Rev Account Stud (2015) 20:82109

DOI 10.1007/s11142-014-9294-7

Analysts choice of peer companies

Gus De Franco Ole-Kristian Hope
Stephannie Larocque

Published online: 22 May 2014

Springer Science+Business Media New York 2014

Abstract This is the first large-scale study to examine the peer companies used by
sell-side equity analysts in their research reports. Using a unique hand-collected data
set, we investigate the relation between peer valuation and peer choice by analysts. We
find that analysts on average select peer companies with high valuations and that this
effect varies systematically with analysts incentives and ability. Our evidence provides partial support for the idea that analysts choose peers strategically. We also find
some support for the idea that the selection of peers with high valuations is used in part
to justify optimistically skewed target prices and stock recommendations.
Keywords

Analysts  Comparable  Peer  Valuation

JEL Classification

G30  G31  G32  G34  M41

1 Introduction
This is the first study to systematically examine sell-side equity analysts choice of
peer companies in a large sample. Analysts are sophisticated users of financial
statement information who provide, among other things, earnings estimates,
valuations (including target prices), and stock recommendations for the firms they
cover. These analysts often use peers in their research to compare performance and
G. De Franco  O.-K. Hope
Rotman School of Management, University of Toronto, Toronto, Canada
e-mail: gus.defranco@rotman.utoronto.ca
O.-K. Hope
e-mail: okhope@rotman.utoronto.ca
S. Larocque (&)
Mendoza College of Business, University of Notre Dame, Notre Dame, IN, USA
e-mail: larocque.1@nd.edu

123

Analysts choice of peer companies

83

valuations across firms as well as to estimate the valuation of the firms they cover.
Moreover, analysts within-industry comparisons comprise a large part of their
value-added (Boni and Womack 2006).
We hand-collect analysts reports from Thomson Investext and manually extract
the information on comparable or peer companies (peers hereafter) from the
reports. Using this unique data set, we can examine the peers used by analysts in
their research reports. In our primary analyses, we consider the relation between
peer valuation and peer choice by analysts. If an analyst is induced to estimate a
high valuation for a firm she covers, one possible strategy would be to choose peer
companies with high valuation multiples so as to make the firm in question appear
relatively undervalued. We test for this effect using the valuation multiples most
widely used in practice: earnings to price and book to price (Barker 1999; Block
1999, 2010). We show that analysts tend to select peers with higher valuations,
consistent with analysts choosing peers optimistically.
In establishing a link between peer firm valuation and peer firm choice, we
control for numerous firm characteristics. Specifically, we control for both the level
of the potential peer firm characteristics and the similarity between the peer firm and
the firm being analyzed. Documenting how analysts choose peer firms in practice is
new to the literature.
We further examine whether the relation between peer valuation and peer choice
varies systematically with two analyst-specific factors: analyst incentives and analyst
ability. First, we investigate whether the relation between peer valuation and peer
choice is increasing in analyst incentives, and we find that this is the case: our results
indicate that the effect is larger for analysts with investment banking affiliations,
suggesting analysts choose peers to increase the chances of attracting investment
banking business. Second, previous research suggests that analysts who are ranked by
Institutional Investor magazine (II-ranked analysts) exhibit superior performance in a
variety of settings. We find that the relation between peer valuation and peer choice is
reduced for II-ranked analysts, suggesting that higher analyst ability significantly
reduces the optimism in peer selection. We further provide some evidence supporting
the idea that analysts may choose peers with high valuations to justify their
optimistically skewed stock recommendations and target prices.
Although the above evidence is suggestive of strategic behavior on analysts part,
it is difficult to fully rule out alternative explanations. Although we include
numerous control variables and also employ partition tests that lend further credence
to our primary analyses (i.e., we find that the results are strongest in the subsamples
for which we predict a stronger effect), the firms being analyzed in fact may have
greater potential that is not captured in our models. Another potential explanation is
that analysts may believe that the firm being analyzed has similar growth
opportunities as the selected peers, such that the analysts believe the peers to be
appropriate.1 Our findings should be interpreted with these caveats in mind. We
attempt to address these issues by examining the future (i.e., realized) profitability
and growth of the actually selected peers chosen relative to the report firms. We find
evidence of similar sales growth, earnings per share growth, and book value per
1

McNichols and OBrien (1997) provide a further discussion of such potential self-selection biases.

123

84

G. Franco et al.

share growth but stronger return on assets and return on equity for the selected peers
relative to report firms; this provides some evidence that analysts choose peers that
they perceive to have greater future profitability.
Our research, which is the first to employ a comprehensive data set of peers actually
used by sell-side analysts in their research reports, offers a unique contribution to the
literature. Our study responds to Bradshaws (2011) call for research into analysts
activities beyond their earnings forecasts and stock recommendations. In a survey
reported in Institutional Investor (October 2010), investors were asked to rate the
importance they place on a dozen sell-side equity analyst research attributes. The
highest ranked attribute was industry knowledge (see also Kadan et al. 2012). In
contrast, earnings estimates ranked last out of the 12 attributes. One way that analysts
impart their industry knowledge is through their written reports, which include
analyses employing peer firms as benchmarks. Whereas prior research recommends
how peer firms should be selected for valuation purposes, our study investigates
whether and how peer valuation affects the selection of peers, shows how this varies
cross-sectionally, and describes which peers are used by analysts. Financial analysis
textbooks commonly recommend the use of peers. Yet there is little theory to guide
the choice of peers in valuation and performance comparisons. While empirical
researchers have studied which multiples best explain stock price as well as how to
choose peers (Boatsman and Baskin 1981; Alford 1992; Bhojraj and Lee 2002; Liu
et al. 2002, 2007), the literature has not yet examined how peers are chosen in practice.
Our paper also relates to an emerging stream of research that examines peers or
benchmarks. Peers are used by financial analysts to support their valuation multiples,
earnings forecasts, and overall stock recommendations (e.g., Bradshaw et al. 2010),
by investors to judge the merits and comparability of investments (e.g., De Franco
et al. 2011), by fund managers in structuring their investment portfolios (e.g., Chan
et al. 2007), by compensation committees in setting executive compensation (e.g.,
Albuquerque 2009; Albuquerque et al. 2013), in determining valuation multiples
(e.g., Bhojraj and Lee 2002), by auditors in applying analytical procedures (e.g.,
Hoitash et al. 2006; Minutti-Meza 2013), and by researchers in choosing estimation
samples to detect earnings management (e.g., Ecker et al. 2013). To use peer firms as
benchmarks or comparables requires relying on the comparability of financial
information and its mapping into valuation, compensation, or other variables of
interest. Our study investigates the manner in which analysts choose peer firms.
The next section reviews the relevant literature and develops our research
questions. Sections 3 and 4 present the research design and sample selection,
respectively. Section 5 discusses the main and corroborative empirical findings.
Section 6 concludes.

2 Background and research questions

2.1 Brief background on peer firms in equity valuation
Financial analysis textbooks commonly recommend the use of peer firms in
valuation (Healy and Palepu 2007; Stickney et al. 2007; Damodaran 2009). Valuing

123

Analysts choice of peer companies

85

firms using multiples of their financial or operating drivers is a simple and popular
approach to valuation (e.g., Baker and Ruback 1999; Imam et al. 2008; Block 2010).
Using multiples to estimate a firms value assumes the firm should be valued by
capitalizing earnings (or cash flows, sales, or other metrics) at a given price-earnings
(or price-cash flow, enterprise value-sales, or other) multiple, such as the mean or
median multiple for a set of comparable firms. Unlike the discounted cash flow and
residual income valuation approaches, the use of multiples (at least seemingly)
avoids the problems of estimating the required return (i.e., the discount rate or cost
of capital) and of forecasting terminal values. However, implementing a multiples
approach has its challenges. As Baker and Ruback (1999) outline, these challenges
include choosing the appropriate value driver (i.e., cash flow, earnings, or other),
choosing the peers against which to compare the firm, and measuring the peer
multiple associated with these peers.
There is little theory to guide the selection and use of peers in valuation and
performance comparisons. Empirical researchers have considered which multiples best
explain stock price, as well as how to choose peer firms. Liu et al. (2002, 2007) compare
the performance of numerous value drivers (forward earnings, historical earnings, cash
flows, book value, and sales) and find that multiples derived from forward earnings best
explain stock price. Boatsman and Baskin (1981) test the price-to-earnings (P/E)
valuation method using as a firms peer each of a random firm from the same industry
and the firm from the same industry with the most similar earnings growth rate. They
conclude that the second approach is more accurate. In a study of P/E multiples, Alford
(1992) finds that comparable firms from the same industry or with similar risk and
earnings growth outperform comparable firms chosen from the broader market or with
similar size or analyst long-term growth forecasts. Kim and Ritter (1999) and
Roosenboom (2007) investigate the use of multiples by underwriters for firms
undergoing initial public offerings. In a recent study, Ecker et al. (2013) examine how
the criteria for choosing estimation samples (i.e., peers) affect the ability to detect
discretionary accruals. Bhojraj and Lee (2002) argue that the choice of comparable
firms used should be a function of the variables that drive cross-sectional variation in a
given valuation multiple and demonstrate the efficacy of their approach.
2.2 Potential effect of bias in peer selection
Although we expect analysts to pick peers based on attributes described in financial
statement analysis texts and elsewhere, we know from the literature that analysts
have incentives other than to merely maximize investor wealth. (Bradshaw (2011)
provides an overview of the literature on biased analysts). In other words, given the
discretion in peer selection, analysts may choose peers in a biased manner.
Specifically, if, based on her incentives the analyst wants to recommend that
investors buy a stock, one way to justify the higher value would be to pick peer
firms with high valuations.2 However, as there are incentives for analysts to also be
2

As mentioned in the introduction, although our findings are consistent with analysts behaving
strategically, it is clearly difficult to rule out alternative explanations. We explore some alternative
interpretations of our results in Sect. 5.4. Thus we suggest that our use of terms such as bias and
strategic should be interpreted cautiously.

