contain a violation related to the Duties and Conflicts category, which is the parent classification for the Business/Bad Faith Misconduct rule.
18
Specifically, the Business/Bad Faith Misconduct rule requires firms to observe high standards of commercial honor and just and equitable
principles of trade in their business practices. Examples of violations of this rule include instances in which firms make “false, misleading or
exaggerated statements or claims” or “omit material information in advertisements and sales literature directed to the public.”
19
One potential limitation of this data is that the date of the actual violation is rarely referenced. Instead, FINRA only reports the date that the
financial institution is sanctioned. Thus, my measures rely on the sanction date and assume that culture remains essentially unchanged between the
event date and the sanction date.
126
Table 1
Sample of financial institutions. This table lists the sample of Financial Institutions included in the sample. Panel A lists the financial institutions and
their respective security subsidiaries. Panel B describes financial institutions’ business activities as observed in BrokerCheck reports.
Panel A: Sample and subsidiaries
Financial institution Security subsidiary
Allianz SE Commerz Markets LLC

BNP Paribas BNP Paribas Investment Services, LLC
BNP Paribas BNP Paribas Securities Corp.
BPCE Natixis Bleichroeder LLC
BPCE Natixis Securities North America, Inc.
Banco Santander Santander Investment Securities, Inc.
Banco Santander Santander Securities Corp.
Bank of Montreal BMO Capital Markets Corp.
Bank of Nova Scotia Scotia Capital (USA), Inc.
Canadian Imperial Bank of Commerce CIBC World Markets Corp.
Capital One Financial Corp. Capital One Southcoast, Inc.
Cera Ancora VZW KBC Financial Products USA, Inc.
Citigroup, Inc. Citigroup Global Markets, Inc.
Citigroup, Inc. Morgan Stanley Smith Barney, LLC
Credit Suisse Group Credit Suisse Securities (USA), LLC
DZ Bank AG DZ Financial Markets, LLC
Deutsche Bank AG Deutsche Bank Securities, Inc.
DnB NOR ASA DnB NOR Nor Markets, Inc.
Goldman, Sachs Group Epoch Securities, Inc.
Goldman, Sachs Group Goldman Sachs Execution & Clearing, L.P.
Goldman, Sachs Group Goldman Sachs JBWere Inc.
Goldman, Sachs Group Goldman, Sachs and Company
HSBC Holdings PLC Capital Financial Services, INC.
HSBC Holdings PLC HSBC Securities (USA), Inc.
JPMorgan Chase & Co. Chase Investment Services Corp.
JPMorgan Chase & Co. J.P. Morgan Securities, Inc.
Keycorp KeyBanc Capital Markets
Morgan Stanley Morgan Stanley & Co. Incorporated
National Bank of Canada National Bank of Canada Financial, Inc.
Rabobank Nederland Rabo Securities USA, Inc.
Regions Financial Corp. Morgan, Keegan & Company, Inc.
Royal Bank of Canada RBC Capital Markets Corp.
Societe Generale Newedge USA, LLC
Societe Generale SG Americas Securities, LLC
Stifel Financial Corp. Stifel Nicolaus & Company, Inc.
Stifel Financial Corp. Thomas Weisel Partners LLC
SunTrust Banks, Inc. SunTrust Investment Services, Inc.
SunTrust Banks, Inc. SunTrust Robinson Humphrey, Inc.
Toronto-Dominion Bank, The TD Ameritrade Clearing, Inc.
Toronto-Dominion Bank, The TD Ameritrade Inc.
Toronto-Dominion Bank, The TD Securities (USA), LLC
UBS AG UBS Financial Services, Inc.
UBS AG UBS Securities, LLC
Wells Fargo & Company H.D. Vest Investment Securities, Inc.
Wells Fargo & Company Wells Fargo Advisors Financial Network, LLC
Wells Fargo & Company Wells Fargo Advisors, LLC
Wells Fargo & Company Wells Fargo Institutional Securities, LLC
Wells Fargo & Company Wells Fargo Securities, LLC
Panel B: Financial institution business activities
Financial Institution Brokerage Investment advisory Underwriter Other
Allianz SE Yes No Yes Yes

BNP Paribas Yes No Yes Yes
BPCE Yes Yes Yes Yes
Banco Santander Yes Yes Yes Yes
Bank of Montreal Yes Yes Yes Yes
Bank of Nova Scotia Yes No Yes Yes
Canadian Imperial Bank of Commerce Yes No Yes Yes
Capital One Financial Corp. Yes No Yes Yes
Cera Ancora VZW Yes No Yes Yes
Citigroup, Inc. Yes Yes Yes Yes
Credit Suisse Group Yes Yes Yes Yes
DZ Bank AG Yes No Yes Yes
(continued on next page)
127
Table 1 (continued)
Panel B: Financial institution business activities
Financial Institution Brokerage Investment advisory Underwriter Other
Deutsche Bank AG Yes Yes Yes Yes

DnB NOR ASA No No Yes Yes
Goldman, Sachs Group Yes Yes Yes Yes
HSBC Holdings PLC Yes Yes Yes Yes
JPMorgan Chase & Co. Yes Yes Yes Yes
Keycorp Yes No Yes Yes
Morgan Stanley Yes Yes Yes Yes
National Bank of Canada Yes Yes Yes Yes
Rabobank Nederland Yes Yes Yes Yes
Regions Financial Corp. Yes Yes Yes Yes
Royal Bank of Canada Yes Yes Yes Yes
Societe Generale Yes No Yes Yes
Stifel Financial Corp. Yes Yes Yes Yes
SunTrust Banks, Inc. Yes Yes Yes Yes
Toronto-Dominion Bank, The Yes Yes Yes Yes
UBS AG Yes Yes Yes Yes
Wells Fargo & Company Yes Yes Yes Yes
% of Sample Financial Institutions 97% 66% 100% 100%
correlations range from approximately 55% to 66%, confirming that violations are persistent, which is consistent with corporate
culture generally being stable over time (Heskett and Kotter, 1992).
4.3. Empirical models
Baseline tests of the catering hypotheses H1 through H4 employ several datasets. Analyst characteristics and forecasts are ob-
tained from I/B/E/S. Returns data are obtained from CRSP and firm fundamentals are obtained from Compustat. Tests examining
broker-hosted conferences (H3) rely on conference data obtained from the Bloomberg Corporate Events database. Tests examining
commission revenues (H4) utilize data obtained from Ancerno. I discuss the models employed to test each hypothesis in turn. All
variables are described in Appendix B.
4.3.1. Tests examining forecast accuracy

H1 tests the association between violations (i.e., my proxy for weak corporate culture) and forecast accuracy. Following
Clement (1999), I construct measures of relative absolute forecast error as follows:
AFEijt ¯ jt
AFE
RAFErrorijt = ,
¯
AFEjt
where AFEijt is the absolute forecast error for analyst i’s forecast for firm j in year t and AFE
¯ jt is the mean absolute forecast error for
firm j in year t across all analysts providing forecasts for the firm in the sample. Following prior studies, forecast error is calculated
using the last forecast issued in the first eleven months of the fiscal year, thus reducing the sample to an analyst-firm-year panel. By
construction, RAFError partially reflects analysts’ objectivity as it controls for important firm-year differences, the characteristics of
covered firms’ information environment, within the sample. I multiply RAFError by minus one and create a variable RAccuracy:
higher values of this variable indicate more accurate forecasts.
To test H1, I employ the following regression model:
RAccuracyijt = 0 + 1 Violationsft + 2 Horizon ijt + 3 AnalystControlsijt + 4 FIControlsft + Yeart + ijt ,

