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PRELIMINARY – Please do not cite or quote without permission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].