The Impact of Management Earnings Forecasts
on Firm Risk and Firm Value
Stephen R. Foerster
Professor of Finance
Richard Ivey School of Business
University of Western Ontario
1151 Richmond Street North
London, Canada N6A 3K7
Tel: 519 661 3726; Fax 519 661 3431
Email: sfoerster@ivey.ca
Stephen G. Sapp
Associate Professor of Finance
Richard Ivey School of Business
University of Western Ontario
1151 Richmond Street North
London, Canada N6A 3K7
Tel: 519 661 3006; Fax 519 661 3485
Email: ssapp@ivey.ca
Yaqi Shi*
Assistant Professor of Accounting
Richard Ivey School of Business
University of Western Ontario
1151 Richmond Street North
London, Canada N6A 3K7
Tel: 519 661 4097; Fax 519 661 3485
Email: yshi@ivey.ca
Current Version: March 15, 2010
*
Yaqi Shi is the corresponding author. We wish to thank Sean Davis, Bruce Johnson, Si Li, Larry Menor, and
participants at the Northern Finance Association (2008) Annual Conference, Financial Management Association
European (2009) Conference and Financial Accounting and Reporting Section (FARS) 2010 Mid-Year Meeting for
helpful comments. Yaqi Shi acknowledges financial support from Richard Ivey School of Business, University of
Western Ontario. All errors remain our own responsibilities.
The Impact of Management Earnings Forecasts
on Firm Risk and Firm Value
Abstract
This study investigates whether voluntary disclosure of management earnings forecasts influences investors’
assessment of both firm risk and firm value. We control for possible endogeneity between various firmspecific characteristics and the voluntary issuing of different types of management earnings forecasts by
utilizing a two-stage Heckman treatment analysis. We find a significant negative relationship between the
issuance of management earnings forecasts and a variety of measures of firm risk (idiosyncratic risk, stock
return volatility, beta, and bid-ask spreads) suggesting that information quality is an important determinant
of both diversifiable risk and nondiversifiable systematic risk. We also demonstrate that management
earnings forecasts are positively associated with firm value as captured by Tobin’s Q when firms have
established their forecasting reputation by consistently releasing more frequent, precise and accurate
forecasts. Overall, our results suggest that releasing high-quality management earnings forecasts is
associated with significant capital market benefits.
JEL Codes: G12, G14, G32, G34, M41
Keywords: management earnings forecasts, voluntary disclosure, firm risk, firm value, information quality
1
1. Introduction
Capital market participants rely on a steady stream of information to assess risk and judge future
prospects in order to accurately value a firm’s equity. Firm management assists in this process by providing
information through a variety of channels, such as regulatory filings as well as voluntary communication
with outside investors and analysts (Healy and Palepu, 2001). Forecasts related to the firm’s anticipated
earnings per share (EPS) are one of the primary voluntary disclosure mechanisms through which managers
can provide additional information and signals to outside stakeholders about the expected future
performance of their firm. In fact, Anilowski, Feng and Skinner (2007) document that the number of firms
releasing voluntary earnings guidance has increased from 10-15 percent in the 1990s to approximately 30
percent in 2004. In this study, we evaluate the aggregate economic implications of management earnings
forecasts for U.S. firms. Specifically, we investigate how different characteristics of management earnings
forecasts influence investors’ assessment of both firm risk and firm value.
This study extends recent research which has demonstrated that better governance and better
disclosure practices are associated with higher firm valuation (Durnev and Kim, 2005, find this for firms
across 27 countries), with lower cost of capital (Botoson, 1997; Sengupta, 1998; Francis et al., 2004;
Lambert, Leuz and Verrechia, 2007; and Francis, Nanda and Olsson, 2008) and with a decrease in both
idiosyncratic risk as well as the level of private information (Brown and Hillegeist, 2007; and Ferreire and
Laux, 2007). Prior literature investigating the consequences of voluntary management earnings forecasts
largely focuses on short-term results such as stock market reactions immediately following the issuing of the
forecasts. For instance, using short-term event windows, previous work finds that management earnings
forecasts influence firm value (e.g., Pownall, Wasley and Waymire, 1993) and lead to a reduction in bid-ask
spreads (Coller and John, 1997). Starting to move towards a longer-term perspective, Ng, Tuna and Verdi
(2008) explore how forecasting credibility moderates the under-reaction to management earnings forecast
news using 3-month and 12-month hedge portfolio returns.
2
In another recent study, Rogers, Skinner and Buskirk (2009) go beyond looking at the stock price
reaction to earnings forecasts. They examine a sample of firms that voluntarily disclose earnings forecasts
and also have exchange-traded equity options. Using an event study approach, they compare implied
volatilities from these options before and after the release of earnings forecasts as well as after the release of
earnings announcements. These implied volatility measures are meant to capture potential changes in firm
risk as captured by investor uncertainty about firm value. They find that the forecasts themselves are
associated with increased short-run volatility when bad news is released, particularly if it is released on a
sporadic basis while there is a modest decline in volatility when news is good and especially when released
on a more regular basis.
Rogers et al. (2009) provide interesting results demonstrating how investor risk perception is
impacted by earnings forecasts. Our analysis extends their work by exploring the overall effects of
management earnings forecasts on both firm risk and firm value, particularly beyond short-term market
reactions and treating management earnings forecasts as isolated events. While short-term market reactions
can provide important information, we are interested in understanding the aggregate impact of management
issuing earnings forecasts at the firm level. For example, firm valuation measures such as Tobin’s Q may be
better able to capture the expected changes in future cash flows and thus reflect the aggregate impact of
actions by management (Lang and Stulz, 1994; Daske, Hail, Leuz and Verdi, 2008). The survey of senior
managers performed by Graham, Harvey and Rajgopal (2005) found that these managers believe disclosing
reliable and precise information can reduce information risk about a company’s stock. Consequently, the
consistent disclosure of management earnings forecasts should have tangible effects on both the apparent
riskiness of a stock to investors as well as on firm value. In addition, Anilowski et al. (2007) posit that given
the pervasiveness of management earnings forecasts, aggregate forecast news provides information about the
expected future cash flows of firms in general. However, none of the prior studies directly examines the
overall impact of management earnings forecast strategies on both firm risk and firm value (e.g., no such
3
studies are reported by Hirst, Koonce, and Venkataraman, 2008).We address this void and extend the
existing literature by empirically testing the relationship between earnings forecasts and firm risk as well as
firm value.
Furthermore, managers face a broad array of choices regarding the release of forward-looking
earnings information (King, Pownall and Waymire, 1990; Hirst et al., 2008). The first stage of our analysis
focuses on the impact of a firm deciding to voluntarily issue earnings forecasts. By deciding to provide
guidance to outside stakeholders, management appears to be making a clear statement regarding its
commitment to provide investors with information beyond that required in regulatory filings, yet the
decision to simply disclose such forecast information may not be sufficient to impact the perception of firm
risk or firm value. It is also important to consider other characteristics of these forecasts. Having chosen to
issue an earnings forecast, the manager faces various choices regarding the characteristics of that forecast.
These choices include the precision of the forecast (e.g., point estimate, range, open-interval or qualitative),
the frequency (e.g., quarterly versus annual, or the number of forecasts), and the horizon (e.g., next quarter,
next year or further). In addition, because management earnings forecasts can be verified ex post, the
accuracy of the forecasts, i.e., how the forecasted earnings deviate from the actual realized earnings, is
another important forecasting characteristic (Hirst et al., 2008). Further, 92 percent of the executives
interviewed by Graham et al. (2005) claim that developing a reputation for transparent reporting is a key
driver for voluntary disclosure. As a result, we view forecasting reputation as another important
characteristic of management earnings forecasts. Our proxy for forecasting reputation is a composite score of
forecasting frequency, precision and accuracy. We, therefore, are able to measure the aggregate impact of
management earnings forecasts on firm risk and firm value.
Our tests focus on the impact on both firm risk and firm value of five of the most important
characteristics of management earnings forecasts: occurrence, frequency, precision, accuracy and reputation.
Specifically, using 52,206 firm-year observations for 12,610 firms from 1998 to 2007, we estimate mixed
4
models to examine the effects of the above-mentioned forecast characteristics on firm risk and firm value.
We adopt four proxies to measure firm risk: idiosyncratic risk, beta, total firm risk, and bid-ask spread. We
consider a broad cross-section of risk measures because the estimation risk literature suggests that a lack of
information will not only influence idiosyncratic risk, but also nondiversifiable systematic risk (Barry and
Brown, 1985; Lambert et al., 2007). Our proxy for firm value is Tobin’s Q, which is based on an investor’s
assessment of the present value of anticipated future cash flows (e.g., see Lang and Stulz, 1994), and is a
common valuation measure in the literature to capture long-term effects (e.g., see Doige, Karolyi and Stulz,
2004 and Litvak, 2008) rather than short-term market reactions to specific events.