t (1)
where i denotes analyst, j denotes firm, t denotes time, and f denotes the financial institution (for which analyst i is employed in year
t). The proxy for weak corporate culture, Violations, is measured as the natural log of one plus the number of disclosure events a
financial institution receives in a year. Horizon is the number of days between the forecast date and the fiscal period end.
AnalystControls is a vector that includes analyst characteristics relating to forecast accuracy (Clement, 1999; Jacob et al., 1999).
Experience is the relative firm-specific forecast experience of the analyst providing the forecast.20 Coverage is the approximate number
of firms covered by the analyst. To control for differences across firm-years, Horizon, Experience, and Coverage are relative to the firm-
year and are constructed similarly to RAFError by differencing out and scaling by the firm-year mean of each measure.
The model also includes a vector of financial institutional characteristics (FIControls) that may influence the quality of analysts’
forecasts. FISize is the relative number of analysts employed at the financial institution. FIBusinessLines is the total number of business
20
Inferences remain unchanged if I measure experience based on general as opposed to firm-specific experience, as well as if I augment the tests
with both measures of experience.
128
Table 2
Characteristics of security code violations. This table describes the security code violations obtained from the FINRA
BrokerCheck reports. Panel A provides the sample selection process. Panel B presents the most frequently occurring security
code violations in the sample. Panel C presents frequencies by year. Panel D presents frequencies by financial institution.
Panel E presents autocorrelation coefficients for security code violations for up to 3 prior years.
Panel A: Sample selection of security code violations
Completed Disclosure Events for Financial Institutions with available I/B/E/S Data (2005–2012) 1,448
Less: Events with missing (or duplicate) case numbers (235)
Less: Events with no fines indicated (56)
Less: Events issued by state agencies (665)
Less: Events containing research violation (20)
Final Sample 472
Panel B: Top 5 security code violations
Security Code Description # of Violations % of Violations
NASD Rule 2110 Standards of Commercial Honor and Principles of Trade 198 56%
NASD Rule 3010 Supervision of Employees 147 42%
NASD Rule 6955 Order Data Transmission Requirements 58 17%
FINRA Rule 2010 Standards of Commercial Honor and Principles of Trade 55 16%
NYSE Rule 342 Offices - Approval, Supervision and Control 49 14%
Panel C: Frequency by year
Year Violations
2005 51
2006 47
2007 69
2008 48
2009 56
2010 77
2011 59
2012 65
Total 472
Panel D: Frequency by financial institution
Financial Institution Violations
Allianz SE 4
BNP Paribas 11
BPCE 8
Banco Santander 3
Bank of Montreal 9
Bank of Nova Scotia 0
Canadian Imperial Bank of Commerce 14
Capital One Financial Corp. 1
Cera Ancora VZW 4
Citigroup, Inc. 48
Credit Suisse Group 21
DZ Bank AG 0
Deutsche Bank AG 28
DnB NOR ASA 0
Goldman, Sachs Group 62
HSBC Holdings PLC 21
JPMorgan Chase & Co. 34
Keycorp 13
Morgan Stanley 34
National Bank of Canada 4
Rabobank Nederland 0
Regions Financial Corp. 10
Royal Bank of Canada 25
Societe Generale 18
Stifel Financial Corp. 13
SunTrust Banks, Inc. 12
Toronto-Dominion Bank, The 22
129
Table 2 (continued)
Panel D: Frequency by financial institution
Financial Institution Violations
UBS AG 50
Wells Fargo & Company 3
Total 472
Panel E: Persistence of violations
Violationst
Violationst-1 0.55
Violationst-2 0.62
Violationst-3 0.66
lines in the financial institution (as observed in the FINRA BrokerCheck report). FIPrestige, a proxy for reputation, is an indicator
variable that takes the value of one if the financial institution is one of the top five Institutional Investor Ranked financial institutions,
and zero otherwise (as measured in 2012). All models include year-fixed effects. Continuous variables are winsorized at the 1st and
99th percentiles. H1 predicts α1 < 0.
4.3.2. Tests examining report informativeness
H2 examines the relation between violations and return informativeness using the following regression model:
Informativenessijt = 0 + 1 Violationsft + 2 AnalystControlsijt + 3 FIControlsft
+ 4 FirmControlsjt + 5 MktControlsjt + Firmj + Yeart + ijt

j t (2)
where i denotes analyst, j denotes firm, t denotes time, and f denotes the financial institution (for which analyst i is employed in year
t). Informativeness is calculated as the absolute value of abnormal returns measured over the 3-day period centered on all analyst
report dates (i.e., forecast, recommendation, or target price revision dates). Abnormal returns are adjusted from the portfolio return
from 125 benchmark portfolios based on size, book-to-market, and momentum (5 × 5 × 5) using the procedure outlined in
Daniel et al., (1997). I retain all analyst report dates based on an analyst issuing an earnings forecast, recommendation or target price
on I/B/E/S. Violations, AnalystControls and FIControls are as defined in the prior tests. To control for heterogeneity across covered
firms, the vector FirmControls includes variables LogAssets (the natural log of assets), LogMTB (the natural log of the market-to-book
ratio), and InstHold (percent of outstanding shares held by institutional investors).21 The model also controls for market conditions
through the inclusion of four variables: MktMomentum, MktVolatility, FirmMomentum, and FirmVolatility. MktMomentum and Firm-
Momentum control for momentum and are constructed as cumulative monthly returns over the prior six months for the market and
firm, respectively. MktVolatility and FirmVolatility control for volatility and are measured as the standard deviation of returns over the
prior six months for the market and firm, respectively. The model includes time-fixed effects and varies the inclusion of firm-fixed
effects to control for unobserved heterogeneity across covered firms. H2 predicts β1 < 0.
4.3.3. Tests examining institutional product quality
H3 and H4 examine the relationship between violations and institutional product quality using the following regression model:
InstitutionalProductijt = 0 + 1 Violationsft + 2 AnalystControlsijt + 3 FIControlsft
+ 4 FirmControlsjt + Firmj + Yeart + ijt ,

j t (3)
where i denotes analyst, j denotes firm, t denotes year, and f denotes the financial institution (for which analyst i is employed in year t).
InstitutionalProduct is one of two measures: Conferences or Commissions. Conferences is the number of broker-hosted investor conferences that
the analyst's employer hosts. Commissions is the natural log of the dollar value of trading commissions routed through an analyst's firm. This
data is obtained from Ancerno, formally Abel-Noser, a consulting firm that tracks institutional clients’ trades and commissions. Each ob-
servation in the Ancerno dataset corresponds to a trade and indicates the stock traded, the commission paid to the broker, and the broker
21
Controlling for these covered firm controls in Model 1 does not affect my inferences.
130
Table 3
Descriptive statistics. This table presents descriptive statistics for the variables used in the various samples testing the catering hypotheses. Panel A
presents descriptive statistics for the dependent variables. Panel B presents descriptive statistics that are calculated based on the 79,287 analyst-
firm-year observations in the RAccuracy regressions for all variables except MktMomentum, MktVolatility, FirmMomentum, and FirmVolatility, which
are calculated based on the 259,014 analyst-firm-day observations used in the Informativeness regressions. All variables are defined in Appendix B.
Panel A: Dependent variables
Variable N Mean Std. Dev. 25th Median 75th
Dependent Variables
RAccuracy 79,287 0.01 0.77 −0.27 0.13 0.54
Informativeness 259,014 0.03 0.03 0.01 0.02 0.04
Conferences 77,207 0.27 0.5 0 0 0
Commissions 65,746 41,365 157,645 1,950 9,228 33,676
Panel B: Independent variables
Variable Mean Std. Dev. 25th Median 75th
Violations 3.81 2.80 2 3 6

Horizon 94.73 70.97 57 66 82
Experience 3.45 3.72 1 2 5
Coverage 17.13 10.21 11 16 21
FISize 108.15 43.92 83 115 132
FIBusinessLines 20.77 4.44 20 21 24
FIPrestige 0.37 0.48 0 0 1
Assets 17,353 44,155 1,142 3,622 12,841
MTB 2.94 3.77 1.34 2.17 3.58
InstHold 0.74 0.21 0.63 0.78 0.90
MktMomentum 0.03 0.15 −0.03 0.05 0.1
MktVolatility 0.04 0.02 0.03 0.04 0.05
FirmMomentum 0.01 0.24 −0.13 −0.01 0.13
FirmVolatility 0.09 0.05 0.05 0.07 0.11
responsible for facilitating the transaction. Other variables are as defined above. The model includes firm-fixed effects to control for un-
observed time-invariant heterogeneity across covered firms. H3 and H4 predict positive coefficients on γ1.
4.4. Descriptive statistics
Table 3 provides descriptive statistics for the variables used in Equations 1 through 3. Panel A provides descriptive statistics for
the dependent variables used in the catering analyses. Sample sizes vary due to the unit-of-observation and the availability of data.
The mean (median) values for RAccuracy is 0.01 (0.13) and the mean (median) absolute return reaction to analyst reports (In-
formativeness) is 3% (2%). The average covered firm generates $41,635 in commission revenue for an analyst's broker and is involved
in 0.27 conferences.22 Panel B provides sample statistics for the independent variables used in the study. The mean (median) number
of unlogged violations is 3.81 (3.00). Other variables are generally consistent with prior studies. For example, analysts have ap-
proximately 3 years of experience and cover approximately 17 firms. Firms in the sample tend to be large and diverse: the average
firm employs approximately 108 analysts and operates across nearly 21 business lines.
5. Results
5.1. Forecast accuracy
Table 4 provides the results from estimates of Eq. 1, which tests the relation between corporate culture and forecast accuracy.
Panel A presents the results for the one-year ahead earnings forecast, which represents one of the most frequently issued analyst
outputs. Column 1 presents the univariate results, and Column 2 includes all the controls specified in Eq. 1. Consistent with H1, the
results indicate negative and statistically significant associations between Violations and RAccuracy (p < 0.01), indicating that weak
corporate culture is associated with less accurate forecasts.23 The loadings on control variables are also generally consistent with
prior research: longer horizon forecasts are less accurate as evidenced by the negative coefficient on Horizon, while forecasts made by
22
Some firms have more than one conference per year. The results are also robust to using an indicator variable for conference activity.
23
The univariate model explains little variation in accuracy as it tests how a time-varying bank characteristic (Xbt) influences an individual
analyst's forecast for a given firm in a particular year (Yibxt), where i indexes analyst, b indexes bank, x indexes firm, and t indexes time. Covariates
included in the multivariate models (such as horizon) can naturally explain more variation in the outcome variable than measures of culture since
they are measured at a more granular level, i.e., the analyst-firm-level (Xixt) as opposed to the bank-year level (Xbt). Consistent with this explanation,
prior studies also provide evidence indicating that bank characteristics explain little variation in forecast properties (Cowen et al., 2006)
131
Table 4
Security code violations & forecast accuracy. This table provides the results of OLS regressions of RAccuracy on weak corporate culture, as measured
by security code violations. Panel A presents the results when accuracy is measured using one-year ahead forecasts. Panel B presents the results for
two-year, three-year, one-quarter, two-quarter, three-quarter, and four-quarter ahead forecasts. Standard errors are clustered by financial institu-
tion-year. ***, **, and * denote 1%, 5% and 10% level of significance, respectively. All variables are defined in Appendix B.
Panel A: Annual earnings forecast accuracy
(1) (2)
VARIABLES RAccuracy RAccuracy
Violations −0.0475*** −0.0457***

