Upload
letuyen
View
216
Download
0
Embed Size (px)
Citation preview
0
Investor Relations and Information Assimilation
Kimball L. Chapman
Washington University in St. Louis
Gregory S. Miller
University of Michigan
Hal D. White
Penn State University
PRELIMINARY – Please do not cite or quote without permission
Abstract
Research has shown the importance of corporate disclosure and dissemination in reducing
information asymmetry and improving market efficiency. However, even though investors and
analysts might receive corporate disclosures, they often need help with assimilating the
information to better understand its implications for firm value. This paper examines whether
investor relations (IR) teams provide value by facilitating the assimilation of firm information by
the market. We find that firms with IR officers have lower stock price volatility, lower analyst
forecast dispersion, higher analyst forecast accuracy, and quicker price discovery, consistent with
IR professionals aiding market participants in their assimilation of firm information. We also
show that our findings are stronger for firms with longer-tenured IR officers. Finally, we find that
when firms transition from a long-tenured IR officer to a new IR officer, stock price volatility
increases, analyst forecasts become more disperse and less accurate, and the price discovery
process slows. Collectively, these findings suggest that in-house IR teams, particularly those with
greater experience, help facilitate information assimilation by the market, which has positive
market effects.
We would like to thank Michelle Hanlon, John Core, Rodrigo Verdi, Chandra Kanodia and
workshop participants at the Massachusetts Institute of Technology and the University of
Minnesota for their helpful feedback on the paper.
1
1. Introduction
This paper examines whether in-house investor relations (IR) teams provide value to firms by
facilitating the assimilation of firm information by the market.1 Prior research has shown the
importance of disclosure (Leuz and Verrecchia 2000; Shroff, Sun, White and Zhang 2013) and
dissemination (Bushee, Core, Guay and Hamm 2010; Blankespoor, Miller and White 2013; Twedt
2016) in reducing information asymmetry and improving the price discovery process by
providing firm information to market participants. However, even though investors and analysts
might receive corporate disclosures (or other news) about the firm, they often need help
assimilating the information to better understand its implications for firm value. We contend that
IR professionals play a crucial role in helping the market better understand the firm by engaging
in regular, ongoing interactions with investors and analysts to resolve uncertainties they might
have about the firm, thereby improving market efficiency.2 Our results are consistent with IR
professionals playing such a role.
Information assimilation has become a more salient issue for firms in recent years. In
particular, there has been a large increase in the frequency and length of firm disclosures, which
can overload investors’ processing abilities (Simon 1978; Merton 1987; Hirshleifer and Teoh 2003).
1 We define assimilation as the process of developing a comprehensive and contextual understanding of information.
In our setting, this understanding relates to the firm and its prospects, and thus firm value. Note that assimilation is
related to, but distinct from, the concept of readability. For example, a firm might disclose plans to increase capital
expenditures by 10% in the following year. Investors that receive this disclosure likely understand it from a readability
perspective, i.e., they understand the words and concepts; however, they might not fully appreciate the implications
of the disclosure for firm value—i.e., they need help assimilating the information. 2 The National Investor Relations Institute (NIRI) defines investor relations as the integration of “finance,
communication, marketing and securities law compliance to enable the most effective two-way communication
between a company, the financial community, and other constituencies, which ultimately contributes to a company’s
securities achieving fair valuation” (NIRI 2016). We argue that this two-way communication is precisely the mechanism
through which IR professionals provide a more robust and tailored discussion with firm stakeholders by tying together
pieces of information, summarizing information, correcting misinformation, and clarifying details.
2
Additionally, shareholder activism and firm-related discussions in online forums, newsrooms
and social media have become much more prevalent, as investors and pundits challenge
management’s assertions and strategies. These oft-times unvetted communications and opinions
about firms can travel quickly across a broad set of market participants, which can have a
significant impact on public perceptions, and thus valuation, of the firm (Lee, Hutton and Shu
2015). As a result, a growing number of firms are retaining IR officers to communicate with
market participants on a regular basis to help them assimilate information about the firm.3
We contend that although firm disclosures can be very useful in reducing uncertainty about
the firm, even the most carefully crafted firm disclosures can be insufficient, as it is very difficult
for firms to fully anticipate and effectively address all market participant demands for
information via disclosure. As a result, IR officers spend a lot of time with investors and analysts,
tying together pieces of information, providing clarifications, correcting misunderstandings,
summarizing information, etc., which helps reduce uncertainty about the firm and its prospects.
The importance of these interactions to investors and analysts is demonstrated by survey
evidence in Brown, Call, Clement and Sharp (2015), which indicates that private communications
with the firm are valued by analysts even more than public firm disclosures, such as earnings
guidance, conference calls and financial reports. Note that despite the significant value of these
private discussions, they do not run afoul of financial reporting regulations, even after the
3 As shown in our Table 1, Panel D, the number of micro-cap (mid-cap) firms with IR programs increased from 22%
(44%) in 2003 to 45% (59%) in 2013, and roughly 80% of the largest firms currently have an IR officer.
3
adoption of Regulation FD, as the SEC recognizes investors’ diverse information demands and
the market’s need for help with assimilation (SEC 2000).4
It is also important to note that these discussions between IR professionals and market
participants are nontrivial in both duration and frequency. In fact, IR officers dedicate more of
their time to direct communications with investors and analysts than to any other single task.
According to a 2004 National Investor Relations Institute (NIRI) study, IR departments spend
83% of their time, on average, interacting with analysts and institutional investors. These ongoing
interactions address information demands from analysts and investors as well as mitigate any
misinformation they may have, thereby helping them better understand the firm and its prospects
on a timely basis; that is, IR officers can reduce the market’s uncertainty about the firm.
Accordingly, we predict that firms with IR officers will have lower stock price volatility and that
analysts covering those firms will have less disperse forecasts with improved forecast accuracy.
To examine the impact of IR on information assimilation, we conduct our analyses using an
entropy balancing technique, which is a quasi-matching approach that weights each observation
such that post-weighting distributional properties of treatment and control observations are
virtually identical, thereby ensuring covariate balance (Hainmueller, 2012; McMullin and
Schonberger, 2015).5 Given the decision to initiate an IR program is a firm choice, it is important
to control for factors that drive firms’ decision to hire IR officers. As such, we match on firm size,
market-to-book ratio, leverage, scaled earnings and earnings volatility, as these factors have been
4 The SEC has expressed its desire to preserve mechanisms whereby investors and analysts can build their own unique
"mosaic” of understanding through discussions with the firm. In particular, the SEC states “an issuer is not prohibited
from disclosing a non-material piece of information to an analyst, even if…that piece helps the analyst complete a
"mosaic" of information that, taken together, is material.” (SEC 2000). See Section 2 for more detail. 5 See Section 4 for a detailed discussion of entropy balancing and its advantages over propensity score matching.
4
shown to be related to the decision to hire an IR team (Kirk and Vincent 2014; Bushee and Miller
2012). We also control for the amount of firm disclosure, since firms with IR tend to provide more
press releases (Kirk and Vincent 2014; Bushee and Miller 2012). In addition to controlling for other
firm-specific characteristics, we also test our predictions using firm fixed effect regressions.
Using this approach for a sample of firms from 2002 to 2013, we find that firms with in-house
IR officers have lower stock price volatility, lower analyst forecast dispersion and higher analyst
forecast accuracy, consistent with IR professionals aiding market participants in their assimilation
of firm information. We also examine whether IR officers are more effective at reducing
uncertainty when they have longer tenure, as they will both be more familiar with the operations
of the firm and have established closer relationships with investors and analysts, allowing for
more effective communications between parties. Consistent with our expectations, we show that
our findings are stronger for firms with longer-tenured IR officers.
In economic terms, our results suggest that firms with IR officers (with tenures of at least one
year) have 2.3% lower volatility, a 2.8% quicker price discovery process, and analyst forecasts for
these firms are 7.8% more accurate and 6.3% less dipserse than those for firms without IR officers.
When restricting our sample to firms with IR officers that have more experience (at least three
years), we observe even greater benefits. In particular, these firms show 3.8% lower volatility, a
3.1% quicker price discovery process, and analysts’ forecasts for firms with experienced IR are
10.4% more accurate and 12.5% less disperse than those for firms without an IR officer.
To mitigate potential concerns that an omitted variable is influencing our results, we also
perform the impact threshold of the confounding variable analysis recommended by Larcker and
Rusticus (2010). While this analysis cannot rule out the possibility that an omitted variable
5
influences our results, it does suggest that such a variable would need to have a significantly
larger effect than any of our current control variables to overturn our main inferences. In
particular, a correlated omitted variable would need to be at least 1.5, 2.9, 3.3 or 22.2 times larger
than the most impactful control variable in our tests of volatility, forecast accuracy, forecast
dispersion and price formation, respectively.
We then turn our analysis to three event-specific settings involving IR: (i) earnings
announcements, (ii) major accounting restatements by other firms in the same industry, and (iii)
the loss of a long-tenured IR officer. Each of these settings allows us to examine the role of IR
professionals in the assimilation process. In the first setting, we examine earnings announcements
because these are highly salient information events, where analysts have a particularly strong
interest in talking with management in an attempt to get clarification regarding earnings news
(Soltes 2014). We find that firms with IR officers experience a quicker price discovery process,
consistent with more efficient assimilation of earnings news by the market.
In our second setting, we use an exogenous shock to uncertainty to examine the impact of IR
on information assimilation. In particular, we examine peer firm accounting restatements
following Gleason et al. (2008), who show that when a firm has a restatement, there is not only a
significant price drop for that firm, but also for peer firms within the same industry due to
increased uncertainty about the content and credibility of non-restating firms’ disclosures. We
find that although both IR and non-IR firms suffer price declines when a peer firm has a
significant restatement, the stock price decline for IR firms is less severe and rebounds much more
quickly, consistent with IR officers more efficiently addressing uncertainties held by the market.
6
In our final setting, we examine a firm’s loss of a long-tenured IR officer. Unlike the previous
two settings, in this setting, we expect the information environment to deteriorate. In particular,
long-tenured IR officers have established relationships with investors and analysts that allow for
more effective communications, so their loss should negatively impact the assimilation process.
Consistent with our expectations, we find that stock price volatility increases, analyst forecasts
become more disperse and less accurate, and the price discovery process slows when firms lose
a long-tenured IR officer. These results help rule out competing explanations related to
contemporaneous changes in traditional disclosure, as firms are unlikely to discontinue effective
disclosures when a long-tenured IR officer leaves the firm.