Briefly, our results reveal the following. First, as predicted, our cross-sectional tests indicate that
firms disclosing management earnings forecasts exhibit lower idiosyncratic risk, systematic risk (i.e., beta),
and total firm risk compared to firms not releasing forecasts. In addition, other forecasting characteristics
also help explain investors’ perceptions about firm risk. Firms releasing more frequent, precise, or accurate
management earnings forecasts are also associated with reduced idiosyncratic risk, systematic risk and total
firm risk. Finally, with respect to the impact of forecasting reputation on firm risk, we find that when firms
have established a reputation by disclosing more frequent, precise and accurate forecasts, their forecasts are
associated with incremental benefit on firm risk. In particular, their disclosure practices are not only related
to lower idiosyncratic, systematic, and total firm risk, but also associated with higher market liquidity (i.e.,
lower bid-ask spreads). Collectively, our results on firm risk present consistent evidence that information
quality, as proxied by the multiple characteristics of management earnings forecasts, has a significant impact
on the market’s assessment of both the idiosyncratic risk and systematic risk of our sample firms.
We also investigate the effects of management earnings forecasts on firm value (Tobin’s Q). We first
examine the impact of forecasting occurrence on firm value. Surprisingly, our results show that the act of
releasing management earnings forecasts per se is not associated with higher Tobin’s Q. This suggests that
the occurrence of management earnings forecasts does not necessarily have a positive effect on investors’
5
perception about firms’ future cash flows. Consequently, we explore the influence of other forecasting
characteristics (forecasting frequency, precision, accuracy and reputation) on firm value. We find that more
frequent management earnings forecasts are associated with higher Tobin’s Q in our models after controlling
for idiosyncratic risk and total firm risk. Similar results are obtained with respect to the impact of forecasting
accuracy on firm value. However, we find consistent evidence that more precise management earnings
forecasts are related to higher Tobin’s Q in all of our tested models. Finally, when firms have built-up their
forecasting reputation by releasing more frequent, precise and accurate information, the associations between
voluntary disclosure and Tobin’s Q are clearly more positive and significant. In sum, our findings suggest
that firms enjoy marginal benefits when disclosing more frequent earnings forecasts, and the benefits
become more obvious when these forecasts are precise and accurate. Moreover, for ―stronger reputation‖
forecasters the aggregate impact of their forecasting strategies on firm value is clearly positive.
To mitigate the concern of potential causality and of the possible endogeneity between various firmspecific characteristics and the voluntary issuing of different types of management earnings forecasts, we
also perform a two-stage Heckman Treatment analysis and find that this procedure strengthens our findings.
Our work contributes to the literature on the effects of information quality on firm risk. While most
of the prior studies focus on how information quality influences information risk or idiosyncratic risk
(Botosan, 1997; Coller and Yohn, 1997; Brown and Hillegeist, 2007; and Ferreire and Laux, 2007), a new
stream of literature emerges to investigate the impact of information quality on both idiosyncratic risk and
systematic risk (Lambert et al., 2007; Ashbaugh-Skaife et al., 2009). More specifically, Lambert et al. (2007)
model the direct and indirect effects of information quality on the cost of capital and they show that the
quality of information not only influences market’s assessment of variance of a firm’s cash flows (i.e.,
idiosyncratic risk) but the assessed covariance with other firms’ cash flows (i.e., systematic risk). In a similar
fashion, recent studies on the aggregate impact of earnings news at the market level indicate that firm-level
earnings news may not diversify away when aggregated (e.g., Anilowski et al., 2007; Ball, Sadka and Sadka,
6
2009). Nonetheless, there is scant empirical evidence on how information quality influences firm systematic
risk even though this is a fundamental issue because it influences the interaction between firms and investors
with respect to capital allocation choices (Lambert et al., 2007). By examining the impact of management
earnings forecasts on various firm risk measures, our research adds a new perspective to this literature. More
importantly, our research design offers several advantages to gauge the impact of information quality. First,
relative to other measures of disclosure quality that are aggregate or subjective metrics (Welker, 1995;
Botosan, 1997), management earnings forecasts provide a unique setting – they are a less uncertain measure
of information quality as they can be validated when the actual realized earnings are released. Second, our
proxy for forecasting reputation directly measures the quality of information related to management earnings
forecasts. Our findings that management earnings forecasts impact both idiosyncratic risk and beta support
the recent theoretical work by Lambert et al. (2007) and suggest that information quality is an important
determinant of both idiosyncratic risk and nondiversifiable systematic risk which will ultimately be
impounded in the market’s assessment of firm value.
Our research also makes two contributions to the literature on management earnings forecasts. First,
we contribute to the literature on the economic consequences of management earnings forecasts. Different
from prior studies that focus on short-term stock market reactions using even-study approaches (e.g.,
Pownall et al., 1993 and Coller and John, 1997), we assess the overall impact of management earnings
forecasts on both firm risk and firm value. In addition, managers have great discretion and control over
forecast characteristics, yet only nascent understanding exists regarding the impact on markets of issuing
forecasts with different characteristics (Hirst et al., 2008). By focusing on five forecasting characteristics, we
advance the literature by adding evidence on the aggregate economic implications of these multidimensional characteristics. Our results suggest that different forecast characteristics have different impact
on firm risk and firm value. Managers may benefit from our research because they can better understand the
7
choices they make related to their earnings forecasts and how they may as a result garner desirable benefits
from market participants.
Second, we extend the literature on management earnings forecasts by examining how forecast
antecedents and characteristics interact to impact both firm risk and value.
In their review on the
management earnings forecast literature, Hirst et al. (2008) call for future research to explore the interactions
among forecast antecedents, characteristics and consequences, or between two (or more) forecast
antecedents or characteristics (p317, p334). Our paper heeds the call by Hirst et al. (2008) because our
reputation construct is the interaction among three forecast characteristics: frequency, precision and
accuracy. By forming a composite score for firm’s forecasting reputation, we can test the aggregate effects
of management earnings forecasts. In addition, by employing a Heckman two-step model, we explicitly link
the influence of antecedents to the characteristics (i.e., the first stage model) and then the impact of the
characteristics on investors’ perception of firm risk and firm value (i.e., the second stage model).
The paper is organized as follows.
Section 2 reviews the relevant literature and presents a
development of our hypotheses. Section 3 describes the data as well as our empirical models. Results are
presented in section 4 and section 5 presents a summary and conclusions.
2. Literature Review and Hypotheses Development
Lack of information about firms and their securities can affect firm risk and value through several
channels. Our work first builds on the estimation risk literature (Klein and Bawa, 1976; Miller, 1977; Barry
and Brown, 1985, 1986; Lambert et al., 2007). Barry and Brown (1985) develop a model of differential
information in which the amount of information available differs across securities. They show that investors
demand a risk premium for the lack of information, and this information risk is nondiversifiable, so it affects
the systematic risk of securities, as measured by betas or covariances. Along the same line, recent theoretical
work by Lambert et al. (2007) models the direct and indirect effects of accounting information quality on the
8
cost of equity in a multi-security Capital Asset Pricing Model (CAPM) setting and concludes that
information risk increases market participants’ estimation of the variance of a firm’s cash flow (i.e.,
idiosyncratic risk) and covariances with other firms’ cash flows (i.e., systematic risk). Empirical support for
this line of theoretical work also exists in the literature (Guay, 1999; Sorescu & Spanjol, 2008; AshbaughSkaife et al., 2009; Bartram, Brown and Conrad, 2009; Godfrey, Merrill and Hansen, 2009). Specifically,
Ashbaugh-Skaife et al. (2009) reveal that firms that disclose internal control deficiencies have significantly
higher idiosyncratic risk, systematic risk and cost of equity.
