Analyst Coverage Networks and Corporate Financial Policies

This article was downloaded by: [200.89.68.
196] On: 23 December 2024, At: 05:12

Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA
Management Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org
Analyst Coverage Networks and Corporate Financial

Policies
Armando Gomes, Radhakrishnan Gopalan, Mark T. Leary, Francisco Marcet
To cite this article:

Armando Gomes, Radhakrishnan Gopalan, Mark T. Leary, Francisco Marcet (2024) Analyst Coverage Networks and
Corporate Financial Policies. Management Science 70(8):5016-5039. https://doi.org/10.1287/mnsc.2023.4891
Full terms and conditions of use: https://pubsonline.informs.org/Publications/Librarians-Portal/PubsOnLine-

Terms-and-Conditions
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use or
systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher approval,
unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or support
of claims made of that product, publication, or service.
Copyright © 2023, INFORMS
Please scroll down for article—it is on subsequent pages
With 12,500 members from nearly 90 countries, INFORMS is the largest international association of operations
research (O.R.) and analytics professionals and students. INFORMS provides unique networking and learning
opportunities for individual professionals, and organizations of all types and sizes, to better understand and use
O.R. and analytics tools and methods to transform strategic visions and achieve better outcomes.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
MANAGEMENT SCIENCE
Vol. 70, No. 8, August 2024, pp. 5016–5039
https://pubsonline.informs.org/journal/mnsc ISSN 0025-1909 (print), ISSN 1526-5501 (online)
Analyst Coverage Networks and Corporate Financial Policies

Armando Gomes,a Radhakrishnan Gopalan,a Mark T. Leary,a,b Francisco Marcetc,*
Downloaded from informs.org by [200.89.68.196] on 23 December 2024, at 05:12 . For personal use only, all rights reserved.
a
Olin Business School, Washington University in St. Louis, St. Louis, Missouri 63130; b National Bureau of Economic Research, Cambridge,
Massachusetts 02138; c School of Economics and Business, University of Chile, Santiago 8330015, Chile
*Corresponding author
Contact: gomes@wustl.edu, https://orcid.org/0000-0002-1755-1154 (AG); gopalan@wustl.edu, https://orcid.org/0000-0001-9446-8509
(RG); leary@wustl.edu, https://orcid.org/0000-0002-3199-4736 (MTL); fmarceto@fen.uchile.cl, https://orcid.org/0000-0002-0035-1965
(FM)
Received: March 29, 2021 Abstract. We use the setting of analyst coverage networks to shed light on the nature of
Revised: June 16, 2022 peer effects in financial policies. First, we use the “friends-of-friends” approach and exploit
Accepted: September 3, 2022 the fact that analyst coverage networks partially overlap to identify endogenous peer
Published Online in Articles in Advance: effects, in which firms respond directly to the capital structure choices of their peers, sepa
November 15, 2023 rately from contextual effects, in which they respond to their peers’ characteristics. We
further show evidence that analysts facilitate these peer effects through their role as infor
https://doi.org/10.1287/mnsc.2023.4891 mational intermediaries. Analyst network peer effects are distinct from industry peer effects
and are more pronounced among peers connected by analysts that are more experienced
Copyright: © 2023 INFORMS
and from more influential brokerage houses. Finally, the analyst peer effects become weaker
after exogenous reductions in common coverage as a consequence of brokerage closures.
History: Accepted by Victoria Ivashina, finance.

Funding: F. Marcet acknowledges funding from ANID/CONICYT Proyecto FONDECYT Iniciacion
[Grant 11221187].
Supplemental Material: The data files and online appendix are available at https://doi.org/10.1287/mnsc.
2023.4891.
Keywords: analyst network • friends of friends • peer effects • equity shock • capital structure
1. Introduction its peer group, giving rise to contextual effects. In a linear

A growing literature shows that a number of corporate model, when the expected behavior of the group is
policies, including security issuance, investment, pay a function of the average group characteristics, these
out policy, and stock splits, are influenced by a firm’s effects cannot be separately identified, which Manski
peers.1 Although these results have been illuminating, terms the reflection problem. Distinguishing between
several challenges remain. First, although the empirical exogenous and endogenous effects is relevant because,
approaches suggest the existence of some form of social for example, policy interventions such as targeted indus
interaction effect, separating the effect of peers’ policy try tax subsidies for debt financing, may influence the
choices from the effect of their characteristics remains financial policies of peers while leaving their fundamen
challenging. Second, to the extent that firms do respond tals unchanged. These policies may generate multiplier
to the policy choices of their peers, the mechanisms effects through endogenous peer effects (Glaeser et al.
through which firms learn about, and are influenced 2003) but are unlikely to if only contextual effects are
by, the choices of their peers are still not fully under present.
stood. In this paper, we use the setting of security ana Previous approaches to identifying peer effects in cor
lyst networks to shed light on both these issues. porate financial policies have used either random group
As discussed by Manski (1993), a positive association assignment (Shue 2013) or exogenous characteristics of
between a firm’s policy and that of its peers can arise peer firms as instruments (Leary and Roberts 2014, Kaus
from multiple sources. First, peer firms may receive com tia and Rantala 2015) to separate social effects from corre
mon shocks or share common unobserved characteristic lated effects. Yet, neither approach can fully break the
that make them follow similar policies (what Manski reflection problem, because it is unclear whether firms
calls “correlated effects”). Second, there are two types are responding to their peers’ financial policy choices
of potential social interaction effects, what Manski refers themselves or to the peer characteristics that drove those
to as “endogenous effects” and “contextual effects.” choices. For example, suppose a firm’s peer receives a
Endogenous effects arise when firms respond directly to shock to its investment opportunities and issues equity
the financial policy choices of their peers. Alternatively, a to fund it. If the focal firm issues equity in response, it
firm’s choices may be influenced by the characteristics of is not clear whether they are responding to the peer’s
5016
Gomes et al.: Analyst Coverage Networks and Corporate Financial Policies
Management Science, 2024, vol. 70, no. 8, pp. 5016–5039, © 2023 INFORMS 5017
equity issuance or to the change in peer’s investment 2013). Because analysts typically cover a portfolio of
opportunities that prompted it.2 firms, the two-way communication between an analyst
We exploit the structure of analyst coverage networks and management can also result in the propagation of
(i.e., groups of firms covered by a common analyst) to financial policies across the firms in an analyst’s portfo
better identify the presence of endogenous social effects lio. We show evidence consistent with this mechanism.4
in financing decisions, separately from contextual effects. We focus on financial policies such as leverage, debt
Kaustia and Rantala (2015, 2021) argue that peer groups issuance, and equity issuance. We classify all firms that
can be defined by the set of firms covered by the same share a common analyst with a firm as its “analyst peers”
sell-side equity analysts because these common coverage and relate the firm’s financial policy to the weighted
choices reflect economic linkages among firms. An addi average financial policy of its analyst peers. We use the
tional advantage of this setting, however, is that unlike number of common analysts between the firm and its
other peer group definitions such as industries, groups peer firms as the weights. This methodology gives rise to
defined by common analyst coverage partially overlap. a network, which we refer to as the analyst coverage net
To distinguish endogenous peer effects from exogenous work, that is, the graph where the firms are the nodes
peer effects, we exploit the fact that we can observe and the weighted edges between two firms are the num
intransitive triads in the analyst network. That is, we can ber of common analysts between the firms.
observe firm triads i, j, and k, such that firms i and j and We begin by documenting a positive association between
firms j and k have common analysts, whereas firms i and a firm’s financial policy and that of its analyst peers. The
k do not have any. As shown by Bramoullé et al. (2009) association holds for leverage in levels and changes and
and Goldsmith-Pinkham and Imbens (2013), this is a key debt and equity issuance decisions and is robust to control
property of the analyst coverage network that allows for ling for firm characteristics, analyst peer characteristics, the
identification of peer effects by exploiting the “friends of average policies of industry peers, and industry peer char
friends” approach. By using an exogenous characteristic acteristics. When we differentiate between within industry
of the peers of a firm’s peers, we can construct an instru analyst peers and outside industry analyst peers, we find
ment that is correlated with peer firm’s financial policies that the association extends to both sets of peers.
but is arguably uncorrelated with their characteristics, Firms in the same analyst network may pursue similar
enabling identification of endogenous effects separately financial policies simply because they share similar
from contextual effects. Thus, our instrument is con characteristics or face a similar environment. This is
structed using shocks to characteristics of firms that do especially likely given that analysts cover firms with
not share a common analyst with the firm in question. common underlying economic features. Firms with com
This makes it hard to attribute the correlations we mon analysts may also follow similar policies due to the
uncover to analysts choosing to cover firms with com preferences of their common analysts (Degeorge et al.
mon unobserved characteristics. 2013). We control for such correlated effects following
A second advantage of the analyst network setting the procedure of Leary and Roberts (2014). Specifically,
is that it enables us to study the role security analysts we use idiosyncratic equity return shocks as an exoge
play in transmitting financial policy-relevant infor nous source of variation in peer firm financial policy
mation across firms, highlighting a new mechanism (and possibly characteristics) and relate it to a firm’s
facilitating peer effects in financial policies. The role of financial policy. A large prior literature in finance shows
sell-side analysts in acquiring, analyzing, and dissemi that firms change their leverage and equity issuance
nating information for investors has been much studied.3 decision in response to changes to their stock price
In their role as information intermediaries, analysts may (Marsh 1982, Baker and Wurgler 2002), which supports
also facilitate the flow of information between firms. For the relevance of our instrument. To the extent we are
example, Martens and Sextroh (2021) show that firms are able to isolate idiosyncratic shocks to peer firms’ equity
more likely to cite each other’s patents when they share a values, the shocks are unlikely to be correlated with the
common analyst, suggesting that analysts facilitate inter characteristics of the focal firm, and thus any peer effects
firm information transfers. Analysts not only learn about we document are unlikely to include correlated effects.
firm performance and prospects from company disclo In constructing the idiosyncratic equity shock, we esti
sures and conversations with managers, but also pro mate an augmented market model controlling for the
duce processed information in the form of research average returns of analyst peers.
reports and recommendations. To the extent firm man We estimate reduced-form regressions that establish a
agers pay attention to analyst reports, they may convey significant association between a firm’s financial policy
information about the analyst’s assessment of market and the idiosyncratic return shocks of its analyst network
conditions and preferred firm policies, and so on. Man peers. This association is robust to controlling for the
agers may pay attention to analyst reports either because financial policies, characteristics, and return shocks of
they believe it to contain value relevant information or in the firm’s industry peers, suggesting that the analyst net
their effort to cater to analyst preferences (Degeorge et al. work effect is distinct from an industry effect.5 These
5018 Management Science, 2024, vol. 70, no. 8, pp. 5016–5039, © 2023 INFORMS
results apply to leverage, net and gross equity issuance, (b) not reflected in firm j’s stock return, a scenario we
and the debt-equity choice. We also show that these consider unlikely.
results extend to seasoned equity offerings (SEOs) in We find robust evidence for endogenous peer effects
particular: Higher idiosyncratic returns of analyst cover using the friends of friend approach. The effects we docu
age network peers increase the probability that the focal ment are economically significant. A 1-standard-deviation
firm issues new seasoned equity in the next year. We fur increase in peer firm average leverage is associated with a
ther find that SEOs that follow those of analyst network 0.482-standard-deviation increase in a firm’s leverage.
peers are associated with higher announcement returns, Peer effects are also present in a firm’s decision to issue
consistent with previous evidence that analysts reduce equity. A 1-standard-deviation increase in gross equity
information asymmetry around security issuances and issuance likelihood by peers leads to a 0.520-standard-
of information spillovers across firms in the analyst’s deviation increase in net gross equity issuance likelihood.
coverage network (Chan and Chan 2014). We also use Overall, after controlling for the endogeneity in the net
the idiosyncratic shock to peer firms’ stock prices as an work formation, we find that peer firms in the same ana
instrument for peer firm financial policies in a two-stage lyst coverage network affect each other.
least-squares (2SLS) specification and document a posi Having established the presence of endogenous peer
tive association between a firm’s financial policy and effects in analyst coverage networks, we next explore the
analyst peer policies that is distinct from correlated potential mechanisms behind this behavior. Firms may
effects, suggesting the existence of at least one form of respond to the financial policy choices of those with com
social effects among analyst networks. mon analysts for several reasons. First, as argued by
To separate endogenous effects from contextual effects, Kaustia and Rantala (2015, 2021), firms within the ana
we define a firm’s indirect peers as those firms that share lyst network share economic linkages and face similar
a common analyst with other firms in the focal firm’s ana market frictions and investment opportunities. Thus,
lyst network but do not have an analyst in common with they may look to each other to learn about optimal finan
the focal firm itself (i.e., “friends of friends”). To illustrate, cial policy. In addition, the analyst may play a role in the
suppose we observe firms i, j, and k, such that firms i and j propagation of information about financial policy. Apart
and firms j and k have common analysts, whereas firms i from regularly communicating with the firms, analysts
and k do not have any. We then use the idiosyncratic also use common models to value the firms and bench
equity shock of firm k as an instrument for the financial mark them with one another. During the course of their
policy of firm j to document its influence on i’s financial communication and valuation, analysts obtain informa
policy. The exclusion restriction for this approach is that tion that other firms may find valuable. Such information
the indirect peer’s (firm k’s) return shock should influence can be about the state of financial markets, growth
firm i’s financial policy only through its influence on the opportunities, a particular financial policy, and so on,
direct peer’s (firm j’s) financial policy and not otherwise. and may originate either from a particular portfolio firm
One necessary condition is that firm i does not respond or from the analyst. If learning about this information
directly to its indirect peers’ financial policies. This is rea from analyst reports the firm managers choose to act on
sonable given that the set of firms two analyst connections that information, then we expect financial policies to
away is large and heterogeneous, consisting of firms that propagate from one firm to another within the analyst’s
not only have no common analysts but are also mostly in portfolio.7 We follow a two-pronged strategy to show
different industries.6 evidence consistent with this mechanism.
Identification also requires that firm k’s equity shock We first do a battery of tests to ensure that our results
should not be correlated with firm j’s and firm i’s charac are specific to analyst networks and not solely a result of
teristics. To the extent we are able to isolate shocks to common industry linkages. First, as noted earlier, in all
equity values that are idiosyncratic to firm k, it is reason our tests, we control for industry average policies, either
able to assume that it is uncorrelated with firm j’s charac directly or through industry average return shocks. Sec
teristics. To the extent firm k and firm i do not even have ond, we find similar results when we focus on the firms
common analysts, it is even more reasonable to expect in the analyst network that are not from the same indus
idiosyncratic shocks to firm k’s equity to be uncorrelated try as the firm in question. Third, we estimate a placebo
with firm i’s characteristics. One potential threat to the test in which we define pseudo peer groups as firms in
former is if firm j is a supplier or customer to firm k, in the same industry as a firm’s direct analyst peers but that
which case economic shocks may have spillover effects. do not have a common analyst with the firm in question.
However, a major advantage of using shocks to firms We find no evidence of peer effects in this sample. These
outside of firm i’s analyst network as instruments is that results are robust across various industry definitions
we can control for the stock returns of the firms within such as Hoberg-Phillips industry peers and Global In
the analyst network. In that case, any identification dustry Classification Standard (GICS) code. Fourth, we
threat would have to come from a shock to firm j that is exploit a quasi-natural experiment to shock the network
(a) significant enough to solicit a response from firm i but structure. Specifically, we explore the closure and mergers
of brokerage houses as an exogenous reduction in common capital structure policies and that analyst networks influ
analyst coverage (Hong and Kacperczyk 2010, Kelly and ence the way firms interact with one another. As such,
Ljungqvist 2012, Israelsen 2016, Martens and Sextroh 2021). our study is related to two main streams of literature.
After identifying pairs of firms that suffered an exogenous The first is the literature exploring the role of peer effects,
reduction in common analysts, we find that the relation or the interaction among agents, in corporate finance.
between the focal firm’s leverage and equity issuance deci Most of this literature provides evidence either for infor
mation transmission through networks of firms and/or
sions and those of its peers become weaker after the shock.
These results suggest that analysts play an important role managers or of the resultant correlated behavior among
in propagating information from one firm to another and members of the network. Evidence of network effects
the analyst network peer effect is not entirely driven by eco in corporate policies is shown by Shue (2013), Leary
nomic linkages that we do not observe. and Roberts (2014), Kaustia and Rantala (2015), Fracassi
Finally, we document cross-sectional variation in our (2017), and Grennan (2019), whereas Matvos and
estimated social effects that are unique to information Ostrovsky (2010) and Pool et al. (2015) document peer
transmission through the analyst network by testing effects among mutual fund managers.
whether more influential analysts are more effective at Our paper is closest in spirit to Leary and Roberts
transmitting information across firms. Consistent with (2014), which shows evidence that firms are influenced
this idea, we find stronger peer effects among firms con by their peers when making capital structure decisions.
nected by analysts from brokerage houses with more “all- Although instructive, Leary and Roberts (2014) note sev
star” rated analysts and by more experienced analysts. eral unresolved issues from their analysis, which they
We make a number of important contributions. First, leave for future research. First, their empirical approach
we add to the growing literature on peer effects in corpo is not fully able to distinguish the nature of the social
rate policies. Although previous studies have shown evi interaction effects among firms. That is, do firms respond
dence of social interaction effects in capital structure directly to the financial policy choices of their peers, or
(Leary and Roberts 2014), dividends (Grennan 2019), and do they respond to the change in peers’ characteristics
investment decisions (Fracassi 2017), the mechanisms and (e.g., investment opportunities, etc.) that prompted the
economic forces behind these effects are less well under financial policy choice? Second, they call for future work
stood. We make two contributions relative to these earlier to better understand the mechanisms driving inter
studies. First, we use the friends of friends approach to dependence in financial policies. That is, how does in
formation that informs financial policy choices get
separately identify firms’ responses to their peers’ policy
transmitted across firms?
choices from the influence of peer characteristics, a dis
We address both these issues by studying peer effects
tinction that was difficult to make with previous ap
in financing decisions within analyst coverage networks.
proaches. This approach can be productively used to
We propose analyst-firm interactions as one mechanism
document endogenous peer effects in other networks that
through which relevant information could transmit from
partially overlap such as board networks and supply
one firm to another, and we further use the partially
chain networks. Second, we show evidence suggesting
overlapping nature of analyst coverage networks to dif
that analysts may play an important role in propagating
ferentiate endogenous peer effects from exogenous or
financial policies across firms. This information trans
contextual effects. Our main model is an extended ver
mission represents an important underexplored channel
sion of the Manski-type linear-in-means model similar to
through which peer effects may arise. We also contribute
those studied in Goldsmith-Pinkham and Imbens (2013)
to the literature on the effects and implications of shared and Bramoullé et al. (2009) (see also the survey by Blume
analyst coverage networks. Our results highlight an et al. 2010).8
underappreciated role of security analysts in facilitating At the same time, we build on the large literature that
information flow across firms (not just between firms and studies the role of sell-side analysts in financial markets.
investors). Although other studies have shown that com Focusing on the traditional role of analysts as informa
mon analyst coverage is associated with return comove tion intermediaries between firms and outside investors,
ment (Muslu et al. 2014, Israelsen 2016), we demonstrate some studies have shown evidence that analysts pro
that it can also lead to commonality in financial policies. duce value-relevant information about the firms they
The rest of the paper is organized as follows. Section 2 cover (Womack 1996, Barber et al. 2001, Jegadeesh et al.
discusses the related literature. Section 3 discusses our 2004). Although we are not aware of any prior studies
data and empirical methodology. Section 4 discusses the that show direct evidence of analysts recommending
empirical evidence, and Section 5 concludes. policy changes based on actions of other firms in their
network, the combination of previous results suggest
2. Related Literature this is a plausible channel. First, Hilary and Shen (2013)
Our paper’s main contribution is to show that analysts show that analysts with more experience covering a
play an important role in facilitating peer effects in given firm are better able to use information from that
firm’s management forecast to improve forecast accu than or equal to 9000). We then match the CRSP-
racy for other firms they cover. This suggests that some Compustat sample to IBES and identify all firms that are
of the information analysts generate in covering one connected to at least one other firm in the sample
firm, such as industry-specific information, is relevant through a common analyst. We identify an analyst as fol
for understanding and valuing other firms in their net lowing a firm in a fiscal year if that analyst makes at least
work. This information transfer affects not only analyst one earnings forecast during the year, and the forecast is
made at most six months before and three months after
forecasts but also investor behavior (Piotroski and Roul