123

86

G. Franco et al.

accurate, it is an empirical question whether analysts on average choose peers with

higher valuations.
2.3 Conditional analyses to explore variation in bias
Given that the literature suggests there are both costs as well as benefits to being
optimistic, we expect optimistic bias to be more prevalent in certain contexts. In the
following, we explore the effects of analyst incentives and analyst ability on
optimistic bias in peer choice.
2.3.1 Effects of analyst incentives
Research shows that sell-side analysts do not always act in the best interests of
investors. Rather, they may be serving personal objectives, such as increasing their
compensation, improving relations with management, garnering investment banking
business for their brokerage firm, promoting the stock to garner brokerage trading
volumes, and increasing the value of shares personally owned (e.g., Lin and
McNichols 1998; Michaely and Womack 1999, 2005; Heflin et al. 2003; Ertimur et al.
2007; Barniv et al. 2009, 2010). Prior research demonstrates that affiliated analysts
(i.e., those with direct investment banking business with the firm) issue more
optimistic forecasts (Dugar and Nathan 1995; Lin and McNichols 1998; Dechow et al.
2000).3 Based on these prior findings, we predict that the relation between peer choice
and peer valuation, if any, in peer selection will increase in analysts incentives,
proxied for using the investment banking affiliation data from Guan et al. (2012).
2.3.2 Effects of analyst ability
We next examine whether high-ability analysts, defined to include those analysts
who are ranked by Institutional Investor (II) magazine, exhibit less bias than other
analysts. This ranking is a first-order determinant of an analysts compensation
(Groysberg et al. 2011). Research suggests that II-ranked analysts exhibit higher
ability as evidenced by their greater forecasting accuracy, stock recommendation
profitability, and report readability (e.g., Stickel 1990, 1992; De Franco et al. 2014).
As a specific example, Gleason and Lee (2003) find that forecasts made by
Institutional Investor analysts (1) elicit a stronger immediate price response and (2)
are followed by a less pronounced subsequent price drift. We further expect that
institutional investors demand less biased (or more neutral) advice. For example,
Clarke et al. (2007, 728) state that one explanation for our results is that the
reputational concerns of all-star analysts make them less likely to succumb to
pressure from their investment bank to alter their earnings forecasts and
recommendations to increase deal flow. Accordingly, we test whether the relation
between peer choice and peer valuation is mitigated for higher-ability analysts.
3

Das et al. (1998) and Lim (2001) suggest that forecast optimism is used to increase access to
management, especially in cases where the information asymmetry between management and investors is
high.

123

Analysts choice of peer companies

87

2.4 Do analysts choose high-valuation peers to justify target prices and stock
recommendations?
Although the primary purpose of this study is not to explain prior research findings
about optimism in analysts recommendations, we explore whether the observed
relation between peer valuation and peer choice could partially explain the
mechanism through which analysts justify their outputs such as recommendations or
target prices. To be precise, we consider a situation in which the analyst has already
decided to issue a favorable report for reasons explored in prior literature.4 The
analyst is aware that she is reaching somewhat by issuing the particular
recommendation or target price. Thus one tool the analyst might resort to is to
benchmark against a firm with high valuationmaking the covered firm appear
relatively undervalued, which supports the positive recommendation. In contrast,
such a justification would be less necessary if she did not issue a favorable report. In
other words, the relation between peer choice and peer valuation would be stronger
for positive reports.
To provide empirical evidence on this important issue, we consider both the
increases in and the levels of analysts stock recommendations and target prices.
Specifically, using the same specification as for our primary analyses, we are
interested in testing whether the interaction between the peer firm valuation multiple
and the recommendation or target price (using both levels and changes specifications) is negative and significant. Finding evidence consistent with our prediction
would be consistent with analysts making use of their peer choice to justify their
recommendations and target prices.

3 Research design
We use a unique hand-collected data set to examine which peers are used by
analysts in their research reports. Our primary interest is in the possible effect of
peers valuation multiples on analysts choice of peers. Using a probit specification,
we test for a peer valuation effect after controlling for numerous alternative
determinants of peer firm choice.
PeerChoiceijk a c Multiple Peerj R bm Economic Controlm;ij eijk :

In Eq. 1, PeerChoiceijk is an indicator variable that equals one if analyst ks

report on firm i contains peer j in determining valuation and zero otherwise. The
treatment sample (i.e., PeerChoice = 1) for this test includes peers chosen by
analysts in their reports. To estimate the model, we require a sample of peers not
chosen by analysts. To provide observations in which PeerChoice = 0, we
randomly match each selected peer j with a firm from the universe of I/B/E/Scovered companies in the same GIC two-digit industry as the firm i, with available
4

A large literature finds that equity analysts buy recommendations are (much) more prevalent than hold
or sell recommendations (e.g., Barber et al. 2001; Malmendier and Shanthikumar 2007; Ke and Yu 2006;
Barniv et al. 2009). Similarly, Brav and Lehavy (2003) find that analysts set target prices that are, on
average, 28 percent higher than contemporaneous stock prices.

123

88

G. Franco et al.

data, and not chosen by the analyst in the report. Thus we have an equal number of
selected and nonselected peers in the sample, for each firm i. We limit the matched
sample of peers not chosen by analysts to those firms in the I/B/E/S universe,
because analysts are more likely to choose peer companies that are followed by
analysts. We further limit the sample of peers chosen and not chosen by analysts to
those in the same GIC two-digit industry because, for the analyst reports in our
sample, we find that 92 % of the peer firms chosen by analysts are from the same
two-digit GIC industry. Similar to Boni and Womack (2006), we note that GIC
industry definitions seem to better represent analyst choices than SIC classifications
as only 72 % of peers are from the same two-digit SIC industry.
Multiple Peer is a peers valuation multiple and is measured in the reciprocal,
that is, with enterprise valuation or price in its denominator, so as to reduce the
impact of extreme values (Liu et al. 2002; De Franco et al. 2011). We consider the
following two widely used valuation multiples: earnings to price (E/P) and book to
price (B/P). A negative association between a potential peers valuation multiple
and the probability that the company is chosen as a peer, after controlling for the
economic factors, is consistent with peers being chosen in an optimistically biased
manner.
Economic Control is a set of M control variables expected to explain the choice
of peers within a similar industry. This set designates factors not associated with any
bias by the analyst. To the best of our knowledge, no prior large-scale empirical
evidence exists on how peers are chosen in practice by sell-side equity analysts. As
Litov et al. (2012) argue, analysts make substantial investments in understanding
complementarities within groups of similar firms. Litov et al. note that the more
unusual the strategy relative to peer companies in an industry, the less coverage the
firm receives by analysts. We are motivated by these observations as well as by
arguments in widely used financial statement analysis texts and well-cited academic
studies in picking variables to construct our set of control variables. Thus, although
we acknowledge that there is limited theory to guide the empirical analyses, our
approach is grounded in prior research. Generally speaking, we consider comparable firms to be those in the same industry that have similar size, similar accounting
return drivers, and similar share characteristics. In our empirical analyses, we
control for both the level of these variables as well as the similarity between firms
and their potential peers.
Our controls include standard economic factors, such as similar size and industry,
on which firms are often matched by academic researchers (e.g., Barber and Lyon
1996) as well as several other potential dimensions. From a practitioner perspective
for example, Koller et al. (2005) recommend choosing peers based on similarity in
risk, growth, and after-tax return on capital. It is an empirical question as to which
economic factors better explain peers actually used in practice.
Based on a review of leading financial analysis texts and prior academic research,
we start with the levels of and similarity in revenues (i.e., size), revenue growth,
leverage, asset turnover, profit margin, stock price volatility, trading volume
turnover, and stock return. LogSales is the natural logarithm of total revenues in
year t - 1. SalesGrowth is year t - 1 revenues less year t - 2 revenues scaled by
year t - 2 revenues. Leverage is total long-term debt divided by total assets at the