(−5.47) (−5.68)
Horizon −0.5753***
(−68.61)
Experience 0.0125***
(3.43)
Coverage −0.0072
(−0.45)
FISize 0.0113
(0.77)
FIBusinessLines 0.0338
(1.32)
FIPrestige −0.0177
(−1.42)
Year Fixed Effects? Yes Yes

Observations 79,287 79,287
R-squared 0.0017 0.1707
Panel B: Earnings forecast accuracy at other horizons
(1) (2) (3) (4) (5) (6)

VARIABLES RAccuracy2Yr RAccuracy3Yr RAccuracy1Qtr RAccuracy2Qtr RAccuracy3Qtr RAccuracy4Qtr
Violations −0.0108** −0.0145** −0.0143** −0.0093** −0.0113* −0.0039

(−2.12) (−2.25) (−2.57) (−1.98) (−1.92) (−0.71)
Horizon −0.6398*** −0.5905*** −0.0168 −0.1351*** −0.1441*** −0.2470***
(−26.82) (−15.06) (−0.82) (−9.61) (−6.25) (−7.33)
Experience 0.0068** −0.0019 0.0149*** 0.0072** 0.0008 −0.0053*
(2.42) (−0.54) (4.15) (2.49) (0.25) (−1.84)
Coverage −0.0023 −0.0256*** −0.0333*** 0.0206*** 0.0072 0.0084
(−0.22) (−2.69) (−2.63) (2.76) (0.98) (1.43)
FISize −0.0067 −0.0101 −0.0151 −0.0059 −0.0073 −0.0111
(−0.74) (−0.89) (−1.48) (−0.67) (−0.82) (−1.19)
FIBusinessLines 0.0197 0.0195 0.0227 0.0089 0.0187 0.0078
(1.19) (1.19) (1.45) (0.64) (1.00) (0.54)
FIPrestige −0.0100 0.0106 −0.0071 0.0046 0.0055 0.0015
(−1.39) (1.23) (−0.74) (0.64) (0.65) (0.18)
Year Fixed Effects? Yes Yes Yes Yes Yes Yes

Observations 66,274 43,171 74,359 75,609 72,639 67,136
R-squared 0.0387 0.0107 0.0011 0.0024 0.0009 0.0013
more experienced analysts are more accurate as evidenced by the positive coefficient on Experience.
Table 4 Panel B expands the forecast accuracy analysis to examine forecasts of alternative horizons. Columns 1 and 2 present the
results when accuracy is measured based on annual two-year ahead and three-year ahead forecasts, respectively. Columns 3 through
6 present the results when accuracy is measured based on quarterly earnings forecasts up to the four-quarter-ahead forecast. With the
sole exception of the fourth-quarter ahead earnings forecast, the results continue to indicate negative and statistically significant
associations between Violations and RAccuracy. The control variables also generally load consistently with prior studies. For example,
Horizon continues to generate negative loadings in most of the models, indicating that forecasts issued further away from the fiscal
period end are less accurate. One exception is Coverage, which loads positively and significantly for two-quarter ahead forecasts.
5.2. Report informativeness
Table 5 provides the results from estimates of Eq. 2 which tests the relation between corporate culture and report informativeness.
Column 1 presents the univariate results. The results indicate a negative and significant loading on Violations, which supports H2. In
Column 2, I include the set of analyst, financial institution, firm, and market controls specified in Eq. 2 and generate similar
132
Table 5
Security code violations & return informativeness. This table provides the results of OLS regressions of Informativeness on weak corporate
culture, as measured by security code violations. Standard errors are clustered by financial institution-year. ***, **, and * denote 1%, 5%
and 10% level of significance, respectively. All variables are defined in Appendix B.
(1) (2) (3)
VARIABLES Informativeness Informativeness Informativeness
Violations −0.0032*** −0.0007*** −0.0004**

(−15.41) (−3.44) (−2.16)
Experience 0.0003*** 0.0003***
(4.24) (3.56)
Coverage −0.0002 0.0002
(−0.88) (1.10)
FISize −0.0004 −0.0001
(−1.31) (−0.43)
FIBusinessLines 0.0015*** 0.0009*
(3.00) (1.72)
FIPrestige −0.0002 −0.0003
(−0.95) (−1.15)
LogAssets −0.0029*** −0.0041***
(−55.71) (−13.55)
LogMTB −0.0030*** −0.0034***
(−19.26) (−9.72)
InstHold −0.0048*** −0.0118***
(−10.83) (−10.93)
MktMomentum −0.0204*** −0.0199***
(−21.72) (−21.44)
MktVolatility 0.0379*** 0.0699***
(4.18) (7.83)
FirmMomentum −0.0090*** −0.0070***
(−22.55) (−17.28)
FirmVolatility 0.1488*** 0.0799***
(71.71) (32.77)
Firm Fixed Effects? No No Yes

Year Fixed Effects? No Yes Yes
Observations 259,014 259,014 259,014
R-squared 0.0041 0.1578 0.2037
inferences. Column 3 further controls for firm fixed effects. The results continue to indicate a negative and significant coefficient on
Violations (p < 0.05), suggesting that unobserved time invariant firm heterogeneity does not unduly influence my inferences. Other
variables generally load consistently with prior studies. The variable Experience loads positively and significantly, which is consistent
with the notion that more experienced analysts produce, on average, higher quality research (Clement, 1999). Consistent with
Kadan et al., (2009), higher levels of firm and market momentum are also associated with less pronounced market reactions while
higher levels of firm and market volatility are associated with more pronounced market reactions.
Overall, the results from the report informativeness analysis provide support for H2. Taken together with the results from Table 4,
the findings suggest that analysts employed by weak culture firms produce lower quality written research, as evidenced by less accurate
forecasts and less informative reports (i.e., lower absolute return reactions for forecasts, target prices, and recommendation revisions).
5.3. Institutional product quality
Table 6 provides the results from estimates of Eq. 3, which tests the relation between corporate culture and institutional product
quality. Panel A presents the results for broker-hosted conferences, and Panel B presents those for commission revenue. In each panel,
Column 1 presents the univariate results, Columns 2 and 3 vary the inclusion of controls and firm fixed effects, and Columns 4 and 5
test the sensitivity of the models to alternative specifications. Column 4 includes firm-year fixed effects to control for unobservable
time-varying heterogeneity of covered firms. Column 5 examines a relative quality measure constructed by demeaning conferences or
commissions by the firm-year average. Across all tests, the results indicate positive and significant associations between Violations and
both Conferences (Panel A) and Commissions (Panel B). The sensitivity analyses in Columns 4 and 5 further indicate that the models do
not appear to be unduly influenced by unobservable time-varying heterogeneity of covered firms not captured by the covered firm
characteristics or firm fixed effects.24 The loadings on the control variables indicate that larger, more prestigious firms are more likely
24
For example, results from Column 4 indicates that the models offer relatively high explanatory power (i.e., R-squared of 47% in Panel A and R-
squared of 65% in Panel B). The coefficient estimates on Violations also exhibit small changes in magnitude moving from the model controlling for
firm characteristics (Column 2) to the those controlling for firm and firm-year fixed effects (Columns 3 and 4).
133
Table 6
Security code violations & institutional product quality. This table examines the relationship between security code violations and institutional
product quality. Panel A provides the results of OLS regressions of Conferences on weak corporate culture, as measured by security code violations.
Panel B provides the results of OLS regressions of Commissions on security code violations. Standard errors are clustered by financial institution-year.
***, **, and * denote 1%, 5% and 10% level of significance, respectively. All variables are defined in Appendix B.
Panel A: Security code violations & conferences
(1) (2) (3) (4) (5)

VARIABLES Conferences Conferences Conferences Conferences RConferences
Violations 0.0623*** 0.0381*** 0.0345*** 0.0377*** 0.0323***