stone 2004, Muslu et al. 2014, Brochet et al. 2018). the end of the fiscal period. We also require the analyst to
Other studies suggest that, in addition to communicat follow the pair of firms for at least two years in the entire
ing information to investors, analysts also influence out sample for us to consider them to be connected through
comes at the firms they cover. This can happen through the analyst. Our final sample spans the period 1993–2015
analysts’ monitoring role (Chen et al. 2015), catering to and includes 41,411 firm-year observations.9
analysts’ preferred policies (Degeorge et al. 2013), or
learning from analyst feedback (Bae et al. 2022). Thus, 3.2. Empirical Methodology
the existing literature provides evidence that (i) analysts We begin by documenting the extent to which financial
produce information that is relevant for firms across policies of analyst peers are associated with a firm’s
their coverage network, and (ii) communication runs financial policy. We do that by estimating the following
both from the firm to the analyst and from the analyst regression:
back to the firm. This combination suggests that analysts yijt � α + β1 yACN IND ′ ACN ′ IND
�it + β2 y�ijt + γ1 X�it�1 + γ2 X�ijt�1
may serve a role in facilitating information flow across
the firms they cover. Some recent work supports this + γ′3 Xijt�1 + δ′ ui + φ′ vt + ɛijt ,
role. For example, Martens and Sextroh (2021) show that (1)
pairs of firms are more likely to cite each other’s patents where the indices i, j, and t refer to firm, industry, and
if they have common analysts, which they interpret as year, respectively. The dependent variables that we
evidence that analysts facilitate the flow of relevant com model are Leverage, ∆Leverage, Net Debt, Net Equity, and
petitive information between firms. Brown et al. (2019) Gross Equity. All variables we use in our analysis are
survey corporate investor relations officers on what fac defined in Appendix A. Net Debt, Net Equity, and Gross
tors are important for managing their company’s narra Equity are dummy variables that identify debt and equity
tive. A majority of respondents agree that “knowledge issuances. These take a value one if the firm issues debt
about industry trends and/or competitors” is a very (equity) in excess of 1% of total assets and zero other
important service provided by sell-side analysts. Thus, wise10; Xijt�1 is the set of firm-specific controls. Following
there is some precedent for the idea that analysts facili Leary and Roberts (2014), we include lagged (one
tate information transfer across firms. The one element period) values of Log(Sales), Market to book, Tangibility,
of our proposed channel that does not have direct prece Profitability, and Log(1 + Coverage). We also include the
dent in the literature is whether analysts are influenced current size of the firm’s network, Log(1 + Connections),
by the financial policy choices made by firms they cover as our controls. Coverage is defined as the number of ana
or, at a minimum, communicate these choices to other net lysts following a firm in a given year. Connections is the
work firms. Whether this channel operates is ultimately an size of a firm’s network (degree: number of firms con
empirical question to which we turn next. Our results pro nected to firm i by common analysts); yACN �it represents
vide additional evidence supporting information transfer the weighted average value of the outcome variable for
across firms covered by a common analyst, and further all the firms that are connected to firm i through common
show that this information transfer plays a role in propa analysts in year t. The weights for each firm l in the ana
gating financial policies across connected firms. lyst network equals the number of common analysts
between firm l and firm i. Specifically,
3. Data and Empirical Methodology PI
nilt ylt
3.1. Sample Selection y�it � Pi≠lI
ACN
, (2)
We obtain our data from standard sources: financial i≠l nilt
information from Compustat, stock price information where nilt represents the number of common analysts
from the Center for Research in Security Prices (CRSP), between firm i and firm l. In calculating yACN
�it , we use the
and analyst coverage information from the Institutional financial policies of peer firms in the current year along
Brokers’ Estimate System (IBES). From the overall CRSP- with the current network structure. We use a weighted
Compustat merged data set, we exclude financial firms average instead of a simple average to give more weight
(Standard Industrial Classification (SIC) codes between to peer firms with more common analysts. Such peers
6000 and 6999), utilities (SIC codes between 4900 and may have a stronger influence on a firm’s financial policy
4949) and government companies (SIC codes greater because there is a greater likelihood that one or more
analysts will transmit information across the firms. Our using a five-year rolling window.13 We then calculate
coefficient of interest is β1. We also include a set of Equity shock for firm i in year t as the difference between
ACN the return on the firm’s stock in year t and the predicted
weighted average peer firm characteristics (X�it�1 ) as
controls. These are the same set of characteristics as return based on the market and peer portfolio excess
included in Xijt�1 and discussed previously, but we returns during the year and the loadings estimated using
include the contemporaneous size of firm’s analyst net the data from the prior five years. We require firms to
ACN have at least 24 months of historical data to estimate the
work (Log(1 + Connections)t ). To calculate X�it�1 , we use

the current network structure and lagged peer firm previous model. Equity shock represents the idiosyncratic
characteristics. shock to a firm’s stock return. We then calculate the
To distinguish the effect of analyst network peers from weighted average equity shock for the analyst network,
that of industry peers (Leary and Roberts 2014), we also Equity shockACN
�it , using the number of common analysts
control for the average value of the outcome variable for as the weights and the industry average equity shock,
all other firms in the same industry (based on three-digit Equity shockIND
�ijt , as the simple average equity shock for
SIC code), yIND all firms in the same industry as firm i.
�ijt (excluding the firm i), and their average
IND
characteristics, X�ijt�1 , as additional controls.11 In all the We use Equity shockACN ACN
�it as an instrument for y�it and
regressions, except for those with changes in Leverage as use a reduced form model and 2SLS to estimate its effect
the outcome variable, we include firm and year fixed on firm i’s financial policy after controlling for industry
effects. For the regressions with change in leverage as the corporate policy and industry characteristics. To the
outcome variable, we include industry and year fixed extent that Equity shockACN �it captures idiosyncratic shocks
effects. We also include the independent variables in first to the stock price and consequently leverage of analyst
difference form in this specification, except for the equity peer firms, it is unlikely to be correlated with firm i’s
shock, analyst coverage and firm’s network. The standard characteristics. To this extent, the reduced-form model
errors we estimate are robust to heteroskedasticity and and the 2SLS will isolate the social effects and exclude
clustered at the firm level. correlated effects. The specific identifying assumptions
As shown in Manski (1993), a significant β1 can arise that we make for this are the following. First, for instru
from one of three sources. First, it can reflect the fact that ment relevance, we assume that Equity shockACN �it is corre
there are some unobserved similarities among firms in lated with the peer firm’s financial policy either directly
the same analyst network (correlated effects). These sim or indirectly through one or more characteristic. A large
prior literature documents the important effect stock
ilarities may result in the firms choosing similar financial
prices can have on firm financial policies (Marsh 1982
policies. Alternatively, it can arise from firms responding
and Baker and Wurgler 2002, among others), and stock
to either the behavior (endogenous peer effects) or char
price changes often reflect changes in firm characteristics
acteristics (exogenous peer effects) of the peer firms. To
such as investment opportunities, expected profitability
control for correlated effects, following Leary and Rob
or risk, which in turn have been shown to be important
erts (2014), we use idiosyncratic shocks to the value of
determinants of firm financial policies. This suggests the
the peer firm’s equity as an instrument for their financial
relevance assumption will be satisfied in our setting,
policy (or characteristic). We define expected returns
which is further supported by strong first-stage results
based on a one-factor market model augmented to
below. The second assumption we make to isolate social
include the excess return on the analyst network portfo
effects is that Equity shockACN �it is uncorrelated with firm
lio. We use the equally weighted portfolio returns of all
i’s characteristics. To the extent our procedure for defin
firms that share a common analyst with a firm to calcu ing Equity shockACN �it isolates truly idiosyncratic shocks,
late the excess returns. Although the excess return on the this assumption is likely to be valid.14 Leary and Roberts
analyst network firms does not necessarily represent a (2014) explore the properties of this instrument in depth
priced risk factor, we include it to absorb any common and demonstrate its suitability in this context.
shocks that may affect firms in the same analyst network.12 Our tests using Equity shockACN �it as an instrument will
For example, Muslu et al. (2014) and Israelsen (2016) show not be able to isolate endogenous peer effects from exog
that shared coverage explains comovement and excess enous peer effects because the idiosyncratic shock to
comovement between pairs of stocks with common ana equity values can change, or reflect changes in, some
lysts. Thus, we model the firm’s stock return as (possibly unobserved) peer firm characteristic and firms
rit � αit + βM ACN ACN may respond to the changes to peer firm characteristic as
it (rmt � rft ) + βit (r �it � rft ) + ηit ,
opposed to the changes in peer firm behavior.
where the subscript t refers to time in months, rmt and rft To help isolate the endogenous peer effects from exog
are the monthly return on the market and risk free asset, enous peer effects, we exploit the fact that analyst net
respectively, and r ACN
�it is the equally weighted average works partially overlap with each other, and we can find
return of all firms in the analyst network of firm i. We intransitive triads in the analyst network. Bramoullé et al.
estimate this regression individually for each firm-year (2009) show that one can identify peer effects when there
are intransitive triads, even when the analyst network is policy, yet not be captured by firm j’s stock return. Finally,
endogenous, say because analysts choose to cover firms if firm i responds directly to firm k’s financial policy or
with common (unobserved) features. In other words, in characteristics, then we may not capture peer effects oper
our network there are triads i, j, and k, such that firms i ating through analyst networks with this approach. How
and j and firms j and k have common analysts, whereas ever, we argue this is unlikely, given that the indirect
firms i and k do not have any common analyst. Following peers consist of firms that do not share any common
the friends-of-friends approach outlined in Bramoullé analysts with firm i and (in some specifications) are also in
et al. (2009), we use the characteristic of firm k (namely different industries.16 Thus, any social interaction effects
Equity shock) as an instrument for the financial policy of should be much stronger among direct peers than be
firm j to identify its influence on firm i’s financial policy. tween indirect peers. Furthermore, we perform cross-
In our subsequent discussion we refer to firm k as an indi sectional tests later to demonstrate the relevance of analyst
rect peer of firm i. We use a slightly modified and in some connections in transmitting policy choices across firms.
senses a stricter version of the friends-of-friends approach
proposed by Bramoullé et al. (2009). To identify endoge 3.3. Summary Statistics
nous peer effects, they only require that some of the indi Table 1 provides descriptive statistics for the analyst
rect peers not be direct peers of the firm in question. If networks. On average, a firm is connected to 42 other
that is true, then one can use the characteristics of all the firms through common analysts. Interestingly, only 10.58
indirect peers as instruments for peer firm behavior. In (29%) of these connections are from the same three-digit
our tests, we use the Equity shock of only the indirect peers SIC code industry. The low percentage of within industry
that are not direct peers of the firm in question to instru connections helps us independently estimate peer effects
ment for peer firm behavior. By construction, there are no arising from both industry and analyst networks. Even
analysts in common between firms i and its indirect peers. with alternate definitions of industry (Fama-French and
Therefore, the specific instrument we employ is the sim GICS codes), we find that the percentage of within indus
ple average Equity shock of the indirect peers.15 try analyst peers is uniformly less than 50%. We find that
The identifying assumptions necessary for us to isolate on average, two connected firms in our sample have 1.91
the endogenous peer effects are as follows. First, we analysts in common. Surprisingly, this number does not
require that the Equity shock of firm k be correlated with vary much in the sample. The 25th percentile of the aver
the behavior of firm j. This will happen as long as there age number of common analysts by firm-year is 1.11,
are some social effects in analyst networks, which our whereas the 75th percentile is 2.38. We find that firms
earlier results confirm. Our second assumption has two within an industry are likely to have more common ana
parts to it. First, we require firm k’s equity shock to not be lysts as compared with firms across industries. Two firms
correlated with firm j’s characteristic. This is exactly the within the same industry have on average 3.13 common
same as the exclusion restriction in our instrumental var analysts, whereas this number is only 1.56 for two firms
iable (IV) estimate and is also similar to the assumption from different industries.
in Leary and Roberts (2014). The second part of our We exclude from our analysis firms that are not con
assumption is that firm k’s equity shock not be correlated nected to any other firm through common analysts. The
with firm i’s characteristic. This assumption is more eas variable Connected Firms identifies the percentage of
ily satisfied than the previous assumption as, in addition firms that are connected to at least one other firm each
to our instrument being idiosyncratic, firm i and firm k year in the overall CRSP-Compustat-IBES sample. We
do not have any common analysts and they are often not find that about 94% of the firms in the overall sample are
even from the same industry. connected to at least one other firm. Thus, the uncon
Additionally, because firm k is in a different analyst nected firms, which we exclude, constitute only 6% of
network than firm i, we can control for the average the CRSP-Compustat-IBES merged sample.
equity return of firm i’s analyst network in this specifica The average (median) number of indirect connections,
tion, which further rules out any confounding influence defined as the pairs i and k, such that firms i and j and
from common shocks to fundamentals that are not cap firms j and k have common analysts, whereas firms i and
tured by the asset pricing model or spillovers from firm k k do not have any, are 414.72 (382), and the 25th percen
to firm j and allows for identification even when the ana tile of the number of indirect connections is 222, whereas
lyst network is endogenous. For example, if firm j is a the 75th percentile is 576. Most of the indirect connec
supplier to firm k then a shock to firm k may impact firm tions are to firms in different three-digit SIC code indus
j’s characteristics. This raises the possibility that we could tries. The mean (median) number of across industry
still be picking up a contextual effect (i.e., firm i respond indirect connections is 394.18 (359).
ing to firm j’s characteristic) in the friends-of-friends Table 2 reports the average value of the outcome vari
approach. However, because we can partial out the ables we use in our analysis. We find that the average
return shocks of the direct peers, such a spillover would Leverage (change in Leverage) for the firms in our sample is
have to be significant enough to impact firm i’s financial 22% (1%). When we identify debt issuances as instances
Note. This table presents the descriptive statistics for the analyst coverage network in terms of number of direct and indirect peers, the average number of common analysts by firm-year broken
when there is a more than 1% increase in the book value
2.38
3.29
1.62
3.48
P75
1.8
1.5
4
of total debt relative to the book value of total assets, we
find that firms issue debt during 36% of the firm-years in
Average number of analysts in common
our sample period. We use two variables to identify