Our work also draws on prior studies that link disclosure to liquidity and the cost of capital (e.g.,
Diamond and Verrecchia, 1991; Leuz and Verrecchia, 2000; O’ Hara, 2003; Easley and O’Hara, 2004). This
line of research maintains that voluntary disclosure reduces the information asymmetry between uninformed
and informed investors, and thus increases a firm’s liquidity and decreases a firm’s cost of capital. O’Hara
(2003) asserts that liquidity can affect the risk of holding an asset, and Merton (1987) shows that in
equilibrium the value of a firm is always lower when there is incomplete information. Empirical work also
substantiates this stream of theory (e.g., Botosan, 1997; Francis et al., 2008). Therefore, it is reasonable to
grant that voluntary disclosure may reduce firm risk and enhance firm value through the mechanism that
increases liquidity. Managers choose to issue forward-looking earnings forecasts to reduce information
asymmetry between managers and analysts and investors (Ajinkya and Gift, 1984; Coller and Yohn, 1997;
Healy and Palepu, 2001). As argued above, lowering information asymmetry is viewed as desirable because
it is associated with lower estimation risk, higher liquidity and lower cost of capital. Ultimately, the
voluntary disclosure of management earnings forecasts can lower uncertainties of a firm’s future prospects
and thus enhance firm value.
We use multiple measures to capture the uncertainty or riskiness associated with a firm by using
characteristics of the firm’s share price, such as systematic risk, overall stock return volatility, idiosyncratic
stock volatility (as estimated using CAPM), and average bid-ask spreads. We use Tobin’s Q to proxy for
9
firm value. As discussed above, Tobin’s Q is a common valuation measure in the literature to capture longterm effects (e.g., see Doige et al., 2004 and Litvak, 2008) rather than short-term market reactions to specific
events. Taking all the above reasoning together, we advance our first hypothesis:
Hypothesis H1a: Firms that release voluntary earnings forecasts are associated with lower firm risk
(as captured by overall stock return volatility as well as idiosyncratic stock volatility, beta, and bidask spreads).
Hypothesis H1b: Firms that release voluntary earnings forecasts are associated with an enhanced firm
valuation (as captured by Tobin’s Q).
Other characteristics of management earnings forecasts may also impact firm risk and firm value. As
in the first hypothesis, disclosure occurrence is an important aspect of a firm’s overall disclosure strategy;
but so is the disclosure frequency. Prior research shows that there are large variations in how frequently
firms choose to release earnings forecasts. For instance, Rogers and Stocken (2005) find that for their
sample firms during the period 1997-2000, 63 percent of firms released only one forecast and only one
percent of firms provide earnings forecasts every year. The forecast frequency has also been found to
depend on environmental factors. For example, Waymire (1985) shows that firms disclosing more frequent
forecasts are associated with lower earnings volatility. Graham et al. (2005) posit that forecast frequency is
related to the firm’s probability of meeting or beating earnings benchmarks.
The above evidence suggests that the frequency of forecasts that a firm chooses may be associated
with the firm manager’s uncertainty about future cash flows (i.e., firm risks or other firm-specific
antecedents).
Furthermore, O’Hara (2003) shows that the frequency of new information arrival is a
dimension of the risk associated with holding a stock. Easley and O’Hara (2004) demonstrate that the
quantity of information affects asset prices. Economic theory also predicts that commitment to persistent
disclosure reduces the information risk of a firm, and thus reduces the cost of capital and enhances firm
10
valuation (Diamond and Verrecchia, 1991; Leuz and Verrecchia, 2000). Empirically, Botosan and Harris
(2000) find that managers normally signal their commitment to disclosure by providing disclosures more
frequently. Increasing disclosure frequency can improve both the content and the timeliness of the
information. Therefore, by providing earnings forecasts more frequently, management may supply more
pertinent information to the market and the information revealed is timelier for the investors’ decisions.1 All
these benefits should result in a decrease in the risk or uncertainty of future cash flows. This leads to our
second hypothesis:
Hypothesis H2a: Firms releasing more frequent earnings forecasts are associated with a greater
reduction in firm risk.
Hypothesis H2b: Firms releasing more frequent earnings forecasts are associated with a greater
enhancement in firm valuation.
Another important characteristic of management earnings forecasts is their precision: point forecasts,
specific ranges (i.e., of a closed-interval nature), open-interval ranges (i.e., minimums or maximums), and of
a qualitative nature (i.e., providing general impressions about the firm’s earnings prospects).2 Ajinkya,
Bhojraj and Sengupta (2005) indicate that 78 percent of their sample firms during the period 1997-2002
release point or specific range forecasts.
Empirical studies suggest that forecast precision captures
managers’ beliefs about future cash flow (King et al., 1990) and that, in essence, more precise forecasts are
viewed as indicating greater managerial certainty and ability relative to less-precise forecasts (Hughes and
Pae, 2004). In addition, management earnings forecasts with different precisions have different information
1
On the other hand, the recent literature on managerial myopia suggests that issuing more frequent quarterly forecasts may hinder managers’
inclination to think about long-term growth and instead focus on beating quarterly targets (See a summary in Graham et al., 2005). However, in
the period from 2000 to 2006, Chen, Matsumoto and Rajgopal (2006) only identify 72 firms that announced they would stop issuing quarterly
forecasts; in addition, they find that the driving force to stop quarterly forecasts is undesirable firm performance rather than pursuing long-term
goal. Given the evidence, we do not consider that managerial myopia provoked by more frequent earnings forecasts is the dominant mechanism.
2
Point estimates are those whereby a specific estimate is disclosed such as ―Earnings will be X this period.‖ Range estimates are closed-interval
forecasts of the form ―Earnings will be between X1 and X2 this period.‖ Open-interval estimates are lower or upper bound forecasts of earnings.
A minimum estimate is in the form ―Earnings will be greater than X1 this period‖ whereas a maximum estimate is disclosed such as ―Earnings
will be no more than X2 this period.‖ Qualitative estimates are general impressions in the form ―Earnings will be favorable this year compared
with last year.‖
11
content and rational investors may discount qualitative earnings projections (Pownall et al., 1993). Using
short-term event studies, prior studies provide evidence that management forecast precision affects the
beliefs of investors and financial analysts. For example, Ajinkya and Gift (1984) develop an Expectation
Adjustment Hypothesis and posit that managers will credibly label their forecasts as to precision. Baginski,
Conrad, and Hassell (1993) examine the effects of information precision on short-term equity pricing, and
they support a positive relation between forecast precision and the importance of forecasts on security prices.
Theoretical work suggests that a signal’s precision is important in belief development (Kim and
Verrecchia, 1991a, 1991b; and Morse, Stephan, and Stice, 1991). Specifically, Kim and Verrecchia (1991b)
examine a two-period rational expectations model whereby traders are assumed to be differentially informed
and vary in the precision of their private prior information.
They find that the price reaction to the
unexpected portion of a disclosure is an increasing function of its relative importance across the posterior
beliefs of traders. The relative importance is positively related to the precision of the announcement and
inversely related to the precision of preannouncement information. The study by Kim and Verrecchia
(1991b) implies that the price reaction to the public reaction is a positive function of the information’s
precision. Moreover, Kim and Verrecchia (1994) show that as the precision of public information increases,
liquidity increases. This suggests that a firm can garner benefit by providing adequate disclosure. Easley and
O’Hara (2004) also imply that firms can influence their cost of capital by selecting the precision of
information to investors. Building on this line of research, we expect that consistent disclosure of precise
forecasts will have a positive effect on firm risk and valuation. Thus, we present our third hypothesis:
Hypothesis H3a: Firms disclosing more precise earnings forecasts are associated with a greater
reduction in firm risk.
Hypothesis H3b: Firms disclosing more precise earnings forecasts are associated with a greater
enhancement in firm valuation.
12
The accuracy of the forecast is another characteristic that may influence the economic consequences
of the forecast. Managers may have incentives to issue self-serving forecasts and thus not all management
earnings forecasts are accurate as measured by how actual earnings deviate from forecasted earnings (Hutton
and Stocken, 2009; and Hirst et al., 2008). However, only accurate information will enhance the resource
allocation in capital markets and reduce uncertainties related to the future prospects of a firm (Healy and
Palepu, 2001). Prior studies document that forecast accuracy influences how analysts react to a forecast.
For example, Williams (1996) finds that analysts revise their forecasts more for firms with high prior
forecasting accuracy. In recent studies, Hutton and Stocken (2009) and Ng et al. (2008) show that when a
firm releases accurate information in previous periods, investors respond more to current management
earnings forecast news. In addition, CFOs surveyed by Graham et al. (2005) believe that more accurate
voluntary disclosure helps to eliminate the information risk of a firm, and potentially reduce the risk
premium that investors require. Rogers and Stocken (2005) advance that investors adjust their valuation of a
firm according to the credibility of management earnings forecasts and that firms may suffer from a liar’s
discount when their voluntary disclosure is proven to be deceptive. Taken as a whole, we predict that if firms
consistently provide accurate voluntary earnings forecasts (i.e., the forecasted earnings are relatively close to
the actual earnings), they will benefit from reducing uncertainty about their future prospects. This reasoning
leads to our fourth hypothesis:
Hypothesis H4a: Firms releasing more accurate earnings forecasts are associated with a greater
reduction in firm risk.