(4.84) (3.27) (3.12) (2.95) (3.16)
Experience 0.0304*** 0.0306*** 0.0307*** 0.0295***
(8.44) (9.64) (9.14) (10.17)
Coverage 0.0098 0.0087 0.0082 0.0091
(0.90) (1.16) (1.06) (1.35)
FISize 0.1652*** 0.1804*** 0.1820*** 0.1766***
(7.90) (9.08) (8.49) (9.98)
FIBusinessLines −0.0652 −0.1223*** −0.1273*** −0.1128***
(−1.48) (−3.36) (−3.25) (−3.48)
FIPrestige 0.0285 0.0340* 0.0299 0.0277
(1.49) (1.80) (1.46) (1.60)
LogAssets −0.0233*** 0.0211*** −0.0009
(−7.12) (2.75) (−0.13)
LogMTB 0.0486*** 0.0255*** −0.0001
(7.78) (2.73) (−0.01)
InstHold 0.2636*** 0.0902*** 0.0014
(18.36) (5.04) (0.09)
Firm Fixed Effects? No No Yes No Yes
Year Fixed Effects? No Yes Yes No Yes
Firm-Year Fixed Effects? No No No Yes No
Observations 77,207 77,207 77,207 77,207 77,207

R-squared 0.0070 0.0504 0.3285 0.4744 0.0391
Panel B: Security code violations & commission revenues
(1) (2) (3) (4) (5)

VARIABLES Commissions Commissions Commissions Commissions RCommissions
Violations 0.6778*** 0.3530*** 0.3570*** 0.3771** 0.1302**

(3.90) (2.85) (2.65) (2.51) (2.01)
Experience 0.0112 0.0131 0.0135 0.0209**
(0.58) (0.76) (0.76) (2.42)
Coverage 0.1656*** 0.1606*** 0.1604*** 0.0757***
(3.81) (3.98) (3.86) (4.26)
FISize 1.5412*** 1.4507*** 1.4423*** 0.5365***
(8.61) (6.84) (6.18) (5.63)
FIBusinessLines −0.3400 −0.1179 −0.0958 −0.0350
(−0.87) (−0.28) (−0.21) (−0.21)
FIPrestige 0.3208** 0.3439** 0.3248* 0.1024
(1.98) (1.99) (1.77) (1.21)
LogAssets 0.5462*** 0.4880*** −0.0032
(27.28) (12.57) (−0.17)
LogMTB 0.5532*** 0.1820*** 0.0009
(18.21) (4.72) (0.06)
InstHold 2.4757*** 1.7481*** 0.0232
(25.10) (14.12) (0.40)
Firm Fixed Effects? No No Yes No Yes

Year Fixed Effects? No Yes Yes No Yes
Firm-Year Fixed Effects? No No No Yes No
Observations 65,746 65,746 65,746 65,746 65,746

R-squared 0.0341 0.3019 0.5161 0.6473 0.0716
to provide specialized research services, and that these services are more likely to be provided for firms with higher market-to-book
ratios and firms with high institutional holdings. Interestingly, more complex institutions appear to provide fewer conferences, as
evidenced by the negative loading on FIBusinessLines in the Conferences regressions (Panel A). This is potentially due to diversified
financial institutions investing less in hosting conferences as they have a wider array of products to offer.
134
Overall, the results from the analyses thus far provide support for H1-H4. Analysts employed by weak corporate culture financial
institutions issue lower quality written research as evidenced by the negative and significant associations between Violations and
RAccuracy and Informativeness. At the same time, these analysts also appear to provide a higher quality institutional product as
evidenced by the positive and significant associations between Violations and Conferences and Commissions. Taken together, these
results indicate that analysts employed by weak culture firms issue a more catered research product.
5.4. Robustness
I conduct a battery of robustness tests to further test my claim that weak corporate culture is associated with higher levels of
catering. Table 7 provides the results from these analyses, with each panel considering a different concern. In each panel, Column 1
presents the robustness results for RAccuracy (i.e., estimates of Eq. 1), Column 2 presents the results for Informativeness (i.e., estimates
of Eq. 2), and Columns 3 and 4 present the results for Conferences and Commissions, respectively (i.e., estimates of Eq. 3). Sample sizes
vary across these tests due to varying availability of data.
In Panel A, I consider the effects of other dimensions of corporate culture, using data collected from Glassdoor.com and the KLD
Corporate Responsibility Index. GreatPlaceToWork is an indicator variable that takes the value of one if a financial institution's overall
Glassdoor ranking is above the sample median, and zero otherwise.25 PoorProductQuality is an indicator variable that takes the value
of one if KLD surveys indicate that a financial institution has net product weaknesses that exceed its strengths, and zero otherwise.
The results suggest that controlling for GreatPlaceToWork and PoorProductQuality does not affect the relation between culture and
analysts’ catering. Moreover, the results also provide some evidence to suggest that firms with high levels of employee satisfaction
produce lower quality written research but higher quality institutional products. The coefficient on GreatPlaceToWork is negative and
significant in the Informativeness test (Column 2) and positive and significant in the Commissions test (Column 4). This finding is
consistent with the anecdote presented by Guiso et al. (2015) which suggests Goldman Sachs has both a high level of employee
satisfaction but a weak culture, as per former executive director Greg Smith's memoir describing the deteriorating culture at Goldman
(Smith, 2012). As discussed in Section 3, employee satisfaction surveys may not adequately capture misconduct.
In Panel B, I control for analysts’ cultural background, given recent evidence suggesting that ancestry influences economic behavior and
corruption (DeBacker et al., 2015; Liu, 2016; Brochet et al., 2016; Merkley et al., 2018). I map analysts’ surnames back to their country of
origin using data collected from three ancestral dictionaries: Oxford University Press's Dictionary of American Family Names, Forebears’
genealogical records, and Ancestry.com. Analysts with available countries of origin are then mapped into one of ten culture clusters following
House et al., (2002).26 I re-examine the baseline analyses after including fixed effects for each of these ten cultural clusters and find that my
inferences remain unchanged. Thus, my findings are unlikely to be influenced by heterogeneity in analysts’ cultural origins.
I next consider whether rogue employee behavior explains my findings by making two modifications to the sample. First, I remove
any violations that do not directly indicate that the financial institution is at fault.27 Second, to reduce the influence of rogue analysts,
I collect and analyze individual-level BrokerCheck reports (when available) for analysts in my sample and confirm that these analysts
do not have prior regulatory infractions.28 Panel C reexamines the baseline findings after introducing these two modifications. I retain
only analysts whom I can confirm do not have a prior infraction and construct a new variable, CompanyViolations, based on only
FINRA reports that confirm an allegation against the firm. The results indicate that the relation between security code violations and
research quality persists within this subsample, suggesting that rogue employees do not appear to drive my findings.
I also consider whether heterogeneity in financial institutions’ business models explains my findings. In Panel D, I control for
differences in financial institutions’ client-base, by augmenting the model with measures for the importance of start-up clients
(PercentStartup) and high growth clients (HighGrowth) to the firm. My results continue to hold in this test, indicating that the im-
portance of different clientele does not appear to explain my findings.29
Finally, I validate the catering measures using a case study related to the Goldman Sachs “research huddle” regulatory event.
From 2006 to 2011, Goldman Sachs formally implemented a program known as the “Asymmetric Service Initiative”, in which the
firm selectively provided institutional clients with information not available to its broader client-base. This case represents the only
instance in the sample in which catering was detected and sanctioned by a regulator. In Panel E, I create an indicator variable for
Goldman Sachs and examine the differential accuracy, informativeness, and institutional product quality of Goldman Sachs compared
25
Garrett et al. (2014) and Guiso et al. (2015) use data obtained from the Great Place to Work Institute. Since this data is not publicly available, I
use Glassdoor overall job satisfaction as an alternative proxy for employee satisfaction.
26
These clusters include Confucian Asia, Southern Asia, Latin America, Nordic Europe, Anglo, Germanic Europe, Latin Europe, Sub-Saharan
Africa, Eastern Europe and Middle East.
27
Specifically, I require the allegations to disclose the firm's name (as opposed to an individual) or to disclose a supervisory failure (e.g., NASD
Rule 3010). Since my primary data is based on firm-level violations, 97% of violations indicate a supervisory failure suggesting the low probability
that a rogue employee caused the infraction.
28
These analyses are based on a reduced sample of 755 analysts to whom I am able to match individual FINRA BrokerCheck reports. Analysts are
matched to BrokerCheck reports based on the last name, first initial, and employer name recorded in I/B/E/S. To ensure an accurate match, I require
the analyst to list the Series 86 or 87 Research Analyst Examination on his or her FINRA profile.
29
In untabulated analyses, I also consider heterogeneity in bank type (Cowen et al., 2006) and firm location which may relate to how far a firm is
located from a regulator (Kedia and Rajgopal, 2011; Kubick et al., 2016). The vast majority of the sample banks are full-service investment banks,
and I limit my analysis to these banks. In addition, my results also persist when I control for country-fixed effects or subsample on only firms
headquartered in the United States, suggesting that distance from a regulator does not influence my findings.
135
Table 7
Robustness tests. This table presents robustness tests for the catering hypotheses. In each Panel, Column 1 provides the results for OLS regressions of
RAccuracy on Violations, Column 2 provides results for regressions of Informativeness on Violations, Column 3 provides results from regressions of
Conferences on Violations, and Column 4 provides results from regressions of Commissions on Violations. Panel A controls for other corporate culture
effects. Panel B adds controls for analyst ethnicity. Panel C removes rogue employees from the sample. Panel D controls for financial institution
client type. Panel E examines an event study based on the Goldman Sachs Huddle Case. Standard errors are clustered by financial institution-year.
***, **, and * denote 1%, 5% and 10% level of significance, respectively. All variables are defined in Appendix B.
Panel A: Control for other corporate culture effects
Retail product quality Institutional product quality