1.52
1.27
1.83
1.17
P50
equity issuances. Our first variable defines equity issu

1.2
1.9
2
ances as instances when the difference between cash flow

from equity issues less cash flow from equity repurchases

1.11
1.19
1.14
1.18
is greater than 1% of the book value of total assets, Net
P25
1
equity. Based on this definition, firms issue equity in 22%
of firm-year. When we define gross equity issuances as
years when the cash flow from equity issues is more than
1.07
2.77
0.73
1.88
0.65
1.99
0.56
SD
1% of the book value of total assets, Gross equity, we find

that equity issuances occur in 36% of the firm-years.
Table 2 also provides the summary information for all
Mean
1.91
3.13
1.56
2.53
1.45
2.65
1.38
the control variables in our sample. The summary values

are similar to those for the full CRSP-Compustat-IBES
merged sample. We winsorize all our variables of interest
41,411
35,497
40,030
38,122
39,503
38,207
39,993
at the 1st and 99th percentiles.

N
In Table 3, we provide the summary information for

Equity shock. Although the average value of the own
0.46
0.96
0.68
0.85
0.64
0.83
0.09
0.21
0.97
0.12
0.98
P75
firm’s equity shock (Equity shockOWN ) in our sample is

close to zero at �0.03, it has sufficient dispersion with a
0.16
0.84
0.38
0.62
0.39
0.61
0.01
0.99
0.08
0.92
0.05
0.95
P50
standard deviation of 0.49. As expected, the dispersion

Connections (%)
Connections (%)
declines when the equity shock is averaged over either

0.04
0.54
0.15
0.32
0.17
0.36
0.91
0.03
0.79
0.02
0.88
P25
the industry or analyst peer firms.

0
0.31
0.31
0.29
0.29
0.13
0.13
0.15
0.15
0.11
0.11
SD
0.3
0.3
4. Empirical Results
In this section, we discuss our empirical results. The dis
Mean
0.29
0.71
0.42
0.58
0.42
0.58
0.07
0.93
0.14
0.86
0.09
0.91
cussion is divided into four parts. First, we document a

positive association between a firm’s financial policy and
0.96
P75
that of its analyst peers. Second, we use Equity shock as an

576
555
528
551
58
15
46
26
37
25
37
25
61
38
exogenous peer firm characteristic to establish the exis

tence of social effects distinct from correlated effects. We
0.94
Number of connections
P50
382
359
335
356
38
26
13
19
14
19
31
21
6
then use the friends-of-friends approach to isolate endog

Number of connections
enous peer effects from exogenous peer effects. Finally,

0.92
P25
222
202
182
196
21
11
12
2
4
8
5
8
we provide a series of robustness and placebo tests to dis

Table 1. Analyst Coverage Network Summary Statistics
tinguish peer effects operating through analyst networks

0.02
27.33
12.74
26.04
22.42
14.19
22.79
32.72
41.69
20.32
238.07
239.32
235.09
238.32
SD
16
from those operating within industries and perform sev

eral cross-sectional tests that provide a richer picture of
the mechanism underlying the peer effects.
0.94
42.09
10.58
17.06
25.02
16.65
25.44
20.54
43.36
25.13
414.72
394.18
371.36
389.59
Mean
31.5
4.1. Baseline Regressions

41,411
41,411
41,411
41,411
41,411
41,411
41,411
41,411
41,411
41,411
41,411
41,411
41,411
41,411
23
We provide the results of estimating Equation (1) in our

N
down into within and across industries.
full sample in Table 4. The outcome variable in columns

(1) and (3) is Leverage in first difference and level, respec
Across industries (3 Digit-SIC code)
Across industries (3 Digit-SIC code)

Within industry (3 Digit-SIC code)
Within industry (3 Digit-SIC code)

Across industries (GICS Industry)
Across industries (GICS Industry)

Within industry (GICS Industry)
Within industry (GICS Industry)
tively. The positive and significant coefficient on Industry

Across industries (F-F Industry)
Across industries (F-F Industry)

Within industry (F-F Industry)
Within industry (F-F Industry)
average highlights the positive association between a

firm’s leverage and average leverage of other firms in its
industry (Welch 2004, Frank and Goyal 2008). Coeffi
cients on the firm-specific control variables are consistent
Connected firms
Indirect peers
with prior studies (Rajan and Zingales 1995). From the

Direct peers
coefficients on the industry average characteristics, we

Overall
Overall
find that only industry average Profitability is consistently

related to firm leverage. Consistent with the findings in
Table 2. Summary Statistics: Outcome and Control Variables
Panel A: Firm-specific variables
Focal Firms Industry average (Three-Digit SIC Code) Peer firm simple average
N Mean SD Median Mean SD Median Mean SD Median
∆Leverage 41,411 0.01 0.1 0 0.01 0.05 0.01 0.01 0.04 0.01
Leverage 41,411 0.22 0.22 0.15 0.23 0.14 0.2 0.21 0.11 0.19
Net debt 41,411 0.36 0.48 0 0.33 0.17 0.31 0.38 0.17 0.37
Net equity 41,411 0.22 0.42 0 0.22 0.15 0.21 0.22 0.18 0.18
Gross equity 41,411 0.36 0.48 0 0.3 0.18 0.3 0.38 0.23 0.33
Log(Sales) 41,411 6.74 1.79 6.64 5.84 1.15 5.7 7.11 0.95 7.16
Market to book 41,411 1.68 1.27 1.28 1.62 0.67 1.44 1.83 0.76 1.66
Profitability 41,411 0.12 0.11 0.13 0.09 0.07 0.1 0.14 0.05 0.14
Tangibility 41,411 0.29 0.23 0.22 0.28 0.19 0.22 0.29 0.16 0.26
∆Log(Sales) 41,411 0.1 0.22 0.08 0.09 0.11 0.1 0.11 0.1 0.11
∆Market to book 41,411 �0.04 0.8 0 �0.06 0.39 �0.04 �0.06 0.44 �0.01
∆Profitability 41,411 0 0.07 0 �0.01 0.03 0 0 0.02 0
∆Tangibility 41,411 0 0.04 0 0 0.02 0 0 0.01 0
Log(1+Coverage) 41,411 2.16 0.73 2.2 1.92 0.32 1.92 2.42 0.31 2.45
Panel B: Peer firm weighted averages
Full sample Same industry Different industry
N Mean SD Median Mean SD Median Mean SD Median
∆Leverage 41,411 0.01 0.05 0.01 0.01 0.06 0 0.01 0.05 0

Leverage 41,411 0.21 0.12 0.19 0.18 0.17 0.13 0.2 0.11 0.19
Net debt 41,411 0.38 0.18 0.37 0.32 0.3 0.28 0.36 0.2 0.36
Net equity 41,411 0.22 0.18 0.17 0.19 0.25 0.06 0.2 0.19 0.14
Gross equity 41,411 0.38 0.24 0.34 0.34 0.34 0.25 0.35 0.25 0.31
Log(Sales) 41,411 7.28 1.05 7.3 6.23 2.82 6.96 7.09 1.7 7.36
Market to book 41,411 1.86 0.83 1.66 1.61 1.17 1.41 1.72 0.78 1.61
Profitability 41,411 0.14 0.05 0.15 0.12 0.08 0.13 0.14 0.05 0.14
Tangibility 41,411 0.3 0.18 0.26 0.26 0.23 0.18 0.27 0.15 0.25
∆Log(Sales) 41,411 0.11 0.11 0.11 0.1 0.14 0.09 0.1 0.11 0.1
∆Market to book 41,411 �0.06 0.46 �0.01 �0.05 0.54 0 �0.05 0.45 0
∆Profitability 41,411 0 0.03 0 0 0.03 0 0 0.03 0
∆Tangibility 41,411 0 0.01 0 0 0.02 0 0 0.01 0
Log(1+Coverage) 41,411 2.52 0.36 2.55 2.17 0.99 2.49 2.39 0.58 2.49
Notes. This table reports the summary statistics for the outcome and control variables that we employ in our analysis. All variables used in the
regression analysis are winsorized at the 1st and 99th percentiles and are defined in Appendix A.
Leary and Roberts (2014), firms from more profitable coefficient on Peer average is positive and significant. The
industries have higher leverage. coefficient on Peer average is significantly larger than that
In columns (2) and (4), we augment the model with on Industry average, and inclusion of Peer average reduces
Peer average, the weighted average leverage (in first dif the size of the coefficient on Industry average in first differ
ference and level) of all firms in the analyst network. We ence (level) from 0.507 (0.435) to 0.285 (0.302). This is con
also include the weighted average characteristics of the sistent with analyst peer firm leverage having a large
analyst peer firms in the regressions. We find that the effect on a firm’s leverage. Focusing on the peer firm
Table 3. Summary Statistics: Equity Shocks
N Mean SD P25 Median P75

OWN
Equity shock 41,411 �0.03 0.49 �0.32 �0.09 0.14
Equity shockIND (three-digit SIC code) 41,411 �0.03 0.16 �0.12 �0.04 0.04
Equity shockACN (weighted average) 41,411 �0.04 0.12 �0.1 �0.04 0.02
Equity shockACN (simple average) 41,411 �0.03 0.11 �0.09 �0.04 0.02
Equity shockACN (indirect peer) 41,411 �0.03 0.05 �0.06 �0.03 �0.01
Equity shockACN (indirect peer excl. same industry) 41,411 �0.03 0.05 �0.06 �0.03 �0.01
Notes. This table shows the summary statistics of the equity shock for the individual firms, industry peers; and direct and indirect peers. All
variables used in the regression analysis are winsorized at the 1st and 99th percentiles.
Table 4. Baseline Specification I. Analyst Peer Firms vs. Industry Peers
∆Leverage Leverage Net Equity Gross Equity Net Debt
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Peer average 0.547 0.391 0.301 0.278 0.216