Hypothesis H4b: Firms releasing more accurate earnings forecasts are associated with a greater
enhancement in firm valuation.
Reputation is an important incentive mechanism shaping the behaviors of market participants (Chen,
Francis and Jiang, 2005; Graham et al., 2005; Barnett, Jermier and Lafferty, 2006). For example, Graham et
13
al. (2005) survey and interview more than 400 CFOs to identify the factors that motivate managers to
provide voluntary disclosure and 92 percent of the executives claim that promoting a reputation for
transparent reporting is a key driver. In the context of management earnings forecasts, Williams (1996)
shows that financial analysts response to management earnings forecasts depends greatly on previous
forecast accuracy (i.e., reputation for accuracy). In a recent study, Hutton and Stocken (2009) explore the
effect of firms’ forecasting reputation on investors’ reaction to the voluntary forecasts. They find that
investors are more responsive to a management earnings forecast when a firm has developed a good
reputation for forecast accuracy.
However, no prior study has investigated the aggregate effect of
forecasting reputation on both firm risk and valuation in a multiple-period context.
Our construct for forecasting reputation incorporates three dimensions of forecasting characteristics,
i.e., forecast accuracy, frequency and precision. Prior studies find that forecast accuracy helps firm establish
a forecast reputation (Stocken, 2000; Hutton and Stocken, 2009). Forecast frequency also matters because
firms have to release an adequate number of forecasts to allow investors to judge firms’ forecasting skills
and strategies. In the context of analyst earnings forecasts, Chen et al. (2005) posit that prior forecasting
accuracy and frequency help financial analysts build their forecasting reputation. Finally, forecast precision
may have an impact on forecast reputation for two reasons. First, in rational expectation models, rational
investors discount information signals when the signal is associated with greater uncertainty (Verrecchia,
2001). Second, precise forecasts increase the likelihood that a forecast can be verified ex post, so more
precise forecasts are more accurate than less precise forecasts (King et al., 1990; Baginski et al., 1993).
Our construct of forecasting reputation is consistent with the three dimensions of corporate reputation
defined in Barnett et al. (2006): awareness, assessment and asset. Here, more frequent forecasts can bring
awareness, more precise and accurate forecasts can facilitate assessment while the aggregate effect of
reputation can be ultimately impounded in firm asset (i.e., valuation). Furthermore, Hirst et al. (2008)
advocate testing the interaction effects between forecast antecedents and characteristics, or between two (or
14
more) forecast antecedents or characteristics. Our paper responds accordingly by investigating the
interactions of various forecast antecedents and characteristics where our reputation construct can be viewed
as the interaction among three forecast characteristics: frequency, precision and accuracy.
Reputation for corporate financial transparency plays an important role in determining equity prices
and distribution of returns. Using S&P disclosure and transparency scores, Durnev and Kim (2005) find that
firms with a better reputation for transparent disclosures have higher firm valuation. Klapper and Love
(2004) show similar results using CLSA disclosure and governance data. Additionally, Ferreira and Laux
(2007) indicate that corporate governance reputation is inversely related to idiosyncratic risk. Extending this
research, we expect that a stronger forecast reputation can make the information disclosed in the
management earnings forecasts more valuable to investors by decreasing information asymmetry and
estimation risk. This reasoning leads to our fifth hypothesis:
Hypothesis H5a: Firms with higher forecasting reputation are associated with a greater reduction in
firm risk.
Hypothesis H5b: Firms with higher forecasting reputation are associated with a greater enhancement
in firm valuation.
3. Data and Empirical Model
3.1 Sample Selection and Data Source
We create our sample by merging Compustat files (accounting data), CRSP (stock price data), and
the First Call Corporate Investor Guideline (CIG) database (management earnings forecast data) over the
period from 1998 to 2007. We focus on the post-1998 period because prior studies find that the CIG
database appears to be incomplete prior to 1998 (Anilowski et al., 2007). The CIG database has been utilized
by a number of researchers in accounting to investigate a variety of research questions (e.g., Ajinkya et al.,
2005; Rogers and Stocken, 2005; Anilowski et al., 2007; Wang, 2007; Rogers et al., 2009). Ajinkya et al.
15
(2005) show that the CIG database represents a more complete source of management earnings forecasts
than the Dow Jones News Retrieval System; however, the CIG database may omit management forecasts of
certain firms so a zero observation does not necessarily indicate the absence of a forecast. To alleviate this
concern, we follow Ajinkya et al. (2005) and only incorporate sample firms that appear in the First Call
Analyst Forecast Database which appears to contain more representative coverage.
In our study, we focus on EPS forecasts as these are the most common form of forecast and the most
anticipated by analysts. Consequently, EPS forecasts are the most logical forecasts to consider and focusing
on only one type of forecast allows us to have a ―clean‖ data set, so we eliminate non-EPS forecasts and
earnings pre-announcements released after the end of fiscal period. The CIG database carries both annual
and quarterly forecasts. Our forecast frequency variable is defined as an indicator variable that takes on a
value of 1 if a firm issue both annual and quarterly forecasts3. For our primary tests we focus on annual
forecasts. This is consistent with the arguments in Hutton and Stocken (2009), where the significance of
forecasting accuracy and reputation is enhanced by an increase in the horizon of a forecast, so annual
forecasts in comparison with quarterly forecasts offer a more powerful setting for exploring the influences of
forecast accuracy and reputation.4
We include only firms with total assets greater than $1 million. (We exclude firms whose total assets
are below $1 million to avoid possible extreme values for Tobin’s Q.) A breakdown of management earnings
forecast observations is presented in Table 1. Among our total sample of 52,206 firm-year observations,
12,815 (24.6%) are defined as forecasters, i.e., firms releasing annual management earnings forecasts. This
is consistent with the evidence in Anilowski et al. (2007) that 25% of firms issued management earnings
forecasts in the period of 2001-2003. Among the 12,815 forecasting observations, 81.6% of the firms
provide quarterly forecasts at the same time. Point forecasts account for 20.4% of our sample, whereas range
3
We also use the number of total management earnings forecasts in our sample period as an alternative proxy for forecast frequency. However,
our results are not sensitive to the two measures of forecast frequency.
4
We use both quarterly and annual forecasts in our robustness checks, and the results are similar.
16
forecasts account for 67.2% and the remaining 12.4% are either open-interval or qualitative forecasts.
Finally, 6.8% of the forecasting sample is defined as reputational forecasters, i.e., firms releasing point
forecasts, more frequent, and more accurate forecasts.
[Insert Table 1 Here]
3.2 Tests on Firm Risk
To test our hypotheses on firm risk, we rely on four risk measures that are standard in the finance
literature: (i) idiosyncratic or firm-specific risk (Goyal and Santa-Clara, 2003), (ii) stock return volatility
(Guay, 1999), (iii) beta or market risk (Guay, 1999), and (iv) bid-ask spreads on quoted share prices (Coller
and Yohn, 1997).. The calculation and motivation for the first three measures are described below in more
detail. The spreads are calculated as the average of the end of month closing bid and ask prices over the past
year using prices from CRSP. The bid-ask spread is viewed as a proxy for market liquidity and the
asymmetry of information (See Welker, 1995).
We estimate the idiosyncratic or firm-specific risk as well as beta from the following regression of
excess stock returns on the market risk premium:
Ri,t – Rf,t = αi + βi(Rm,t – Rf,t) + εi,t,
i = 1, …, t = 1,…,T,
(1)
where Ri,t is the monthly stock return for firm i in month t, Rf,t is the monthly return from holding a 30-day
risk-free treasury-bill provided by CRSP, Rm,t is the monthly return from the CRSP value-weighted market
index, αi (or alpha) is the intercept term, βi (or beta) is the slope coefficient, and εi,t is an error term. We run
this regression on a rolling 3-year basis for each firm in our sample. The idiosyncratic risk is the standard
deviation of the residuals from the regression using model (1). The volatility is the annualized monthly
standard deviation of each firm’s return series.
We then estimate mixed models to test the effects of management earnings forecasts on firm risk.
The mixed models we adopt contain both fixed and random effects. We include both effects since our
sample firms will have some common and some unique features – our mixed models can capture both
17
effects. To control for the systematic variation for each firm that may not be captured by our explanatory
factors, we model firm-specific effects as random effects. Modeling firm effects as random effects is more
appropriate than modeling them as fixed effects because our data are a sub-sample of all U.S. firms and our
set of control factors contain several measures which are relatively time invariant.