(1) (2) (3) (4)
VARIABLES RAccuracy Informativeness Conferences Commissions
Violations −0.0573*** −0.0005* 0.0374*** 0.3169***

(−6.78) (−1.85) (3.13) (3.00)
GreatPlaceToWork 0.0169 −0.0018*** −0.0026 0.3523***
(1.14) (−4.96) (−0.16) (3.14)
PoorProductQuality −0.0295* −0.0009** 0.0617** 1.1180***
(−1.76) (−1.97) (2.02) (4.06)
Baseline Controls? Yes Yes Yes Yes

Observations 39,527 125,721 43,551 32,812
R-squared 0.1724 0.2040 0.3940 0.6072
Panel B: Control for analysts' cultural backgrounds

(1) (2) (3) (4)
Violations −0.0468*** −0.0004* 0.0357*** 0.3529***

(−5.23) (−1.88) (3.18) (2.61)

Country of Origin FE? Yes Yes Yes Yes
Observations 76,143 252,514 74,436 62,905
R-squared 0.1676 0.2041 0.3293 0.5173
Panel C: Remove Rogue Employees

(1) (2) (3) (4)
CompanyViolations −0.0413*** −0.0006** 0.0445*** 0.3827***

(−3.94) (−2.44) (3.52) (2.91)

Observations 30,499 102,694 30,831 26,064
R-squared 0.1787 0.2104 0.3793 0.5638
Panel D: Control for Financial Institution Client Type
Retail Product Quality Institutional Product Quality

(1) (2) (3) (4)
Violations −0.0449*** −0.0004** 0.0297*** 0.3086**

(−5.33) (−2.09) (2.71) (2.27)
Percent Startup −0.1857 0.0016 −0.0980 −9.4936***
(−1.57) (0.43) (−0.56) (−3.33)
High Growth −0.0101 −0.0000 0.0413 0.4128**
(−0.58) (−0.10) (1.38) (2.25)

Observations 79,287 259,014 77,207 65,746
R-squared 0.1708 0.2037 0.3289 0.5222
136
Table 7 (continued)
Panel E: Event Study Using Goldman Sachs Huddle Case (2006−2011)
Retail Product Quality Institutional Product Quality

(1) (2) (3) (4)
GoldmanSachs −0.1288*** −0.0012** 0.0780*** 0.6573***

(−9.80) (−2.49) (3.17) (2.80)

Observations 49,402 162,061 48,225 38,741
R-squared 0.1676 0.2127 0.3585 0.4942
to its peers between 2006 and 2011. The results suggest that Goldman provided less accurate and less informative written research,
but higher quality institutional products during this period, thus providing additional support for the measures I use to test catering.30
Overall, the robustness analyses provide additional support for the catering hypotheses. The findings indicate that the relation
between corporate culture and analysts’ catering activity is not easily explained by other cultural forces, rogue employee behavior, or
heterogeneity in firms’ business models and locations.
6. Additional analysis examining analyst report bias
The results thus far have established a robust relation between corporate culture and analysts’ catering behavior. In this section, I
further explore the relation between corporate culture and analyst report bias (i.e., optimism bias reflected in earnings forecasts,
recommendations, or target prices). Prior studies generally provide evidence that analysts may optimistically bias their research in order
to please prospective or recent investment banking clients (Lin and McNichols, 1998; Michaely and Womack, 1999; Bradshaw et al.,
2006), curry favor with management (Francis and Philbrick, 1993; Ke and Yu, 2006), or generate trading revenue (Jackson, 2005;
Cowen et al., 2006). However, the Global Settlement and related regulatory reforms eliminated many of the conflicts of interest
associated with analyst bias, thus reducing the extent to which analysts optimistically bias their reports (Barber et al., 2006; Kadan et al.,
2009; Clarke et al., 2011). The heightened regulatory attention placed on bias may have also created incentives for some brokerage
firms to be particularly careful to avoid the appearance of being optimistically biased. Weak culture firms are likely to be particularly
vigilant about avoiding the appearance of being optimistically biased if conflicts of interest compromise research quality in other ways.
To examine the relation between corporate culture and bias, I consider the following models of analyst report bias, based on
analysts’ earnings forecasts, target prices, and recommendations:
BiasEForecastijt = 0 + 1 Violationsft + 2 ForecastHorizon ijt + 3 AnalystControlsijt
+ 4 FIControlsft + Yeart + ijt

t (4a)
BiasLTGijt or BiasTPijt = 0 + 1 Violationsft + 2 AnalystControlsijt + 3 FIControlsft + Yeart + ijt

t (4b)
Recijt = 0 + 1 Violationsft + 2 AnalystControlsijt + 3 FIControlsft + 4 TotalBuysjt
+ 5 TotalHoldsjt + 6 TotalSellsjt + Yeart + ijt