(0.020)*** (0.024)*** (0.022)*** (0.020)*** (0.021)***
Industry average 0.507 0.285 0.435 0.302 0.287 0.199 0.321 0.238 0.204 0.152
(0.017)*** (0.016)*** (0.022)*** (0.021)*** (0.024)*** (0.023)*** (0.023)*** (0.023)*** (0.022)*** (0.022)***
Own characteristics
Log(Sales) 0.026 0.021 0.040 0.039 �0.092 �0.090 �0.075 �0.075 �0.053 �0.055
(0.003)*** (0.003)*** (0.004)*** (0.004)*** (0.008)*** (0.008)*** (0.009)*** (0.009)*** (0.008)*** (0.009)***
Market to book 0.001 0.002 �0.019 �0.018 0.064 0.061 0.068 0.067 0.010 0.009
(0.0005)*** (0.0005)*** (0.001)*** (0.001)*** (0.004)*** (0.004)*** (0.004)*** (0.004)*** (0.004)** (0.004)**
Profitability �0.028 �0.023 �0.299 �0.294 �0.167 �0.170 0.159 0.159 0.334 0.318
(0.009)*** (0.009)** (0.018)*** (0.018)*** (0.041)*** (0.042)*** (0.042)*** (0.042)*** (0.046)*** (0.046)***
Tangibility 0.093 0.087 0.069 0.071 0.100 0.103 0.074 0.073 0.299 0.305
(0.013)*** (0.013)*** (0.024)*** (0.023)*** (0.044)** (0.043)** (0.047) (0.047) (0.047)*** (0.047)***
Log(1+Coverage) 0.008 0.008 0.001 �0.003 �0.056 �0.051 �0.048 �0.041 0.021 0.022
(0.0009)*** (0.001)*** (0.003) (0.003) (0.007)*** (0.007)*** (0.008)*** (0.008)*** (0.009)** (0.009)**
Log(1+Connections) �0.005 �0.003 �0.016 �0.016 0.036 0.040 0.040 0.040 0.028 0.029
(0.0009)*** (0.0009)*** (0.003)*** (0.003)*** (0.006)*** (0.006)*** (0.006)*** (0.006)*** (0.007)*** (0.007)***
Peer characteristics
Log(Sales) 0.013 �0.011 0.012 0.020 �0.0009
(0.007)* (0.003)*** (0.007)* (0.007)*** (0.008)
Market to book �0.004 0.005 0.0009 �0.013 0.0006
(0.001)*** (0.002)** (0.007) (0.007)* (0.007)
Profitability �0.036 0.078 0.063 �0.075 0.086
(0.026) (0.036)** (0.089) (0.091) (0.101)
Tangibility 0.036 �0.060 0.024 0.056 �0.080
(0.043) (0.024)** (0.049) (0.053) (0.060)
Log(1+Coverage) �0.004 0.018 �0.031 �0.034 �0.007
(0.002)** (0.006)*** (0.014)** (0.016)** (0.018)
Industry characteristics
Log(Sales) �0.003 �0.007 �0.010 �0.008 0.009 0.005 �0.003 �0.006 �0.008 �0.007
(0.006) (0.006) (0.005)** (0.005)* (0.009) (0.009) (0.011) (0.011) (0.012) (0.012)
Market to book �0.006 �0.003 �0.005 �0.005 0.014 0.006 0.011 0.012 0.017 0.016
(0.002)*** (0.002) (0.003) (0.003) (0.007)** (0.008) (0.008) (0.009) (0.008)** (0.009)*
Profitability 0.031 0.024 0.135 0.114 �0.036 �0.017 �0.117 �0.098 0.120 0.052
(0.020) (0.022) (0.028)*** (0.029)*** (0.067) (0.070) (0.076) (0.079) (0.079) (0.082)
Tangibility �0.076 �0.068 0.049 0.036 0.105 0.070 0.116 0.080 �0.055 �0.019
(0.036)** (0.037)* (0.043) (0.042) (0.074) (0.073) (0.088) (0.087) (0.098) (0.098)
Log(1+Coverage) 0.006 0.005 0.006 0.003 �0.047 �0.031 �0.061 �0.047 0.001 0.004
(0.002)*** (0.002)** (0.007) (0.007) (0.013)*** (0.013)** (0.015)*** (0.015)*** (0.017) (0.017)
Observations 41,411 41,411 41,411 41,411 41,411 41,411 41,411 41,411 41,411 41,411
R2 0.153 0.179 0.775 0.781 0.394 0.399 0.457 0.461 0.234 0.237
Notes. The table presents the OLS coefficients for the following regression:
′ ′
yijt � α + β1 yACN IND ′ ACN ′ IND ′
�it + β2 y�ijt + γ1 X�it�1 + γ2 X�ijt�1 + γ3 Xijt�1 + δ ui + φ vt + ɛijt :
The outcome variables are ∆Leverage (columns (1) and (2)), Leverage (columns (3) and (4)), Net Equity (columns (5) and (6)), Gross Equity (columns
(7) and (8)), and Net Debt (columns (9) and (10)). The main independent variables are Peer average and Industry average, which measure the
contemporaneous averages of the respective outcome variable for analyst and industry peers, respectively. All other control variables are lagged
one period, except the number of firm’s connections. In the specification with ∆Leverage as the dependent variable, all the control variables are in
first difference form, except the number of firm’s connections and analyst coverage; and we include year and industry fixed effects (three-digit
SIC code). The remaining specifications include firm and year fixed effects. Standard errors are clustered at the firm level. All variables are
winsorized at the 1st and 99th percentile and are defined in Appendix A. For brevity, we suppress the constant. Standard errors are in
parentheses. Statistical significance at the 10%, 5% and 1% levels are denoted by *, ** and ***, respectively.
characteristics, we find that only the coefficients on peer issuances by a firm and equity issuances by analyst peer
firm average Log(Sales), Market to book and analyst cover firms in the same year. The coefficients on both Peer aver
age are significant in both columns. age and Industry average are of similar magnitude. In col
In columns (5) to (8), we focus on equity issuances umns (9) and (10), we repeat our tests with Net debt as the
and, irrespective of our measure of equity issuance, we dependent variable, and from column (10), we find that
find that there is a positive association between equity there is a positive association between the probability of
debt issuances by a firm in a year and debt issuances of From columns (1) and (2), we find that all three equity
analyst-connected peer firms. Here again we find that shock variables (lagged one period), Equity shockOWN , Equity
the coefficient on Peer average is larger than that on In shockIND , and Equity shockACN are negatively associated
dustry average. Interestingly, we find that none of the with a firm’s market leverage (first difference and level).
industry or analyst peer characteristics are significantly The negative and significant coefficient on Equity shockACN
related to a firm’s decision to issue debt. Overall, our is consistent with the presence of social effects within
results in Table 4 show that firm financial policies are the analyst network. When we model leverage (col
positively related to the financial policies of firms that umn 2), our coefficient estimates on Equity shockIND
are connected through common analysts. The magni and Equity shockOWN are similar to those reported in
tude of the association is greater than that between firm Leary and Roberts (2014) (table 4). In the change speci
financial policy and industry average financial policies. fication (column (1)), however, the industry average
In Table IA-2 of the online appendix, we differentiate shock becomes statistically insignificant once we con
between analyst peers that are from the same industry trol for Equity shockACN .
and those that are from different industries to see if these Columns (3) and (4) indicate a strong positive associa
two groups have a similar effect on firm financial deci tion between Equity shockACN in a year and the probabil
sions. We do this by replacing Peer average with two ity of a firm making equity issues the next year. This
variables: Peer average (same industry) and Peer average suggests that the leverage results in the first two columns
(different industry). These are the weighted averages of are driven, at least in part, by the influence of analyst net
the outcome variable for same and different industry work peers on firms’ equity issuance decisions. In col
analyst peers. We calculate the weighted average using umn (5), we estimate a similar specification for net debt
the methodology outlined in Section 3. From columns (1) issuances. Consistent with the leverage and equity issu
and (2) of Table IA-2, we find that the coefficients on ance results, we find a negative coefficient on the peer
both same and different industry peer averages are posi firm equity shock, although the statistical significance is
tive and significant. The coefficients are also of similar weak. In addition, we find an insignificant relation
size. This indicates that firm leverage is related similarly between Net Debt and the firm’s own equity shock,
to the leverage of analyst peers from both the same and which makes the coefficient on the peer average equity
different industries. In unreported tests we find that the shock difficult to interpret. The results in column (6) indi
coefficients on same and different industry peer averages cate that this weak relation between debt issues and
are not statistically distinguishable. These results further equity shocks is due largely to the comparison set. That is,
reinforce the conclusion that the analyst network may column (5) compares observations with a debt issuance to
have an independent effect on firm leverage apart from all other observations. In column (6), we restrict the sam
the industry effect documented in Leary and Roberts ple to periods of active financing decisions, that is, where
(2014). We find similar results for net debt issuance and there is either a debt issuance or an equity issuance, to test
both net and gross equity issuance. It is noteworthy that for the influence of peer firms on the debt-equity choice.
the different industry analyst peers have a larger influ The dependent variable is an indicator equal to one (zero)
ence on a firm’s decision to issue equity as compared if the issuance is debt (equity). The results indicate that,
with same industry analyst peers. conditional on some form of issuance, the choice between
debt and equity is strongly related both to the firm’s own
4.2. Reduced Form and Structural Regressions equity shock and that of its analyst network peers.18 Over
Having established a positive association between ana all, our evidence in Table 5 shows that there appear to be
lyst peers’ financial policies and a firm’s own financial strong social effects within analyst networks for leverage,
policy, we now go to our next set of tests wherein we use equity issuance, and security choice decisions.
Equity shock as an exogenous peer firm characteristic to Agency-based considerations may play a role in man
control for correlated effects. In Table 5, we report the agerial responses to peer effects, because managers may
results of a reduced form estimation wherein we include prefer leverage-decreasing versus leverage-increasing
Equity shockACN and Equity shockIND instead of peer and decisions to reduce the disciplining role of debt (Jensen
industry average financial policy and repeat our tests.17 1986, Hart and Moore 1994, Novaes 2002). In Table IA-3
We perform the reduced form analysis to provide evi in the online appendix, we test whether we observe any
dence of social effects (endogenous or exogenous). How difference in the economic significance of the peer effect
ever, as discussed previously, this specification cannot between positive and negative peer equity shocks. We
distinguish endogenous from exogenous peer effects. In add a dummy variable, Positive, when peers’ idiosyncratic
this table, we also include Equity shockIND to highlight that returns are positive. Our results provide no evidence
the effect of Equity shockACN is robust to controlling for of relative differences between leverage-increasing and
industry characteristics, suggesting that our peer effect leverage-decreasing peer effects, suggesting that agency
results are not only due to peer firms from the same indus conflicts may not be a first-order driver of firms’ responses
try. We explore this issue further in subsequent tests. to their peers.
Table 5. Reduced Form Regression with Idiosyncratic Equity Shock
∆Leverage Leverage Net Equity Gross Equity Net Debt Net Debt (Issuer sample)
(1) (2) (3) (4) (5) (6)
Equity shockACN �0.026 �0.024 0.056 0.079 �0.039 �0.099

(0.005)*** (0.006)*** (0.019)*** (0.019)*** (0.023)* (0.030)***
Equity shockIND �0.004 �0.020 0.002 0.019 0.007 �0.024
(0.004) (0.004)*** (0.013) (0.014) (0.017) (0.023)

Equity shockOWN �0.006 �0.017 0.057 0.072 �0.006 �0.062
(0.001)*** (0.002)*** (0.005)*** (0.005)*** (0.005) (0.007)***
Own characteristics
Log(Sales) 0.025 0.042 �0.090 �0.075 �0.053 0.034
(0.003)*** (0.004)*** (0.008)*** (0.009)*** (0.009)*** (0.012)***
Market to book 0.003 �0.016 0.056 0.060 0.010 �0.028
(0.0006)*** (0.001)*** (0.004)*** (0.004)*** (0.004)** (0.005)***
Profitability �0.023 �0.295 �0.179 0.142 0.327 0.061
(0.009)** (0.018)*** (0.042)*** (0.043)*** (0.046)*** (0.060)
Tangibility 0.085 0.060 0.121 0.092 0.300 0.085
(0.013)*** (0.024)** (0.044)*** (0.047)* (0.047)*** (0.068)
Log(1+Coverage) 0.009 �0.004 �0.038 �0.027 0.022 0.041
(0.001)*** (0.003) (0.008)*** (0.008)*** (0.009)** (0.012)***
Log(1+Connections) �0.004 �0.016 0.036 0.038 0.029 �0.016
(0.0009)*** (0.003)*** (0.006)*** (0.006)*** (0.007)*** (0.009)*
Peer characteristics
Log(Sales) 0.043 0.0005 �0.005 0.003 0.003 �0.006
(0.007)*** (0.003) (0.007) (0.007) (0.008) (0.011)
Market to book �0.004 �0.010 0.021 0.007 0.002 �0.009
(0.002)*** (0.002)*** (0.007)*** (0.007) (0.007) (0.009)
Profitability �0.059 �0.099 �0.083 �0.068 0.154 0.100
(0.028)** (0.037)*** (0.088) (0.091) (0.101) (0.137)
Tangibility 0.125 0.042 0.037 0.027 �0.041 �0.010
(0.044)*** (0.024)* (0.049) (0.053) (0.060) (0.080)
Log(1+Coverage) �0.004 0.003 �0.029 �0.015 �0.014 �0.005
(0.002)** (0.007) (0.015)* (0.015) (0.018) (0.023)
Industry characteristics Yes Yes Yes Yes Yes Yes
Observations 41,411 41,411 41,411 41,411 41,411 20,150
R2 0.118 0.768 0.395 0.458 0.231 0.613
Notes. The table presents the results of the following OLS regression:
yijt � α0 + α1 Equity shockACN IND OWN ′ ACN ′ IND ′ ′ ′
�it�1 + α2 Equity shock�ijt�1 + α3 Equity shockijt�1 + γ1 X�it�1 + γ2 X�ijt�1 + γ3 Xijt�1 + δ ui + φ vt + ɛijt :
The outcome variables are ∆Leverage (column (1)), Leverage (column (2)), Net Equity (column (3)), Gross Equity (column (4)), and Net Debt (column
(5)). Column (6) isolates the subsample of observations in which either an equity or debt issuance, but not both, occurred (Net Debt (Issuer
sample)). The main independent variables are Equity shockACN and Equity shockIND which measure the lagged averages of the idiosyncratic equity
shock for analyst peers and industry peers, respectively. All control variables are lagged one period, except the number of firm’s connections. In
the specification with ∆Leverage as the dependent variable, all the control variables are in first difference form, except the number of firm’s
connections and analyst coverage; and we include year and industry fixed effects (three-digit SIC code). The remaining specifications include
firm and year fixed effects. Standard errors are clustered at the firm level. All variables are winsorized at the 1st and 99th percentile and are
defined in Appendix A. For brevity we suppress the constant. Standard errors are in parentheses. Statistical significance at the 10%, 5% and 1%
levels are denoted by *, ** and ***, respectively.
In Table IA-4 of the online appendix, we use alternate statistical significance, in unreported tests, we repeat
thresholds to define the net (gross) equity issuance the estimation after replacing Equity shockACN with
dummy (1%, 3%, and 5% of total assets). In all cases, we Equity shockACN (simple average). We find that the coeffi
find that our results are robust to using different thresh cients on Equity shockIND continue to be insignificant
olds to identify net and gross equity issuance. (and the coefficient on Equity shockACN (simple average)
Once we include Equity shockACN , the coefficients on significantly positive) in that specification. This high
Equity shockIND are insignificant in all the columns. lights that it is the fact that Equity shockACN averages over
Equity shockACN is different from Equity shockIND along a specific set of peers that is responsible for soaking up
two dimensions. First, it averages across firms connected the effect of Equity shockIND .
through common analysts irrespective of their industry SEO decisions are another corporate financial policy
affiliation. Second, it is a weighted average with the that can be influenced by peer effects. The influence of
weights equal to the number of common analysts. To see analysts on equity issuance decisions has been previ
which of these is responsible for Equity shockIND losing ously documented: Chang et al. (2006) find that firms
covered by fewer analysts are less likely to issue equity, information asymmetry around security issuances and
and Bowen et al. (2008) and Chan and Chan (2014) find of information spillovers across firms in the analyst’s
that analyst coverage can mitigate the information asym coverage network (Chan and Chan 2014).
metry associated with SEO discounts. In Table 6, we In Table 7, we provide the results of the two-stage least
investigate peer effects on SEOs. Specifically, we test squares estimation that uses Equity shockACN as an instru
whether a firm is more likely to have an SEO event when ment for the average financial policies of peer firms. In
peer firms have positive idiosyncratic returns in the pre all the specifications we also include the average finan
vious year and whether SEOs that follow peers’ deci cial policies of firms in the same industry as an additional
sions are associated with better announcement returns. control. At the top of Table 7, we provide the coefficients
We obtain SEO events from Securities Data Company on the instruments from the first stage regression. Esti
(SDC) Platinum and use a linear probability model with mating the 2SLS has advantages and disadvantages rela
the same econometric specification for equity issuances tive to the reduced form. The advantage is that it allows
used in Table 5. The dependent variable in our panel us to estimate the magnitude of the impact of analyst
data analysis is an indicator variable, SEO, that takes the peer firm policies on firms’ financial decisions. The limi
value of one when a firm has an SEO in a given year and tation, however, is that interpreting the magnitude in
zero otherwise. In column (1), we find that peer equity this way requires us to assume that the peer firms’ equity
shocks are positively associated with future SEOs of shock influences firm i’s financial policy only through its
firms; higher idiosyncratic returns of analyst coverage effect on peers’ financial policies. As discussed earlier, it
network peer firms increase the probability that the focal is possible that peers’ equity shock influences firm i’s pol
firm issues new seasoned equity in the next year. icies because it is a shock to the peers’ characteristics,
In columns (2) to (4), we analyze SEO announcement such as investment opportunities or competitive posi
returns using raw returns, market-adjusted returns, and tion. This would represent an exogenous peer effect, in
buy and hold abnormal returns, respectively, within a 6 which case we would be wrong to attribute the entire
three-day window around the filing date. Across all magnitude to endogenous peer effects, that is, the effect
three specifications, we find that equity shocks of peer of peers’ policies on firm i’s policies.
firms are positively and significantly associated with Despite this caveat, the results in Table 7 are instruc
SEO announcement returns. These results suggest, first, tive. The first stage results indicate that Equity shockACN
that firms are influenced by their analyst network peers is significantly related to peer firm leverage (columns (1)
also when making SEO decisions. Furthermore, SEOs and (2)) and equity issuance (columns (3) and (4)) deci
that follow those of analyst network peers appear to sions. Furthermore, the F values for weak instrument
be associated with higher announcement returns, con tests shown at the bottom of the table are all large and
sistent with previous evidence that analysts reduce greater than the threshold of 10.
Table 6. SEOs and Idiosyncratic Equity Shock
SEO CRET[-3,3] CAR[-3,3] BHAR[-3,3]