Riski,t = a + b DMEF,i,t + Σ γ Fk,i,t + ei,t
i = 1, …,N and t = 1,…,T
(2)
We model the firm-specific effects by decomposing the residual, ei,t, into a firm-level effect and a white
noise component. Consequently, the residual is decomposed as: ei,t = μi + vi,t because we model the firmspecific effects as a random effect, the residual is decomposed into a firm-level effect (μi) and a white noise
component (vi,t).
In our specification, DMEF,i,t refers to the five characteristics of management earnings forecasts
related to our five hypotheses: (i) for H1, it represents the occurrence of management earnings forecasts,
i.e., takes on a value of 1 when a firm releases management earnings forecasts, and 0 otherwise; (ii) for H2,
it denotes the frequency of management earnings forecasts, i.e., takes on a value of 1 if a firm also releases
quarterly forecasts, and 0 otherwise;5 (iii) for H3, it stands for the precision of management earnings
forecasts, i.e., 4 for point forecasts, 3 for range forecasts, 2 for open-interval, 1 for qualitative, and 0
otherwise; (iv) for H4, it means the accuracy of management earnings forecasts, i.e., takes on a value of 1 if
the absolute value of the difference between the forecasted earnings and the actual realized earnings is below
the median 0.29, and 0 otherwise;6 (v) for H5, it refers to the reputation of management earnings forecasts,
i.e., takes on the value of 1 if a firm issued point forecast (more precise) and also issued quarterly forecasts
(more frequent) and if the absolute value of the difference between the forecasted earnings and the actual
realized earnings is below the median 0.29 (more accurate), and 0 otherwise.
5
We also use the number of management earnings forecasts in our sample period to test the hypothesis on frequency, and the results remain
similar.
6
For the accuracy variable, we also use the absolute value of the difference between the forecasted earnings and the actual realized earnings,
deflated by the beginning-of-period stock price, and the results remain similar.
18
Our control variables, Fk, are motivated from the existing literature and are described as follows (note
that the i,t subscript is suppressed but each variable is measured on a firm-year basis). SIZE represents the
log of the market value of equity as of the firm’s fiscal year-end. Since larger firms tend to be less risky, we
expect the coefficient to be negative. LIABILITY is the ratio of total liabilities to total assets measured at
the fiscal year-end. Since firms with more leverage tend to be riskier, we expect the coefficient to be
positive. INTANGIBLE is the ratio of total value of intangible assets to total assets measured at the fiscal
year-end. The variable captures the degree of information asymmetry – more intangible assets imply there is
more intellectual capital and thus assets which are harder to value by investors. Firms with more intangible
assets will have a higher degree of information asymmetry so we expect the coefficient to be positive.
GROWTH is the percentage increase in sales over the past 3 years. The variable captures the firm’s future
prospects assuming that larger growth indicates increased future prospects. We expect firms which are
growing more rapidly to have a higher degree of information asymmetry so we expect the coefficient to be
positive. We also include YEAR and INDUSTRY dummies, not reported in the tables. We use a series of
ten industry classifications based on the standard break-down according to two-digit SIC codes. Specifically,
we define the industries as agriculture/forestry/fishing, mining, construction, manufacturing, transportation,
wholesale trade, retail trade, finance/insurance/real estate, services and other.
In unreported tests, we replace the year dummies with a REG_FD dummy that takes on the value of
1 for years 2000 and beyond and 0 otherwise to capture the effects of the introduction of the Security and
Exchange Commission’s (SEC) Regulation Fair Disclosure (Reg FD) in October 2000. The regulation
prohibits the selective disclosure of material nonpublic information, which would include management
earnings forecasts. Research such as Kwong and Small (2007) has found that analysts tend to be less
accurate in predicting earnings after the introduction of Reg FD, while Chiyachantana et al. (2004) found
that it led to improved liquidity and a decrease in information asymmetry. We find that our results reported
below do not change with the inclusion of this dummy variable.
19
3.3 Tests on Firm Value
In our tests of the hypotheses related to firm value, we measure firm value using Tobin’s Q.7 Tobin’s
Q is computed as the market value of equity plus the book value of the total liabilities in the numerator and
book value of assets in the denominator using the data from the firm’s fiscal year end obtained from
Compustat. Using Tobin’s Q, we estimate mixed models similar to those discussed in the previous section
to formally evaluate its relationship with the voluntary disclosure of management earnings forecasts while
controlling for other factors believed to influence firm value.
Tobin’s Qi,t = a, + b DMEF,i,t + Σ γ Fk,i, t + ei,t
i = 1, …,N and t = 1,…,T
(3)
where DMEF,i,t are the same as those used in equation (2). For the control variables Fk,i, t, we include SIZE,
LIABILITY, GROWTH, YEAR and INDUSTRY dummies as before but also include ROA (return on
assets), which is calculated as the operating income divided by the total assets as of the firm’s fiscal yearend. It is expected that firms with higher profitability will have higher value.8
3.4 Controlling for Self-Selection Bias
Since disclosing management earnings forecasts is discretionary and may depend on the same factors
included in our tests on firm risk and firm value, we control for this potential endogeneity by using a twostage model. Specifically, we use the Heckman two-step model (Heckman, 1979). With the Heckman
model, the first stage of the model investigates how different factors influence the likelihood that a firm will
voluntarily issue management earnings forecasts. In the second stage, the estimated likelihood that a firm
will voluntarily issue such guidance is inserted into the model to offset both the discrete nature of issuing
management earnings forecasts (i.e., a firm either does or does not issue them) and the factors which may
7
Note: we also considered other measures of firm value including the market-to-book ratio and alpha from the CAPM but found that they did
not provide any further insights beyond the more commonly considered Tobin’s Q. Consequently, these results are not presented.
8
We also perform additional sensitivity checks following previous literature (e.g., Lang and Stulz, 1994; Durnev and Kim, 2004) by controlling
for: 1) R&D expenses (deflated by total assets) in our Tobin’s Q model; and 2) dividend yield (deflated by total assets). Our results are not
sensitive to these corrections. However, adding these control variables substantially reduced our sample size, and thus these results are
unreported.
20
influence that decision and therefore may also influence the measures of firm risk and valuation. In addition,
Hirst et al. (2008) calls for exploring the potential interactions between forecasting characteristics,
antecedents and consequences. By adopting the two-stage Heckman models which link antecedents to
characteristics and consequences, we explicitly model these interactions. The model consists of the
following two stages:
Probability (Voluntary Disclosure of MEF) = f(ωFk,i, t)
(4a)
where a probit model is estimated using maximum likelihood to obtain estimates of ω. For each observation
 (ˆFi ,t )
in our sample we compute: ˆi ,t 
where φ is the normal probability distribution function and Φ is
(ˆFi ,t )
the normal cumulative distribution function. This provides an estimate of the non-selection hazard or the
probability that a firm with non-censored outcomes having those characteristics, Fi,t, would have issued a
management earnings forecast.
The characteristics we include in our first stage model are common in the literature and control for
the majority of factors believed to influence the management forecasting decision from previous studies,
especially measures of firm risk. The first variable is SIZE (as defined above). ANALYST is the number of
analysts following the firm. Following prior studies (e.g., Ajinkya et al., 2005), we expect the coefficient to
be positive. GOOD NEWS is a dummy variable taking on a value of one if the firm’s earnings per share
have increased from last year to this year. Previous literature (e.g., Skinner, 1994) documents that firms are
more likely to release earnings forecasts to mitigate litigation costs when facing bad news, so we predict a
negative coefficient on this variable. LOSS is a dummy variable taking on the value of 1 if the firm reported
a loss during the previous 12 months, and 0 otherwise. It is possible that management’s ability to forecast
earnings be constrained when firms making losses, so we expect the coefficient to be negative (Ajinkya et
al., 2005). SURPRISE is the absolute value of the differences between the current period’s and previous
period’s earnings per share deflated by the share price at the current fiscal year-end. We also control for
21
YEAR and INDUSTRY dummies in our first-stage probit model. To ensure the robustness of our results, we
also estimate a model in which we replace the YEAR and INDUSTRY dummies with REG_FD and
LITIGATE dummy variables. REG_FD takes on the value of 1 post- Regulation FD period (after October,
2000), and 0 otherwise. LITIGATE refers to industries with higher litigation risks, i.e., 1 for all firms in the
biotechnology (2833-2836 and 8731-8734), computers (3570-3577 and 7370-7374), electronics (36003674), and retail industry (5200-5961), 0 otherwise (Ajinkya et al., 2005) Our results remain the same for the
models with year and industry dummies or those controlling for REG_FD and LITIGATE.