It is important to note that despite all our empirical analyses, we recognize that the behavior
of IR officers is largely unobservable, and thus difficult to capture. Our inferences rely heavily on
the fact that IR officers’ primary role, which takes the vast majority of their time, is to engage
investors and analysts in two-way dialogue to help them better understand the firm and its
prospects (NIRI 2016). However, another role of IR is to oversee the public disclosure of the firm.
Although public firms already release a significant amount of public information, prior research
has shown that IR engagement leads to even greater disclosure (Kirk and Vincent 2014; Bushee
and Miller 2012). Despite controlling for the amount of disclosure in our analyses, it’s reasonable
to assume that as experienced IR officers learn to better communicate with the market, they
improve the firm’s disclosures. The resulting improvements in disclosure may be nuanced and
therefore not picked up by our disclosure control variables. Thus, the impact of IR that we find
in the paper might not solely be the direct result of IR officers’ interactions with market
participants, but rather a combination of their interactions and changes in disclosures resulting
7
from those private interactions. In the end, irrespective of the mechanism, our results provide key
insights into the role of IR in facilitating information assimilation.
Our findings contribute along two major dimensions. First, we contribute to the large literature
that focuses on the role of disclosure (Leuz and Verrecchia 2000; Shroff et al. 2013) and
dissemination (Bushee et al. 2010; Blankespoor et al. 2013; Twedt 2016) in reducing information
asymmetry and improving market efficiency. 6 Collectively, this literature focuses on
management’s ability to get information to relevant market participants via corporate disclosures.
We build on this literature by providing evidence that firms can improve market efficiency and
reduce uncertainty by hiring an IR officer to help the market properly assimilate the information
it receives.
Second, despite the growing popularity of IR programs, surprisingly little research has been
conducted on the value of IR. This nascent literature has generally focused on the role of IR in
attracting visibility for the firm (i.e., greater analyst coverage, news media attention and
institutional ownership), which leads to increased liquidity, and thus higher firm valuations
(Bushee and Miller, 2012; Kirk and Vincent, 2014). However, it has been unclear whether the IR
function adds value along other dimensions. We contribute to this literature by showing that the
IR function can provide additional value to firms by reducing investors’ uncertainty about the
firm through enhanced information assimilation.
The next section discusses the motivation. Section 3 discusses the sample; Section 4
discusses the research design and empirical results. We conclude in Section 5.
6 See Beyer, Cohen, Lys and Walther (2010) for a review of the disclosure literature.
8
2. Motivation
2.1. Assimilation in the capital market
A well-functioning economy relies on capital markets for efficient capital allocation (Rajan and
Zingales 1998). However, information asymmetry between managers and investors exists in these
markets, as there is a separation between ownership and control (Jensen and Meckling 1976). This
information asymmetry induces market frictions by introducing concerns of adverse selection,
resulting in less efficient markets (Akerlof 1970). To combat these frictions, capital markets must
have a rich information environment that provides market participants with the information
necessary to determine firm valuation. Consistent with this notion, prior research has shown the
importance of corporate disclosure (Leuz and Verrecchia 2000; Beyer et al. 2010; Shroff et al. 2013)
and broad dissemination of firm news (Bushee et al. 2010; Blankespoor et al. 2013; Twedt 2016) in
reducing information asymmetry and improving market efficiency.
It is important to note, however, that even when investors and analysts receive corporate
disclosures or other news about the firm, they often need help understanding the implications of
the information for firm value. That is, they need help assimilating the information. For example,
an investor might learn through a firm’s disclosure that a particular firm plans to increase its
capital expenditures, adjust its product mix, invest more in R&D, or increase its presence in a new
region in the coming years. Similarly, an investor might learn about changes in the firm’s
economic environment, e.g., product market innovations, industry shifts in risk, financing
availability, or competitor actions. With each of these items, there will likely be some uncertainty
as to the exact nature of the event or action and ultimately its implications for the firm. Thus, once
9
investors and analysts receive information, they need help assimilating that information to
understand it in a broader, more comprehensive context with respect to firm value.
Although there has always been a need for information assimilation support, it has become a
more important issue for firms in recent years. In particular, there has been a significant increase
in the frequency and length of firm disclosures as a result of major disclosure regulation, most
notably the Sarbanes-Oxley Act of 2002. Given that investors have limited resources and
processing abilities, large volumes of information can ‘overload’ investors, which reduces their
ability to fully process the information (Simon 1978; Merton 1987; Bloomfield 2002; Hirshleifer
and Teoh 2003; Blankespoor, Miller and White. 2014). This can result in differential
interpretations of information across investors, thereby increasing the need and demand for help
with assimilation. In addition to the deluge of corporate disclosure, there have been large
increases in shareholder activism and firm-related discussions from external parties in online
forums, newsrooms and social media, as investors and pundits challenge management’s
assertions and strategies. These oft-times unvetted communications and opinions about firms can
travel quickly across a broad set of market participants, which can have a significant impact on
public perceptions, and thus valuation, of the firm (Lee, Hutton and Shu 2015).
2.2. Role of investor relations in assimilation
To address these information challenges, firms are interacting more consistently with the
market. As a result, managers must decide how they want to interact with investors. Many firms
simply rely on top management (e.g., CFO) to conduct IR activities. The downside of this
approach is that IR activities are time-consuming and require specialized expertise. As an
10
alternative, some firms hire external IR consultants to perform IR activities. The advantage of IR
consultants is that they are typically much less costly than hiring a dedicated IR professional and
are therefore a better fit for firms that are financially constrained or for those with lower demand
for IR services. However, one possible limitation of hiring an external IR firm is that they are often
shared across multiple firms. While a shared IR firm may have a strong overall expertise in IR,
they are less likely to have as detailed an understanding about the firm or have established
relationships with the firm’s investors and analysts relative to an internal IR officer, so they are
much less helpful in facilitating information assimilation. 7
Internal IR officers are typically senior-level managers who are authorized spokespersons for
the firm and are also the primary point of contact for investors (Thomson, 2009). IR officers most
often report directly to the CFO of the firm and frequently provide capital-market intelligence to
the CEO and Board of Directors (Mellon, 2013). We contend that these IR professionals not only
provide considerable help as experts in disclosure compliance, but they can play a critical role in
helping the market better understand the firm by engaging in regular, ongoing dialogue with
investors and analysts to resolve uncertainties they might have about the firm, thereby improving
market efficiency.8
7 Note that the focus of our study is on the impact of internal IR officers, whereas prior studies, such as Bushee and
Miller (2012) and Solomon (2012), are focused on external IR consulting firms. As described above, there are clear
differences between the two types of IR professionals, which should be considered when comparing studies. 8 In considering whether to implement an in-house IR program, managers must also consider the costs of doing so. IR
programs are costly to maintain, with a mean annual IR budget of nearly $900,000 (BNY, 2013). Additionally, to the
extent that IR programs lead to increased interactions with investors, they are likely to increase time required by other
managers and costs of generating information to address investors’ inquiries. Implementing and maintaining costly IR
programs is made more difficult by the fact that it is often hard to judge their efficacy. That said, an increasing
proportion of firms appear to believe an internal IR program is worth the cost, as evidenced by the increasing
proportion of firms that have internal IR programs (see Table 1, Panel D).
11
IR officers can help investors assimilate information through several means. First, they can
point investors to additional information that should be considered in parallel to contextualize
the disclosure. This additional information might be public already, but investors either did not
have these facts readily accessible or simply didn't appreciate the relations. Second, IR officers
can summarize a large amount of information, as market participants are often time-constrained,
particularly when they follow many firms. Third, they can address any incorrect information held
by investors. Finally, IR officers can clarify details to investors regarding their disclosures. As one
analyst describes in Brown et al. (2015), “It’s not nonpublic material information; it’s clarification
of points. They help you digest the information a little bit better.” This can also include making
sense of the firm’s economic environment. As indicated by another analyst in Brown et al. (2015),
“We ask for qualitative thoughts and insights into industry trends or specific business lines, just
so that we’re also double-checking our own thought processes and that our models are solid.”
Collectively, these communications provide a more complete vision of the firm.
2.3. Importance of IR experience in facilitating assimilation
With more job experience, IR officers are likely to grow more effective at information
assimilation through repeated interactions with investors and analysts. Findings from social
psychology and organizational behavior describe how repeated interactions generate
interpersonal familiarity, which makes communications more efficient and effective. Steiner
(1972) suggests that interpersonal familiarity allows individuals to have more productive
interactions because: (i) they can spend more time and effort on the task at hand, rather than
acquiring the interpersonal information that is a necessary precondition for productive
12
coordinated effort, and (ii) there is a lowered risk of mistrust, role ambiguity and
miscommunication. Further, Jehn and Shah (1997) show that familiarity makes behavior between
individuals more predictable and future interactions less uncertain, and therefore less costly.
Inter-personal familiarity also allows individuals to distinguish interpersonal conflict from
non-personal disagreement, making it easier for individuals to disagree in constructive ways
(Shah and Jhen, 1993). Moreover, familiarity can lead to friendship, and as relationships shift
towards friendship, levels of task-oriented communication and coordination improve (Jhen and
Shah, 1997). Collectively, these findings suggest that when IR officers gain experience through
repeated interactions with investors, they can reduce uncertainties and misperceptions about the
firm by improving the effectiveness and efficiency of the communications, and thus improve
information assimilation.
2.4. Contrasting perspective on the role of IR in facilitating assimilation
One might argue that IR can only play a trivial role in reducing uncertainty about the firm, as
firms are already required to publicly disclose all relevant material information. However,
corporate disclosures are often insufficient, as it is very difficult for firms to convey information
in a comprehensive manner that addresses all market participant needs without overloading
them with disclosure. Moreover, there is typically large variation across investors’ information
demands. As such, there is significant demand from market participants for private discussions
with the firm, and this information appears to be very useful. In fact, in a survey of 365 analysts,
Brown et al. (2015) find that private communication with the firm “is a more important input to
13
analysts’ earnings forecasts and stock recommendations than even management earnings
guidance, earnings conference calls, and the recent 10-K or 10-Q report.”