(1) (2) (3) (4)
Equity shockACN 0.028 0.050 0.045 0.045

(0.010)*** (0.023)** (0.022)** (0.021)**
Equity shockIND 0.007 0.005 �0.0003 �0.0008
(0.008) (0.019) (0.018) (0.018)
Equity shockOWN 0.028 0.003 �0.006 �0.006
(0.003)*** (0.004) (0.003)* (0.003)*
Own characteristics Yes Yes Yes Yes
Peer characteristics Yes Yes Yes Yes
Industry characteristics Yes Yes Yes Yes
Observations 41,411 2,341 2,341 2,341
R2 0.186 0.097 0.105 0.106
Mean SEO announcement returns �0.008*** �0.021*** �0.022***
Notes. The table presents the results of the effect of peer equity shocks on the probability of a SEO and the influence on the SEO announcement
returns. The outcome variables are SEO (column (1)), which is a dummy that takes the value of one if a firm in a given year had a SEO, and it is
zero otherwise. CRET[-3,3] (CAR[-3,3]) and BHAR[-3,3] are the cumulative raw (abnormal) returns and buy-and-hold abnormal returns,
respectively, in a three-day window surrounding the SEO announcement (filling) day. The main independent variables are Equity shockACN and
Equity shockIND, which measure the lagged averages of the idiosyncratic equity shock for analyst peers and industry peers, respectively. All
control variables are lagged one period, except the number of firm’s connections. In column (1) standard errors are clustered at the firm level,
whereas columns (2) to (4) are standard errors double clustered at the industry level (three-digit SIC code) and year level. Column (1) includes
firm and year fixed effects. Columns (2) to (4) include industry and year fixed effects. All variables are winsorized at the 1st and 99th percentile
and are defined in Appendix A. For brevity we suppress the constant. Standard errors are in parentheses. Statistical significance at the 10%, 5%
and 1% levels are denoted by *, ** and ***, respectively.
Table 7. Two-Stage Least-Square Instrumentals Variable Regression Using Equity Shock
∆Leverage Leverage Net Equity Gross Equity Net Debt (Issuer sample)
(1) (2) (3) (4) (5)
First stage
Equity shockACN �0.013 �0.020 0.078 0.113 �0.040
(0.002)*** (0.003)*** (0.007)*** (0.008)*** (0.011)***
Second stage
Instrumented peer average 1.723 1.043 0.672 0.660 2.117
(0.410)*** (0.302)*** (0.224)*** (0.163)*** (0.842)**
Industry average �0.185 0.086 0.095 0.122 �0.020
(0.165) (0.100) (0.066) (0.052)** (0.111)
Own characteristics
Equity shockOWN �0.005 �0.017 0.057 0.072 �0.060
(0.001)*** (0.001)*** (0.004)*** (0.004)*** (0.007)***
Log(Sales) 0.017 0.037 �0.088 �0.074 0.014
(0.003)*** (0.004)*** (0.008)*** (0.008)*** (0.013)
Market to book 0.004 �0.016 0.054 0.059 �0.028
(0.0007)*** (0.001)*** (0.004)*** (0.004)*** (0.004)***
Profitability �0.010 �0.289 �0.196 0.133 0.045
(0.010) (0.017)*** (0.039)*** (0.040)*** (0.058)
Tangibility 0.074 0.078 0.115 0.087 0.124
(0.014)*** (0.022)*** (0.040)*** (0.044)* (0.072)*
Log(1+Coverage) 0.007 �0.012 �0.033 �0.014 0.023
(0.001)*** (0.004)*** (0.007)*** (0.008)* (0.014)
Log(1+Connections) �0.002 �0.013 0.040 0.035 �0.014
(0.001)** (0.003)*** (0.006)*** (0.006)*** (0.011)
Peer characteristics Yes Yes Yes Yes Yes
Industry characteristics Yes Yes Yes Yes Yes
Observations 41,411 41,411 41,411 41,411 19,109
Kleibergen-Paap F value 52.651 53.391 112.661 208.026 12.801
Cragg-Donald F value 88.724 90.738 250.692 410.395 23.639
Anderson-Rubin F value 21.398 14.1 9.313 17.348 10.853
Anderson-Rubin p value 3.83 e-06 0.0002 0.002 0.00003 0.001
Notes. The table presents the results of the 2SLS regression that relates peer firm average financial policy to firm financial policy. The outcome
variables are ∆Leverage (column (1)), Leverage (column (2)), Net Equity (column (3)), and Gross Equity (column (4)). Column (5) isolates the
subsample of observations in which either an equity or debt issuance, but not both, occurred (Net Debt (Issuer sample)). The main endogenous
variables are peer averages of the outcome variables. The instrument in the first stage is Equity shockACN, which measures the lagged averages of
the idiosyncratic equity shock for analyst peers. All control variables are lagged one period, except the number of firm’s connections. In the
specification with ∆Leverage as the dependent variable all the control variable sare in first difference form, except the number of firm’s
connections and analyst coverage; and we include year and industry fixed effects (three-digit SIC code). The remaining specifications include
firm and year fixed effects. Standard errors are clustered at the firm level. All variables are winsorized at the 1st and 99th percentile and are
defined in Appendix A. For brevity, we suppress the constant. Standard errors are in parentheses. Statistical significance at the 10%, 5% and 1%
levels are denoted by *, ** and ***, respectively.
Focusing on the results of the second stage, we find Tamer 2014, Leary and Roberts 2014). Nonetheless, we
that the coefficient on the instrumented peer average can get a sense of the relative magnitude of different fac
leverage is positive and significant in columns (1) and tors by comparing scaled coefficients across the indepen
(2), consistent with the presence of peer effects in lever dent variables. By this measure, the leverage decisions of
age decisions that propagate through analyst networks. network peers seems to have a larger impact on firms’
Our estimates are also economically significant. The own leverage choice than the standard included firm
coefficient on Peer average in column (2) indicates that a characteristics. Log(Sales) is closest in magnitude, with a
1-standard-deviation increase in peer firm weighted scaled coefficient of 0.3.
average leverage is associated with a 0.569-standard- From columns (3) and (4), we find that the decision of
deviation increase in the firm’s leverage (0.569 � 1.043 × peer firms to issue equity in a year is associated with the
(0.12/0.22)). However, this is only a partial effect. In a lin own firm’s decision to issue equity. We find that the
ear model of social interactions, any impact on firm i’s effect of analyst peers is greater than the effect of indus
leverage will feed back to that of its peers (given a non try peers. Our estimates are also economically signifi
zero β1 in Equation (1)). In the presence of these feedback cant. Scaling by the relevant standard deviations, the
effects, the coefficients on y�it (and X�ijt ) do not necessar coefficients on instrumented peer equity issuance are
ily give the full marginal effect of a change in the inde large relative to those on the included firm characteristics
pendent variable on the outcome variable (Kline and (only firm size has a larger scaled coefficient in the net
equity specification). Results in column (5) similarly find instruments are uncorrelated with the characteristics of
a significant positive relation between the debt-equity the direct peers. This is likely to be true for idiosyncratic
choice of analyst network peers (conditional on an issu return shocks as they isolate value-relevant events that
ance) and a firm’s own debt-equity choice. are unique to the indirect peers (i.e., this is the identifica
tion assumption used in the previous section).
4.3. Indirect Peer Approach The first row of Table 8 presents the coefficients on the
In Table 8, we use the friends-of-friends methodology to indirect peer average equity shock from the first stage.
help isolate exogenous peer effects from endogenous peer We find that the equity shocks of indirect peer firms are
effects. Specifically, we identify indirect peer firms for significantly related to the level and change in leverage,
every firm. These are firms that are not directly con net and gross equity issuances, and the debt-equity
nected to a firm through common analysts but are con choice of direct peers. Furthermore, the F values indicate
nected to one or more of its analyst network peers. We that for these policy variables the instrument easily
then estimate a two-stage least squares model in which passes the weak instrument test.
we use the average equity shock of these “indirect Regarding the second stage, in the odd numbered col
peers” as an instrument for the financial policies of a umns, which include all indirect peers, we find a signifi
firm’s direct peers to identify endogenous peer effects cant relation between a firms’ financial policies and
in financial policy.19 those of their direct peers for both the change and the
As discussed in Section 3, to separate contextual from level of leverage, gross equity issuances, and the debt-
endogenous peer effects, the key identification assump equity choice (though the coefficient on the instrumen
tion is that the characteristics of the indirect peers used as ted average peer debt-equity choice is significant only at
Table 8. Two-Stage Least Square Instrumentals Variable Regression Using Indirect Peers
Indirect peers All Diff. Ind. All Diff. Ind. All Diff. Ind. All Diff. Ind. All Diff. Ind.
∆Leverage Leverage Net equity Gross equity Net debt (Issuer sample)
Dependent variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
First stage
Equity shockACN �0.038 �0.041 �0.077 �0.060 0.096 0.107 0.170 0.144 �0.119 �0.136
(indirect peers) (0.007)*** (0.007)*** (0.011)*** (0.011)*** (0.026)*** (0.027)*** (0.028)*** (0.028)*** (0.041)*** (0.039)***
Second stage
Instrumented peer average 0.914 1.110 0.964 1.019 0.517 0.499 1.083 1.054 1.530 1.384
(0.418)** (0.375)*** (0.274)*** (0.339)*** (0.668) (0.587) (0.412)*** (0.474)** (0.917)* (0.749)*
Industry average 0.069 0.042 0.037 0.034 0.045 0.045 0.036 0.036 0.039 0.042
(no overlap) (0.059) (0.053) (0.014)*** (0.017)** (0.024)* (0.022)** (0.016)** (0.017)** (0.026) (0.023)*
Peer average stock return �0.002 �0.003 0.002 0.004 0.019 0.020 �0.026 �0.024 �0.055 �0.053
(0.004) (0.004) (0.008) (0.010) (0.060) (0.053) (0.042) (0.048) (0.021)*** (0.019)***
Own characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
(incl. equity shock)
Peer characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Ind. characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Obs. 41,411 41,411 41,411 41,411 41,411 41,411 41,411 41,411 16,293 16,293
Kleibergen-Paap F-value 34.032 39.157 52.06 31.411 13.779 15.435 35.447 25.577 8.404 12.081
Cragg-Donald F-value 63.974 78.905 96.329 62.595 32.953 43.272 81.166 62.740 17.077 24.586
Anderson-Rubin F-value 4.56 8.788 13.497 10.722 0.614 0.794 7.879 5.927 3.395 4.027
Anderson-Rubin p-value 0.033 0.003 0.0002 0.001 0.433 0.373 0.005 0.015 0.065 0.045
Notes. The table presents the results of the 2SLS regression that relates peer firm average financial policy to firm capital structure. The outcome
variables are ∆Leverage (columns (1) and (2)), Leverage (columns (3) and (4)), Net Equity (columns (5) and (6)) and Gross Equity (columns (7) and
(8)). Columns (9) and (10) isolate the subsample of observations in which either an equity or debt issuance, but not both, occurred (Net Debt
(Issuer sample)). The main endogenous variables are peer averages of the outcome variables. The instrument in the first stage is Equity shockACN
(indirect peers) which measures the lagged averages of the idiosyncratic equity shock for indirect analyst peers. The odd columns show the
results of the indirect peer approach using the Equity shock of all indirect peers (All). Even columns show the results of the indirect peer approach
using the Equity shock of indirect peers that are only in different industries as firm i (Diff. Ind.). All control variables, but excluding Industry
average (no overlap) and the number of firm’s connections, are lagged one period. In the specification with ∆Leverage as the dependent variable all
the control variables are in first difference form, except the number of firm’s connections and analyst coverage; and we include year and industry
fixed effects (three-digit SIC code). The remaining specifications include firm and year fixed effects. Standard errors are clustered at the firm
level. All variables are winsorized at the 1st and 99th percentile and are defined in Appendix A. For brevity we suppress the constant. Standard
errors are in parentheses. Statistical significance at the 10%, 5% and 1% levels are denoted by *, ** and ***, respectively.
the 10% level). The positive and significant coefficients of social effects in Table 8. In the second stage, we now
on Peer average for those corporate policies suggest that find a significant relation between the instrumented peer
the average financial policy choice of analyst peer firms average policy and own-firm policies for the change in
has a causal effect on a firm’s financial policy choice. Our leverage, as well as for gross equity issues. Importantly,
results are also economically significant. From the coeffi these results are consistent whether we consider at least
cient in column (2), we find that a 1-standard-deviation two (columns (4) to (6)) or four (columns (7) to (9)) ana
increase in peer firm average leverage is associated with lysts in common with a direct peer to identify indirect
an initial 0.482-standard-deviation increase in a firm’s peers.20 Overall, these results support the findings in
leverage (0.482 � 0.964 × (0.11/0.22)). For equity issu Table 8 of endogenous peer effects working through ana
ances, we find that a 1-standard-deviation increase in lyst networks and show that these results are robust to
gross equity issuance likelihood by peers leads to a varying definitions of indirect peers.
0.520-standard-deviation increase in gross equity issu
ance likelihood. 4.4. Role of Analysts in Peer Effects
In the even numbered columns, we repeat our tests Results in the previous section provide evidence for
using an instrument that excludes all indirect peers in endogenous peer effects in analyst coverage networks.
the same industry as firm i. In these columns, we find Peer effects may be present among firms with common
similar peer effects for the change and level in firm analysts for several potential reasons. For one, these
leverage, gross equity issuance, and the debt-equity firms are often in the same or similar industries or share
choice. Although the coefficient for the level of net other economic linkages (e.g., suppliers and customers).
equity issuances is insignificant, it is of similar mag As such, they may be the firms that peers are most likely
nitude as that in column (5). Overall, these results to look to for comparison on matters such as financial
suggest that our results are robust to excluding same- policy (Kaustia and Rantala 2015, 2021). Additionally,
industry indirect peers. analysts may play a more direct role in facilitating peer
It is important to remark that the coefficients associ effects through their role as information intermediaries.
ated with industry averages of the outcome variables Analysts are constantly engaging in two-way communi
are also positive and statistically significant, but they are cation with firms in their coverage network. Recent
substantially smaller in comparison with peer firm endoge research (Martens and Sextroh 2021) suggests that this
nous variables. Our results thus suggest that analyst net communication helps transmit information across firms
works are likely an important source for industry peer and not just between firms and investors. To the extent
effects. that some of this information is relevant to financial deci
One concern with the test in Table 8 is that the number sion making, analysts may play a role in propagating
of indirect peers in each group is considerably larger financial policies across firms. In this section, we show
than the number of direct peers (or industry peers). This evidence consistent with analyst information transmis
diminishes variation in the average equity shock across sion as one mechanism behind our documented peer
indirect peer groups. Although the results in Table 8 effects. We begin by showing that the peer effects in ana
indicate that enough power remains to identify the lyst coverage networks are distinct from those in indus
endogenous peer effects, in Table IA-5 of the online tries. That is, the peer effects we document do not arise
appendix, we report results of a robustness check to only because firms with common analysts often are in
address this concern. Specifically, we limit the set of indi the same industry (Leary and Roberts 2014). Rather, the
rect peers to those with at least two (and four) analysts in analyst coverage network itself is a particular setting in
common with a direct peer (while still imposing that which peer effects are prevalent. We then show cross-
they have no analysts in common with the firm in ques sectional variation in the strength of peer effects that is
tion). This limits the size of the indirect peer groups and consistent with analysts playing a role in transmitting
produces cross-group dispersion in average equity information across firms. On both fronts, our strategy is
shock only slightly below that of the direct peer groups. to compare estimated peer effects across subsets of firms
The results are similar to those in Table 8, except for the within the analyst network. Because there is likely to be a
case of leverage in levels (we restricted our sample to large degree of overlap between the sets of indirect peers
include only firms with nonmissing indirect peers’ of these partitions of direct peers, we return for these
equity shock). In the first stage, we find a considerable tests to the specification in Table 5, which uses the idio
reduction in the size of the coefficient on the indirect syncratic return shocks of direct peers as instruments for
peers’ equity shock when the dependent variable is the their financial policy choices.
level of leverage, likely because we are now relying on
only a subset of the peers of the direct peers. However, 4.4.1. Distinguishing Analyst Network Effects from
we continue to find strong first-stage coefficients for the Industry Peer Effects. Although we control for industry
other three policy variables. We do not display the averages in all our tests and show that the effects are pre
results for net equity issuances as we do not find results sent for analyst peers not from the same industry (Table
IA-2 in the online appendix), this still raises the question Figure 1. (Color online) Example of Analyst Coverage
of whether analyst network effects that we document are Network
simply capturing industry peer effects. Our control for
industry averages may prove inadequate if the number
of analysts in common (which we use to form our
weighted average peer equity shock) between pairs of
firms in the same industry is higher in comparison with
pairs of firms across industries. To the extent that firms