123

Analysts choice of peer companies

89

end of year t - 1. AssetTurnover is total revenues divided by total assets in year

t - 1. ProfitMargin is income before extraordinary items divided by total revenues
in year t - 1. Volatility is the standard deviation in monthly stock returns during the
4 years ending in year t - 1. Turnover is the volume of shares traded divided by
shares outstanding in year t - 1. Return is the stock return during year t - 1.
To measure the similarity, we take the absolute value of the difference between
firm is and peer js respective variables at time t - 1 and multiply the difference by
-1. We then add the prefix Sim_ to the variable. For example, Sim_LogSales is our
proxy for similar size and equals the absolute value of the difference between the
logarithm of total revenues for firm i and the logarithm of total revenues for peer j,
both measured at the end of fiscal year t - 1. We multiply the variable by -1 so
that greater values indicate more similarity. We also control for the relative
valuation multiple of firm j by including the negative of the absolute value of the
difference between the valuation multiples of firm i and peer j (Sim_Multiple).5
Note that we focus on the similarity in accounting return components (i.e., sales,
sales growth, leverage, asset turnover, and profit margin) between firms to capture
closeness in firm fundamentals. At the same time, we extend our analysis to stockbased components including volatility, turnover, volume, and returns. This allows
our analysis to capture similarities in firm fundamentals as well as in a firms
information and trading environments. Because stock returns reflect both realized
and anticipated performance, tests that encompass stock-related components might
capture similarities in longer-term economic performance. These measures can also
capture similarity in risk.
Our economic factors further include comparability in earnings and returns as
employed by De Franco et al. (2011) and Bhojraj et al. (2003). ComparabilityEarn
is the R2 from a regression of firm is quarterly earnings on the quarterly earnings of
firm j during the 16 quarters ending in year t - 1. ComparabilityRet is the R2 from a
regression of firm is monthly returns on the monthly returns of firm j during the
4 years ending in year t - 1. These measures are defined so that greater values
indicate more comparability. The intuition underlying these measures is as follows.
Changes in input prices or customer demand for firms with similar business models
should translate into similar changes in accounting and economic profitability,
which in turn translates into stock returns and earnings performance covarying over
time for similar peers.
Last, we include variables that indicate whether the peers and firms are in the
same industry using more finely partitioned industry categories. This is partly
motivated by results in Kadan et al. (2012), who conclude that industry expertise is
an important aspect of sell-side research. SameGIC3digit is an indicator variable
that equals one if the firm is in the same three-digit GIC code as the potential peer
and zero otherwise. SameGIC5digit is an indicator variable that equals one if the
firm is in the same five-digit GIC code as the potential peer and zero otherwise. Note
that all variables used in the study are defined in an Appendix.
Next, we test whether peer selection bias varies with the two conditioning factors
using the following model:
5

Note that inferences are not affected by the inclusion of Sim_Multiple or other control variables.

123

90

G. Franco et al.

PeerChoiceijk a c1 Multiple Peerj  CondFactor c2 Multiple Peerj

c3 CondFactor R bm Economic Controlm;ij eijk :

where CondFactor is either II-Rank or Affiliation. A positive coefficient on the

interaction between II-Rank and Multiple Peer would be consistent with II-ranked
analysts being less optimistically biased than other analysts. A negative coefficient
on the interaction between Affiliation and Multiple Peer would suggest that analysts
are more likely to be optimistically biased when investment banking ties exist.
We measure II-Rank as an indicator variable that takes the value of one if the
analyst is II-ranked in a given year and zero otherwise. Affiliation is an indicator
variable proxying for investment banking ties using data from Guan et al. (2012).
Finally, we test whether analysts choose peers with high valuations to justify
positively skewed target prices and buy recommendations. We include both the
levels of and the changes in these variables as it is not clear which dominates for our
purpose.
PeerChoiceijk a c1 Multiple Peerj  Recommendation c2 Multiple Peerj
c3 Recommendation R bm Economic Controlm;ij eijk :
3a
PeerChoiceijk a c1 Multiple Peerj  Target c2 Multiple Peerj
c3 Target R bm Economic Controlm;ij eijk :

3b

where Recommendation is either (1) an indicator variable equal to one for analyst
reports containing a recommendation upgrade and zero otherwise (i.e., a changes
specification), or (2) an indicator variable equal to one for buy or strong buy
recommendations and zero otherwise (i.e., a levels test).6 Similarly, Target is also
tested in changes (a raised target from the previous report) and in levels (a continuous variable equal to the percentage difference between the analysts target
price and the current price).

4 Data, sample selection, and descriptive statistics

The peers that an analyst uses in her analysis are not available in machine-readable
databases. We thus hand-collect a sample of analysts reports from Thomson
Investext and manually extract peer information from the reports. To the extent that
Thomson Investext does not contain reports from all brokerage houses (e.g.,
Goldman Sachs), our sample may have selection bias. Our results should be
interpreted accordingly.
We collect a sample of analyst reports from 2005 to 2008. The high cost of
collecting the peer information limits our analysis to 2 years of data. (Inferences
throughout this study are unchanged when we limit our sample to the 2005 or the
2008 reports).

In our sample, 59 % of analyst reports are coded as Buy.

123

Analysts choice of peer companies

91

The reports are chosen as follows. We begin with all firms (i.e., firm is) in our
sample with data available on Compustat and CRSP. For a randomly selected subset
of these firms, we search Investext to find up to three reports per firm i, each written by
a different analyst and each mentioning the word comparable or peer firms (i.e.,
potential peer js). Peer information is manually collected from the analysts reports.
We read through the analyst report to ensure that the peer companies listed are the
ones used for benchmarking. This is important because peers can also be used in other
contexts, such as in written discussions of firm is industry dynamics. We find that
about 98 % of the analyst reports that we identify explicitly use peers to benchmark
valuation. We collect financial information from CRSP and Compustat for each of the
peer j firms mentioned in the analysts reports. As discussed below, we also collect
information for a sample of nonselected peers within the same industry. The sample is
further reduced by the requirement of financial variables for the peer j firms, plus the
random matching process we employ. The final sample comprises 2,547 report-year
observations (for 1,368 unique firms) and 13,575 peers (3,165 unique peers).7 To
mitigate the influence of outliers, all continuous variables are winsorized at the top
and bottom 1 %. (Inferences are unaffected if we do not winsorize variables).
Table 1, Panel A, presents descriptive statistics for the number of peers
mentioned per analyst report. For the final sample of firms, the average firm has
approximately 6.5 peers with available data to perform the analyses. In addition,
there is considerable cross-sectional variation in the number of peers selected per
analyst report. Specifically, the 5th percentile equals one peer company per analyst
report whereas the 95th percentile equals 19 peer companies.
We compare the characteristics of firms covered by analyst reports and
companies chosen by analysts as peers in their reports. Table 1, Panel B, provides
these descriptive statistics. We find that the firms covered in analyst reports and
those chosen as peers in the analyst reports are similar in terms of sales growth,
leverage, asset turnover, volatility, turnover, and returns.
In the sample of analyst reports for which we can identify II-rank, 14 % of those
reports are written by an II-ranked analyst. Similarly, in the sample of analyst
reports for which we can identify the analysts affiliation, 31 % of those reports are
written by an affiliated analyst (untabulated).
Before we proceed to our empirical analyses, we explore the extent of variation in
analysts choice of comparables over time and across analysts. That is, given the lack of
prior evidence on analysts choice of peer firms, analysts may have limited choice in
picking comparable firms. We first focus on the same firm followed by the same analyst
in two different years in our sample. For this (relatively small) subsample, an average of
19.1 peers are used in each report, with a standard deviation of 13.3 and a minimum
(maximum) of 5 (58). These (untabulated) statistics suggest considerable variation in
peer choice. Furthermore, of these 19.1 peers (on average), 15 or 81 % are unique to that
analyst-firm-year, while the other 19 % are used by the analyst in both years. In other
words, analysts do change the comparables they employ for a firm over time.
7

In other words, our analyses are conducted at the analyst-report-firm-peer level. As a robustness test, we
have conducted the analyses at the firm-peer level. No inferences are affected using this alternative
specification.