t (4c)
where i denotes analyst, j denotes firm, t denotes year, and f denotes the financial institution (for which analyst i is employed in year
t). BiasEForecast is earnings forecast bias, measured as analyst i’s forecast error for firm j in year t, less the mean forecast error for all
analysts covering firm j in year t, scaled by the absolute value of the mean forecast error. BiasLTG and BiasTP are the long term growth
and target price biases, respectively. These variables are constructed by demeaning the analyst's forecast by the average forecast
across all analysts providing forecasts, and scaling by the standard deviation of all forecasts. Rec is recommendation bias, coded one
for sell recommendations (sell or strong-sell), two for hold recommendations and three for buy recommendations (buy or strong-buy).
The earnings forecast models (Eq. 4a) consider both annual and quarterly forecasts and control for horizon. The recommendations
models control for the total buy, hold and sell recommendations issued by analysts covering the firm. Eqs. (4a) and (4b) are im-
plemented using OLS, whereas Eq. (4c) uses an ordered probit model (Cowen et al., 2006). All other variables are as defined in Eq. 1.
Higher levels of BiasEForecast, BiasLTG, BiasTP, and Rec indicate more optimism bias in the respective analyst output.
Table 8 provides the results from estimates of these models. Columns 1 through 3 provide the results for annual forecasts up to three-
years ahead. Columns 4 through 7 provide the results for quarterly forecasts up to four-quarters ahead. Columns 8, 9, and 10 provide the
results for long-term growth forecasts, target price forecasts, and recommendations, respectively. Overall, the results indicate a negative
30
In untabulated analyses, I also include Violations in this regression and find qualitatively similar results. This suggests that the relationship
between culture and product quality during this time period is not unduly influenced by Goldman Sachs.
137
Table 8
Security code violations & analyst report bias. This table presents results from tests examining the relationship between security code violations and
analyst report bias. Columns 1 through 7 provide the results from OLS regressions of earnings forecast bias at various horizons on weak corporate
culture, as measured by Violations. Columns 8 and 9 provide results from regressions of long-term growth and target price bias on Violations,
respectively. Column 10 provides the results from an ordered probit regression of recommendations on Violations. Standard errors are clustered by
financial institution-year. ***, **, and * denote 1%, 5% and 10% level of significance, respectively. All variables are defined in Appendix B.
ForecastBias Other Research Outputs
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
VARIABLES Bias1Yr Bias2Yr Bias3Yr Bias1Qtr Bias2Qtr Bias3Qtr Bias4Qtr BiasLTG BiasTP Rec
Violations −0.0538* −0.0254 −0.0390 −0.0361 −0.0591* −0.0321 −0.0679** −0.0709* −0.0546** −0.0984***
(−1.69) (−1.12) (−1.17) (−1.30) (−1.87) (−0.94) (−2.01) (−1.96) (−2.09) (−2.75)
Horizon 0.3259*** 0.9993*** 0.9047*** 0.4530*** 0.4323*** 0.4715*** 0.7084*** N/A N/A N/A
(5.73) (6.93) (4.61) (3.18) (5.09) (3.89) (3.98) (N/A) (N/A) (N/A)
Experience −0.0181 0.0014 0.0137 −0.0171 −0.0538*** −0.0107 0.0121 0.0369*** 0.0434*** 0.0148*
(−0.92) (0.09) (0.85) (−1.00) (−2.82) (−0.61) (0.62) (3.15) (5.15) (1.86)
Coverage 0.0892 0.0336 0.1261*** 0.1001** 0.0375 0.0275 0.0477 0.0134 −0.0032 −0.0418*
(1.60) (0.72) (2.64) (1.99) (0.78) (0.70) (1.33) (0.40) (−0.15) (−1.95)
FISize −0.0163 −0.0267 0.0046 0.0237 −0.0641 −0.0521 −0.0327 −0.0013 −0.2237*** −0.1857***
(−0.31) (−0.64) (0.09) (0.44) (−1.07) (−0.95) (−0.60) (−0.04) (−5.46) (−3.60)
FIBusinessLines 0.1215* 0.1642** 0.1009 −0.0101 0.1042 0.0710 −0.0153 −0.0658 0.1946*** 0.1940***
(1.69) (2.47) (1.28) (−0.14) (1.16) (1.01) (−0.20) (−1.01) (2.74) (2.91)
FIPrestige −0.0149 −0.0092 0.0108 −0.0157 −0.0523 −0.0321 0.0042 −0.0164 0.0867** −0.0255
(−0.32) (−0.25) (0.25) (−0.36) (−1.16) (−0.66) (0.09) (−0.32) (2.23) (−0.42)
Buy/Hold/Sell No No No No No No No No No Yes
Controls?
Year FE? Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 79,027 66,217 43,139 73,972 75,333 72,425 66,954 7,053 65,242 43,750
R-squared 0.0025 0.0047 0.0019 0.0004 0.0009 0.0003 0.0006 0.0064 0.0105 0.2126
relation between Violations and bias, suggesting that weak culture is associated with less optimistically biased reports.31 These findings
suggest that analysts at weak culture banks appear to be careful in avoiding the appearance of being optimistically biased in the
presence of increased regulatory scrutiny.32 Taken together with the main catering analyses, my results indicate that regulatory reforms
may have altered the nature of analysts’ conflicts of interest, but not eliminated underlying cultural problems.
7. Conclusion
In this study, I provide empirical evidence consistent with weak corporate culture influencing analysts’ research production and
resulting in higher levels of catering. Using security code violations collected from the FINRA BrokerCheck service, I measure the
weakness of financial institutions’ corporate culture. I demonstrate that violations arising in divisions unrelated to equity research are
associated with less accurate and less informative reports, thereby indicating that analysts employed by weak culture firms produce a
lower quality written product for their retail investors. At the same time, I also find that violations are positively associated with
broker-hosted conferences and commission revenues, illustrating that analysts employed by weak culture firms produce a higher
quality soft product for their institutional clients. In additional analyses, I find that weak culture analysts issue less optimistically
biased reports, potentially to avoid regulatory scrutiny.
My findings contribute to a burgeoning literature examining the role of corporate culture and its effects on economic outcomes. While I
focus on research analysts, I expect my findings to have broader implications for the financial services industry in general. Recent concerns from
the Federal Reserve Board and SEC have suggested that many regulatory and consumer protection problems that have arisen in recent years
including LIBOR rigging, tax evasion, and money laundering, are rooted in cultural problems in financial institutions (Tarullo, 2014). Further,
regulators have expressed concerns regarding the limitations of regulation such as the Global Settlement. My findings corroborate these concerns
and suggest that weak corporate cultures can limit financial institutions’ roles as informational intermediaries in capital markets.
While this study provides evidence for the impact of cultural forces on financial analysts, my findings are not without limitations.
First, given the persistence of cultural forces, especially over relatively short horizons, it is an empirical challenge to generate causal
evidence through the use of exogenous variation in corporate culture. I attempt to circumvent these challenges by conducting a series
of tests that control for alternative explanations for my findings. Second, while my findings provide evidence consistent with financial
31
In untabulated analyses, I also conduct an out-of-sample analysis examining the relationship between weak corporate culture and re-
commendation bias prior to the Global Settlement. Regulators claimed that the issuance of overly optimistic recommendations was partly driven by
weak corporate culture. My analyses confirm this conjecture: the results indicate that firms with higher levels of violations produced the most
optimistic research in the two-year period preceding the Global Settlement.
32
Malmendier and Shanthikumar (2014) also examine patterns in bias across multiple analyst forecast outputs. They find that some analysts
strategically distort their reports by issuing both upwardly biased recommendations and downwardly biased earnings forecasts. My results are
inconsistent with weak culture analysts strategically distorting their forecasts as they issue less optimistically biased earnings forecasts and re-
commendations.
138
institutions’ sharing common values and norms regarding the importance of protecting their clients, my findings cannot describe or
explain how these values and norms are communicated through an organization. These values may be reflected (explicitly or im-
plicitly) through a variety of channels within an organization including recruitment, training, promotion, performance reviews, and
other policies that are not easily observable. Future research can provide a better understanding of the channels through which
cultural forces permeate an organization. Such findings can better inform regulators and practitioners regarding solutions for the
improvement of financial institutions’ corporate culture.
Appendix A. Sample Disclosure Events
This figure provides sample disclosure events obtained from the FINRA BrokerCheck online tool, available at http://brokercheck.
finra.org. Example 1, Example 2.
139
140
Appendix B. Variable Definitions
This table provides definitions of the variables used in this study.
Variable Definition Data Provider
BiasEForecast Analyst i's forecast error (FE) for firm j in year t, less the mean FE across all analysts covering firm j in year I/B/E/S
t, scaled by the absolute value of the mean FE across all analysts covering firm j in year t. FE is the forecast
minus actual earnings, and is based on the last forecast issued in the first 11 months of the year.
BiasLTG Analyst i's long-term growth forecast for firm j in year t less the mean long-term growth forecast for firm j I/B/E/S
in year t across all analysts providing long-term growth forecasts, divided by the standard deviation of
long-term growth forecasts for firm j in year t across all analysts providing long-term growth forecasts.
BiasNQtr BiasEForecast for the N-quarter forward forecast, where N ranges from one to four. I/B/E/S
BiasNYr BiasEForecast for the N-year forward forecast, where N ranges from one to three. I/B/E/S
BiasTP Analyst i's target price forecast for firm j in year t less the mean target price forecast for firm j in year t I/B/E/S
across all analysts providing target price forecasts, divided by the standard deviation of target price
forecasts for firm j in year t across all analysts providing target price forecasts.
Commissions The natural log of the sum of all fees paid for trades of a firm through an analyst's firm in a year. Ancerno
CompanyViolations The natural log of one plus the number of security code violations that are confirmed to be company-wide FINRA
violations. Company-wide violations include either a NASD Rule 3010 violation or disclose the firm's name
or one of the following phrases in the allegation: "FIRM", "SUPERV", "LLC", "INC.", "COMP.", "COMPANY",
or "CO."
Conferences The number of broker-hosted conferences held by an analyst's employer for the covered firm in a year. Bloomberg Corporate
Events
Coverage The number of firms covered by an analyst in a given year less the average number of firms covered by all I/B/E/S
analysts covering a firm in a given year, scaled by this average.
FIBusinessLines The natural log of one plus the number of business lines. FINRA
FIPrestige An indicator variable that takes the value of one if a financial institution's brokerage house is rated in the Institutional Investor
top five Institutional Investor rankings in 2012, or zero otherwise.
Experience The number of years in which analyst i has issued forecasts for firm j as of year t less the mean number of I/B/E/S
years in which all analysts covering firm j have issued forecasts for firm j as of year t, scaled by the mean
number of years in which all analysts covering firm j have issued forecasts for firm j as of year t.
FirmMomentum Cumulative firm returns over the prior six months. CRSP
FirmVolatility Standard deviation of firm returns over the prior six months. CRSP
FISize The number of analysts employed by the financial institution in which analyst i issues forecasts for firm j in I/B/E/S
year t less the mean number of analysts employed by financial institutions across all analysts issuing
forecasts for firm j in year t, divided by the mean number of analysts employed by financial institutions
across all analysts issuing forecasts for firm j in year t.
GoldmanSachs An indicator variable that takes the value of one for forecasts issued by analysts employed by Goldman I/B/E/S
Sachs, and zero otherwise.
GreatPlaceToWork An indicator variable that takes the value of one if a financial institution's overall Glassdoor ranking is Glassdoor
above the sample median, and zero otherwise.
HighGrowth An indicator variable that takes the value of one if the average market-to-book ratio of a financial Compustat
institution's research clients is above the sample median, and zero otherwise.
HighViolations An indicator variable that takes the value of one if the total number of violations a financial institution FINRA
receives over the 2005–2012 time period exceeds the sample median, and zero otherwise.
Horizon The number of days until the fiscal period end for analyst i's forecast of firm j in year t less the average I/B/E/S
number of days until the fiscal period end for all analysts issuing forecasts for firm j in year t, divided by the
average number of days until the fiscal period end for all analysts issuing forecasts for firm j in year t.
Informativeness The absolute value of abnormal returns measured over the 3-day period centered on all analyst report dates CRSP
(i.e., earnings forecast, recommendation, or target price revisions). Abnormal returns are adjusted from the
portfolio return from 125 benchmark portfolios based on size, book-to-market, and momentum (5 × 5 × 5)
using the procedure outlined in Daniel et al., (1997).
InstHold The percentage of a firm's shares held by institutional investors. Thomson One
LogAssets The natural log of assets. Compustat
LogMTB The natural log of the market-to-book ratio. Compustat
MktMomentum Cumulative market returns over the prior six months. CRSP
MktVolatility Standard deviation of market returns over the prior six months. CRSP
PercentStartup The percentage of a financial institution's research clients that are less than two years old. Compustat
PoorProductQuality An indicator variable that takes the value of one if a financial institution has net product quality KLD
weaknesses that exceed its strengths, and zero otherwise.
RAccuracy Minus one times the absolute relative forecast error (RAFError). RAFError is computed as analyst i's I/B/E/S
absolute forecast error (AFE) for firm j in year t, less the mean AFE across all analysts covering firm j in year
t, scaled by this mean. AFE is the absolute value of the forecast minus actual earnings, and is based on the
last forecast issued in the first 11 months of the year.
RAccuracyNQtr RAccuracy for the N-quarter forward forecast, where N ranges from one to four. I/B/E/S
RAccuracyNYr RAccuracy for the N-year forward forecast, where N ranges from one to three. I/B/E/S
RCommissions Commissions less the average commissions across all analysts covering a firm in a given year. Ancerno
141
RConferences Conferences less the average number of broker-hosted conferences across all analysts covering a firm in a Bloomberg Corporate
given year. Events
Rec A categorical variable that takes the value of one for sell recommendations (strong-sell or sell), two for hold I/B/E/S
recommendations, and three for buy recommendations (buy and strong-buy).
TotalBuys Total number of outstanding buy recommendations across all analysts covering the firm in the calendar I/B/E/S
year.
TotalHolds Total number of outstanding hold recommendations across all analysts covering the firm in the calendar I/B/E/S
year.
TotalSells Total number of outstanding sell recommendations across all analysts covering the firm in the calendar I/B/E/S
year.
Violations The natural log of one plus the number of annual disclosure events reported in BrokerCheck reports. FINRA
References
ABA, 2012. FINRA examinations: Regulatory priorities and firm best practices. American Bar Association, September 18.
Barber, B., Lehavy, R., McNichols, M., Trueman, B., 2006. Buys, holds, and sells: The distribution of investment banks’ stock ratings and the implications for the
profitability of analysts’ recommendations. J. Account. Econ. 41, 87–117.
Barniv, R., Hope, O.K., Myring, M.J., Thomas, W.B., 2009. Do analysts practice what they preach and should investors listen? Effects of recent regulations. Account.
Rev. 84, 1015–1039.
Becker, G.S., 1973. A theory of marriage: Part I. J. Pol. Econ. 81, 813–846.
Black, B., 2013. Punishing Bad Brokers: Self-Regulation and FINRA Sanctions. Brooklyn Journal of Corporate, Financial, & Commercial. Law 8, 23–55.
Bradshaw, M.T., Richardson, S.A., Sloan, R.G., 2006. The relation between corporate financing activities, analysts’ forecasts and stock returns. J. Account. Econ. 42,
53–85.
Bradshaw, M.T., 2011. Analysts’ forecasts: what do we know after decades of work? Unpublished working paper.
Bradshaw, M., Ertimur, Y., O'Brien, P., 2017. Financial Analysts and Their Contribution to Well-Functioning Capital Markets. Foundations Trends Account. 11,
119–191.
Brochet, F., Miller, G.S., Naranjo, P.L., Yu, G., 2016. Managers’ cultural background and disclosure attributes. .
Brown, L.D., Call, A.C., Clement, M.B., Sharp, N.Y., 2015. Inside the “Black Box” of sell‐side financial analysts. J. Account. Res. 53, 1–47.
Chen, C.Y., Chen, P.F., 2009. NASD Rule 2711 and changes in analysts’ independence in making stock recommendations. Account. Rev. 84, 1041–1071.
Clarke, J.E., Khorana, A., Patel, A., Rau, P.R., 2011. Independents’ day? Analyst behavior surrounding the Global Settlement. Annals Finance 7, 529–547.
Clement, M.B., 1999. Analyst forecast accuracy: Do ability, resources, and portfolio complexity matter? J. Account. Econ. 27, 285–303.
Clement, M.B., Tse, S.Y., 2003. Do investors respond to analysts' forecast revisions as if forecast accuracy is all that matters? Account. Rev. 78, 227–249.
Cohn, A., Fehr, E., Maréchal, M.A., 2014. Business culture and dishonesty in the banking industry. Nature 516, 86–89.
Corwin, S.A., Larocque, S.A., Stegemoller, M.A., 2017. Investment banking relationships and analyst affiliation bias: The impact of the global settlement on sanctioned
and non-sanctioned banks. J. Financ. Econ. 124 (3), 614–631.
Cowen, A., Groysberg, B., Healy, P., 2006. Which types of analyst firms are more optimistic? J. Account. Econ. 41, 119–146.
Craig, S., 2009. Goldman's trading tips reward its biggest clients. The Wall Street Journal, August 24.
Crémer, J., 1993. Corporate culture and shared knowledge. Industr. Corporate Change 2, 351–386.
Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997. Measuring mutual fund performance with characteristic‐based benchmarks. J. Finance 52, 1035–1058.
Davis, A., 2004. Increasingly, stock research serves the pros, not ‘little guy;’ in the wake of Spitzer Pact, Wall Street and upstarts are catering to elite few; ordering
'bespoke' reports. The Wall Street Journal, March 5.
DeBacker, J., Heim, B.T., Tran, A., 2015. Importing corruption culture from overseas: Evidence from corporate tax evasion in the United States. J. Financ. Econ. 117,
122–138.
Drake, M.S., Joos, P.R., Pacelli, J., Twedt, B., 2018. Analyst forecast bundling. Unpublished working paper.
Du, Q., Yu, F., Yu, X., 2017. Cultural proximity and the processing of financial information. Journal of Financial and Quantitative Analysis, forthcoming.
Edwards, B.P., 2016. The dark side of self-regulation. University of Cincinnati Law Review, forthcoming.
Egan, M., Matvos, G., Seru, A., 2016. The market for financial adviser misconduct. Unpublished working paper.
Ertimur, Y., Sunder, J., Sunder, S.V., 2007. Measure for measure: the relation between forecast accuracy and recommendation profitability of analysts. J. Account. Res.
45, 567–606.
FINRA, 2003. Ten of nation's top investment firms settle enforcement actions involving conflicts of interest between research and investment banking. Financial
Industry Regulatory Authority, April 28.
FINRA, 2013. Report on conflicts of interest. Financial Industry Regulatory Authority, October.
FINRA, 2017. Series: what to expect. Preparing for a FINRA cycle examination. http://www.finra.org/sites/default/files/14_0453%201_What%20to%20Expect_Cycle
%20Exam.pdf (accessed March 8, 2017).
FINRA, 2017. Annual regulatory and examination priorities letters. http://www.finra.org/industry/annual-regulatory-and-examination-priorities-letters (accessed
March 8, 2017).
FINRA, 2017. Top NASD rule violations. http://www.finra.org/investors/top-nasd-rule-violations (accessed March 8, 2017).
Firth, M., Lin, C., Liu, P., Xuan, Y., 2013. The client is king: do mutual fund relationships bias analyst recommendations? J. Account. Res. 51, 165–200.
Fisch, J.E., 2007. Does analyst independence sell investors short? UCLA Law Rev. 55, 39–97.
Francis, J., Philbrick, D., 1993. Analysts' decisions as products of a multi-task environment. J. Account. Res. 31, 216–230.
Frankel, R., Kothari, S.P., Weber, J., 2006. Determinants of the informativeness of analyst research. J. Account. Econ. 41, 29–54.
Gao, F., Lisic, L.L., Zhang, I.X., 2014. Commitment to social good and insider trading. J. Account. Econ. 57, 149–175.
GAO Report, 2012. Securities research. Additional actions could improve regulatory oversight on analyst conflicts of interest. United States Government Accountability
Office, January 12.
Garrett, J., Hoitash, R., Prawitt, D.F., 2014. Trust and financial reporting quality. J. Account. Res. 52, 1087–1125.
Gleason, C.A., Lee, C.M., 2003. Analyst forecast revisions and market price discovery. Account. Rev. 78, 193–225.
Goldstein, M.A., Irvine, P., Kandel, E., Wiener, Z., 2009. Brokerage commissions and institutional trading patterns. Rev. Financ. Studies 22, 5175–5212.
Green, T.C., Jame, R., Markov, S., Subasi, M., 2014. Broker-hosted investor conferences. J. Account. Econ. 58, 142–166.
Grossman, S.J., Hart, O.D., 1986. The costs and benefits of ownership: A theory of vertical and lateral integration. J. Political Econ. 94, 691–719.
Groysberg, B., Healy, P.M., Maber, D.A., 2011. What drives sell‐side analyst compensation at high‐status investment banks? J. Account. Res. 49, 969–1000.
Groysberg, B., Healy, P., Serafeim, G., Shanthikumar, D., 2013. The stock selection and performance of buy-side analysts. Manage. Sci. 59, 1062–1075.
Guiso, L., Sapienza, P., Zingales, L., 2015. The value of corporate culture. J. Financ. Econ. 117, 60–76.
Harvard Law, 2016. 2015 FINRA enforcement actions. Harvard Law School Forum on Corporate Governance and Financial Regulation, January 10.
Hermalin, B.E., 1999. Economics and corporate culture. Unpublished working paper.
Heskett, J.L., Kotter, J.P., 1992. Corporate culture and performance. Business Rev. 2, 83–93.
House, R., Javidan, M., Hanges, P., Dorfman, P., 2002. Understanding cultures and implicit leadership theories across the globe: an introduction to project GLOBE. J.
142
World Business 37, 3–10.