Despite their idiosyncratic nature and value to investors, these discussions do not run afoul of
Regulation Fair Disclosure (Reg FD) rules. Interpretative guidelines published by the SEC
regarding Reg FD allow for private communications related to clarification and interpretation:
“[A]n issuer ordinarily would not be conveying material nonpublic information if it corrected historical facts
that were a matter of public record. An issuer also would not be conveying such information if it shared
seemingly inconsequential data which, pieced together with public information by a skilled analyst with
knowledge of the issuer and the industry, helps form a mosaic that reveals material nonpublic information. It
would not violate Regulation FD to reveal this type of data even if, when added to the analyst's own fund of
knowledge, it is used to construct his or her ultimate judgments about the issuer.”9
In additional guidance, the SEC notes its broader perspective on private communications with
the firm and information assimilation, particularly with respect to analysts:
“[A]n issuer is not prohibited from disclosing a non-material piece of information to an analyst, even if,
unbeknownst to the issuer, that piece helps the analyst complete a "mosaic" of information that, taken
together, is material. Similarly, since materiality is an objective test keyed to the reasonable investor,
Regulation FD will not be implicated where an issuer discloses immaterial information whose significance is
discerned by the analyst. Analysts can provide a valuable service in sifting through and extracting information
that would not be significant to the ordinary investor to reach material conclusions. We do not intend, by
Regulation FD, to discourage this sort of activity. The focus of Regulation FD is on whether the issuer discloses
material nonpublic information, not on whether an analyst, through some combination of persistence,
knowledge, and insight, regards as material information whose significance is not apparent to the reasonable
investor.”10
Accordingly, we contend that IR communications can have a significant impact on information
assimilation, even when they are entirely consistent with Reg FD.
It is also important to highlight that IR spends a significant amount of time engaging with
market participants. According to a 2004 study conducted by the National Investor Relations
Institute, IR departments spend 83% of their time, on average, interacting with analysts and
9 http://www.sec.gov/divisions/corpfin/guidance/regfd-interp.htm 10 https://www.sec.gov/rules/final/33-7881.htm
14
institutional investors. Moreover, the 2012 BNY Mellon Global Trends in IR survey indicates that
IR officers typically undertake an average of 145 one-on-one meetings with investors and analysts
each year. This does not include one-on-one meetings conducted between other managers and
investors, in which the IR officer is also likely to participate (Mellon 2012). Combined with the
large volume of phone discussions, in-person investor events, company site visits, non-deal road
shows, etc, it is clear that IR can play a significant role in helping the market assimilate firm
information. While the CEOs and CFOs can also help with the assimilation process, they generally
have neither the time to adequately address idiosyncratic information demands nor the same
degree of specialized training and frequent practice in judging which private communications
are permissible, and may therefore hesitate to engage in private communication.11
3. Sample
To conduct our analyses, we start by collecting all available earnings conference call transcripts
from the Reg FD newswire service available through Factiva for the years 2002 (the earliest year
conference call transcripts are available) through 2013 (the last full year for which transcripts were
available at the time we collected data for this study). We then identify firms with IR officers and
infer the tenure of the IR officer from the names and titles of participants on the earnings
conference call transcripts. We identify IR officers as managers whose titles do not include words
indicative of either the CEO or CFO but do include at least one of these words: “IR”, “investor
11 In contrast to the information assimilation role of IR, some evidence suggests that IR may have a role in “hyping the
stock”. For example, Solomon (2012) finds that outsourced IR firms generate greater media coverage of their clients’
good news press releases relative to their bad news press releases, and that this positive media coverage increases
returns around news announcements, but reverses at the upcoming earnings announcements. To the extent that this
finding is indicative of a broader attempt by IR teams to mislead investors in general, we should not observe reductions
in investor uncertainty for firms with IR. Thus, it is an empirical question as to whether IR does indeed positively
impact firms’ information environment.
15
relations”, or “investor”. Because some IR officers are also responsible for public relations
activities, we also include managers whose titles include the words “public relations,” “external
relations,” or “strategic” when no IR officers are identified using the previous list. We categorize
firms with no IR officer participating on the conference call transcript as non-IR firms.12
We infer IR officer tenure by observing when each IR officer was first listed as a participant on
a conference call. To improve the accuracy and relevance of our tenure measure, we exclude (i)
firms with fewer than 15 observations spread across the entire unadjusted sample, as it is less
clear how long the IR officers of these firms have been serving in their positions, and (ii)
observations where IR officer tenure is less than one year (as the benefits of IR officers are not
immediate because time is required to establish a credible relationship with investors). We match
our adjusted sample of IR officer firm-quarters from the Reg FD newswire to Compustat by firm
name and the proximity of the conference call date from the Reg FD newswire to the earnings
announcement date recorded in Compustat.
Table 1 presents sample descriptive statistics. In particular, Panel A of Table 1 summarizes our
sample construction methodology. Our primary sample consists of 144,822 firm-quarters across
both IR and non-IR firm-quarters. This is the maximum sample size used in our empirical tests.
Tests excluding IR observations where tenure is less than three years have a maximum sample
size of 132,748 observations. Empirical tests using properties of analyst forecasts (accuracy or
12 IR officers need not speak on the call to be included in the list of participants, as firms hosting conference calls provide
the names and titles of participants in advance of the call to the hosting operator. Also, participant names and titles are
often provided during the introductory remarks. To mitigate incorrect classifications, in firm-quarters in which we do
not observe an IR officer, but the same IR officer is listed in previous and subsequent quarters and no other IR officer
is observed in the interim, we assume the IR officer was in the job during the missing quarter and adjust the sample
accordingly.
16
dispersion) use the largest possible number of observations from the primary sample where data
used to construct these measures are available. Panel B of Table 1 shows that our sample consists
of firms across a broad range of industries, with a heavier concentration of firms in the financial
and business equipment industries. Panel C of Table 1 shows a relatively uniform, albeit slightly
decreasing trend in the number of observations between the years 2002 (12,811 observations) and
2013 (10,393 observations). Finally, Panel D of Table 1 documents the increasingly common use
of IR officers across firms of different sizes during our sample period.
4. Research design and empirical results
The primary function of IR professionals is to interact with investors and analysts on an
ongoing basis to address their information needs as well as mitigate any misinformation they
may have, thereby helping them better understand the firm and its prospects on a timely basis.
That is, IR officers help reduce the market’s uncertainty about the firm. In this section, we discuss
our empirical approach to examine the relation between IR programs and uncertainty. More
specifically, for our main analyses, we examine the relation between the presence of IR and stock
price volatility, analyst forecast dispersion, and analyst forecast accuracy. We then conduct
several additional analyses to support our inferences drawn from our main analyses. Table 2
provides summary statistics and definitions of the dependent variables, variables of interest, and
control variables used in our analyses.
Throughout our main analyses, we estimate our regressions using two approaches. First, we
use firm fixed effects to account for any unobservable time-invariant factors related to the firm,
which can bias our inferences. Second, we employ an entropy balancing technique, which is a
17
quasi-matching approach that weights each observation such that post-weighting distributional
properties of treatment and control observations are virtually identical, thereby ensuring
covariate balance (Hainmueller, 2012; McMullin and Schonberger, 2015).13 To see the intuition
behind entropy balancing, consider it in the context of the traditional propensity score matching
(PSM) approach. Under PSM, observations are essentially assigned a weight of 0 or 1; that is, they
are either included or excluded from the matched sample based on the outcome of the first-stage
model. In comparison, entropy balancing weights observations on a continuous scale, thereby
preserving the entire sample and ensuring covariate balance by identifying the precise weights
of control observations that allow for an optimal weighted match with treatment observations.
Our use of entropy balancing rather than the commonly used PSM approach is motivated by
two primary considerations. First, when we employ the PSM approach, we fail to achieve
covariate balance due to differences between treatment and control observations along several
determinants of IR, which raises concerns about our ability to use PSM for causal inferences
(Drake, 1993). Second, using the PSM approach reduces our sample size by 65% to 75% in our
main analyses because of the imbalance between IR and non-IR observations in our un-matched
sample. (See Table 3 for distributional properties of the matching variables for the original
sample, the sample after propensity score matching and the sample after entropy balancing.)
13 In particular, the entropy balancing method works by first determining the distributional properties (mean and
variance) of the treatment observations. These distributional properties become the target distributional properties of
the post-weighting control sample (also known as the “balance conditions”). The algorithm proceeds by first assigning
possible weights to control observations and then testing whether the balance conditions have been met (i.e.,
distributional properties of treatment and post-weighted control observations are identical). The algorithm repeats this
process over multiple iterations until a set of weights for control observations are found such that the balance
conditions are met. Treatment observations are not re-weighted, meaning they retain their default weighting of one
while control observations are assigned a positive weight that may be greater or less than one. After the algorithm
finishes assigning weights to each observation, these weights are used in subsequent regression analyses.
18
The entropy balancing technique preserves our full sample and ensures covariate balance
between treatment (IR) and control (non-IR) observations by re-weighting observations such that
the post-weighting mean and variance for IR and non-IR observations are virtually identical
along the following dimensions: firm size, market-to-book ratio, leverage, scaled earnings and
earnings volatility. These factors have been shown to be related to the decision to hire an IR team
(Kirk and Vincent 2014; Bushee and Miller 2012). In addition to these benefits, entropy balancing
also has higher model efficiency and less first-stage model dependency than PSM (Hainmueller,
2012). Accordingly, we report our results using entropy balancing.14
4.1 Main Analyses
4.1.1 IR and stock return volatility
Our first main analysis examines the relation between IR and investor uncertainty. We posit
that if IR professionals help reduce investors’ uncertainty about the firm by helping them better
assimilate information, we should observe reduced stock return volatility. Our intuition draws
from prior theoretical and empirical work (Barry 1978; Brown 1979; Dye 1985; Pastor and
Veronesi 2003; Billings, Jennings and Lev 2015). The idea is that investors are uncertain about the
parameters of the distribution of firms’ future cash flows and earnings. However, they learn
about these parameters over time as they receive information about the firm. Given investors’
uncertainty is positively correlated with stock return volatility (Barry 1978; Brown 1979; Billings
et al. 2015), to the extent IR professionals can lower uncertainty through interactions with
investors, there should be a decrease in return volatility.
14 Despite the advantages of entropy balancing over PSM, in untabulated analyses, we also conduct our tests using
PSM and our results are qualitatively similar.