123

92

G. Franco et al.

Table 1 Descriptive statistics

Panel A: Number of peers per analyst report
Mean

6.5

SD

6.2

Percentiles
1st

5th

25th

Median

75th

95th

99th

1.0

1.0

2.0

5.0

9.0

19.0

30.0

Panel B: Firm characteristics

Firm reports (N = 2,547)

Analyst-chosen peers (N = 13,575)

Mean
(1)

Median
(2)

Mean
(3)

Median
(4)

Sales ($mm)

6,176

969

9,479

2,781

SalesGrowth

0.186

0.131

0.176

0.142

Leverage

0.187

0.131

0.175

0.140

AssetTurnover

0.792

0.664

0.774

0.654

ProfitMargin

-0.251

0.076

-0.080

0.084

Volatility

0.227

0.179

0.204

0.167

Turnover

2.501

1.975

2.100

1.842

Return

0.243

0.146

0.199

0.161

Panel C: Number of unique peers across multiple analyst reports on the same firm
Number of peers per report

Number of unique peers per report

% of unique peers per report

10.6

7.8

70.8

This table presents means and medians of peers per analyst report, firm characteristics for firms covered
by analysts and the peers actually chosen by these analysts for valuation purposes, and the number of
unique peers per analyst report. ***, **, and * denote significance at the 1, 5, and 10 % levels. Variables
are defined in Appendix

We next consider the same firm followed by different analysts in the same year.
For this subsample, an average of 10.6 peers is used in each report, with a standard
deviation of 7.6 and a minimum (maximum) of 1 (44). Of these 10.6 peers (on
average), 7.8 or 71 % are unique to the firm report, and 29 % are used by another
analyst in our sample following the same firm in the same year (Panel C of Table 1).
These numbers indicate substantial variation across analysts for the same firm.
Third, we investigate whether certain firms are repeatedly selected by analysts as
benchmark or gold-standard peers in their respective industry. Indeed, we find that
there is a large variation in the number of mentions a given firm receives in each
GIC two-digit industry, with peer firms being mentioned anywhere from once to
over 20 times across analyst reports on different firms in a given industry. The firms
most commonly picked as peers include Allergan, Allstate, Ameristar Casinos,
Arctic Cat, Cisco, Dow Chemical, FedEx, Kraft, McDonalds, Microsoft, Walgreens,
and Verizon, among others. In untabulated sensitivity analyses, we find that our
inferences hold after excluding such benchmark peer firms from our peer sample.

123

Analysts choice of peer companies

93

5 Test results
This section describes the results of our main tests. The first subsection provides
analyses of the relation between peer selection and peer valuation multiples. We
next report conditional analyses in which we consider the effects of analyst
incentives and analyst ability. We further provide some preliminary evidence on the
relation between optimistic biases (or high valuation effects) in peer choice and
stock recommendations and target prices. Reported standard errors are clustered at
the firm level because the estimation of Eqs. 1 through 3a could suffer from timeseries dependence. (Inferences throughout are similar when we instead cluster
standard errors by analyst).
5.1 Regression analyses
5.1.1 Tests of peer valuation and peer selection
Table 2 shows the probit estimation results of Eq. 1. As a benchmark, the first
column shows the model with control variables only.8 The next two columns
include one of the two test variables as well as all controls.9 Model-fit statistics
indicate that the number of companies correctly classified as chosen peers is
approximately 88 %, and the pseudo-R2 is approximately 54 %, suggesting that the
model captures much of the variation in peer selection.
Our focus is on the effects of peer valuation multiples on peer selection. In both
columns (2) and (3), the coefficient on Multiple Peer is negative and significantly
different from zero (at the 1 % level using two-sided tests). In column 2, the
coefficient on Multiple Peer-E/P is -3.51, while in column 3, the coefficient on
Multiple Peer-B/P is -0.34.10 These results are consistent with analysts choosing
peers with higher valuation multiples. These findings are also consistent with the
idea that analysts choose peers in an optimistically biased manner.11 Again, we note
that alternative explanations for these results may exist. Our conditional analyses
below should provide some comfort on this issue, and we additionally explore
realized profitability and growth in a later section.
To assess the economic significance of the results, in untabulated analyses, we
rerun the analyses with each continuous variable standardized to mean zero and unit
variance. Using this approach, in column 3, for every one standard deviation
increase in Multiple-B/P Peer, there is an approximate 19 % change in the
probability that firm j is chosen as a peer. For Multiple Peer-E/P, the equivalent
economic effect is 15 %. As a comparison, for a one standard deviation change in
8

If we exclude the levels of the similarity variables, the coefficients on all but one of the similarity
variables are positive and significant as expected (untabulated).

There is no evidence of serious multicollinearity in our empirical tests. Specifically, the highest
variance inflation factor is four.

10

The number of observations is lower for Multiple Peer-E/P as the use of this multiple requires positive
earnings.

11
In untabulated analyses, we have also used sales to enterprise value and EBITDA to enterprise value as
valuation proxies and find consistent results.

123

94

G. Franco et al.

Table 2 Probit analysis of peer choice

Predicted sign

Multiple Peer

(1)

Multiple Peer is
E/P
(2)

B/P
(3)

-3.510***

-0.342***

(0.474)

(0.057)

0.00001***

0.00001***

LogSales

0.00001***
(0.000)

(0.000)

(0.000)

SalesGrowth

0.075*

0.114*

0.055

(0.042)

(0.067)

(0.042)

0.015

0.159*

0.065

(0.074)

(0.096)

(0.078)

AssetTurnover

0.170***

0.232***

0.150***

(0.030)

(0.034)

(0.030)

ProfitMargin

0.016

1.263***

0.020

(0.014)

(0.178)

(0.015)

Volatility

-0.548***

-0.435***

-0.545***

(0.120)

(0.157)

(0.122)

Turnover

0.142***

0.137***

0.145***

(0.009)

(0.011)

(0.009)

Return

-0.018

-0.116***

-0.106***

Leverage

(0.026)

(0.036)

(0.028)

2.036***

-0.030

Sim_Multiple

(0.471)

(0.047)

Sim_LogSales

0.00004*

0.001*

0.00004*

(0.000)

(0.0005)

(0.000)

Sim_SalesGrowth

0.081*

0.091*

0.089**

(0.052)

(0.070)

(0.051)

0.480***

0.598***

Sim_Leverage

0.407***
(0.095)

(0.141)

(0.097)

Sim_AssetTurnover

0.525***

0.479***

0.532***

(0.050)

(0.051)

(0.051)

Sim_ProfitMargin

0.001

1.500***

0.002

(0.009)

(0.221)

(0.009)

Sim_Volatility

0.009

0.036

0.152

(0.155)

(0.203)

(0.151)

Sim_Turnover

0.084***

0.060***

0.075***

(0.009)

(0.010)

(0.009)

ComparabilityEarn

0.201***

0.197***

0.231***

(0.062)

(0.071)

(0.063)

ComparabilityRet

1.332***

1.245***

1.294***

(0.116)

(0.133)

(0.119)

123

Analysts choice of peer companies

95

Table 2 continued
Predicted sign

SameGIC3digit
SameGIC5digit

?
?

(1)

Multiple Peer is
E/P(2)

B/P(3)

0.787***

0.768***

0.778***

(0.060)

(0.068)

(0.060)

1.100***

1.131***

1.098***

(0.046)

(0.052)

(0.047)

% Concordant

87.4

87.8

87.8

Pseudo-R2 (%)

52.5

53.8

53.5

Observations

27,150

19,147

26,684

This table reports the results of estimating probit models predicting the analysts peer choice based on
financial factors (column 1) and the potential peers valuation (columns 2 and 3). Intercepts are included
in all regressions but not tabulated for brevity. Standard errors are in parentheses and are clustered at the
firm level. Significance levels are based on one-tailed tests where there is a prediction for the sign of the
coefficient and based on two-tailed tests otherwise. ***, **, and * denote significance at the 1, 5, and
10 % levels, respectively. Variables are defined in Appendix

Sales, there is an approximate 1618 % change in the probability that firm j is

chosen as a peer.
With respect to the control variables, firm size, asset turnover, stock turnover,
and stock returns are all statistically significant. In addition, similarity in leverage,
asset turnover, and stock turnover also explain peer choice. Finally, comparability
of both earnings and returns relate positively to peer choice, as do same three- and
five-digit GIC industry membership.
In summary, Table 2 provides evidence that, controlling for numerous factors,
analysts choose peers with higher valuation multiples. These are new findings in the
literature.
5.1.2 Robustness tests
We first test the sensitivity of reported results to our sample selection choices. In
untabulated analyses we relax the requirement that potential peers be covered by
I/B/E/S. Inferences are similar with this alternative sample selection choice.
Next, we consider the possibility that analyst k is more likely to choose company
j as a peer for firm i when the analyst also formally provides research coverage of
company j. We re-estimate the test to control for this possibility by including in our
regressions a variable that indicates whether the analyst also covers company j. We
find that analyst coverage of a company increases the probability that it will be used
as a peer (untabulated). More importantly, the inference with respect to the peer
valuation multiples is unaltered.12
12

The number of observations is smaller in these robustness tests as we cannot match all the analysts in
our sample to I/B/E/S individual analysts/codes. The percent concordant and the pseudo R2 are higher in
these smaller samples and the estimated coefficients on the valuation measures are all negative and
significant at the 1 % level.