Hoi, C.K., Wu, Q., Zhang, H., 2013. Is corporate social responsibility (CSR) associated with tax avoidance? Evidence from irresponsible CSR activities. Account. Rev.
88, 2025–2059.
Irvine, P.J., 2000. Do analysts generate trade for their firms? Evidence from the Toronto stock exchange. J. Account. Econ. 30, 209–226.
Irvine, P., Lipson, M., Puckett, A., 2006. Tipping. Rev. Financ. Stud. 20, 741–768.
Jacob, J., Lys, T.Z., Neale, M.A., 1999. Expertise in forecasting performance of security analysts. J. Account. Econ. 28, 51–82.
Jackson, A.R., 2005. Trade generation, reputation, and sell-side analysts. J. Finance 60, 673–717.
Kadan, O., Madureira, L., Wang, R., Zach, T., 2009. Conflicts of interest and stock recommendations: the effects of the global settlement and related regulations. Rev.
Financ. Studies 22, 4189–4217.
Ke, B., Yu, Y., 2006. The effect of issuing biased earnings forecasts on analysts' access to management and survival. J. Account. Res. 44, 965–999.
Kedia, S., Rajgopal, S., 2011. Do the SEC's enforcement preferences affect corporate misconduct? J. Account. Econ. 51, 259–278.
Kreps, D.M., 1990. Corporate Culture and Economic theory: Perspectives On Positive Political Economy. Cambridge University Press, New York, N.Y.
Kubick, T.R., Lynch, D.P., Mayberry, M.A., Omer, T.C., 2016. The effects of regulatory scrutiny on tax avoidance: an examination of SEC comment letters. Account.
Rev. 91, 1751–1780.
Lin, H.W., McNichols, M.F., 1998. Underwriting relationships, analysts' earnings forecasts and investment recommendations. J. Account. Econ. 25, 101–127.
Liu, X., 2016. Corruption culture and corporate misconduct. J. Financ. Econ. 122, 307–327.
Maber, D.A., Groysberg, B., Healy, P.M., 2016. Client service, compensation, and the sell-side analyst objective function: an empirical analysis of relational incentives
in the investment-research industry. Unpublished working paper.
Malmendier, U., Shanthikumar, D., 2014. Do security analysts speak in two tongues? Rev. Financ. Stud. 27, 1287–1322.
Mehran, H., Stulz, R.M., 2007. The economics of conflicts of interest in financial institutions. J. Financ. Econ. 85, 267–296.
Merkley, K.J., Michaely, R., Pacelli, J., 2017. Does the scope of sell-side analyst industry matter? An examination of bias, accuracy and information content of analyst
reports. J. Finance 72, 1285–1334.
Merkley, K.J., Michaely, R., Pacelli, J. 2018. Cultural diversity on Wall Street: evidence from sell-side analysts’ forecasts. Unpublished working paper.
Michaely, R., Womack, K.L., 1999. Conflict of interest and the credibility of underwriter analyst recommendations. Rev. Financ. Stud. 12, 653–686.
Stanley, Morgan, 2018. Research conflict management policies. https://www.morganstanley.com/institutional-research/management-policies (accessed February 23,
2018).
Moyer, L., 2011. Goldman settles probe over research ‘huddles. The Wall Street Journal, June 9. Online.
O'Reilly, C.A., Chatman, J., 1996. Culture as social control: corporations, cults, and commitment. Res. Organ. Behav. 18, 157–200.
NASAA, 2003. Wall Street analyst conflicts of interest global settlement. North American Securities Administrators Association, May 7.
Pearce, 2017. A stockbroker's introduction to FINRA examinations and investigations. https://www.secatty.com/a-stockbroker-s-introduction-to-finra-examinations-
and-investiga.html (accessed March 8, 2017).
Popadak, J.A., 2013. A corporate culture channel: how increased shareholder governance reduces firm value. Unpublished working paper.
Richards, Lori, 2006. Speech by SEC staff: the process of compliance. Securities and Exchange Commission, October 19.
Richards, Lori, 2007. Speech by SEC staff: working towards a culture of compliance: some obstacles in the path. Securities and Exchange Commission, October 18.
Schein, E.H., 1990. Organizational culture. Am. Psychol. 45, 109–119.
Shimer, R., Smith, L., 2000. Assortative matching and search. Econometrica 68, 343–369.
Smith, G., 2012. Why I am leaving Goldman Sachs. The New York Times, March 14.
Tarullo, D., 2014. Good compliance, not mere compliance. Board of Governors of the Federal Reserve System, October 20.
Tuch, A.F., 2014. The self-regulation of investment bankers. Unpublished working paper.
U.S. Senate, 2003. Hearing before the committee on banking, housing and urban affairs. United States Senate One Hundred Eighth Congress, May 7.
143