We estimate different first-stage probit models (equation 4a) to test our five hypotheses. For H1, our
dependent variable is MEF (i.e., taking on a value of 1 if a firm discloses a management earnings forecast,
and 0 otherwise). To test H2, we estimate a probit model with D_Frequency (i.e., taking on a value of 1 if a
firm also issues quarterly forecasts, and 0 otherwise) as the dependent variable. With respect to H3, the
dependent variable for the first-stage model is D_Precision (i.e., taking on a value of 1 for point forecasts,
and 0 otherwise).9 For H4, the dependent variable is D_Accuracy (i.e., taking on a value of 1 if the absolute
value of the difference between the forecast earnings and the actual realized earnings is below the median
0.29, and 0 otherwise). Finally, to test the hypothesis on forecasting reputation H5, we estimate a probit
model with D_Reputation as the dependent variable (i.e., taking on a value of 1 if D_Frequency =1 and
D_Precision=1 and D_Accuracy = 1, and 0 otherwise). The estimated probability of a firm voluntarily
issuing earnings forecasts, ̂i,t , (also called the Inverse Mills Ratio) is inserted into the model estimated
above to help control for the endogeneity and the censoring of the outcomes:
Riski,t = a, + b ̂ i,t + Σ γ Fk,i,t + ei,t i = 1, …,N
t = 1,…,T
Tobin’s Qi,t = a, + b ̂ i,t + Σ γ Fk,i,t + ei,t i = 1, …,N
9
t = 1,…,T
(4b)
(4c)
We also estimated this model using the PRECISION variable defined as 4 for point forecasts, 3 for range forecasts, 2 for open-interval
forecasts, 1 for qualitative forecasts, and 0 otherwise. Therefore, we also estimated four probit models for point, range, open-interval and
qualitative forecasts, respectively. In the second stage, we inserted the four Inverse Mills Ratios into the firm risk or firm value models. Our
results remain similar so we only report the results for the more parsimonious model focusing on point forecasts.
22
The interpretation of the results and the hypothesized directions are the same for model (4) as for
model (3).10 However, as Larcker and Rusticus (2008) point out, in using instrumental variables in
accounting research, it is important to assess the validity of the first stage model. A simple way to do so is to
look at the first stage F-test. For probit regressions, the likelihood ratio tests are similar to the F-tests in
ordinary regressions (Zivot, Startz and Nelson, 1998). For our probit models, the likelihood ratios show that
the first-stage regressions are highly significant, so our models do not appear to suffer from weak
instrumental variables and thus from a bias due to significant unobservable factors.
4. Results
Before formally testing our hypotheses, we examine our data. In Table 2 we present some summary
statistics. As mentioned earlier, our data set contains 52,206 firm-year observations for 12,610 U.S. firms
over the period from 1998 to 2007. We find that our firms represent a broad cross-section of the U.S.
market. Our set of firms ranges from very small firms to very large firms (average market capitalization of
$283 million with the 25th and 75th percentiles being $63 million and $1,129 million respectively), and from
very profitable to less profitable (with a median annual return on assets of 1.7% over our sample period and
an interquartile range from -3.1% to 6.4% per annum) over all industries. Median liability-to-assets is 54%.
Median aggregate 3-year sales growth is just under 30%. There were no explicit screens on our sample
firms beyond requiring a minimum $1 million asset size and that they have data for at least one complete
year during our sample period.
[Insert Table 2 Here]
To test our hypotheses, we start by using the models in equations (2) and (3) above. In Tables 3 and
4, we present the results for the estimations of models (2) and (3) respectively. We begin our analysis with
10
Although there are different ways of implementing this Heckman model, our approach is similar to that employed in studies such as Lang,
Lins, and Miller (2003) whereby the Inverse Mills Ratio is used in place of the original variable. As described in Greene (1997), this approach
focuses on the self-selection component of the first stage estimation. If the Inverse Mills Ratio is included along with the original variable, the
interpretation of the estimated coefficients is different as the model will capture both the self-selection as well as the potential influence of
unobservable factors.
23
Table 3 by examining the role played by voluntary management earnings forecasts in explaining differences
in our measures of firm risk. Panel A presents the results from the estimation of the mixed-effect model
(model 2), while Panels B and C summarize the results from the two-stage Heckman treatment model
(models 4a and 4b). In Panel A we see that the level of a firm’s idiosyncratic risk, stock return volatility,
beta, and bid-ask spread decrease when firms voluntarily issue management earnings forecasts – the
estimated coefficient on MEF is significantly negative as we predicted in H1. Although not the focus of our
analysis, we find that firms with higher growth are associated with higher risk consistent with prior studies
(Guay, 1999; Ashbaugh-Skaife et al., 2009). In addition, similar to Ashbaugn-Skaife et al. (2009), we find
the surprising result that highly levered firms are associated with lower firm risk. Since our control variables
may also influence the decision to voluntarily issue management earnings forecasts, we try to correct for this
potential endogeneity using a Heckman two-step procedure.
In Panel B, we present the estimation of model 4a, i.e., the first stage of the Heckman model.
Consistent with prior literature (Ajinkya et al., 2005; Karamanou and Vafeas, 2005), we find that larger
firms, firms with more analysts following, and firms with bad news are more likely to release management
earnings forecasts. Simply losing money does not increase the likelihood of issuing management earnings
forecasts—it must be bad or unexpected news.
In Panel C we present the results from the estimation of model 4b. We find that the estimated
coefficients on the Inverse Mills Ratio (our estimated value for the likelihood a firm will issue a MEF
correcting for external factors) continue to be negative for a firm’s idiosyncratic risk, stock return volatility,
beta, and bid-ask spread (all significant except for the last one). In general, our findings therefore indicate
that firms disclosing management earnings forecasts are associated with lower systematic risk, idiosyncratic
risk and total risk, which supports H1a suggesting that firms issue management earnings forecasts to reduce
information asymmetry and thus reduce risk.
[Insert Table 3 Here]
24
In Table 4, Panel A we present the results from equation 3 (mixed models) and in Panel B (and
subsequent tables) we present the results from models 4a (implicitly) and 4c (explicitly). To capture the
important role played by firm risk and thus the asymmetry of information between management and
investors, we include our measures of risk in the model as well. Our measure of firm value is Tobin’s Q, a
measure which provides a broad, aggregate perspective on how investors are valuing a firm. It measures the
entire value of the firm (i.e., its enterprise value) relative to the assets that it employs so an increase in this
value is related to an increase in how investors are valuing the firm. In Panel A, all coefficients on the
issuing of MEFs are negative and significant. This is inconsistent with our hypothesis that firms voluntarily
releasing earnings forecasts are valued more highly by investors. We also find that larger firms and firms
that grow faster are associated with higher firm value. In Panel B, the results from the Heckman treatment
model indicate that the estimated coefficients on the Inverse Mills Ratios are negative and significant in all
models. These results imply, somewhat surprisingly, that firms that release management earnings forecasts
are associated with lower valuation after controlling for a number of variables such as size and growth,
which is opposite to our predictions in our H1b.
There are a number of potential explanations for these apparently conflicting results. The major
concern is that the management earnings forecast dummy variable is simply one measure (and a very coarse
measure as it is a dichotomous variable) to capture the degree to which management is able to decrease the
asymmetry of information between management and investors. There may be other characteristics and
antecedents which play a significant role but we are only considering whether or not the firm issues
forecasts. This will be investigated in subsequent tests. It is, however, also possible that endogeneity
between the decision to voluntarily disclose a forecast of earnings per share and other factors is masking
subtle differences in the firms. We investigate the possible influence of other factors on our results by
investigating Hypotheses 2 through 5 which deal with the impact of different characteristics of the earnings
forecasts issued by management on firm value.
25
[Insert Table 4 Here]
Table 5 presents the results from the estimation of models 4b and 4c for the Heckman 2-step model
in which we have further refined our measure of management earnings forecasts to consider the frequency of
management earnings forecasts on firm risk and valuation. Panel A presents the results on firm risk while
Panel B demonstrates the results on firm value.11 In Panel A, the coefficients on the Inverse Mills Ratio are
negative and significant for models with idiosyncratic risk, stock return volatility and beta as dependent
variables. This implies that firms issuing more frequent forecasts are associated with lower idiosyncratic
risk, systematic risk and total risk. However, the coefficient on the Inverse Mills Ratio for the average
spreads is positive albeit insignificant, suggesting that more frequent forecasts do not reduce average spread
in the long run.12
Results in Panel B are somewhat mixed. In the models with idiosyncratic risk and stock return
volatility as our risk measures, the coefficients on the Inverse Mills Ratios are positive and significant.