123

96

G. Franco et al.

Finally, our tabulated results are based on a pooled sample using observations
from 2005 to 2008. In untabulated analyses, we run the analyses separately for the
2 years. We find that both valuation multiples are significant (at the 1 % level) in
the predicted direction for both years.
5.1.3 Conditional analyses: analyst incentives and analyst skill
Table 3 presents the results of the conditional analyses. Panel A tabulates the results
relating to analyst incentives, operationalized as the analysts investment banking
ties. Using either Multiple Peer-E/P and Multiple Peer-B/P, the interaction is
negative as predicted and statistically significant at the 1 % level. We conclude that
analysts are more likely to pick peer firms with high valuations when they or their
employers are more likely to benefit through new fees for investment banking
services.
Panel B shows the results of tests that examine whether the relation between peer
valuation and peer choice is mitigated by analyst skill, proxied for by II-rank. The
interaction term is positive and statistically significant at the 1 % level for Multiple
Peer-E/P in column 1 but not significant for Multiple Peer-B/P in column 2. These
results provide some support for our hypothesis and are consistent with the most
skilled and successful analysts being less optimistically biased than other analysts.
In untabulated F tests, we find that II-ranked analysts do not exhibit any bias.
The conditional analyses provide further support for our primary teststhat is,
we find that the optimistic bias in peer selection is most pronounced in subsamples
for which we have reasons to believe that such a bias is more likely to exist. In other
words, these contextual analyses also help rule out alternative explanations (beyond
what our control variables capture) for why analysts choose peer firms with high
valuations.
5.2 Empirical evidence on peer valuations and target prices/stock
recommendations
Results in Panel A of Table 4 show that the coefficient on the interaction term with
recommendation increases is negative for both valuation multiples and also
statistically significant. Using the level of stock recommendation, we find a
significant effect for Multiple Peer-E/P but not Multiple Peer-B/P. Panel B further
shows that the interactions with Target are negative and significant for both
valuation multiples in the changes specification but are not significant in the levels
analyses.
Overall, we find support for our prediction that choosing peer firms with high
valuations may be a mechanism through which analysts justify higher recommendations and target prices in five out of eight cases. These findings provide partial
support for the notion that analysts rationalize target price increases and positively
skewed stock recommendations by selecting peer firms with high valuation. We
caution that our tests are exploratory and conclusions should be made carefully, but
this issue is important to both researchers and practitioners.

123

Analysts choice of peer companies

97

Table 3 Peer choice and analyst incentives and analyst skill

Panel A: Effect of analyst incentives
Predicted sign

Multiple Peer 9 Affiliation

Multiple Peer

Affiliation

Multiple Peer is
E/P
(1)

B/P
(2)

-4.083***

-0.766***

(1.136)

(0.141)

-2.983***

-0.237***

(0.475)

(0.061)

0.263***

0.368***

(0.070)

(0.078)

LogSales

0.00001***

0.00001***

(0.000)

(0.000)

SalesGrowth

0.116*

0.116*

(0.067)

(0.067)

Leverage

0.172*

0.169*

(0.096)

(0.095)

AssetTurnover

0.236***

0.160***

(0.034)

(0.034)

ProfitMargin

1.303***

0.533***

(0.180)

(0.170)

Volatility

-0.500***

-0.422***

(0.166)

(0.161)

Turnover

0.139***

0.132***

(0.011)

(0.011)

Return

-0.111***

-0.148***

(0.036)

(0.037)

Sim_Multiple

2.063***

2.784***

(0.470)

(0.496)

Sim_LogSales

0.001*

0.001*

(0.0004)

(0.0005)

Sim_SalesGrowth

0.101*

0.069

(0.069)

(0.069)

Sim_Leverage

0.477***

0.612***

(0.137)

(0.121)

Sim_AssetTurnover

0.478***

0.473***

(0.051)

(0.051)

Sim_ProfitMargin

1.484***

1.226***

(0.225)

(0.212)

Sim_Volatility

0.039

0.127

(0.204)

(0.199)

Sim_Turnover

0.060***

0.058***

(0.010)

(0.010)

123

98

G. Franco et al.

Table 3 continued
Panel A: Effect of analyst incentives
Predicted sign

Multiple Peer is
E/P
(1)

B/P
(2)
0.031

Sim_Return

0.049*
(0.038)

(0.037)

ComparabilityEarn

0.193***

0.179***

(0.071)

(0.072)

ComparabilityRet

1.242***

1.239***

(0.132)

(0.132)

SameGIC3digit

0.767***

0.747***

(0.068)

(0.068)

SameGIC5digit

1.133***

1.143***

% Concordant

(0.052)

(0.052)

87.8

87.9

Pseudo-R2 (%)

53.9

54.1

Observations

19,147

18,978

Panel B: Effect of analyst skill

Predicted sign

Multiple Peer 9 II-Rank

Multiple Peer

?
-

Multiple Peer is
E/P
(1)

B/P
(2)

4.067***

0.143

(0.851)

(0.144)

-4.510***

-0.339***

(0.601)

(0.077)

II-Rank

-0.442***

-0.256***

(0.073)

(0.093)

LogSales

0.00001***

0.00001***

(0.000)

(0.000)

SalesGrowth

0.126*

0.125*

(0.075)

(0.074)

Leverage

0.129

0.106

(0.101)

(0.102)

AssetTurnover

0.241***

0.173***

(0.036)

(0.036)

ProfitMargin

1.373***

0.623***

(0.192)

(0.183)

Volatility

-0.502***

-0.427**

(0.174)

(0.173)

Turnover

0.128***

0.122***

(0.012)

(0.012)

123

Analysts choice of peer companies

99

Table 3 continued
Panel B: Effect of analyst skill
Predicted sign

Return
Sim_Multiple
Sim_LogSales
Sim_SalesGrowth
Sim_Leverage
Sim_AssetTurnover
Sim_ProfitMargin
Sim_Volatility
Sim_Turnover

?
?
?
?
?
?
?
?

Sim_Return

ComparabilityEarn

ComparabilityRet
SameGIC3digit
SameGIC5digit

?
?
?

Multiple Peer is
E/P
(1)

B/P
(2)

-0.133***

-0.166***

(0.036)

(0.037)

1.907***

2.648***

(0.483)

(0.526)

0.001*

0.001*

(0.0001)

(0.001)

0.123**

0.099*

(0.070)

(0.070)

0.397***

0.499***

(0.139)

(0.131)

0.452***

0.451***

(0.055)

(0.054)

1.489***

1.209***

(0.233)

(0.220)

-0.020

0.075

(0.218)

(0.216)

0.065***

0.061***

(0.010)

(0.010)

0.019

-0.0005

(0.040)

(0.039)

0.221***

0.202***

(0.076)

(0.077)

1.315***

1.283***

(0.137)

(0.137)

0.820***

0.798***

(0.070)

(0.070)

1.113***

1.127***

(0.055)

(0.055)

% Concordant

87.9

87.8

Pseudo-R2 (%)

54.0

53.9

Observations

15,558

15,419

This table reports the results of estimating probit models predicting the analysts peer choice based on
financial factors, the potential peers valuation, and either the analysts incentives (Panel A) or the
analysts skill level (Panel B). Intercepts are included but not tabulated. Standard errors are in parentheses
and are clustered at the firm level. Significance levels are based on one-tailed tests where there is a
prediction for the sign of the coefficient and based on two-tailed tests otherwise. ***, **, and * denote
significance at the 1, 5, and 10 % levels, respectively. Variables are defined in Appendix

123

100

G. Franco et al.

Table 4 Peer choice and analyst stock recommendations and target prices
Panel A: Effect of analyst recommendation
Predicted sign

Multiple Peer 9 RaiseRec

Multiple Peer 9 BUY
Multiple Peer

Multiple Peer is
E/P
(1)

B/P
(2)

-2.869*

-0.406*

(1.868)

(0.285)

RaiseRec

SalesGrowth
Leverage
AssetTurnover
ProfitMargin
Volatility
Turnover
Return
Sim_Multiple
Sim_LogSales
Sim_SalesGrowth
Sim_Leverage

?
?
?
?