19
To test the effects of IR programs on stock return volatility, we estimate the following OLS
regression across the primary sample of IR and non-IR firm-quarters, as indicated:
ReturnVolatilityit = β0+ β1IRFirmit + Controlsit + εit (1)
where ReturnVolatility is the standard deviation of daily abnormal returns during the quarter
multiplied by 100; IRFirm is an indicator variable set equal to one for firm-quarters in which we
observe an IR officer. Controls is a vector of control variables intended to absorb variation in stock
return volatility attributable to firm characteristics, capital market conditions and/or
characteristics of the firm’s information environment likely to be associated with stock return
volatility over the quarter. In particular, LnAtq controls for firm size. Growth controls for the
change in sales. ScaledEarnings and UnexpectedEarnings control for firm profitability and
performance relative to expectations. We control for M&A activity, as it may increase differences
of opinion among investors about the future prospects of the firm. We control for variability in
earnings (EarningsVolatility) because investor uncertainty is likely to be higher for firms with more
variable earnings. Prior12MonthReturn controls for the effect of stock price momentum on
investor uncertainty. Leverage controls for the financial risk of the firm. MB controls for the
growth prospects of the firm. Forecasts and NumPressReleases are used to proxy for disclosure, as
management forecasts and other press releases convey important information about firm value
(Healy and Palepu 2001). MediaMentions captures the amount of media coverage the firm receives.
NumAnalysts and InstOwnership capture the degree of external monitoring and analysis. Detailed
variable definitions are provided in the footnotes to Table 2.
To the extent IR programs help investors assimilate information, we expect that IR firms will
have lower stock price volatility relative to non-IR firms. Thus, we predict a negative coefficient
20
estimate on IRFirm (β1<0) in Equation (1). Column 1 (2) of Table 4 reports the results from
estimating Equation 1 using firm fixed effects (entropy balancing), excluding observations where
IR tenure is less than one year. Consistent with our prediction, the coefficient estimate for β1 is
negative and significant at the 1% level across both estimation methods. In Columns 3 and 4 of
Table 4, we repeat our estimation of Equation (1) excluding observations where IR tenure is less
than three years based on the intuition that longer-tenured IR officers are likely to be more
effective at helping investors assimilate information. The results are similar to those in Columns
1 and 2; however, the coefficient estimate for β1 is slightly more negative and statistically
significant at the 1% level. As shown at the bottom of Table 4, a test of differences in the IRFirm
coefficients across the two groups suggests that longer tenured IR officers have a larger impact
on stock volatility. The economic magnitude of these results implies that firms employing an IR
officer who has a minimum tenure of at least one (three) year(s) have 2.3% (3.8%) lower stock
price volatility relative to firms without an IR officer. Collectively, the findings in Table 4 provide
evidence consistent with our prediction that IR firms have lower stock price volatility than do
non-IR firms, as IR officers reduce uncertainty about the firm by helping investors assimilate
information. Moreover, this effect is more pronounced for firms with longer tenured IR officers.
4.1.2 IR and analyst forecast properties
We next examine the impact of IR on analyst uncertainty, where we focus our analyses on
analysts’ forecast dispersion and forecast accuracy. By engaging in regular, ongoing dialogue
with analysts, IR professionals can help analysts reduce their uncertainties regarding the firm. By
addressing the idiosyncratic needs across analysts, IR professionals effectively coordinate the
21
collective belief across analysts, thereby decreasing analyst forecast dispersion. In turn, this more
robust understanding of the firm and its prospects should result in analysts providing more
accurate forecasts about a firm’s future performance.
In order to test the effects of IR on analyst forecast dispersion and accuracy, we estimate the
following regression across the primary sample of IR and non-IR firm-quarters, as indicated:
ForecastDispersionit or ForecastAccuracyit = β0+ β1IRFirmit + Controlsit +εit (2)
where ForecastDispersion is the standard deviation of analyst forecasts made before the earnings
announcement date, and ForecastAccuracy is the negative absolute value of the mean analyst
forecast error, deflated by stock price for the latest quarterly EPS forecast made by each analyst
before the earnings announcement date. Controls is defined as before.
Given the frequency with which IR officers interact with analysts to help them assimilate
information by addressing their idiosyncratic information needs, we expect to observe less
disperse and more accurate analyst forecasts for IR firms relative to non-IR firms. As such, we
predict a negative coefficient estimate on IRFirm (β1<0) when ForecastDispersion is the dependent
variable and a positive coefficient estimate on IRFirm (β1>0) when ForecastAccuracy is the
dependent variable in Equation (2).
Panel A (B) of Table 5 reports the results from estimating Equation (2) using ForecastDispersion
(ForecastAccuracy) as the dependent variable. Further, Columns 1 and 2 (3 and 4) indicate
estimations excluding observations where IR tenure is less than one year (three years) using both
firm fixed effects and entropy balancing methods. Consistent with our predictions, analyst
forecasts are less disperse and more accurate for firms with IR. In particular, IRFirm is negative
22
(positive) and statistically significant at the 10% level or better across all columns in Panel A
(Panel B). Further, there is weak evidence that the IR effect is stronger for the sample of firms
with longer tenured IR officers. The economic magnitude of these results implies that firms
employing an IR officer who has a minimum tenure of at least one (three) year(s) have 7.8%
(10.4%) more accurate and 6.3% (12.5%) less disperse analyst forecasts relative to firms without
an IR officer. Collectively, the evidence in Tables 4 and 5 suggest that IR officers help the market
assimilate information, resulting in improved market outcomes.
4.2 Additional Analyses
In our main analyses, we examine the impact of IR officers on investors’ assimilation of firm
news (as captured by reduced stock return volatility) and analysts’ assimilation of firm news (as
captured by decreased forecast dispersion and increased forecast accuracy). In these tests, we
attempt to mitigate coefficient bias along several dimensions. Namely, we include year and firm
fixed effects to difference out unobserved time-specific and/or firm-specific characteristics
affecting market outcomes. We also use an entropy balancing method to address omitted variable
bias. Further, we control for several firm-level characteristics, capital market conditions and
characteristics of the firm’s information environment that could impact market outcomes.
Notwithstanding these factors, we conduct several additional tests to support our inferences
and improve our internal validity. In particular, we examine two event-specific settings in which
the market faces increased uncertainty and IR can play a significant role in facilitating
assimilation: (i) earnings announcements, and (ii) major accounting restatements by other firms
in the same industry. We also examine the loss of a long-tenured IR officer, which should hinder
23
assimilation because long-tenured IR officers have established relationships with investors and
analysts that allow for more effective communications.
4.2.1 IR and the price discovery process
Earnings announcements are very important information events. Soltes (2014) documents that
demand for private communications with managers peaks following earnings announcements as
investors and analysts attempt to understand and interpret earnings news. Given the high degree
of information processing and interpretation occurring around this time, we expect the effects of
information assimilation will be particularly impactful.
To capture the role of IR in this setting, we examine price discovery, which is the process by
which information is impounded into a firm’s public stock price. In particular, we estimate the
speed of price discovery using an intraperiod timeliness (IPT) metric following prior research
(Twedt 2016; Bushman, Smith, and Wittenberg-Moerman 2010; Butler, Kraft, and Weiss 2007;
McNichols 1984). This metric captures the timeliness with which information is impounded into
prices from the earnings announcement date through the tenth day after the earnings
announcement (day 0 to day +10).15 To calculate IPT, we first construct a curve that plots (for each
day in the window) the proportion of the event window’s abnormal return realized up to and
including that particular day. For day m, this value is the cumulative buy-and-hold abnormal
return from day 0 through day m, scaled by the cumulative buy-and-hold abnormal return for
the entire period. The resulting IPT measure estimates the area under the curve for each firm-
15 We also conduct our analyses using 15-day and 30-day windows, and our results are qualitatively similar.
24
quarter as a measure of price discovery, where greater values (areas) signal more timely price
discovery. This process for creating the IPT measure is summarized in the formula below:
IPT0, 10 = ½ Σ10m=0 (BHm-1 + BHm)/BH10 = Σ9m=0 (BHm)/BH10 )+ 0.5 (3)
Using this IPT measure, we then estimate the following regression across the primary sample of
IR and non-IR firm-quarters, as indicated:
IPTit = β0+ β1IRFirmit + Controlsit+ εit,, (4)
where IPT is described above, and Controls is defined as before for our main analyses. If firms
with IR officers are able to help market participants assimilate firm information more efficiently
and effectively, we should observe more timely price discovery—i.e., a greater IPT value.
Accordingly, we predict a positive coefficient estimate on IRFirm (β1>0).
Columns 1 and 2 (3 and 4) of Table 6 report the results of estimating Equation (4). The
coefficient estimate for β1 is positive and statistically significant across all columns of Table 6. The
economic magnitude of these results implies that firms employing an IR officer who has a
minimum tenure of at least one (three) year(s) have a 2.8% (3.1%) quicker price discovery process.
These results support our main inferences that IR officers help capital market participants
assimilate information, resulting in a timelier price discovery process.
4.2.2 IR and peer firm accounting restatements
We next examine the effect of IR when there is a major accounting restatement by a peer firm
in the same industry. In this setting, we expect that the information assimilation role of IR will be
particularly beneficial because a significant accounting restatement by a peer firm serves as an
25
exogenous shock that is likely to increase uncertainty among investors about other firms in the
same industry. As shown in Gleason, Jenkins and Johnson (2008), when a firm has a restatement,
there is not only a significant price drop for that firm, but there is a contagion effect whereby peer
firms within that same industry also experience a decline in their stock price, as there are concerns
about the content and credibility of their information. These events are likely to trigger
considerable discussion with the firm to reduce investors’ uncertainties. Accordingly, we expect
the contagion effect to be less severe for IR firms. In order to test this prediction, we estimate the
following regression across the sub-sample of IR and non-IR firm-quarters in industries in which
a major accounting statement is announced:
Cumret [2, 30] = β0+ β1IRFirmit + Controlsit+ εit, (5)
where Cumret [2, 30] is the cumulative abnormal return measured over days +2 to +30 after the
filing date of an 8K marked as item 4.02 of a firm within the same industry for which the
cumulative abnormal return to the restatement firm was negative 1% or lower, which are the
restatements most likely to lead to industry contagion (Gleason et al. 2008). Equation (5) is
estimated across all firms in the same SIC 4-digit industry as the restatement firm. If the contagion
effect is reduced for IR firms relative to non-IR firms in the same industry, we should observe a
positive coefficient estimate on IRFirm (β1 > 0).