in both the same industry and analyst network are more
similar and more influential, our analyst peer weighted
average may be a more precise measure of industry
effects than the simple industry average. We therefore
perform several additional tests to address this issue.
We start by re-estimating the regressions from Table 5,
but partitioning the set of peer firms to three different
groups: (i) firms that are both in the same industry and in
the analyst network; (ii) firms that are in the analyst net
work but not in the same industry; and (iii) firms that
are in the same industry but not in the analyst network.21
The results, tabulated in the online appendix (Table IA-6),
show significant evidence of social interaction effects
among all three groups of firms and, somewhat surpris Notes. In this figure, we present a hypothetical analyst network for
ingly, even stronger effects among analyst network only illustration purposes. In the figure, the shape-families represent
firms than among firms in the same industry. These industries, whereas the shapes represent individual firms. The lines
connecting the shapes represent common analysts.
results suggest that the peer effects operating through ana
lyst networks do not simply reflect industry peer effects.
A potential limitation with the previous test is that analyst network captures links across firms in different
analysts may choose to cover firms that are economically industries, then we should expect Equity shockACN (pseudo-
connected, even if not in the same industry. For instance, peer) to be significantly related to the corporate policies of
analysts might choose firms in other industries but con the focal firm.
nected through customer-supplier relationships. Thus, The results in Panel A of Table 9 show that there is no
firms that are in the same analyst network, but in differ significant relationship between the equity shocks of
ent industries, may exert influence on one another as a pseudo-peers and a firm’s financial policy. We repeat the
result of their product market connections rather than tests with alternate industry definitions, including two-
the analyst connection. In other words, the connection digit SIC codes (Panel B), GICS codes (Panel C), and
that an analyst creates between firms may proxy for eco the peer classification of Hoberg and Phillips (2010)
nomic linkages between those firms that as researchers according to a product similarity measure (Panel D).
we cannot perfectly observe. Results are again insignificant when we calculate Equity
We address this concern in Table 9 by performing a shockACN (pseudo-peer) over firms that are in the same two-
placebo test. Instead of using the average equity shock of digit SIC code as the analyst peer firms and that do not
firms in the same analyst network, we define a set of have any common analyst with the firm in question. The
pseudo peers that are in the same industry as the peer only coefficients that are significant in the right direction
firms in the analyst network, but do not share a common are obtained when we focus on firms within the same
analyst with firm i. To illustrate, refer to Figure 1, in GICS industries. Here we find that there is a negative
which numbered shapes represent firms, the shapes association between Equity shockACN (pseudo-peer) and
themselves (triangle, circle, etc.) represent an industry, Leverage, but no relation for leverage changes and
and the lines connecting the shapes represent common equity issuances. We also find a positive and signifi
analysts. Thus, circle-1, pentagon-1, square-1, and triangle- cant coefficient for Gross Equity (and weakly significant
1 represent firms that are connected to firm star-0 for the conditional debt-equity choice) when we define
through common analysts but are in a different industry. pseudo-peers using the Hoberg-Phillips industry
To conduct our placebo test, we focus on the firms in the peer definition, but not for leverage (levels or changes)
same industry as these firms but that do not have a com or net equity issues (Panel D). To summarize, we ob
mon analyst with star-0. These are firms pentagon-2 to tain very weak evidence for an association between
pentagon-4, square-2 to square-4, and triangle-2 to triangle-4. Equity shockACN (pseudo-peer) and firm financial poli
We refer to this average as the Equity shockACN (pseudo- cies. This is in contrast to the strong relation between a
peer) and repeat our tests with this simple average. If the firm’s financial policies and those of the firms that are
Table 9. Placebo Test with Pseudo Peers
∆Leverage Leverage Net Equity Gross Equity Net Debt (Issuer sample)
(1) (2) (3) (4) (5)
Panel A: Three-digit SIC code

Equity shockACN (pseudo-peer) �0.010 �0.011 �0.063 �0.016 0.032
(0.012) (0.014) (0.038) (0.042) (0.064)
Observations 40,428 40,428 40,428 40,428 19,606

R2 0.117 0.769 0.393 0.459 0.615
Panel B: Two-digit SIC code
Equity shockACN (pseudo-peer) �0.025 �0.030 �0.132 �0.038 0.131
(0.019) (0.024) (0.066)** (0.070) (0.109)
Observations 40,146 40,146 40,146 40,146 19,484
R2 0.114 0.77 0.394 0.459 0.614
Panel C: Fama-French industry classification
Equity shockACN (pseudo-peer) 0.033 0.006 �0.193 �0.104 0.124
(0.019)* (0.023) (0.066)*** (0.069) (0.105)
Observations 39,865 39,865 39,865 39,865 19,347
R2 0.112 0.771 0.397 0.461 0.617
Panel D: GICS industry classification
Equity shockACN (pseudo-peer) �0.017 �0.053 �0.039 0.086 �0.045
(0.015) (0.017)*** (0.052) (0.054) (0.080)
Observations 40,257 40,257 40,257 40,257 19,583
R2 0.112 0.769 0.399 0.462 0.616
Panel E: Hoberg-Phillips peers
ACN
Equity shock (pseudo-peer) �0.005 �0.007 0.052 0.111 �0.138
(0.012) (0.015) (0.045) (0.048)** (0.073)*
Observations 34,942 34,942 34,942 34,942 17,233
R2 0.117 0.767 0.404 0.467 0.62
Own characteristics (including equity shock) Yes Yes Yes Yes Yes
Industry characteristics (including equity shock) Yes Yes Yes Yes Yes
Pseudo-peer characteristics Yes Yes Yes Yes Yes
Notes. The table presents the results of OLS regressions that relate pseudo peer firm idiosyncratic equity shock to firm capital structure. The
outcome variables are ∆Leverage (column (1)), Leverage (column (2)), Net Equity (column (3)) and Gross Equity (column (4)). Column (5) isolates
the subsample of observations in which either an equity or debt issuance, but not both, occurred (Net Debt (Issuer sample)). The main independent
variable is Equity shockACN (pseudo-peer) that is the simple average equity shock of all firms that are in the same industry as a firm’s analyst peers
but that do not have any common analysts with the firm in question. We use alternate industry definitions to identify the pseudo-peers in the
different panels. All control variables are lagged one period, except the number of firm’s connections. In the specification with ∆Leverage as the
dependent variable all the control variables are in first difference form, except the number of firm’s connections and analyst coverage; and we
include year and industry fixed effects. The remaining specifications include firm and year fixed effects. In all the specifications we include as
control variables individual firm, industry and pseudo-peer characteristics (similar specification as in Table 5). Standard errors are clustered at
the firm level. All variables are winsorized at the 1st and 99th percentile and are defined in Appendix A. For brevity, we suppress the constant.
Standard errors are in parentheses. Statistical significance at the 10%, 5% and 1% levels are denoted by *, ** and ***, respectively.
in the same industries as the pseudo peers, but that are driven largely through a stronger effect on equity issu
in the analyst network of firm i. This suggests that our ance decisions. This suggests that peer effects in equity
previous results were not simply driven by economic issuance decisions are especially salient when informa
connections between industries, but that analyst net tion flows through analyst coverage networks, consis
works play a particular role in propagating financial tent with the analysts’ focus on covering equities and
policies across firms. their role in reducing information asymmetry around
Finally, in Table IA-7 in the online appendix, we esti security offerings.
mate similar regressions as the reduced form models in
Table 5 but include only the average equity shock of 4.4.2. Exogenous Variation in Common Analysts. To
industry peers to compare the strength of peer effects further address the concern that our results are driven
between industry peers and analyst network peers. The by omitted variables associated with analysts choosing
results indicate that the influence of peer firms on issu to cover economically connected firms, we exploit bro
ance choice is substantially larger through analyst co kerage house closures and mergers as a source of exog
verage networks than through industries, and this is enous variation in the analyst network. If common
analysts transmit peer effects, then an exogenous reduc firm-specific Equity shock and characteristics (we also
tion in common analysts should weaken the influence of include firm k′ s characteristics, Xkt�1 ).22 If common ana
peer firms on an individual firm’s decision. lysts play a role in facilitating peer effects, we expect the
We identify the closure and merger of brokerage influence of peer firms’ Equity shock on financial policy
houses following Kelly and Ljungqvist (2012) (see appen decisions of firm i to become weaker after the brokerage
dix A of their paper) and use their methodology to iden closure, given the loss of common analysts. Hence, we
tify the set of analysts (and affected firms) that stopped expect coefficients β1 and β2 to have the opposite sign
their coverage as a consequence of the brokerage shock. from β5. Consistent with our previous results (Table 5),
We follow the pair model methodology proposed by we expect a negative (positive) sign for β5 when we
recent papers that use exogenous shocks such as the consider leverage (equity issuances) as the outcome
death of directors/executives and the brokerage closures variable.
to capture how shocks in connection among firms lead to Table 10 presents the results. In even columns we per
changes in commonalities of corporate policy decisions form the regression using the first definition of treated
(Fracassi 2017), to changes in information spillovers of firm-pairs, whereas in odd columns we restrict the
common analysts (Martens and Sextroh 2021), and to sample to the second definition. In columns (1) to (4), the
changes in stock return comovement (Israelsen 2016). outcome variable is Leverage (change and level). We find
First, we identify and keep pairs of firms with at least that the sign of β1 and β2 associated with interaction
one common analyst that was affected by a brokerage terms Post(t + 2) × Peer Equity shock and Post(t + 3) ×
house closure/merger. Specifically, we classify firm- Peer Equity shock have the opposite sign of β5 and they
pairs as treated in two ways: (1) pairs of firms that were are statistically significant. In addition, the results are
connected by one analyst who stopped the coverage of at stronger when using the second definition (more de
least one of the two firms (i and/or k) as a consequence of manding) of treated firm-pairs. Moreover, we find that
the brokerage shock and (2) pairs of firms that were con Peer Equity shock has the same sign as in Table 5 in all the
nected by one analyst who stopped following firm i and columns. For equity issuances and conditional debt-
firm k at the same time, which is a more demanding equity choice, shown in columns (5) to (10), we also find
definition. that β1 and β2 have the opposite sign of β5, but the results
Because many brokerage closures/mergers take place are only statistically significant for equity issues where a
over a period of several months, and typically the connec common analyst stopped her coverage of both firms at
tion is broken or weakens the year after the merger/ the same time. In sum, our results suggest that after bro
closure of the brokerage house, we define the year of the kerage closures/mergers and the attendant reduction in
shock in the analyst network as the year after the brokerage common analysts, there is a reduction in the influence of
event. Moreover, as a consequence of using the lagged peer firms. This effect is more pronounced when com
peer firms’ equity shock, we exclude the year after of the mon analysts stopped their coverage of both firms simul
brokerage closure from the analysis. Hence, our sample taneously. These results suggest that analyst connections
contains the year before the brokerage event (t � �1), the play a particular role in facilitating peer effects and that
year of the brokerage event (t � 0), and years 2 (t � +2) and our earlier results are not entirely driven by economic
3 (t � +3) after that. linkages that we do not observe.
Following Martens and Sextroh (2021) and Israelsen
(2016), we perform the following regression to exploit 4.4.3. Analyst Influence. If analyst networks are impor
the exogenous changes in common analysts: tant in transmitting corporate policy decisions from one
firm to another, then the characteristics of the analyst
yikt � β0 + β1 Post(t + 2)ikt × Peer Equity shockkt�1
herself may be important for the strength of these peer
+ β2 Post(t + 3)ikt × Peer Equity shockkt�1 effects. More influential analysts should be more effec
+ β3 Post(t + 2)ikt + β4 Post(t + 3)ikt tive at transmitting policy-relevant information across
firms. Firms are also likely to take the comments of such
+ β5 Peer Equity shockkt�1 + α1 Equity shockIND
�ijt�1
analysts more seriously. We construct two measures that
+ α2 Equity shockOWN ′ ′ IND ′
ijt�1 + γ1 Xkt�1 + γ2 X�ijt�1 + γ3 Xit�1 capture the potential influence of analysts. Specifically,
from Institutional Investor magazine we collect the names
+ δ′ uik + φ′ vt + ɛikt : (3)
of the top four analysts (first, second, third, and runner-
Equation (3) is similar to the reduced form regression in up) for each industry during 1990–2015. We classify an
Table 5, although each observation is at the firm-pair analyst as being influential from the first year the analyst
level (firm i and firm k); Post(t + 2) and Post(t + 3) are appears in the Institutional Investor ranking. We classify
dummy variables that takes the value of one in years 2 brokerage houses that employ three or more influential
and 3 after the brokerage shock (and zero otherwise), analysts as All-star brokerage houses. These roughly repre
respectively. Moreover, we include year (vt) and firm- sent about 10% of all brokerage houses in our sample.
pair (uik) fixed effects, and we control for industry and We differentiate between all-star brokerage houses and
Table 10. Pair Model Regression and Brokerage Closures
Net Debt
∆Leverage Leverage Net Equity Gross Equity (Issuer sample)
1 2 1 2 1 2 1 2 1 2
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Post(t + 2) × Peer Equity shock 0.007 0.018 0.014 0.027 �0.025 �0.071 �0.019 �0.065 0.013 0.036
(0.004)* (0.006)*** (0.003)*** (0.006)*** (0.014)* (0.025)*** (0.015) (0.027)** (0.020) (0.039)
Post(t + 3) × Peer Equity shock 0.019 0.026 0.012 0.016 �0.022 �0.083 �0.011 �0.068 0.014 0.005
(0.003)*** (0.006)*** (0.004)*** (0.007)** (0.014) (0.026)*** (0.014) (0.026)*** (0.019) (0.036)
Post(t + 2) �0.007 �0.014 0.006 0.007 0.014 0.011 0.037 0.035 �0.017 �0.002
(0.002)*** (0.003)*** (0.002)*** (0.004)* (0.007)* (0.013) (0.008)*** (0.014)** (0.011) (0.018)
Post(t + 3) �0.003 �0.016 0.008 0.008 0.014 0.014 0.045 0.027 �0.030 �0.018
(0.002)* (0.003)*** (0.003)*** (0.005) (0.008) (0.016) (0.009)*** (0.018) (0.012)** (0.023)
Peer Equity shock �0.008 �0.015 �0.005 �0.007 0.006 0.029 0.014 0.030 �0.005 �0.014
(0.002)*** (0.004)*** (0.002)** (0.004)* (0.009) (0.017)* (0.009) (0.016)* (0.011) (0.021)
Equity shockIND �0.015 �0.036 �0.025 �0.028 0.085 0.153 0.129 0.162 �0.226 �0.228
(0.004)*** (0.008)*** (0.003)*** (0.008)*** (0.015)*** (0.032)*** (0.015)*** (0.033)*** (0.026)*** (0.060)***
Equity shockOWN 0.007 0.006 �0.021 �0.026 0.059 0.082 0.087 0.105 �0.090 �0.113
(0.002)*** (0.003) (0.002)*** (0.003)*** (0.007)*** (0.012)*** (0.007)*** (0.011)*** (0.009)*** (0.015)***
Own characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Peer and Industry characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year and firm-pair fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 29,480 8,816 29,480 8,816 29,480 8,816 29,480 8,816 14,096 4,196
R2 0.296 0.317 0.859 0.859 0.458 0.492 0.586 0.594 0.702 0.7
The table presents the results of the following pair-model regression:
yikt � β0 + β1 Post(t + 2)ikt × Peer Equity shockkt�1 + β2 Post(t + 3)ikt × Peer Equity shockkt�1 + β3 Post(t + 2)ikt + β4 Post(t + 3)ikt
+ β5 Peer Equity shockkt�1 + α1 Equity shockIND OWN ′ ′ IND ′ ′ ′
�ijt�1 + α2 Equity shockijt�1 + γ1 Xkt�1 + γ2 X�ijt�1 + γ3 Xit�1 + δ uik + φ vt + ɛikt :
Notes. The outcome variables are ∆Leverage (columns (1) and (2)), Leverage (columns (3) and (4)), Net Equity (columns (5) and (6)), and Gross
Equity (columns (7) and (8)). Columns (9) and (10) isolate the subsample of observations in which either an equity or debt issuance, but not both,
occurred (Net Debt (Issuer sample)). The main independent variables are Peer Equity shockkt�1 and Equity shockIND, which measure the lagged
averages of the idiosyncratic equity shock for analyst peer firm k and industry peers, respectively. Also, Equity shockOWN measures the lagged
averages of the idiosyncratic equity shock of firm i. All control variables are lagged one period, except the number of firm’s connections. In the
specification with ∆Leverage as the dependent variable all the control variables are in first difference, except the equity shock and the number of
firm’s connections and analyst coverage. We classify firm-pairs as treated in two ways: (1) pairs of firms that were connected by one analyst who
stopped the coverage of at least one of the two firms (i and/or k) as a consequence of the brokerage shock; and (2) pairs of firms that were
connected by one analyst who stopped following firm i and firm k at the same time. Post(t + 2) and Post(t + 3) are dummy variables that takes the
value of one in year two and three after the brokerage shock (and zero otherwise), respectively. In all the specifications we include directed firm-
pair and year fixed effects. Standard errors are clustered at directed firm-pair level. All variables are winsorized at the 1st and 99th percentile and
are defined in Appendix A. For brevity, we suppress the constant. Standard errors are in parentheses. Statistical significance at the 10%, 5% and
1% levels are denoted by *, ** and ***, respectively.
non-all-star brokerage houses to see if there is any differ and level of leverage, net and gross equity issuance), we
ence in the extent of peer effects within their networks. find a larger coefficient on the averages for peers con
Next, we differentiate analysts based on their level nected through analysts from all-star brokerage houses
of experience. Each year, we define analysts to have relative to peers connected through non-all-star broker
more (less) experience if they are above (below) sample age houses, although we find that the coefficients are
median in terms of the number of years since they first statistically different only for leverage and net equity iss
appear on IBES. uances. The one exception is for the conditional debt-
Table 11 examines the impact of all-star brokerage equity choice in column (5), suggesting that connections
houses (Panel A) and analyst experience (Panel B) on the through all-star analysts may be more influential for
strength of the analyst network peer effect. In Panel A, equity issuance decisions than for debt decisions.
we present the results of the reduced form in which we Similar, but stronger results are obtained in Panel B
include the weighted average equity shock for the two where we differentiate analysts based on their experi
groups of peers (All-Star and Non All-Star). The All-Star ence (more experienced versus less experienced).23 In all
average is calculated over peers that share at least one specifications, we find stronger peer effects among firms
analyst from an all-star brokerage house, whereas the that are connected through more experienced analysts.
non-all-star average is calculated over peers connected All these differences are statistically significant, with the
only by analysts not from all-star brokerage houses. For exception of the coefficient for net equity issuances. Inter
outcome variables in columns (1) to (4) (first difference estingly, the peer effect is not statistically different from
Table 11. Cross-Sectional Tests: All-Star Brokerage Houses and Analyst Experience
Net Debt
∆Leverage Leverage Net Equity Gross Equity (Issuer sample)
(1) (2) (3) (4) (5)
Panel A: All-star brokerage houses (all-star vs. non all-star)