Sim_AssetTurnover

Sim_ProfitMargin

123

B/P
(4)

-1.843**

-0.070

(0.812)

(0.090)

-3.471***

-0.333***

-2.410***

-0.296***

(0.698)

(0.069)

(0.544)

(0.062)

0.052

0.048

(0.139)

(0.165)
0.161***

0.094

(0.056)

(0.057)

BUY
LogSales

E/P
(3)

0.00001***

0.00001***

0.00001***

0.00001***

(0.000)

(0.000)

(0.000)

(0.000)

0.052

-0.016

0.082

0.045

(0.117)

(0.067)

(0.070)

(0.044)

0.309**

0.198*

0.181*

0.076

(0.138)

(0.115)

(0.099)

(0.081)

0.298****

0.173***

0.233**

0.150***

(0.053)

(0.050)

(0.035)

(0.031)

0.900***

0.004

1.251***

0.016

(0.268)

(0.023)

(0.184)

(0.015)

-0.687***

-0.475**

-0.434***

-0.514***

(0.231)

(0.207)

(0.156)

(0.125)

0.133***

0.151***

0.136***

0.145***

(0.016)

(0.014)

(0.011)

(0.009)

-0.120**

-0.123***

-0.117***

-0.106***

(0.055)

(0.044)

(0.035)

(0.029)

1.993***

-0.081**

2.001***

-0.035

(0.691)

(0.045)

(0.467)

(0.047)

0.001

0.00005**

0.001**

0.00005*

(0.0005)

(0.000)

(0.001)

(0.000)

-0.001

0.034

0.076

0.083*

(0.128)

(0.080)

(0.073)

(0.053)

0.707***

0.798***

0.463***

0.611***

(0.196)

(0.139)

(0.147)

(0.099)

0.564***

0.528***

0.458***

0.514***

(0.081)

(0.085)

(0.052)

(0.052)

1.328***

-0.001

1.473***

0.0004

(0.325)

(0.014)

(0.224)

(0.009)

Analysts choice of peer companies

101

Table 4 continued
Panel A: Effect of analyst recommendation
Predicted sign

Multiple Peer is
E/P
(1)

B/P
(2)

E/P
(3)

B/P
(4)

0.125

0.141

0.214*

Sim_Volatility

-0.025
(0.288)

(0.227)

(0.194)

(0.150)

Sim_Turnover

0.050***

0.067***

0.059***

0.075***

(0.014)

(0.014)

(0.010)

(0.009)

Sim_Return

0.001

0.004

0.034

0.018

(0.066)

(0.054)

(0.037)

(0.033)

ComparabilityEarn

0.293***

0.344***

0.201***

0.235***

(0.108)

(0.098)

(0.072)

(0.064)

ComparabilityRet

1.357***

1.382***

1.318***

1.341***

(0.173)

(0.168)

(0.137)

(0.123)

SameGIC3digit

0.743***

0.734***

0.794***

0.808***

(0.100)

(0.086)

(0.069)

(0.061)

SameGIC5digit

1.109***

1.105***

1.122***

1.082***
(0.059)

(0.073)

(0.065)

(0.054)

% Concordant

88.0

88.0

87.9

87.8

Pseudo-R2 (%)

54.4

54.2

54.1

53.7

Observations

8,027

11,068

18,065

25,081

Panel B: Effect of analyst target price

Predicted sign

Multiple Peer 9 RaiseTarget

Multiple Peer 9 TargetUpside
Multiple Peer
RaiseTarget

Multiple Peer is
E/P
(1)

B/P
(2)

-3.600**

-0.303**

(1.754)

(0.166)

SalesGrowth
Leverage

B/P
(4)

-0.393

0.037

(1.091)

(0.082)

-2.864***

-0.204***

-3.063***

-0.325***

(0.759)

(0.069)

(0.529)

(0.062)

0.031

-0.042

(0.127)

(0.111)
0.099

0.102

TargetUpside
LogSales

E/P
(3)

(0.110)

(0.069)

0.00001***

0.00001***

0.00001***

0.00001***

(0.000)

(0.000)

(0.000)

(0.000)

-0.044

-0.114

0.123*

0.059

(0.134)

(0.083)

(0.074)

(0.048)

0.405***

0.279**

0.174*

0.074

(0.156)

(0.137)

(0.103)

(0.085)

123

102

G. Franco et al.

Table 4 continued
Panel B: Effect of analyst target price
Predicted sign

AssetTurnover
ProfitMargin
Volatility
Turnover
Return
Sim_Multiple
Sim_LogSales

?
?

Sim_SalesGrowth

Sim_Leverage

Sim_AssetTurnover

Sim_ProfitMargin

Sim_Turnover
Sim_Return
ComparabilityEarn
ComparabilityRet
SameGIC3digit
SameGIC5digit
% Concordant
2

Pseudo-R (%)

123

?
?
?
?
?
?

Multiple Peer is
E/P
(1)

B/P
(2)

E/P
(3)

B/P
(4)

0.296***

0.167***

0.224***

0.155***

(0.056)

(0.050)

(0.037)

(0.031)

1.035***

0.015

1.376***

0.022

(0.300)

(0.030)

(0.196)

(0.017)

-0.944***

-0.688***

-0.500***

-0.724***

(0.296)

(0.209)

(0.173)

(0.128)

0.135***

0.147***

0.133***

0.141***

(0.019)

(0.016)

(0.012)

(0.010)

-0.172***

-0.147***

-0.124***

-0.096***

(0.057)

(0.045)

(0.036)

(0.031)

1.885***

0.016

1.933***

-0.018

(0.721)

(0.057)

(0.480)

(0.047)

0.002

0.0001

0.002**

0.0001

(0.002)

(0.000)

(0.001)

(0.000)

-0.059

0.014

0.086

0.090*

(0.146)

(0.105)

(0.079)

(0.059)

0.648***

0.813***

0.436***

0.593***

(0.237)

(0.187)

(0.155)

(0.106)

0.609***

0.570***

0.442***

0.520***

(0.092)

(0.087)

(0.055)

(0.051)

1.694***

-0.009

1.643***

-0.004

(0.355)

(0.020)

(0.252)

(0.011)

0.078***

0.088***

0.066***

0.080***

(0.015)

(0.015)

(0.011)

(0.009)

-0.095

-0.078

0.021

0.013

(0.075)

(0.059)

(0.038)

(0.035)

0.337***

0.375***

0.195***

0.240***

(0.120)

(0.114)

(0.074)

(0.068)

1.291***

1.351***

1.206***

1.231***

(0.189)

(0.176)

(0.143)

(0.126)

0.668***

0.717***

0.773***

0.810***

(0.116)

(0.098)

(0.073)

(0.063)

1.198***

1.160***

1.146***

1.101***

(0.077)

(0.072)

(0.056)

(0.051)

88.7

88.4

87.8

87.8

56.3

55.4

53.9

53.8

Analysts choice of peer companies

103

Table 4 continued
Panel B: Effect of analyst target price
Predicted sign

Observations

Multiple Peer is
E/P
(1)

B/P
(2)

E/P
(3)

B/P
(4)

6,061

8,188

16,382

22,592

This table reports the results of estimating probit models predicting the analysts peer choice based on
financial factors, the potential peers valuation, and the analysts stock recommendation increases and
level (Panel A) or target price increases and levels (Panel B). Intercepts are included but not tabulated.
Standard errors are in parentheses and are clustered at the firm level. Significance levels are based on onetailed tests where there is a prediction for the sign of the coefficient and on two-tailed tests otherwise. ***,
**, and * denote significance at the 1, 5, and 10 % levels, respectively. Variables are defined in
Appendix

Table 5 Firm complexity and peer choice

Percentage of peers per analyst report from the same industry as the peer firms major industry
Number of SIC four-digit segments exceeding 10 % of total sales
1

[3

SameSIC4digit

35.1

22.0

14.8

0.8

SameGIC5digit

52.0

50.7

47.8

26.1

This table presents descriptive statistics regarding the analysts peer choices by industry and the number
of large segments (i.e., sales in a given SIC four-digit industry that exceed 10 % of firm sales) within the
firm for which the peer is selected. Variables are defined in Appendix

5.3 Additional analysis regarding firm complexity

Peer firm choice may be affected by the complexity of the firm the analyst is
recommending. In particular, for multi-segment firms (i.e., those with productmarket segments spanning more than one industry), the analyst might be compelled
to choose comparable firms from more than one industry. Unlike our primary
analyses, there is no strategic intent (or bias) involved in this context; instead
analysts are likely picking peers optimally given the complexity of the firm being
analyzed.
Table 5 investigates the choice of peer firms relative to the number of large
segments, defined as sales in a given four-digit SIC industry that exceed 10 % of
total firm sales, according to Compustat. For firms with 1 ([3) large segment(s),
analysts peer choices are from the firms main GIC five-digit industry 52.0 %
(26.1 %) of the time and from the firms main SIC four-digit industry 35.1 %
(0.8 %) of the time. These statistics suggest that firm complexity has a significant
impact on the choice of out-of-industry peers to capture the multiple facets of the
firm in question.