Panel A of Table 7 reports the cumulative abnormal return around the 8-K filing of the
restatement (-1, +1) as well as four post-announcement return windows: (+2, +5), (+2, +15), and
(+2, +30). As shown, IR peer firms suffer less of a contagion effect than do non-IR peer firms—i.e.,
-0.24% for IR firms vs. -0.35% non-IR firms (p-value = 0.024). Further, IR peer firms’ stock price
reverts to its previous level within 15 days, whereas non-IR firms appear to continue to suffer
26
from the contagion effect over the first 30 days, consistent with IR officers effectively and
efficiently addressing market participants’ concerns about the firm.
Columns 1 and 2 (3 and 4) of Panel B report the results from estimating Equation (5) excluding
observations where IR tenure is less than one year (three years). Consistent with our prediction
and the univariate results in Panel A, the coefficient estimate for β1 is positive and statistically
significant across all columns, indicating that the contagion effect is less severe for IR firms
relative to non-IR firms within the same industry. Moreover, the coefficient magnitudes are
higher in Columns 3 and 4, where we restrict the sample of IR firms to those with greater tenure.
This analysis provides support for our main inferences using an exogenous shock to the
information environment to capture the effect of IR on assimilation.
4.2.3 Transition from long-tenured to new IR officer
As a final test of the impact of IR on information assimilation, we examine whether the capital
market outcomes we examined in our previous analyses (return volatility, forecast accuracy,
forecast dispersion and price discovery) deteriorate when a firm loses a long-tenured IR officer.
As discussed earlier, as IR officers gain experience, they develop a more robust understanding of
the firm and establish trust and credibility with investors and analysts, which improves the
effectiveness and efficiency of their communications, resulting in improved information
assimilation. Thus, when experienced IR officers leave the firm, assimilation should deteriorate.
An important benefit of this analysis is that it helps rule out competing explanations related
to contemporaneous effects of traditional disclosure. That is, it is possible that IR firms simply
have higher quality disclosures. Although we proxy for the firm’s disclosure policy using
27
earnings guidance and the number of firm-initiated press releases in our main analyses (i.e.,
Forecasts and NumPressReleases), these controls may not capture nuanced differences in disclosure
quality. The advantage of this setting is that disclosures are likely to remain largely unchanged
around a change in the IR officer.
Our intuition for examining IR turnover rather than the initial hiring of an IR officer is that the
hiring of an IR officer might correspond with a broader initiative by the firm to increase
transparency, thereby making it difficult to disentangle which of the two effects caused the
improved information environment. In contrast, it is highly unlikely that the turnover of an IR
officer would be correlated with a firm-initiated reduction in disclosure unrelated to the IR
officer. Thus, if we observe deterioration in the information environment when an experienced
IR officer leaves the firm, we argue that the IR officer must have played a positive role in the
assimilation process over and above the firm’s traditional disclosure policy.
To conduct our test, we eliminate all non-IR firms from the primary sample, which reduces
the sample from 144,822 to 28,744 firm-quarter observations. We then further restrict the sample
to 568 instances in which a long-tenured IR officer (with a minimum tenure of 3 years) leaves the
position and where we observe a new IR officer within two quarters. We collect the 8 firm-
quarters around the quarter in which the long-tenured IR officer departs the position, which
represent 4,544 firm-quarters. As a final restriction, we collapse these observations in average pre
and post period values to mitigate the possibility of serial correlation arising from using a non-
collapsed sample. These steps result in a final sample of 1,136 observations.
Using this reduced sample, we estimate the following regression:
28
Market Outcomeit = β0+ β1Postit + Controlsit + εit, (6)
where Market Outcome is ReturnVolatility, ForecastDispersion, ForecastAccuracy or IPT, and Post is
an indicator variable equal to one for the two years following the quarter in which the new IR
officer is appointed. Controls represents the control variables included in the earlier tests. If
assimilation deteriorates with the loss of a long-tenured IR officer, we should observe a positive
coefficient estimate on Post (i.e., β1 > 0) when the dependent variable is ReturnVolatility or
ForecastDispersion and a negative coefficient estimate on Post (i.e., β1 < 0) when the dependent
variable is ForecastAccuracy or IPT.
Table 8 reports the results from estimating Equation (6). In spite of the large reduction in
sample size, results in Table 8 are consistent with our predictions. These findings indicate that
capital market outcomes deteriorate when firms transition from a long-tenured IR officer to a new
IR officer, suggesting there is better information assimilation by IR officers when they have been
in their position for a longer time. This supports the view that relationships and expertise matter
in information assimilation.
One possible alternative explanation for these results is that there may have been a
contemporaneous turnover in top management (CEO), which caused greater uncertainty about
the firm’s future. To mitigate this concern, we estimate Equation (6) across the sample of
observations for which there is CEO information available on ExecuComp, which reduces our
sample from 1,136 to 908 observations. After controlling for CEO turnover, our results continue
to hold at statistically significant levels for ForecastDispersion, ForecastAccuracy and IPT, but we no
longer find a significant result for ReturnVolatility (t-stat=1.15).
29
Another possible alternative explanation is that other factors at the firm changed
simultaneously to the change in IR officer, such as firm disclosure or profitability, which could
cause the negative future outcomes we observe. However, in untabulated results, we find no
difference between the frequency of forecasts, the level of media coverage or firm earnings when
comparing the pre and post periods. Overall, these additional tests support our main inferences.
4.2.4 Impact threshold of the confounding variable
We test the sensitivity of our results to possible correlated omitted variables using the impact
threshold for a confounding variable (ITCV) procedure recommended by Larcker and Rusticus
(2010) and Frank (2000). This approach calculates the minimum magnitude an omitted variable
would need to be (relative to the most impactful included variable) in order to overturn the main
result. Table 9 reports the results of this analysis, which we conduct for each of our main findings:
stock price volatility, analyst forecast accuracy and dispersion and price formation in columns
(1), (2), (3) and (4), respectively. We find that an omitted (confounding) variable would need to
be at least 1.5, 2.9, 3.3 or 22.2 times larger than the most impactful control variable for tests of
volatility, forecast accuracy, forecast dispersion and price formation, respectively. While these
results cannot rule out the possibility that an omitted variable influences our results, they show
that such a variable would need to be rather large in magnitude in order to overturn our results.
5. Conclusion
The purpose of this paper is to study the role of IR programs in improving information
assimilation. We show that firms with in-house IR programs have lower stock price volatility,
higher analyst forecast accuracy, lower analyst forecast dispersion and quicker price discovery
30
around earnings announcements, and these results are generally stronger for longer-tenured IR
officers. We also see a similar effect consistent with better information assimilation for IR firms
when there is an accounting restatement for other firms in the industry. Finally, we observe a
deterioration in the information environment when firms transition from a long tenured to new
IR officer, as IR experience plays a major role in assimilation.
These results contribute to our understanding of how information assimilation influences the
information environment and capital market outcomes. It also expands upon previous literature
documenting various capital market benefits of IR programs. These results are relevant to
managers given the increasing popularity of IR programs and to capital market participants given
our evidence that IR programs improve information assimilation, particularly during times of
industry uncertainty. Our results also suggest that IR programs provide a potential remedy to
problems arising from limited investor attention and information overload.
31
References
Akerlof, G. A. (1970). The market for Lemons: Quality, Uncertainty and the Market Mechanism.
Quarterly Journal of Economics, 84.
Barry, C. B. (1978). Effects of uncertain and nonstationary parameters upon capital market
equilibrium conditions. Journal of Financial and Quantitative Analysis, 13(03), 419-433.
Beyer, A., Cohen, D., Lys, T., & Walther, R. (2010). The Financial Reporting Environment: Review
of The Recent Literature. Journal of Accounting and Economics, 50(2), 296-343.
Billings, M. B., Jennings, R., & Lev, B. (2015). On guidance and volatility. Journal of Accounting and
Economics, 60(2), 161-180.
Blankespoor, E., Miller, B. P., & White, H. D. (2014). Initial evidence on the market impact of the
XBRL mandate. Review of Accounting Studies, 19(4), 1468-1503.
Blankespoor, E., Miller, G. S., & White, H. D. (2013). The role of dissemination in market liquidity:
Evidence from firms' use of Twitter™. The Accounting Review, 89(1), 79-112.
Bloomfield, R. J. (2002). The “incomplete revelation hypothesis” and financial reporting.
Accounting Horizons, 16(3), 233-243.
Brown, L. D., Call, A. C., Clement, M. B., & Sharp, N. Y. (2015). Inside the “Black Box” of Sell‐Side
Financial Analysts. Journal of Accounting Research, 53(1), 1-47.
Brown, S. (1979). The effect of estimation risk on capital market equilibrium. Journal of Financial
and Quantitative Analysis, 14(02), 215-220.
Bushee, B. J., Core, J. E., Guay, W., & Hamm, S. J. (2010). The role of the business press as an
information intermediary. Journal of Accounting Research, 48(1), 1-19.
Bushee, B. J., & Miller, G. S. (2012). Investor relations, firm visibility, and investor following. The
Accounting Review, 87(3), 867-897.
Bushman, R. M., Smith, A. J., & Wittenberg‐Moerman, R. (2010). Price discovery and dissemination of
private information by loan syndicate participants. Journal of Accounting Research, 48(5), 921-972.
Butler, M., Kraft, A., & Weiss, I. S. (2007). The effect of reporting frequency on the timeliness of earnings:
The cases of voluntary and mandatory interim reports. Journal of Accounting and Economics, 43(2),
181-217.
Drake, C. (1993). Effects of misspecification of the propensity score on estimators of treatment effect.
Biometrics, 1231-1236.
Dye, R. A. (1985). Disclosure of nonproprietary information. Journal of accounting research, 123-145.
32
Frank, K. A. (2000). Impact of a confounding variable on a regression coefficient. Sociological
Methods & Research, 29(2), 147-194.
Gleason, C. A., Jenkins, N. T., & Johnson, W. B. (2008). The contagion effects of accounting
restatements. The Accounting Review, 83(1), 83-110.
Hainmueller, J. (2012). Entropy balancing for causal effects: A multivariate reweighting method
to produce balanced samples in observational studies. Political Analysis, 20(1), 25-46.