ACN
Equity shock (All-Star) �0.018 �0.023 0.051 0.054 �0.048
(0.005)*** (0.006)*** (0.018)*** (0.019)*** (0.028)*

Equity shockACN (Non All-Star) �0.013 �0.007 0.008 0.034 �0.066
(0.004)*** (0.004) (0.014) (0.014)** (0.021)***
(All-Star)-(Non All-Star) �0.005 �0.016** 0.04** 0.020 0.018
Peer characteristics (all-star) Yes Yes Yes Yes Yes
Peer characteristics (non all-star) Yes Yes Yes Yes Yes
Observations 41,411 41,411 41,411 41,411 20,150
R2 0.118 0.768 0.395 0.458 0.614
Panel B: Analyst experience (more experienced vs. less experienced)
Equity shockACN (More experienced) �0.027 �0.026 0.037 0.064 �0.082
(0.005)*** (0.006)*** (0.018)** (0.019)*** (0.028)***
Equity shockACN (Less experienced) �0.004 �0.001 0.011 0.012 �0.028
(0.003)* (0.003) (0.010) (0.011) (0.016)*
(More)-(Less) �0.022*** �0.024*** 0.026 0.052** �0.054*
Peer characteristics (more experienced) Yes Yes Yes Yes Yes
Peer characteristics (less experienced) Yes Yes Yes Yes Yes
Observations 41,411 41,411 41,411 41,411 20,150
R2 0.118 0.768 0.396 0.458 0.613
Notes. The table presents the results of OLS regressions that relate peer firm average financial policy and idiosyncratic equity shock to firm
capital structure. The outcome variables are ∆Leverage (column (1)), Leverage (column (2)), Net Equity (column (3)), and Gross Equity (column (4)).
Column (5) isolates the subsample of observations in which either an equity or debt issuance, but not both, occurred (Net Debt (Issuer sample)).
We classify brokerage houses that employ at least three all-star analysts as all-star brokerage houses. We classify analysts with above median
experience as being more experienced. In Panel A, we differentiate Equity shockACN�it across all-star and non-all-star brokerage houses, whereas in
Panel B, we differentiate it across more and less experienced analysts. All control variables are lagged one period, except the number of firm’s
connections. In the specification with ∆Leverage as the dependent variable, all the control variables are in first difference form, except the number
of firm’s connections and analyst coverage; and we include year and industry fixed effects (three-digit SIC code). The remaining specifications
include firm and year fixed effects. Standard errors are clustered at the firm level. All variables are winsorized at the 1st and 99th percentile and
are defined in Appendix A. For brevity, we suppress the constant. Standard errors are in parentheses. Statistical significance at the 10%, 5% and
1% levels are denoted by *, ** and ***, respectively.
zero for firms connected through less experienced ana these industry classifications (Tables IA-9 to IA-11 in the
lysts, but always significant for firms connected through online appendix).
more experienced ones.
5. Conclusion
4.4.4. Other Robustness Tests. We perform a number Sell-side analysts are important information intermediar
of additional robustness tests whose results are presented ies in financial markets. There is growing evidence that
in the online appendix. We briefly discuss their results they may influence the financial policies of firms that they
here. Recent literature shows that firms with common cover. In this paper, we provide evidence that sell-side
institutional shareholders tend to follow similar financial analysts are an important mechanism underpinning peer
policies (Cronqvist and Fahlenbrach 2008). In Table IA-8 effects in financial policy choices. Building on recent
of the online appendix, we repeat our tests after including empirical methods from the network effects literature
the characteristics of firms that share institutional share to identify peer effects, we find that exogenous changes
holders with the firm in question and find our results on to financial policies of firms covered by an analyst,
analyst peers to be robust. Consistent with prior literature such as leverage and equity issuance, lead other firms
we find evidence for peer effects within institutional covered by the same analyst to make similar changes
shareholder networks. We also repeat our tests with four in policy.
alternate industry definitions: two-digit SIC code, Fama- We use an extended Manski-type linear-in-means
French industry classification, GICS codes, and Hoberg- model and use the idiosyncratic equity shocks of analyst
Phillips peers. We find our results to be robust across peer firms, as well as the return shocks of indirect peers
(“friends of friends”), as instruments for analyst peer • Gross Equity: Dummy variable that takes the value of
firm financial policies. We show that the network effects one if gross equity issuances normalized by book assets at the
that we document are distinct from industry peer effects beginning of the year is greater than a threshold (Compustat
and that these effects are more pronounced among peers items: sstk/at(t � 1) >1%, 3%, 5%).
• Leverage: The ratio of the sum of total long-term debt
connected by analysts that are more experienced and
plus total debt in current liabilities scaled by the market value
from more influential brokerage houses. Moreover, the of assets (Compustat items:(dltt + dlc)/(prcc_f*cshpri + dlc + dltt
peer effects are weaker after the reduction in common + pstkl-txditc)).