123

104

G. Franco et al.

Table 6 Ex post performance of report firms, selected peers, and predicted peers
Report
firms
(1)

Selected
peers
(2)

Predicted
peers
(3)

(1)(2)

(1)(3)

SGt?1

0.065

0.057

0.063

0.009

0.002

SGt?2

0.084

0.075

0.082

0.009

0.002

BVPSGrowtht?1

0.132

0.147

0.147

-0.015

-0.016

BVPSGrowtht?2

0.105

0.094

0.099

0.011

-0.006

EPSGrowtht?1

0.248

0.251

0.222

-0.003

0.026

EPSGrowtht?2

0.133

0.140

0.144

-0.007

-0.011

ROAt?1

0.020

0.036

0.022

-0.016***

-0.002

ROAt?2

0.062

0.090

0.066

-0.028***

-0.005

ROEt?1

0.063

0.091

0.075

-0.028***

-0.012

ROEt?2

0.170

0.225

0.202

-0.055***

-0.032*

This table presents the performance of report firms, peer firms, and predicted peer firms across 2,167
analyst reports for which ex post performance can be measured for years t ? 1 and t ? 2. For report
firms, the mean of sales growth, book value per share growth, EPS growth, ROA, or ROE is presented for
the firm on which the analyst report is written. For peer firms and predicted peer firms, sales growth, book
value growth, EPS growth, ROA, and ROE are averaged at the report level and then across all reports.
Peer firms are the firms selected by the analyst in the analysts report, while predicted peer firms are the
firms most likely to be selected in the analysts report, according to our prediction model. ***, **, and *
denote significance at the 1, 5, and 10 % levels, respectively. Variables are defined in Appendix

5.4 Evidence on subsequent performance of peer-selected firms

As noted in the introduction, in our setting it is challenging to rule out alternative
explanations for our findings. While our peer selection models contain numerous
factors that should help explain why a particular benchmark firm is chosen and our
cross-sectional analyses should help in identification, we certainly acknowledge that
other explanations cannot be ruled out. Specifically, the firms being analyzed may
have greater potential that is not picked up in our models. Another alternative
explanation is that analysts may believe that the firm being analyzed has similar
growth opportunities to the selected peers and thus may see the peers as appropriate.
In an attempt to explore alternative explanations, we examine the future (i.e.,
realized) profitability and growth of the peers chosen relative to peers predicted by
our model.
In Table 6, we compare the performance of report firms, selected peers, and
predicted peers across the 2,167 analyst reports for which ex post performance can
be measured for years t ? 1 and t ? 2.13 For report firms, the mean of sales growth,
book value per share (BVPS) growth, EPS growth, ROA, or ROE is presented for
the 2,167 firms on which the analyst reports are written and for which we have ex
post data for years t ? 1 and t ? 2. For selected peers and predicted peers, these
13
We obtain similar inferences when we limit the sample of report firms to those whose ex post
performance variables can be measured for year t ? 1.

123

Analysts choice of peer companies

105

variables are averaged at the report level and then across all 2,167 reports. Selected
peers are the companies selected in the analysts report, while predicted peers are
the companies most likely to be selected in the analysts report, according to an
analyst prediction model developed using the coefficients from the model presented
in column 1 of Table 2.
In Table 6, sales growth, BVPS growth, EPS growth, ROA, and ROE are
presented for each of report firms, peer firms, and predicted peer firms in columns,
1, 2, and 3, respectively. The last two columns compare the mean of these variables
across the report firms and peer firms and across the report firms and predicted peer
firms. For sales growth, BVPS growth, and EPS growth, we do not find a significant
difference between the ex post performance of report firms and selected peers, nor
between report firms and predicted peers. For ROA and ROE, we find that the ex
post performance of selected peers significantly exceeds that of report firms in both
years, while we find that the ex post performance of predicted peers resembles that
of report firms. Thus our findings provide some evidence that analysts choose peers
with greater future performance than that of the report firm.

6 Concluding remarks
Although a considerable literature on sell-side analysts exists, a large amount of this
research has focused on earnings forecasting, which perhaps is surprising given that
earnings forecasts are either less important than other activities or just one of the
inputs into analysts other activities (Schipper 1991; Bradshaw 2011). We
contribute to a better understanding of analysts work, in particular their choice
of peers for benchmarking purposes, beyond that of well-studied analysts tasks
such as recommendations and forecasts.
Using a unique hand-collected data set, we observe and examine which peers
sell-side equity analysts use in their research reports. Controlling for both the level
of numerous firm characteristics and the similarities between the potential peer firm
and the firm being analyzed, we find that analysts select peers with higher
valuations.
We next investigate whether the bias in peer selection varies systematically with
analyst incentives and analyst reputation (or skill). We find that our result is
especially pronounced when investment banking ties exist and for non-II-ranked
analysts. In addition, there is at least some indication that the previously
documented positive bias in stock recommendations and target prices may be
justified by analysts selecting peers with higher valuations.
Although our results are consistent with analysts choosing peers in an
optimistically biased manner, we acknowledge that it is difficult to completely
control for alternative explanations. For example, the firms being analyzed may
have greater potential that is not fully controlled for in our models. Another
possibility is that analysts may believe that the firm being analyzed has similar
growth opportunities to the selected peers, such that the analysts believe the peers to
be appropriate. Our findings should be interpreted with such caveats in mind.

123

106

G. Franco et al.

Acknowledgments We thank Brian Bratten, Larry Brown, Cristi Gleason, Artur Hugon, Marcus Kirk,
Miguel Minutti-Meza, Partha Mohanram, Rodrigo Verdi, Richard Willis, Kent Womack, Wuyang Zhao,
Yibin Zhou, and Youli Zou as well as workshop participants at University of Notre Dame, University of
Toronto, University of Florida, Norwegian School of Economics, Arizona State University, Purdue
University, University of Illinois at Chicago, University of Iowa, Hong Kong Polytechnic University, and
the AAA Annual (Washington DC), FARS (San Diego), and JCAE (Hong Kong) meetings for comments
and suggestions. We thank Jonathan Chen, Steve El-Hage, Wendy Gu, Megan Jiang, Matthew Literovich,
Jeffrey Ma, Sunyoung Par, Philip Rabenok, Tian Shen, Menghan Yan, and Jimmy Zhu for their excellent
research assistance. We thank Yuyan Guan, Hai Lu, and Franco Wong for data on investment banking
affiliations. We gratefully acknowledge the financial support of the Social Sciences and Humanities
Research Council of Canada, the Rotman School of Management, and the Mendoza College of Business.
Hope further acknowledges the financial support of the Deloitte Professorship.

Appendix
See Table 7.
Table 7 Variable definitions
Affiliation

Indicator variable that equals one if the analysts brokerage firm conducts
investment banking and zero otherwise

AssetTurnover

Total revenues divided by total assets in year t - 1

B/P

Book value per share divided by price per share, at the end of year t - 1

BUY

Indicator variable equal to one for analyst reports containing buy or strong buy
recommendations and zero otherwise

BVPSGrowtht?1

Year t ? 1 revenues less year t revenues scaled by year t revenues

BVPSGrowtht?2

Square root of year t ? 2 revenues less year t revenues scaled by year

t revenues

ComparabilityEarn

R2 from a regression of firm is quarterly earnings on the quarterly earnings of

firm j during the 16 quarters ending in year t - 1

ComparabilityRet

R2 from a regression of firm is monthly returns on the monthly returns of

company j during the 4 years ending in year t - 1

EPSGrowtht?1

Year t ? 1 revenues less year t revenues scaled by year t revenues

EPSGrowtht?2

Square root of year t ? 2 revenues less year t revenues scaled by year

t revenues

E/P

Earnings per share excluding extraordinary items divided by price, at the end
of year t - 1

Leverage

Total long-term debt divided by total assets

LogSales

Natural logarithm of total revenues in year t - 1

Multiple Peer

Peers valuation multiple and is measured in the reciprocal. It takes on one of

four valuation multiples: E/P, B/P, S/EV, and, EBITDA/EV.

P/B

Price per share divided by book value per share, at the end of year t - 1

PeerChoice

Indicator variable that equals one if the analysts report on firm i contains peer
j in determining valuation and zero otherwise

ProfitMargin

Income before extraordinary items divided by total revenues in year t - 1

RaiseRec

Indicator variable that equals one for analyst reports in which the
recommendation is increased from the previous recommendation level and
zero otherwise

123

Analysts choice of peer companies

107

Table 7 continued
RaiseTarget

Indicator variable that equals one for analyst reports in which the target price is
increased from the previous target price and zero otherwise

Return

Stock return during year t - 1

ROAt?1

Net income before extraordinary items plus interest expense in year t ? 1, divided
by assets at the end of year t

ROAt?2

Net income before extraordinary items plus interest expense in years t ? 1 and
t ? 2, divided by assets at the end of year t

ROEt?1

Net income before extraordinary items in year t ? 1 divided by equity at the end
of year t

ROEt?2

Net income before extraordinary items in years t ? 1 and t ? 2 divided by equity

at the end of year t

SalesGrowth

Year t - 1 revenues less year t - 2 revenues scaled by year t - 2 revenues

SameGIC3digit

Indicator variable that equals one if the firm is in the same three-digit GIC code as
the potential peer and zero otherwise

SameGIC5digit

Indicator variable that equals one if the firm is in the same five-digit GIC code as
the potential peer and zero otherwise

SameSIC4digit

Indicator variable that equals one if the firm is in the same four-digit SIC code as
the potential peer and zero otherwise

SGt?1

Year t ? 1 revenues less year t revenues scaled by year t revenues

SGt?2

Square root of year t ? 2 revenues less year t revenues scaled by year t revenues

Sim_

Prefix added to variables that are measured as the absolute value of the difference
between firm is and peer js respective variables, multiplied by -1

TargetUpside

(Target price - current price)/current price

Turnover

Volume of shares traded divided by shares outstanding in year t - 1

Volatility

Standard deviation in monthly stock return during the four years ending in year
t-1