However, in other models the coefficients on the Inverse Mills Ratios are negative and significant. It appears
that the relationship between value and management earnings forecasts works for regressions after correcting
for stock market-related measures of risk. Thus there may be some benefit for companies to issue frequent
earnings forecasts but this perceived benefit is only apparent when certain measures of risk are incorporated
into our models.
[Insert Table 5 Here]
Table 6 presents the results from the Heckman model using the precision of management earnings
forecasts. In Panel A, we show results on firm risk. The coefficients on the Inverse Mills Ratios for our
measures of risk are consistent with our H3a that firms releasing more precise information are associated with
lower risk. The decrease in risk is statistically significant for idiosyncratic risk, systematic risk, and total
11
For brevity, we only report the results on the 2-stage Heckman treatment model. However, we also estimate the 1-stage mixed models, and the
results are similar.
12
In fact, the coefficient on forecast frequency in the mixed model is negative and significant. The inconsistency implies that reducing average
spread may be an antecedent or motivation to release management earnings forecasts, rather than a consequence (Coller and Yohn, 1997).
26
risk. The coefficient on the Inverse Mills Ratio in the average spread model is negative but insignificant.
Panel B presents the results for firm value. The estimated coefficients on the Inverse Mills Ratios are all
positive and significant. The results clearly indicate that firms making more precise forecasts are associated
with an increase in firm valuation. This provides strong support for our hypothesis H3b. It appears that the
precision of the estimates provides more information content and thus impacts the perceived level of risk for
a firm and how investors value the firm.
[Insert Table 6 Here]
The effects of the accuracy of management earnings forecasts on firm risk and firm valuation are
presented in Table 7. Panel A summarizes the results on firm risk. As in previous results, we find that more
accurate forecasts decrease the level of risk. The coefficients on the Inverse Mills Ratio are negative and
significant in the models with idiosyncratic risk, stock return volatility and beta, but the coefficient in the
average spread model is negative and only significant at the 11% level. Taken as a whole, firms release more
accurate management earnings forecasts are associated with lower idiosyncratic risk, systematic risk, and
total risk consistent with our H4a. Panel B presents the results on firm value, which are mixed. In three of
the five models, the coefficients on the Inverse Mills Ratios are negative and significant and in the other two
they are positive and marginally significant. The two results that are consistent with hypothesized results are
when the market-based risk measures are included.
There are several possible explanations for the firm value results. One possibility is that firms
releasing information which better predicts the actual results are not rewarded with an enhancement in firm
valuation. A second possibility is that this ―symmetric‖ measure of accuracy does not accurately capture the
market’s perception of what an accurate forecast is.
Anecdotally, markets may take a somewhat
―asymmetric‖ view and in some cases even missing an earnings forecast by one cent can have a huge,
negative impact on the stock price (for example, see Shefrin (2007, pp. 33-34) for an ―excessive optimism‖
explanation for the severe negative market reaction to the news in January 2005 that eBay missed analysts’
27
consensus forecast by one cent and in response eBay’s stock price declined by 20%). A third possibility is
that accuracy alone, as measured here, is not rich enough to capture the relationship with firm value, which
leads to our reputational measure results below.
[Insert Table 7 Here]
Finally, we consider the influence of the overall reputation of the management earnings forecast on
firm valuation in Table 8. We construct a reputation variable which takes on the value of 1 if a firm releases
point forecasts in a given year and releases more accurate and frequent earnings forecasts (as defined in the
table); and zero otherwise. The results are summarized in Table 8. Panel A shows that in all five model
specifications for our measures of firm risk, the results are consistent with our H5a. Specifically, firms that
have developed a forecasting reputation are associated with lower idiosyncratic risk, systematic risk, total
risk, and enhanced long-run liquidity. Panel B shows the results on firm value. In all five model
specifications, the coefficients on the Inverse Mills Ratios are positive and significant and the significance
levels are higher than in previous tables. These findings provide support for H5b signifying that firms with
strong reputation for quality disclosure are valued more highly than those who do not.
[Insert Table 8 Here]
Taken together with the findings from our tests of the other hypotheses, our results suggest that
although releasing management earnings forecasts per se are not necessarily associated with an increase in
firm valuation, firms are rewarded with their high-quality management earnings forecasts once they have
established a disclosure reputation.
More frequent management earnings forecasts appear to provide
marginal benefits. More precise and accurate forecasts provide more benefits. Frequent, precise and
accurate forecasts provide superior benefits, as captured by our reputation measure.
Our results contrast with Kato, Skinner and Kunimura (2009) who find that market forces such as
reputation are insufficient in Japan to monitor manager’s forecasting behaviors. One plausible explanation is
that the U.S. and Japan have different legal/regulatory environments, so market forces play distinctive roles
28
in these two countries. Our results on the reputational hypothesis also substantiate the view in Lambert et al.
(2007) that information quality can have an indirect effect on the cost of equity capital by changing firms’
real decisions. Once the managers of a firm have built up a forecasting reputation, its forecasting system can
play a positive role in shaping its internal budgeting system and corporate governance. In addition, to protect
the reputation of the firm, managers would try to avoid managerial myopia, which may lead to better
investment decisions. Consequently, forecasting reputation changes firms’ future cash flows, not just
through the perceptions of cash flows by investors, but through real decisions which help better allocate
capitals.
Overall, our results provide evidence that voluntarily providing earnings forecasts may decrease the
asymmetry of information between management and investors. It is, however, not a simple relationship.
Firms must provide precise and accurate guidance for investors to value the information.
5. Summary and Conclusions
In this paper, we study the economic consequences of management earnings forecasts using a large
sample of firms for the period from 1998 to 2007. More specifically, we explore the effects of multiple
management earnings forecast characteristics (i.e., occurrence, frequency, precision, accuracy and
reputation) on firm risk and firm value. We employ four different measures to capture firm risk, i.e.,
idiosyncratic risk, beta, total risk, and bid-ask spread. Compared to firms that do not issue management
earnings forecasts, we find that forecasting firms have lower idiosyncratic risk, market risk (beta) and total
risk. These results also apply to firms that release more frequent, precise or accurate management earnings
forecasts. However, firms providing management forecasts experience an improvement in long-term
liquidity (a decrease in bid-ask spread) only when they have built up a strong forecasting reputation for
consistently disclosing more frequent, precise and accurate information. Taken as a whole, our findings
29
indicate that the quality of information not only influences firms’ idiosyncratic risk, but non-diversifiable
systematic risk.
We use Tobin’s Q in measuring firm value. Our results demonstrate that releasing management
earnings forecasts per se is not associated with an increase in firm value, which implies that a reduction in
market uncertainty does not necessarily translate into valuation. Nevertheless, more frequent management
earnings forecasts appear to provide marginal benefits in firm value. More precise and accurate forecasts
provide more benefits. Finally, frequent, precise and accurate forecasts provide superior benefits, as captured
by the highly significant and positive association between forecasting reputation variable and Tobin’s Q. Our
results advance that firms garner benefits in firm value only when they provide high-quality management
earnings forecasts and have a strong forecasting reputation. An interesting implication which can be gleaned
from our results is that forecasting reputation changes firms’ future cash flows, not just through the
perceptions of cash flows by investors, but through real decisions which help better allocate capital. In
general, our evidence supports the direct and indirect effects of information quality on the cost of capital
advocated in Lambert et al. (2007), and suggests that releasing high-quality management earnings forecasts
is associated with significant capital market benefits
The lack of a clear relationship between reductions in firm risk and enhancements in firm value is not
surprising given that the academic literature on the associations between risk and return is similarly muddled
(Shin and Stulz, 2000a, 2000b; Goyal and Santa-Clara, 2003). Our research suggests that decreases in
market uncertainty are echoed in firm valuation only under certain circumstances, and we leave for future
research the advancement of additional clarification and explanation of the precise relationship between firm
risk and value.
30
References
Ajinkya, B., and M. Gift, 1984. Corporate managers’ earnings forecasts and symmetrical adjustments of market
expectations. Journal of Accounting Research 22: 425-444.
Ajinkya, B., S. Bhojraj, and P. Sengupta. 2005. The association between outside directors, institutional investors and
properties of management earnings forecasts. Journal of Accounting Research 43: 343-376.
Anilowski, C., M. Feng, and D. J. Skinner. 2007. Does earnings guidance affect market returns? The nature of
information content of aggregate earnings guidance. Journal of Accounting and Economics 44: 36-63.
Ashbaugh-Skaife H., D. W. Collins, W. R. Kinney, and R. Lafond. 2009. The effect of SOX internal control
deficiencies on firm risk and cost of equity. Journal of Accounting Research 47: 1-43.