Healy, P. M. & Palepu, K.G. (2001). A review of the empirical disclosure literature. Journal of
Accounting and Economics, 31, 405–440.
Hirshleifer, D., & Teoh, S. H. (2003). Limited attention, information disclosure, and financial
reporting. Journal of accounting and economics, 36(1), 337-386.
Jehn, K. A., & Shah, P. P. (1997). Interpersonal relationships and task performance: An
examination of mediation processes in friendship and acquaintance groups. Journal of
Personality and Social Psychology, 72(4), 775.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs
and ownership structure. Journal of financial economics, 3(4), 305-360.
Kirk, M. P., & Vincent, J. D. (2014). Professional investor relations within the firm. The Accounting
Review, 89(4), 1421-1452.
Larcker, D. F., & Rusticus, T. O. (2010). On the use of instrumental variables in accounting
research. Journal of Accounting and Economics, 49(3), 186-205.
Lee, L. F., Hutton, A. P., & Shu, S. (2015). The role of social media in the capital market: evidence
from consumer product recalls. Journal of Accounting Research, 53(2), 367-404.
Leuz, C., & Verrecchia, R. E. (2000). The economic consequences of increased disclosure (digest
summary). Journal of accounting research, 38, 91-124.
McMullin, J. L., & Schonberger, B. (2015). Entropy-balanced discretionary accruals. Available at
SSRN.
McNichols, M. (1984). The anticipation of earnings in securities markets. University of California,
Los Angeles. Dissertation.
Mellon, BNY. (2012) A Survey Analysis of IR Practices Worldwide – Seventh Edition. Global
Trends In Investor Relations. BNY Mellon.
Mellon, BNY. (2013) A Survey Analysis of IR Practices Worldwide – Eighth Edition. Global
Trends In Investor Relations. BNY Mellon.
33
Merton, R. C. (1987). A simple model of capital market equilibrium with incomplete information.
The journal of finance, 42(3), 483-510.
NIRI (2016). Definition of Investor Relations. The National Investor Relations Institute. Available
at: https://www.niri.org/about-niri
Pástor, L, & Pietro, V. (2003). Stock valuation and learning about profitability. The Journal of
Finance, 58(5), 1749-1790.
Rajan, R. G. & Zingales, L. (1998). Financial Dependence and Growth. The American Economic
Review, 88 (3), 559-586.
SEC (2000). Final Rule: Selective Disclosure and Insider Trading. Securities and Exchange
Comission. Available at: https://www.sec.gov/rules/final/33-7881.htm
Shah, P. P., & Jehn, K. A. (1993). Do friends perform better than acquaintances? The interaction of
friendship, conflict, and task. Group decision and negotiation, 2(2), 149-165.
Shroff, N., Sun, A. X., White, H. D., & Zhang, W. (2013). Voluntary disclosure and information
asymmetry: Evidence from the 2005 securities offering reform. Journal of Accounting Research,
51(5), 1299-1345.
Simon, H. A. (1978). Rationality as process and as product of thought. The American economic
review, 68(2), 1-16.
Solomon, D. H. (2012). Selective publicity and stock prices. The Journal of Finance, 67(2), 599-638.
Soltes, E. (2014). Private Interaction Between Firm Management and Sell‐Side Analysts. Journal of
Accounting Research, 52(1), 245-272.
Steiner, I. D. (1972). Group process and productivity. New York: Academic Press.
Thomson (2009). Thomson Reuters 2009 Investor Relations Practices Survey.
Twedt, B. (2016). Spreading the word: Price discovery and newswire dissemination of
management earnings guidance. The Accounting Review, 91(1), 317-346.
34
Table 1: Sample Descriptive Statistics
Panel A: Sample Construction
N Used In
IR firm/quarter observations excluding obs with IR tenure < 1 year 28,744
Plus: Control firms +116,078
Primary sample used in empirical analysis 144,822 Tables 3, 5
Subsample for which ForecastAccuracy can be calculated 95,476 Table 5
Subsample for which ForecastDispersion can be calculated 71,703 Table 5
Primary sample excluding obs with IR tenure < 3 years 132,748 Tables 3, 5
Subsample for which ForecastAccuracy can be calculated 83,853 Table 5
Subsample for which ForecastDispersion can be calculated 61,138 Table 5
Firm/quarters in the same SIC 4-digit industry as the firm
announcing a major restatement 50,253 Table 7
Subsample excluding obs with IR tenure < 3 years 43,952 Table 7
Firms transitioning from long to short-tenured IR officer 1,136 Table 8
This panel describes how the primary sample was constructed and the composition of sub-samples used in each table.
Panel B: Observations by Industry
Industry N
Business Equipment 23,180
Chemicals 3,151
Consumer Durables 3,007
Consumer Non-Durables 6,740
Energy 6,098
Finance 40,516
Healthcare 13,984
Manufacturing 12,536
Other 16,810
Telecom 3,554
Utilities 3,704
Wholesale/retail 11,542
Total 144,822
This panel describes the number of observations in the primary sample by Fama-French 12-industry categorization.
35
Panel C: Observations by Year
Fiscal Year N
2002 12,811
2003 13,077
2004 13,671
2005 13,691
2006 13,226
2007 12,640
2008 11,994
2009 11,424
2010 10,956
2011 10,566
2012 10,373
2013 10,393
Total 144,822
This panel describes the number of observations in the primary sample by fiscal year.
Panel D: Percentage of Conference Calls with IROs by year and firm size
Firm Size 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Mega-cap (over $25 billion) 72% 74% 78% 81% 77% 78% 78% 79% 81% 81% 79%
Large-cap ($5 billion - $25 billion) 59% 60% 65% 68% 70% 71% 73% 75% 75% 75% 74%
Mid-cap ($1 billion - $4.9 billion) 44% 48% 51% 51% 53% 57% 62% 59% 58% 61% 59%
Small-cap ($150 million - $999 million) 29% 35% 35% 39% 39% 44% 46% 45% 46% 49% 50%
Micro-cap (under $150 million) 22% 23% 27% 30% 35% 36% 39% 38% 38% 42% 45%
This panel describes the percentage of quarterly conference calls with an IR officer listed as a participant as a
percentage of the total number of conference calls available on the FD Newswire service by calendar year and firm
size. Year 2002 is excluded because very few conference call transcripts are available in calendar year 2002 on the
FD Newswire service.
36
Table 2: Variable Summary Statistics
Variable Mean 25% Median 75% n
ReturnVolatility 2.95 1.52 2.30 3.63 144,822
IRFirm 0.19 - - - 144,822
Lnatq 6.15 4.63 6.16 7.55 144,822
Growth 4.38 2.00 4.00 7.00 144,822
ScaledEarnings (0.01) (0.00) 0.01 0.02 144,822
UnexpectedEarnings (0.00) (0.01) 0.00 0.01 144,822
M&A 0.01 - - - 144,822
EarningsVolatility 0.03 0.00 0.01 0.03 144,822
Prior12MonthReturn 0.05 (0.23) (0.02) 0.23 144,822
Leverage 4.12 1.45 2.19 4.63 144,822
MB 2.54 1.05 1.69 2.87 144,822
Forecasts 0.13 - - - 144,822
MediaMentions 3.75 - - - 144,822
NumPressReleases 1.04 - 1.10 1.79 144,822
NumAnalysts 1.09 - 1.10 1.95 144,822
Inst.Ownership 0.38 0.01 0.29 0.70 144,822
ForecastAccuracy (2.13) (0.71) (0.22) (0.07) 95,476
ForecastDispersion 0.57 0.05 0.12 0.31 71,703
IPT[0,10] 4.89 3.00 5.00 7.00 144,822
This table provides summary statistics. ReturnVolatility is the standard deviation of daily abnormal returns
during the quarter multiplied by 100. IRFirm is an indicator variable set equal to one for firm/quarters in
which we observe an IR officer. Lnatq is the log of total assets measured at the end of the quarter. Growth
is the decile ranking of change in revenue from the same quarter in the previous year. ScaledEarnings is
earnings before unusual items scaled by the market value of equity for the firm at the end of the quarter.
UnexpectedEarnings is earnings before unusual items less earnings before unusual items in the same
quarter of the prior year scaled by the market value of equity for the firm at the end of the quarter. M&A
is a dummy variable set equal to one if the firm reported a nonzero amount for acquisitions on its statement
of cash flows. EarningsVolatility is the standard deviation of the firm’s quarterly earnings scaled by the
market value of equity over the prior eight quarters. Prior12MonthReturn is the cumulative abnormal
return (benchmarked by the value-weighted market return) over the prior year measured at the end of the
quarter. Leverage is assets divided by book equity at the end of the quarter. MB is the market value of
equity divided by book equity at the end of the quarter. Forecasts is an indicator variable set equal to one
if the firm provided an earnings forecast during the quarter. MediaMentions is the number of unique news
stories published by the firm during the quarter as captured by the Ravenpack database.
NumPressReleases is the log of one plus the number of firm-initiated press releases as captured by the
Ravenpack database. NumAnalysts is the log of one plus the number of analysts covering the firm.
InstOwnership is the percentage of shares owned by institutional investors. ForecastAccuracy is the
negative absolute value of the mean analyst forecast error, deflated by stock price for the latest quarterly
EPS forecast made by each analyst before the earnings announcement date. ForecastDispersion is the
standard deviation of analyst forecasts made before the earnings announcement date. IPT [0,10] is the
decile-ranked 11-day intraperiod timeliness measure of the speed with which information disclosed in
earnings announcements is impounded into stock prices.