References
Albuquerque, A. (2009). Peer firms in relative performance evaluation. Journal of Accounting and
Economics, 48(1), 6989.
Albuquerque, A., De Franco, G., & Verdi, R. (2013). Peer choice in CEO compensation. Journal of
Financial Economics, 108(1), 160181.
Alford, A. (1992). The effect of the set of comparable firms on the accuracy of the price-earnings
valuation method. Journal of Accounting Research, 30(1), 94108.
Baker, M., & Ruback, R. (1999). Estimating industry multiples. Working paper, Harvard University.
Barber, B., Lehavy, R., McNichols, M., & Trueman, B. (2001). Can investors profit from the prophets?
Security analyst recommendations and stock returns. Journal of Finance, 56, 531563.
Barber, B. M., & Lyon, J. D. (1996). Detecting abnormal operating performance: The empirical power
and specification of test statistics. Journal of Financial Economics, 41, 359399.
Barker, R. G. (1999). Survey and market-based evidence of industry-dependence in analysts preferences
between the dividend yield and price-earnings ratio valuation models. Journal of Business Finance
& Accounting, 26(34), 393418.
Barniv, R., Hope, O.-K., Myring, M. J., & Thomas, W. B. (2009). Do analysts practice what they preach
and should investors listen? Effects of recent regulations. The Accounting Review, 84(4),
10151039.
Barniv, R., Hope, O.-K., Myring, M. J., & Thomas, W. B. (2010). International evidence on analyst stock
recommendations, valuations, and returns. Contemporary Accounting Research, 27(4), 11311167.

123

108

G. Franco et al.

Bhojraj, S., & Lee, C. (2002). Who is my peer? A valuation-based approach to the selection of
comparable firms. Journal of Accounting Research, 40(2), 407439.
Bhojraj, S., Lee, C., & Oler, D. (2003). Whats my line? A comparison of industry classification schemes
for capital market research. Journal of Accounting Research, 41(5), 745774.
Block, S. B. (1999). A study of financial analysts: Practice and theory. Financial Analysts Journal, 55(4),
8695.
Block, S. B. (2010). Methods of valuation: Myth vs. reality. Journal of Investing, 19(4), 714.
Boatsman, J., & Baskin, E. (1981). Asset valuation with incomplete markets. The Accounting Review LVI,
1, 3853.
Boni, L., & Womack, K. (2006). Analysts, industries, and price momentum. Journal of Financial and
Quantitative Analysis, 41(1), 85109.
Bradshaw, M. T. (2011). Analysts forecasts: What do we know after decades of work? Working paper,
Boston College.
Bradshaw, M. T., Miller, G., & Serafeim, G. (2010). Accounting method heterogeneity and analysts
forecasts. Working paper, Boston College, University of Michigan, and Harvard Business School.
Brav, A., & Lehavy, R. (2003). An empirical analysis of analysts target prices: Short-term
informativeness and long-term dynamics. Journal of Finance, 58(5), 19331967.
Chan, L., Lakonishok, J., & Swaminathan, B. (2007). Industry classifications and return comovement.
Financial Analysts Journal, 63(6), 5670.
Clarke, J., Khorana, A., Patel, A., & Rau, P. R. (2007). The impact of all-star analyst job changes on their
coverage choices and investment banking deal flow. Journal of Financial Economics, 84, 713737.
Damodaran, A. (2009). The dark side of valuation: Valuing young, distressed, and complex businesses
(2nd ed.). Upper Saddle River: Pearson Education.
Das, S., Levine, C., & Sivaramakrishnan, S. (1998). Earnings predictability and bias in analysts earnings
forecasts. The Accounting Review, 73(2), 277294.
De Franco, G., Gavious, I., Jin, J., & Richardson, G. (2011a). Do private company targets that hire Big4
auditors receive higher proceeds? Contemporary Accounting Research, 28(1), 215262.
De Franco, G., Hope, O.-K., Vyas, D., & Zhou, Y. (2014). Analyst report readability. Contemporary
Accounting Research. doi:10.1111/1911-3846.12062.
De Franco, G., Kothari, S. P., & Verdi, R. (2011b). The benefits of financial statement comparability.
Journal of Accounting Research, 49(4), 895931.
Dechow, P., Hutton, A., & Sloan, R. (2000). The relation between analysts forecasts of long-term
earnings growth and stock price performance following equity offerings. Contemporary Accounting
Research, 17(1), 132.
Dugar, A., & Nathan, S. (1995). The effects of investment banking relationships on financial analysts
earnings forecasts and investment recommendations. Contemporary Accounting Research, 11(1),
131160.
Ecker, F., Francis, J., Olsson, P., & Schipper, K. (2013). Estimation sample selection for discretionary
accruals models. Journal of Accounting and Economics, 56(2), 190211.
Ertimur, Y., Sunder, J., & Sunder, S. (2007). Measure for measure: The relation between forecast
accuracy and recommendation profitability of analysts. Journal of Accounting Research, 45(3),
567606.
Gleason, C., & Lee, C. (2003). Analyst forecast revisions and market price discovery. The Accounting
Review, 78(1), 193225.
Groysberg, B., Healy, P., & Maber, D. (2011). What drives financial analyst compensation? Financial
Analysts Journal, 67(4), 1829.
Guan, Y., Lu, H., & Wong, F. (2012). Conflict-of-interest reforms and analysts research biases. Journal
of Accounting, Auditing, and Finance, 27(4), 443470.
Healy, P. M., & Palepu, K. G. (2007). Business analysis and valuation: Using financial statements (4th
ed.). Mason, Ohio: Thomson Southwestern.
Heflin, F., Subramanyam, K. R., & Zhang, Y. (2003). Regulation FD and the financial information
environment: Early evidence. The Accounting Review, 78(1), 137.
Hoitash, R., Kogan, A., & Vasarhelyi, M. (2006). Peer-based approach for analytical procedures.
Auditing: A Journal of Practice and Theory, 25(2), 5384.
Imam, S., Barker, R., & Clubb, C. (2008). The use of valuation models by UK investment analysts.
European Accounting Review, 17(3), 503535.
Institutional Investor. (2010 October). The 2010 All-America Research Team (p 52).

123

Analysts choice of peer companies

109

Kadan, O., Madureira, L., Wang, R., & Zach, T. (2012). Analysts industry expertise. Journal of
Accounting and Economics, 54(23), 95120.
Ke, B., & Yu, Y. (2006). The effect of issuing biased earnings forecasts on analysts access to
management and survival. Journal of Accounting Research, 44, 965999.
Kim, M., & Ritter, J. (1999). Valuing IPOs. Journal of Financial Economics, 53(3), 409437.
Koller, T., Goedhart, M., & Wessels, D. (2005). Valuation: Measuring and managing the value of
companies. Hoboken, NJ: McKinsey & Company, Wiley.
Lim, T. (2001). Rationality and analysts bias. Journal of Finance, 56(1), 369385.
Lin, H. W., & McNichols, M. F. (1998). Underwriting relationships, analysts earnings forecasts and
investment recommendations. Journal of Accounting and Economics, 25(1), 101127.
Litov, L., Moreton, P., & Zenger, T. (2012). Corporate strategy, analyst coverage, and the uniqueness
paradox. Management Science, 58(10), 17971815.
Liu, J., Nissim, D., & Thomas, J. (2002). Equity valuation using multiples. Journal of Accounting
Research, 40(1), 135172.
Liu, J., Nissim, D., & Thomas, J. (2007). Is cash flow king in valuations? Financial Analysts Journal,
63(2), 5668.
Malmendier, U., & Shanthikumar, D. (2007). Are small investors naive about incentives? Journal of
Financial Economics, 85, 457489.
McNichols, M., & OBrien, P. (1997). Self-selection and analyst coverage. Journal of Accounting
Research, 35, 167199.
Michaely, R., & Womack, K. (1999). Conflict of interest and the credibility of underwrite analyst
recommendations. Review of Financial Studies, 12(4), 653686.
Michaely, R., & Womack, K. (2005). Brokerage recommendations, Stylized characteristics, market
responses, and biases. In R. H. Thaler (Ed.), Advances in behavioral finance II (pp. 389422).
Princeton, New Jersey: Princeton University Press.
Minutti-Meza, M. (2013). Does using information from peer-firms improve account-level expectation
models? Journal of Accounting Research, 51(4), 779817.
Roosenboom, P. (2007). How do underwriters value initial public offerings? An empirical analysis of the
French IPO market. Contemporary Accounting Research, 24(4), 12171243.
Schipper, K. (1991). Commentary on analysts forecasts. Accounting Horizons, 5(4), 105121.
Stickel, S. (1990). Predicting individual analyst earnings forecasts. Journal of Accounting Research,
28(2), 409417.
Stickel, S. (1992). Reputation and performance among security analysts. Journal of Finance, 47(5),
18111836.
Stickney, C., Brown, P., & Wahlen, J. (2007). Financial reporting, financial statement analysis, and
valuation: A strategic perspective (6th ed.). Mason, Ohio: Thomson/South-Western.

123