Corporate Culture and Analyst Catering

Uploaded by

Copyright:

Available Formats

Corporate Culture and Analyst Catering

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Corporate Culture and Analyst Catering

Uploaded by

Copyright:

Available Formats

Journal of Accounting and Economics 67 (2019) 120–143

Contents lists available at ScienceDirect

Journal of Accounting and Economics

Corporate culture and analyst catering⁎

ARTICLE INFO ABSTRACT

E-mail address: jpacelli@indiana.edu.

2. Background on FINRA enforcement process

3. Related literature and hypothesis development

3.1. Corporate culture

3.2. The Global Settlement and Analysts’ conflicts of interest

4. Data and research design

4.1. FINRA data and sample selection

4.2. Description of FINRA data

Financial institution Security subsidiary

Allianz SE Commerz Markets LLC

Panel B: Financial institution business activities

Financial Institution Brokerage Investment advisory Underwriter Other

Allianz SE Yes No Yes Yes

Panel B: Financial institution business activities

Financial Institution Brokerage Investment advisory Underwriter Other

Deutsche Bank AG Yes Yes Yes Yes

4.3. Empirical models

4.3.1. Tests examining forecast accuracy

RAccuracyijt = 0 + 1 Violationsft + 2 Horizon ijt + 3 AnalystControlsijt + 4 FIControlsft + Yeart + ijt ,

Panel B: Top 5 security code violations

Security Code Description # of Violations % of Violations

Panel C: Frequency by year

Panel D: Frequency by financial institution

Financial Institution Violations

Panel D: Frequency by financial institution

Financial Institution Violations

Panel E: Persistence of violations

4.3.2. Tests examining report informativeness

+ 4 FirmControlsjt + 5 MktControlsjt + Firmj + Yeart + ijt

4.3.3. Tests examining institutional product quality

+ 4 FirmControlsjt + Firmj + Yeart + ijt ,

Variable N Mean Std. Dev. 25th Median 75th

Panel B: Independent variables

Variable Mean Std. Dev. 25th Median 75th

Violations 3.81 2.80 2 3 6

4.4. Descriptive statistics

5.1. Forecast accuracy

Violations −0.0475*** −0.0457***

Year Fixed Effects? Yes Yes

Panel B: Earnings forecast accuracy at other horizons

(1) (2) (3) (4) (5) (6)

Violations −0.0108** −0.0145** −0.0143** −0.0093** −0.0113* −0.0039

Year Fixed Effects? Yes Yes Yes Yes Yes Yes

5.2. Report informativeness

Violations −0.0032*** −0.0007*** −0.0004**

Firm Fixed Effects? No No Yes

5.3. Institutional product quality

(1) (2) (3) (4) (5)

Violations 0.0623*** 0.0381*** 0.0345*** 0.0377*** 0.0323***

Observations 77,207 77,207 77,207 77,207 77,207

Panel B: Security code violations & commission revenues

(1) (2) (3) (4) (5)

Violations 0.6778*** 0.3530*** 0.3570*** 0.3771** 0.1302**

Firm Fixed Effects? No No Yes No Yes

Observations 65,746 65,746 65,746 65,746 65,746

Violations −0.0475* −0.0457*

Violations −0.0108 −0.0145 −0.0143 −0.0093 −0.0113* −0.0039

Violations −0.0032* −0.0007* −0.0004**

Violations 0.0623* 0.0381* 0.0345* 0.0377* 0.0323***

Violations 0.6778* 0.3530* 0.3570* 0.3771 0.1302**

Violations −0.0573*** −0.0005* 0.0374* 0.3169*

Violations −0.0468*** −0.0004* 0.0357* 0.3529*

CompanyViolations −0.0413* −0.0006 0.0445* 0.3827*

Violations −0.0449* −0.0004 0.0297* 0.3086

GoldmanSachs −0.1288* −0.0012 0.0780* 0.6573*