37
Table 3: Pre and Post-Weighting Sample Distributions
Panel A: Original Sample Mean Variance
IR Non-IR Diff IR Non-IR Diff
Lnatq 8.28 5.64 2.64*** 3.18 3.59 -0.41***
ScaledEarnings 0.00 -0.02 0.02*** 0.00 0.01 -0.01***
EarningsVolatility 0.02 0.03 -0.01*** 0.00 0.00 0.00***
Leverage 3.64 4.23 -0.60*** 23.74 29.60 -5.85***
MB 2.89 2.45 0.45*** 13.77 15.25 -1.49***
Observations 28,199 116,623
Panel B: Post PSM Mean Variance
IR Non-IR Diff IR Non-IR Diff
Lnatq 7.49 7.45 0.04** 2.51 2.63 -0.12***
ScaledEarnings 0.00 0.00 0.00 0.01 0.01 0.00***
EarningsVolatility 0.02 0.02 0.00*** 0.00 0.00 0.00
Leverage 3.18 3.61 -0.43*** 19.12 25.93 -6.81***
MB 2.88 2.53 0.35*** 15.08 11.99 3.09***
Observations 17,513 17,513
Panel C: Post Entropy Balancing Mean Variance
IR Non-IR Diff IR Non-IR Diff
Lnatq 8.28 8.28 0.00 3.19 3.18 0.01
ScaledEarnings 0.00 0.00 0.00 0.00 0.00 0.00
EarningsVolatility 0.02 0.02 0.00 0.00 0.00 0.00
Leverage 3.64 3.64 0.00 23.74 23.74 0.00
MB 2.89 2.90 -0.01 13.79 13.77 0.02
Observations 28,199 116,623
This table presents distributional properties (mean and variance) for the IR and non-IR observations across the original
sample (Panel A), the propensity score matched sample (Panel B) and the entropy balanced sample (Panel C).
Differences between the sample means and variances are presented in the columns labeled “Diff”. *, **, *** represent
significance at 10%, 5%, and 1%, respectively. Significance levels are calculated using a t-test for the difference in
means and an F-test for the difference in variances.
38
Table 4: Effects of IR on Return Volatility
Dependent Variable: ReturnVolatility
(Excludes IR tenure < 1yr) (Excludes IR tenure < 3yrs)
(1) (2) (3) (4)
Predicted
Sign Firm FE Entropy Balanced Firm FE Entropy Balanced
IRFirm - -0.155*** -0.075*** -0.189*** -0.121***
(-3.11) (-5.82) (-3.47) (-7.89)
Lnatq -0.424*** -0.179*** -0.436*** -0.167***
(-17.86) (-11.20) (-17.11) (-10.03)
Growth -0.012*** 0.011** -0.013*** 0.010**
(-5.14) (1.99) (-5.24) (1.79)
ScaledEarnings -5.557*** -8.243*** -5.520*** -8.479***
(-45.99) (-12.69) (-44.42) (-12.80)
UnexpectedEarnings 1.813*** 3.043*** 1.844*** 3.276***
(19.42) (4.13) (19.34) (4.29)
M&A 0.380*** 0.419*** 0.385*** 0.475***
(2.93) (2.74) (2.75) (2.72)
EarningsVolatility 3.105*** 6.657*** 3.057*** 7.066***
(11.56) (18.90) (10.86) (15.90)
Prior12MonthReturn 0.122*** -0.074** 0.137*** -0.094***
(8.49) (-2.24) (9.02) (-2.43)
Leverage 0.055*** 0.005 0.056*** -0.000
(13.55) (0.87) (13.12) (-0.04)
MB -0.056*** -0.026*** -0.057*** -0.025***
(-13.09) (-5.79) (-12.49) (-5.38)
Forecasts 0.000 -0.037*** -0.002 -0.027
(0.01) (-2.69) (-0.09) (-1.60)
MediaMentions -0.002*** 0.001 -0.002*** 0.001*
(-2.74) (0.93) (-2.59) (1.83)
NumPressReleases 0.171*** 0.100*** 0.167*** 0.101***
(16.94) (5.27) (15.64) (5.16)
NumAnalysts -0.045*** -0.076*** -0.048*** -0.085***
(-3.73) (-3.15) (-3.67) (-3.41)
Inst.Ownership -0.563*** -0.417*** -0.574*** -0.427***
(-14.18) (-14.16) (-13.25) (-12.82)
N 144,822 144,822 132,748 132,748
𝐴𝑑𝑗 𝑅2 0.624 0.453 0.621 0.458
P-val diff in IR tenure 0.034 0.000 Fixed Effects Firm, Year Industry, Year Firm, Year Industry, Year
This table models the effect of IR on stock price volatility as described in Equation (1). Observations in which IR tenure
is less than one (three) year(s) are excluded from Columns 1 & 2 (3 & 4). The dependent variable, ReturnVolatility, is the
standard deviation of daily abnormal returns during the quarter multiplied by 100. The variable of interest, IRFirm, is an
indicator variable set equal to one for firm/quarters in which we observe an IR officer. Columns 1 & 3 are estimated using
firm and year fixed effects with standard errors clustered by firm. Columns 2 & 4 are estimated using entropy balancing
with year and industry fixed effects. The control variables are defined in Table 2. Continuous variables are winsorized at
the 1% and 99% levels. *, **, *** represent significance at 10%, 5%, and 1%, respectively (one-tailed for predicted
coefficients and two-tailed for non-predicted coefficients).
39
Table 5: Effects of IR on Analyst Forecasts
Panel A: Forecast Dispersion
Dependent Variable: ForecastDispersion
(Excludes IR tenure < 1yr) (Excludes IR tenure < 3yrs)
(1) (2) (3) (4)
Predicted
Sign Firm FE Entropy Balanced Firm FE Entropy Balanced
IRFirm - -0.094** -0.041** -0.173*** -0.082***
(-1.75) (-2.31) (-2.48) (-3.98)
Lnatq -0.040 -0.011 -0.035 0.002
(-1.14) (-1.20) (-0.90) (0.21)
Growth -0.014*** -0.011*** -0.014*** -0.008
(-3.95) (-2.62) (-3.46) (-1.61)
ScaledEarnings -8.587*** -11.732*** -8.240*** -11.845***
(-18.47) (-22.34) (-17.17) (-19.21)
UnexpectedEarnings 2.457*** 3.719*** 2.210*** 3.338***
(7.82) (8.81) (6.69) (6.95)
M&A 0.022 -0.181 0.021 -0.077
(0.16) (-0.91) (0.13) (-0.31)
EarningsVolatility 1.041*** 2.835*** 0.968** 2.879***
(2.63) (6.78) (2.32) (6.11)
Prior12MonthReturn -0.278*** -0.349*** -0.270*** -0.344***
(-14.22) (-13.41) (-13.04) (-10.89)
Leverage 0.047*** 0.022*** 0.052*** 0.020***
(5.44) (4.52) (5.60) (3.32)
MB -0.047*** -0.024*** -0.053*** -0.025***
(-6.48) (-7.68) (-6.52) (-7.14)
Forecasts 0.026 -0.060*** 0.029 -0.050***
(1.63) (-5.97) (1.55) (-4.24)
MediaMentions 0.001 0.003*** 0.002* 0.003***
(1.12) (3.76) (1.66) (3.96)
NumPressReleases 0.048*** 0.077*** 0.040*** 0.057***
(4.39) (6.75) (3.48) (4.57)
NumAnalysts -0.103*** -0.125*** -0.094*** -0.120***
(-3.00) (-6.08) (-2.91) (-5.15)
Inst.Ownership -0.676*** -0.630*** -0.676*** -0.495***
(-8.34) (-15.03) (-7.29) (-9.99)
N 71,703 71,703 61,138 61,138
𝐴𝑑𝑗 𝑅2 0.625 0.295 0.642 0.303
P-val diff in IR tenure 0.456 0.002 Fixed Effects Firm, Year Industry, Year Firm, Year Industry, Year
This panel models the effect of IR on analyst forecast dispersion as described in Equation (2). Observations in which IR
tenure is less than one (three) year(s) are excluded from Columns 1 & 2 (3 & 4). The dependent variable,
ForecastDispersion is the standard deviation of analyst forecasts made before the earnings announcement date. The
variable of interest, IRFirm, is an indicator variable set equal to one for firm/quarters in which we observe an IR officer.
Columns 1 & 3 are estimated using firm and year fixed effects with standard errors clustered by firm. Columns 2 & 4 are
estimated using entropy balancing with year and industry fixed effects. The control variables are defined in Table 2.
Continuous variables are winsorized at the 1% and 99% levels. *, **, *** represent significance at 10%, 5%, and 1%,
respectively (one-tailed for predicted coefficients and two-tailed for non-predicted coefficients).
40
Panel B: Forecast Accuracy
Dependent Variable: ForecastAccuracy
(Excludes IR tenure < 1yr) (Excludes IR tenure < 3yrs)
(1) (2) (3) (4)
Predicted
Sign Firm FE Entropy Balanced Firm FE Entropy Balanced
IRFirm + 0.353* 0.200** 0.613** 0.267***
(1.49) (2.24) (2.31) (2.33)
Lnatq 0.202 -0.041 0.249 -0.070
(1.34) (-0.68) (1.49) (-0.99)
Growth 0.074*** -0.005 0.070*** -0.027
(4.70) (-0.22) (4.12) (-0.91)
ScaledEarnings 29.006*** 34.207*** 28.885*** 36.202***
(17.54) (13.92) (16.58) (12.60)
UnexpectedEarnings -9.363*** -15.562*** -9.077*** -15.933***
(-7.81) (-6.01) (-7.17) (-5.22)
M&A 0.874 1.689** 0.826 1.545
(1.50) (1.69) (1.27) (1.23)
EarningsVolatility -0.617 -10.177*** -0.134 -10.727***
(-0.33) (-4.76) (-0.07) (-4.00)
Prior12MonthReturn 0.979*** 1.236*** 0.995*** 1.360***
(13.20) (6.38) (12.04) (5.32)
Leverage -0.191*** -0.024 -0.222*** -0.020
(-6.35) (-0.77) (-6.91) (-0.51)
MB 0.195*** 0.086*** 0.226*** 0.091***
(7.03) (4.89) (7.23) (4.11)
Forecasts -0.063 0.153*** -0.046 0.143***
(-0.98) (3.53) (-0.60) (2.68)
MediaMentions 0.000 -0.002 -0.004 -0.003
(0.07) (-0.55) (-0.96) (-0.62)
NumPressReleases -0.023 -0.130 0.023 -0.072
(-0.39) (-1.35) (0.37) (-0.64)
NumAnalysts 0.323*** 0.476*** 0.322*** 0.461***
(4.39) (3.60) (3.99) (2.92)
Inst.Ownership 1.191*** 1.784*** 1.091*** 1.516***
(4.19) (6.55) (3.33) (4.32)
N 95,476 95,476 83,853 83,853
𝐴𝑑𝑗 𝑅2 0.566 0.168 0.579 0.174
P-val diff in IR tenure 0.034 0.140 Fixed Effects Firm, Year Industry, Year Firm, Year Industry, Year
This panel models the effect of IR on analyst forecast accuracy as described in Equation (2). Observations in which IR
tenure is less than one (three) year(s) are excluded from Columns 1 & 2 (3 & 4). The dependent variable, ForecastAccuracy
is calculated as the negative absolute value of the mean analyst forecast error, deflated by stock price for the latest quarterly
EPS forecast made by each analyst before the earnings announcement date. The variable of interest, IRFirm, is an indicator
variable set equal to one for firm/quarters in which we observe an IR officer. Columns 1 & 3 are estimated using firm and
year fixed effects with standard errors clustered by firm. Columns 2 & 4 are estimated using entropy balancing with year
and industry fixed effects. The control variables are defined in Table 2. Continuous variables are winsorized at the 1% and
99% levels. *, **, *** represent significance at 10%, 5%, and 1%, respectively (one-tailed for predicted coefficients and
two-tailed for non-predicted coefficients).