coverage as a consequence of brokerage closures. • Log(Sales): Natural logarithmic of sales (Compustat items:
An important question that we leave for future log(sale)).
research is to establish if the propagation that we docu • Market to Book: The ratio of the sum of the total book
ment is value enhancing or value destroying. Prior value of debt plus market value of equity divided by book
research suggests that such mimicking behavior could be value of total assets (Compustat items: (prcc_f*cshpri + dlc + dltt
value-enhancing or not, depending on the mechanism + pstkl-txditc)/at).
driving the behavior. For example, an observational • Market Value of Assets: The sum of the market value of
equity plus total long-term debt plus current liabilities (Com
learning channel may allow firms to get closer to optimal
pustat items: prcc_f*cshpri + dlc + dltt + pstkl-txditc).
capital structure, despite uncertainty about the underly • Net Debt Issuances: The sum of the total long-term debt
ing model (Devenow and Welch 1996). Conversely, plus total debt in current liabilities for the current fiscal year
mimicking that results from irrational herding may not minus the sum of the total long-term debt plus total debt in
bring firms closer to optimal policy. Although our empir current liabilities in the previous fiscal year (Compustat items:
ical approach does not enable us to examine value effects (dltt + dlc � (dltt(t � 1) + dlc(t � 1)))).
directly, our results show a positive association between • Net Debt: Dummy variable that takes the value of one if
peer equity shocks and SEO announcement returns. This net debt issuances normalized by book assets at the beginning
suggests that information relevant for equity issuance of the year is greater than 1%. (Compustat items: (dltt + dlc �
(dltt(t � 1) + dlc(t � 1)))/at(t � 1) >1%).
decisions that is intermediated through analysts may be
• Net Equity Issuances: Difference between equity issu
more likely to be value enhancing. Future research can
ances and equity repurchases (Compustat items: sstk-prstkc).
also explore if there is similar propagation in other corpo Net Equity: Dummy variable that takes the value of one if net
rate policies such as investment and governance provi equity issuances normalized by book assets at the beginning of
sions such as design of executive compensation. Apart the year is greater than a threshold (Compustat items: (sstk-
from research analysts, firms are also connected by other prstkc)/at(t � 1) >1%, 3%, 5%).
channels such as social ties or commonality of board of • Profitability: The ratio of the EBITDA divided by book
directors, executives, commercial/investment bankers value of total assets (Compustat items: oibdp/at).
or other professional advisors, and institutional or active • Stock Return: Annual return for the firm’s stock over the
investors. The methodology used in this paper can be fruit current fiscal year (Compustat items: ((prcc_f/ajex + dvpsx_f/
ajex)/(prcc_f(t � 1)/ajex(t � 1)))�1).
fully used to identify peer effects in these other settings.
• Tangibility: The ratio of the book value of Net Property
Plant and Equipment divided by book value of total assets
Acknowledgments (Compustat items: ppent/at).
The authors thank Victoria Ivashina, the department editor, an
associate editor, and two anonymous referees for valuable
Endnotes
comments and suggestions and Ambrus Kecskés (discussant), 1
Markku Kaustia (discussant), Murray Frank (discussant), Felipe Examples include Matvos and Ostrovsky (2010); Shue (2013);
Leary and Roberts (2014); Kaustia and Rantala (2015); and Fracassi
Cortes, and conference and seminar participants at the 2019
(2017).
FMA Annual Meetings, 2018 Western Finance Association 2
Or perhaps both. For example, a shock to a peer firm’s investment
Meetings, 2017 American Finance Association Meetings, 2017
opportunities that generates a positive return shock may affect the
Edinburgh Corporate Finance Conference, Universidad
peer’s investment behavior and elicit an equity issuance to fund the
Católica de Chile, Universidad Adolfo Ibañez, Central Bank of investment.
Chile and Universidad de Chile for helpful comments and dis 3
See, for example, Frankel et al. (2006); Kadan et al. (2012); Muslu
cussions. This paper is dedicated to the memory of our col et al. (2014); Chang et al. (2006); and Piotroski and Roulstone (2004).
league and friend Radhakrishnan Gopalan, who passed away 4
Recent research has shown that analysts can influence firm poli
in December 2022. He will be dearly missed.
cies either directly by exerting influence (Degeorge et al. 2013, Brad
ley et al. 2017) or indirectly, if managers alter firm policies to meet
analyst forecasts (Hribar et al. 2006, Bhojraj et al. 2009, Gunny 2010).
Appendix A. Variable Definitions We differ from this earlier work in that we study the information
• Coverage: Number of analysts issuing EPS forecasts of flow from one firm to another in an analyst network.
firm i in a given year. 5
We conduct our tests using traditional industry classification mea
• Equity Shock: Idiosyncratic returns defined as the differ sures based on three-digit SIC codes and alternative industry classi
ence between realized and expected returns based on the fications such as the Fama-French industry classifications, the
methodology provided by Leary and Roberts (2014). Hoberg-Phillips industry peers (Hoberg and Phillips 2010, 2016),
20
and S&P GICS codes to allow for the possibility that alternative In columns (1)–(3), we repeat the results in Table 8, but we
industry grouping better capture economic commonalities across restrict our sample to firms with also non-missing indirect peers’
firms. Our results on the analysts’ peer effect are robust across these equity shocks when we impose the restriction of at least two (and
various industry definitions. four) direct peers. Thus, our results are comparable across columns.
6 21
In additional robustness test, we repeat our estimates after exclud Because there are no analyst connections for this latter group, we
ing all indirect peers in the same industry as firm i. use the simple average of their idiosyncratic shocks rather than
7
Although the policies of peer firms may be public knowledge, ana weighting by the number of common analysts.
22
lysts may still play an important role in communicating the nuances Following Martens and Sextroh (2021), standard errors are clus
and value relevance of the policy to other firms, enabling manage tered at the directional firm-pair level.
ment to better assess the suitability of the policy for the firm. 23
We repeat the process and create two weighted average equity
8
Kaustia and Rantala (2015, 2021) also examine peer effects within shocks for the two peer groups: Equity shockACN (More experienced)
the context of analyst coverage networks. However, their focus is and Equity shockACN (Less experienced).
on stock split decisions, and they use analyst networks to identify
groups of related firms rather than study the role of analysts in
transmitting information from one firm to another. References
9 Bae J, Biddle GC, Park CW (2022) Managerial learning from analyst
As we explain later, in reality, our sample starts in 1987 but we
require five years of data to calculate the rolling betas (we use a feedback to voluntary capex guidance, investment efficiency,
window of 60 months). Hence, to estimate the idiosyncratic equity and firm performance. Management Sci. 68(1):583–607.
shock of the year 1992, the first rolling windows is 1987–1991. In Baker M, Wurgler J (2002) Market timing and capital structure. J.
1993, we have all the observations for the lagged equity shock and Finance 57(1):1–32.
control variables and the contemporaneous outcome variables. Barber B, Lehavy R, McNichols M, Trueman B (2001) Can investors
10 profit from the prophets? Security analyst recommendations
In all the regressions, we use the 1% threshold for the gross and
net equity (debt) issuances to define the indicator variable. We and stock returns. J. Finance 56(2):531–563.
Bhojraj S, Hribar P, Picconi M, McInnis J (2009) Making sense of
explicitly identify the cases in which we use a different threshold.
11
cents: An examination of firms that marginally miss or beat
We also create an alternative measure of industry average out analyst forecasts. J. Finance 64(5):2361–2388.
comes that includes only firms that are in the same industry as firm Blume LE, Brock WA, Durlauf SN, Ioannides YM (2010) Identifica
i but are not in the same analyst network as firm i. In other words, tion of social interactions. Preprint, submitted August 16,
we exclude the set of firms that overlap across the analyst coverage https://dx.doi.org/10.2139/ssrn.1660002.
network and industry of firm i. Bowen RM, Chen X, Cheng Q (2008) Analyst coverage and the cost of
12
Leary and Roberts (2014) show evidence that this strategy pro raising equity capital: Evidence from underpricing of seasoned
duces idiosyncratic return estimates that are uncorrelated, both seri equity offerings. Contemporary Accounting Res. 25(3):657–700.
ally and cross-sectionally, within networks. Bradley D, Gokkaya S, Liu X, Xie F (2017) Are all analysts created
13
In each year, we calculate monthly peer returns using the firm’s equal? Industry expertise and monitoring effectiveness of finan
analyst network in that year. To calculate r ACN , we require that a cial analysts. J. Accounting Econom. 63(2–3):179–206.
it
firm has at least one peer firm with valid returns during the time Bramoullé Y, Djebbari H, Fortin B (2009) Identification of peer
period in which we estimate the loadings. effects through social networks. J. Econometrics 150(1):41–55.
14 Brochet F, Kolev K, Lerman A (2018) Information transfer and con
One potential concern could be that analysts may not pass informa
ference calls. Rev. Accounting Stud. 23(3):907–957.
tion about a financial policy change at firm i on to other firms in their
Brown LD, Call AC, Clement MB, Sharp NY (2019) Managing the
network if they know the change was precipitated by a shock that was
narrative: Investor relations officers and corporate disclosure. J.
idiosyncratic to firm i. However, first, although the analyst observes
Accounting Econom. 67(1):58–79.
the policy change itself, the analyst may not be aware at the time
Chan K, Chan Y-C (2014) Price informativeness and stock return
whether it was associated with a purely idiosyncratic shock. Second,
synchronicity: Evidence from the pricing of seasoned equity
Leary and Roberts (2014) show evidence that corporate managers do
offerings. J. Financial Econom. 114(1):36–53.
respond to financing decisions associated with idiosyncratic shocks at
Chang X, Dasgupta S, Hilary G (2006) Analyst coverage and financ
peer firms, suggesting it is plausible that analysts would also.
ing decisions. J. Finance 61(6):3009–3048.
15
Consistent with the simple average idiosyncratic shock of indi Chen T, Harford J, Lin C (2015) Do analysts matter for governance?
rect peers, for the friend-of-friends approach we run the 2SLS with Evidence from natural experiments. J. Financial Econom. 115(2):
the simple average of peer firm outcome and controls variables. 383–410.
However, the results remain unchanged if we use peer-weighted Cronqvist H, Fahlenbrach R (2008) Large shareholders and corpo
averages. rate policies. Rev. Financial Stud. 22(10):3941–3976.
16 Degeorge F, Derrien F, Kecskes A, Michenaud S (2013) Do analysts’
On average, only 7% of indirect peers are in the same industry
(three-digit SIC code) as firm i. preferences affect corporate policies? ECGI - Working Paper
No. 361, Swiss Finance Institute, Zurich.
17
We also include the own firm’s equity shock (Equity shockOWN ) as
Devenow A, Welch I (1996) Rational herding in financial economics.
an additional firm characteristic.
18
Eur. Econom. Rev. 40(3–5):603–615.
In subsequent analyses, we continue to utilize the conditional Fracassi C (2017) Corporate finance policies and social networks.
debt issuance specification from column (6) in place of the specifica Management Sci. 3(8):2420–2438.
tion in column (5). Frank MZ, Goyal VK (2008) Profits and capital structure. Preprint,
19
In these regressions, when we control for industry average finan submitted March 13, https://dx.doi.org/10.2139/ssrn.1104886.
cial policy, we focus on firms that do not have common analysts Frankel R, Kothari S, Weber J (2006) Determinants of the informa
because the policy of firms with common analyst (and in the same tiveness of analyst research. J. Accounting Econom. 41(1):29–54.
industry) is the endogenous variable and is instrumented by the Glaeser EL, Sacerdote BI, Scheinkman JA (2003) The social multi
indirect peer average. plier. J. Eur. Econom. Assoc. 1(2–3):345–353.
Goldsmith-Pinkham P, Imbens GW (2013) Social networks and the Kelly B, Ljungqvist A (2012) Testing asymmetric-information asset
identification of peer effects. J. Bus. Econom. Statist. 31(3):253–264. pricing models. Rev. Financial Stud. 25(5):1366–1413.
Grennan J (2019) Dividend payments as a response to peer influ Kline B, Tamer E (2014) Some interpretation of the linear-in-means
ence. J. Financial Econom. 131(3):549–570. model of social interactions. Working paper, Harvard Univer
Gunny KA (2010) The relation between earnings management using sity, Cambridge, MA.
real activities manipulation and future performance: Evidence Leary MT, Roberts MR (2014) Do peer firms affect corporate finan
from meeting earnings benchmarks. Contemporary Accounting cial policy? J. Finance 69(1):139–178.
Res. 27(3):855–888. Manski CF (1993) Identification of endogenous social effects: The
Hart O, Moore J (1994) A theory of debt based on the inalienability reflection problem. Rev. Econom. Stud. 60(3):531–542.
of human capital. Quart. J. Econom. 109(4):841–879. Marsh P (1982) The choice between equity and debt: An empirical
Hilary G, Shen R (2013) The role of analysts in intra-industry infor study. J. Finance 37(1):121–144.
mation transfer. Accounting Rev. 88(4):1265–1287. Martens T, Sextroh CJ (2021) Analyst coverage overlaps and inter
Hoberg G, Phillips G (2010) Product market synergies and competi firm information spillovers. J. Accounting Res. 59(4):1425–1480.
tion in mergers and acquisitions: A text-based analysis. Rev. Matvos G, Ostrovsky M (2010) Heterogeneity and peer effects in
Financial Stud. 23(10):3773–3811. mutual fund proxy voting. J. Financial Econom. 98(1):90–112.
Hoberg G, Phillips G (2016) Text-based network industries and Muslu V, Rebello M, Xu Y (2014) Sell-side analyst research and
endogenous product differentiation. J. Political Econom. 124(5): stock comovement. J. Accounting Res. 52(4):911–954.
1423–1465. Novaes W (2002) Managerial turnover and leverage under a take
Hong H, Kacperczyk M (2010) Competition and bias. Quart. J. over threat. J. Finance 57(6):2619–2650.
Econom. 125(4):1683–1725. Piotroski JD, Roulstone DT (2004) The influence of analysts, institu
Hribar P, Jenkins NT, Johnson WB (2006) Stock repurchases as an tional investors, and insiders on the incorporation of market,
earnings management device. J. Accounting Econom. 41(1):3–27. industry, and firm-specific information into stock prices. Ac
Israelsen RD (2016) Does common analyst coverage explain excess counting Rev. 79(4):1119–1151.
comovement? J. Financial Quant. Anal. 51(4):1193–1229. Pool VK, Stoffman N, Yonker SE (2015) The people in your neigh
Jegadeesh N, Kim J, Krische SD, Lee CM (2004) Analyzing the borhood: Social interactions and mutual fund portfolios. J.
analysts: When do recommendations add value? J. Finance Finance 70(6):2679–2732.
59(3):1083–1124. Rajan RG, Zingales L (1995) What do we know about capital structure?
Jensen MC (1986) Agency costs of free cash flow, corporate finance, Some evidence from international data. J. Finance 50(5):1421–1460.
and takeovers. Amer. Econom. Rev. 76(2):323–329. Shue K (2013) Executive networks and firm policies: Evidence from
Kadan O, Madureira L, Wang R, Zach T (2012) Analysts’ industry the random assignment of mba peers. Rev. Financial Stud.
expertise. J. Accounting Econom. 54(2):95–120. 26(6):1401–1442.
Kaustia M, Rantala V (2015) Social learning and corporate peer Welch I (2004) Capital structure and stock returns. J. Political
effects. J. Financial Econom. 117(3):653–669. Econom. 112(1):106–132.
Kaustia M, Rantala V (2021) Common analysts: Method for defining Womack KL (1996) Do brokerage analysts’ recommendations have
peer firms. J. Financial Quant. Anal. 56(5):1505–1536. investment value? J. Finance 51(1):137–167.

Analyst Coverage Networks and Corporate Financial Policies

Uploaded by

Copyright:

Available Formats

Analyst Coverage Networks and Corporate Financial Policies

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

Analyst Coverage Networks and Corporate Financial Policies

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [200.89.68.

196] On: 23 December 2024, At: 05:12

Analyst Coverage Networks and Corporate Financial

To cite this article:

Full terms and conditions of use: https://pubsonline.informs.org/Publications/Librarians-Portal/PubsOnLine-

Copyright © 2023, INFORMS

Please scroll down for article—it is on subsequent pages

Analyst Coverage Networks and Corporate Financial Policies

History: Accepted by Victoria Ivashina, finance.

1. Introduction its peer group, giving rise to contextual effects. In a linear

forecasts but also investor behavior (Piotroski and Roul­

work (Log(1 + Connections)t ). To calculate X�it�1 , we use

our sample period. We use two variables to identify

equity issuances. Our first variable defines equity issu­

ances as instances when the difference between cash flow

from equity issues less cash flow from equity repurchases

1% of the book value of total assets, Gross equity, we find

the control variables in our sample. The summary values

at the 1st and 99th percentiles.

In Table 3, we provide the summary information for

firm’s equity shock (Equity shockOWN ) in our sample is

standard deviation of 0.49. As expected, the dispersion

declines when the equity shock is averaged over either

the industry or analyst peer firms.

cussion is divided into four parts. First, we document a

that of its analyst peers. Second, we use Equity shock as an

exogenous peer firm characteristic to establish the exis­

then use the friends-of-friends approach to isolate endog­

enous peer effects from exogenous peer effects. Finally,

we provide a series of robustness and placebo tests to dis­

tinguish peer effects operating through analyst networks

from those operating within industries and perform sev­

4.1. Baseline Regressions

We provide the results of estimating Equation (1) in our

down into within and across industries.

full sample in Table 4. The outcome variable in columns

Across industries (3 Digit-SIC code)

Within industry (3 Digit-SIC code)

Across industries (GICS Industry)

Within industry (GICS Industry)

tively. The positive and significant coefficient on Industry

Across industries (F-F Industry)

Within industry (F-F Industry)

average highlights the positive association between a

with prior studies (Rajan and Zingales 1995). From the

coefficients on the industry average characteristics, we

find that only industry average Profitability is consistently

Table 2. Summary Statistics: Outcome and Control Variables

Panel A: Firm-specific variables

N Mean SD Median Mean SD Median Mean SD Median

Full sample Same industry Different industry

N Mean SD Median Mean SD Median Mean SD Median

∆Leverage 41,411 0.01 0.05 0.01 0.01 0.06 0 0.01 0.05 0

Table 3. Summary Statistics: Equity Shocks

N Mean SD P25 Median P75

Table 4. Baseline Specification I. Analyst Peer Firms vs. Industry Peers

∆Leverage Leverage Net Equity Gross Equity Net Debt

Peer average 0.547 0.391 0.301 0.278 0.216

Table 5. Reduced Form Regression with Idiosyncratic Equity Shock

Equity shockACN �0.026 �0.024 0.056 0.079 �0.039 �0.099

forecasts but also investor behavior (Piotroski and Roul

equity issuances. Our first variable defines equity issu

exogenous peer firm characteristic to establish the exis

then use the friends-of-friends approach to isolate endog

we provide a series of robustness and placebo tests to dis

from those operating within industries and perform sev

(0.005)* (0.006)* (0.018)* (0.019)* (0.028)*