41
Table 6: Effects of IR on Price Discovery
Dependent Variable: IPT [0,10]
(Excludes IR tenure < 1yr) (Excludes IR tenure < 3yrs)
(1) (2) (3) (4)
Predicted
Sign Firm FE Entropy Balanced Firm FE Entropy Balanced
IRFirm + 0.136** 0.095*** 0.149** 0.085**
(1.96) (2.64) (1.88) (1.89)
Lnatq 0.100*** 0.012 0.103*** 0.008
(4.31) (0.93) (4.21) (0.54)
Growth 0.008** 0.006 0.008** 0.008
(2.32) (1.00) (2.24) (1.04)
ScaledEarnings 0.199** 0.401 0.173** 0.316
(2.00) (1.59) (1.70) (1.12)
UnexpectedEarnings -0.087 0.018 -0.089 0.018
(-0.92) (0.07) (-0.93) (0.06)
M&A 0.301 0.548** 0.215 0.377
(1.45) (1.68) (0.98) (1.00)
EarningsVolatility 0.196 -0.351 0.146 -0.629
(0.82) (-0.71) (0.59) (-1.05)
Prior12MonthReturn 0.026 -0.045 0.035** -0.032
(1.60) (-1.27) (2.04) (-0.77)
Leverage 0.000 -0.008* 0.000 -0.008
(0.07) (-1.76) (0.09) (-1.46)
MB -0.003 0.008** -0.004 0.006
(-0.92) (1.67) (-1.20) (0.92)
Forecasts -0.049 0.018 -0.033 0.038
(-1.59) (0.48) (-0.95) (0.83)
MediaMentions -0.002 0.000 -0.000 0.001
(-0.95) (0.32) (-0.04) (0.41)
NumPressReleases 0.004 -0.033 0.004 -0.043**
(0.27) (-1.54) (0.25) (-1.76)
NumAnalysts 0.069*** 0.084*** 0.073*** 0.086***
(3.86) (4.01) (3.85) (3.62)
Inst.Ownership 0.064 0.164*** 0.056 0.191***
(1.17) (3.42) (0.95) (3.36)
N 144,822 144,822 132,748 132,748
𝐴𝑑𝑗 𝑅2 0.061 0.014 0.063 0.013
P-val diff in IR tenure 0.034 0.346 Fixed Effects Firm, Year Industry, Year Firm, Year Industry, Year
This table models the effect of IR on IPT around earnings announcements as described in Equation (4).
Observations in which IR tenure is less than one (three) year(s) are excluded from Columns 1 & 2 (3 & 4). The
dependent variable, IPT[0,10], is the decile-ranked 11-day intraperiod timeliness measure of the speed with which
information disclosed in earnings announcements is impounded into stock prices. Columns 1 & 3 are estimated
using firm and year fixed effects with standard errors clustered by firm. Columns 2 & 4 are estimated using entropy
balancing with year and industry fixed effects. The control variables are defined in Table 2. Continuous variables
are winsorized at the 1% and 99% levels. *, **, *** represent significance at 10%, 5%, and 1%, respectively (one-
tailed for predicted coefficients and two-tailed for non-predicted coefficients).
42
Table 7: The Effect of IR on Returns Following a Major Accounting
Restatement in the Industry
Panel A Cumret [-1, 1] Cumret [2, 5] Cumret [2, 15] Cumret [2, 30]
All Peer firms -0.32% -0.07% 0.03% -0.04%
IR Peer firms -0.24% 0.03% 0.44% 0.79%
Non-IR Peer firms -0.35% -0.10% -0.09% -0.29%
P-value of diff. between IR and Non-IR 0.024 0.0137 0.0001 0.0001
Panel B
Dependent Variable: Cumret [2, 30]
(Excludes IR tenure < 1yr) (Excludes IR tenure < 3yrs)
(1) (2) (3) (4)
Predicted
Sign Firm FE Entropy Balanced Firm FE Entropy Balanced
IRFirm + 0.009** 0.010*** 0.017** 0.012***
(1.76) (4.78) (2.05) (4.21)
N 50,253 50,253 43,952 43,952
𝐴𝑑𝑗 𝑅2 0.167 0.025 0.178 0.030
P-val diff in IR tenure 0.051 0.142 Fixed Effects Firm, Year Industry, Year Firm, Year Industry, Year Controls Included Yes Yes Yes Yes
This table models the effect of IR on stock returns around a major accounting restatement in the industry as described
in Equation (5). Panel A describes univariate cumulative abnormal returns for all, IR and non-IR peer firms. Panel B
provides multivariate results. Observations in which IR tenure is less than one (three) year(s) are excluded in Columns
1 & 2 (3 & 4) in Panel B. Cumret [-1, 1], [2, 5], [2,15] and [2, 30] are the cumulative abnormal returns over days -1 to
1, 2 to 5, 2 to 15 and 2 to 30, respectively relative to the date that the major accounting restatement was announced by
another firm in the same 4-digit SIC industry. Columns 1 & 3 of Panel B are estimated using firm and year fixed effects
with standard errors clustered by firm. Columns 2 & 4 of Panel B are estimated using entropy balancing with year and
industry fixed effects. The control variables are Lnatq, Growth, ScaledEarnings, UnexpectedEarnings, M&A,
EarningsVolatility, Prior12MonthReturn, Leverage, MB, Forecasts, MediaMentions, NumPressReleases, NumAnalysts,
Inst.Ownership, which are defined in Table 2. Continuous variables are winsorized at the 1% and 99% levels. *, **, ***
represent significance at 10%, 5%, and 1%, respectively (one-tailed for predicted coefficients and two-tailed for non-
predicted coefficients).
43
Table 8: Effects of Losing Long-Tenured IR Officer
Dependent Variables
Return Volatility Forecast
Accuracy
Forecast
Dispersion IPT[0,10]
Predicted Sign (1) (2) (3) (4)
Post + / - / + / - 0.066* -0.041** 0.102** -0.088*
(1.36) (-1.68) (1.89) (-1.50)
N 1136 1136 1136 1136
𝐴𝑑𝑗 𝑅2 0.075 0.586 0.056 0.036
Fixed Effects Year, Firm Year, Firm Year, Firm Year, Firm
Controls Included Yes Yes Yes Yes
This table models the effect of losing a long-tenured IR officer on stock return volatility, properties of analyst forecasts
and price formation as described in Equation (6). ReturnVolatility, ForecastAccuracy. ForecastDispersion, and
IPT[0,10] are the dependent variables in Columns (1), (2), (3) and (4), respectively. Post is an indicator variable set
equal to one for the eight quarters following the first quarter in which the new IR officer is appointed. The control
variables are Lnatq, ScaledEarnings, UnexpectedEarnings, M&A, EarningsVolatility, Prior12MonthReturn,
Leverage, MB, Forecasts, MediaMentions, NumPressReleases, NumAnalysts, Inst.Ownership, which are defined in
Table 2. Equation (6) is estimated with firm and year fixed effects and standard errors clustered by firm. Continuous
variables are winsorized at the 1% and 99% levels. *, **, *** represent significance at 10%, 5%, and 1%, respectively
(one-tailed for predicted coefficients and two-tailed for non-predicted coefficients).
44
Table 9: Impact Threshold of Confounding Variable
Impact on Coefficient for IRFirm
Return Volatility Forecast
Accuracy
Forecast
Dispersion IPT[0,10]
(1) (2) (3) (4)
Lnatq 0.07167 0.00614 0.00485 0.00184
Growth 0.00038 0.00093 0.00199 0.00035
ScaledEarnings 0.00241 0.00236 0.00229 0.00005
UnexpectedEarnings 0.00218 0.00184 0.00222 0.00006
M&A 0.00003 0.00015 0.00008 0.00004
EarningsVolatility 0.00273 0.00080 0.00035 0.00014
Prior12MonthReturn 0.00030 0.00087 0.00138 0.00007
Leverage 0.00273 0.00074 0.00102 0.00006
MB 0.00219 0.00013 0.00162 0.00003
Forecasts 0.00006 0.00011 0.00017 0.00009
MediaMentions 0.02557 0.01013 0.01509 0.00054
NumPressReleases 0.00485 0.00074 0.00729 0.00031
NumAnalysts 0.00015 0.00071 0.00084 0.00229
Inst.Ownership 0.00246 0.00168 0.00030 0.00020
Largest impact 0.07167 0.01013 0.01509 0.00229
Impact threshold of confounding variable 0.104 0.029 0.05 0.051
Minimum magnitude of confounding
variable relative to largest impact included
variable required to overturn IRFirm 1.5 2.9 3.3 22.2
This table displays the results of an analysis of the impact threshold of confounding variable (ITCV) for each of the
main results, consistent with Larcker and Rusticus (2010) and Frank (2000). Each column represents the impact on
the IRFirm of each additional independent variable, which is calculated as the partial correlation with that variable
and the dependent variable times the partial correlation of that variable with independent variable of interest, IRFrim.
These partial correlations are not presented for parsimony. The row labeled “Largest impact” simply identifies the
most impactful control variable included in the model. The row labeled “Impact threshold of the confounding variable”
is calculated following Frank (2000). The last row describes the minimum magnitude of the confounding variable
relative to largest impact included variable required to overturn IRFirm, which is calculated as the impact threshold
of the confounding variable divided by the largest impact included variable. For example, Column (4) suggests that a
confounding (omitted) variable would have to be 22.2 times larger than the most impactful included variable
(NumAnalysts) to overturn the observed relationship between IRFirm and IPT [0,10].