Baginski, S., E. Conrad, and J. Hassell. 1993. The effects of management forecast precision on equity pricing and on
the assessment of earnings uncertainty. The Accounting Review 68: 913-927.
Ball, R., G. Sadka and R. Sadka, 2009. Aggregate earnings and asset prices. Journal of Accounting Research 47: 10971133. Barnett, M., J. M. Jermier and B. A. Lafferty. 2006. Corporate reputation: the definitional landscape.
Corporate Reputation Review 9: 26-38.
Barnett, M.L., J. Jermier, and B. A. Lafferty, 2006. Corporate reputation: a definitional landscape. Corporate
Reputation Review 9: 26-38.
Barry, C., and S. Brown. 1985. Differential information and security market equilibrium. Journal of Financial and
Quantitative Analysis 20: 407-422.
Barry, C., and S. Brown. 1986. Limited information as a source of risk. The Journal of Portfolio Management 12: 6672.
Bartram, S. M., G. W. Brown, and J. Conrad. 2009. The effects of derivatives on firm risk and value. Working paper,
University of North Carolina (Available at SSRN: http://ssrn.com/ abstract=1210422).
Botosan, C. 1997. Disclosure level and the cost of equity capital, The Accounting Review 72: 323-350.
Botosan C. and M. S. Harris. 2000. Motivations for a change in disclosure frequency and its consequences: an
examination of voluntary quarterly segment disclosures. Journal of Accounting Research 38: 329-353.
Brown, S. and S. Hillegeist, 2007. How disclosure quality affects the level of information asymmetry, Review of
Accounting Studies 12: 443-477.
Chen, Q., J. Francis, and W. Jiang. 2005. Investor learning about analyst predictive ability. Journal of Accounting and
Economics 39: 3-24.
Chen, S., D. Matsumoto and S. Rajgopal. 2006. Is silence golden? An empirical analysis of firms that stop giving
quarterly earnings guidance in the post Regulation-FD period. Working paper, University of Washington
(Available at SSRN: http://ssrn.com/abstract=820644).
Chiyachantana, C, C. Jiang, N. Taechapiroontong, and R. Wood, 2004. The impact of Regulation Fair Disclosure on
information asymmetry and trading: An intraday analysis. The Financial Review 39: 549-577
31
Coller, M., and T. L. Yohn. 1997. Management forecasts and information asymmetry: An examination of bid-ask
spreads. Journal of Accounting Research 35: 181-191.
Daske H., L. Hail, C. Leuz, and R. Verdi, 2008. Mandatory IFRS reporting around the world: early evidence on the
economic consequences. Journal of Accounting Research 40: 1085-1142.
Diamond, D. W., and R. E. Verrecchia. 1991. Disclosure, liquidity and the cost of capital. The Journal of Finance 46:
1325-1359.
Doidge, C., G. A. Karolyi, and R. Stulz. 2004. Why are foreign firms listed in the U.S. worth more? Journal of
Financial Economics 71: 205-238.
Durnev, A., E. H. Kim. 2005. To steal or not to steal: firm attributes, legal environment, and valuation. The Journal of
Finance 60: 1461–1493.
Easley, D. and M. O’Hara. 2004. Information and the cost of capital. The Journal of Finance 59: 1553-1583.
Ferreira, M. and P. Laux. 2007. Corporate governance, idiosyncratic risk and information flow. The Journal of Finance
62: 951-989.
Francis,J., R. LaFond, P. Olsson and K. Schipper. 2004. Cost of equity and earnings attributes. The Accounting
Review 79: 967-1010.
Francis, J., D. Nanda and P. Olsson, 2008. Voluntary disclosure, earnings quality and cost of capital. Journal of
Accounting Research 46: 53-99
Godfrey, P. C., C. B. Merrill and J. M. Hansen. 2009. The relationship between corporate social responsibility and
shareholder value: an empirical test of the risk management hypothesis. Strategic Management Journal 30:
425-445.
Goyal, A., and P. Santa-Clara. 2003. Idiosyncratic risk matters! The Journal of Finance 58: 975-1007.
Graham, J., C. Harvey and S. Rajgopal, 2005. The economic implications of corporate financial reporting. Journal of
Accounting and Economics 40: 3-73.
Greene, W. 1997. Econometric Analysis, Third Edition. New York: Prentice-Hall.
Guay, W. R. 1999. The impact of derivatives on firm risk: an empirical examination of new derivatives users. Journal
of Accounting and Economics 26: 319-351.
Healy, P. and K. Palepu. 2001. Information asymmetry, corporate disclosure, and the capital markets: A review of the
empirical disclosure literature. Journal of Accounting and Economics 31: 405-440.
Heckman, J. 1979. Sample Selection Bias as a Specification Error, Econometrica 47: 153-161.
Hirst, E., L. Koonce, and S. Venkataraman, 2008. Management earnings forecasts: A review and framework.
Accounting Horizons 22: 315-338.
Hughes, J. S., and S. Pae. 2004. Voluntary disclosure of precision information. Journal of Accounting and Economics:
37: 261-289.
32
Hutton, A. P., and P. C. Stocken. 2009. Prior forecast accuracy and investor reaction to management earnings
forecasts. Working paper, Dartmouth College (Available at SSRN: http://ssrn.com/abstract_id=817108).
Karamanou, I., and N. Vafeas. 2005. The association between corporate boards, audit committees, and management
earnings forecasts: an empirical analysis. Journal of Accounting Research 43: 453-486.
Kato, K., D. J. Skinner and M. Kunimura. 2009. Management forecasts in Japan: an empirical study of forecasts that
are effectively mandated. The Accounting Review, forthcoming.
Kim, O., and R. Verrecchia. 1991a. Trading volume and price reactions to public announcements. Journal of
Accounting Research 29: 302–321.
Kim, O., and R. Verrecchia. 1991b. Market reaction to anticipated announcements. Journal of Financial Economics
30: 273–310.
Kim, O., and R. Verrecchia. 1994. Market liquidity and volume around earnings announcements. Journal of
Accounting and Economics 17: 41-68.
King, R., G. Pownall and G. Waymire. 1990. Expectations adjustment via timely earnings forecast disclosure: Review,
synthesis, and suggestions for future research. Journal of Accounting Literature 9: 113-144.
Klapper, L., and I. Love. 2004. Corporate governance, investor protection and performance in emerging markets.
Journal of Corporate Finance 10: 703-728.
Klein, R., and V. Bawa. 1976. The effect of estimation risk on optimal portfolio choice. Journal of Financial
Economics 3: 215-231.
Kwang, S.-W. and K. Small. 2007. The impact of Regulation Fair Disclosure on earnings management and analyst
forecast bias. Journal of Economics and Finance 31: 87-98.
Lambert, R., C. Leuz and R. Verrecchia. 2007, Accounting information, disclosure and the cost of capital. Journal of
Accounting Research 45: 385-420.
Lang, L., and R. Stulz. 1994. Tobin’s Q, corporate diversification, and firm performance. Journal of Political Economy
6: 1248-1280.
Lang. M, K. Lins, and D. Miller. 2003. ADRs, analysts, and accuracy: Does cross listing in the United States improve
a firm's information environment and increase market value? Journal of Accounting Research 41: 317-345.
Larcker, D. F., and T. O. Rusticus. 2008. On the use of instrumental variables in accounting research. Working paper,
Standford Graduate School of Business (Available at SSRN: http://ssrn.com/abstract=694824).
Litvak, K. 2008. The Long-Term Effect of the Sarbanes-Oxley Act on Cross-Listing Premia. European Financial
Management 14: 875-920.
Leuz, C. and R. Verrecchia. 2000. The economic consequences of increased disclosure. Journal of Accounting
Research 38: 91-124.
Merton, R. 1987. A simple model of capital market equilibrium with incomplete information. Journal of Finance 42:
483–510.
Miller, E.M. 1977. Risk, uncertainty, and divergence of opinion. The Journal of Finance 32: 1151-1168.
33
Morse, D., J. Stephan and K. Stice. 1991. Earnings announcement and the convergence (or divergence) of beliefs. The
Accounting Review 66: 376-388.
Ng, J., I. Tuna and R. Verdi. 2008. Management forecast credibility and underreaction to news. Working paper, MIT
Sloan School of Management (Available at SSRN: http://ssrn.com/ abstract=930697).
O’Hara, M. 2003. Presidential address: liquidity and price discovery. The Journal of Finance 58: 1335-1354.
Pownall, G., C. Wasley, and G. Waymire. 1993. The stock price effects of alternative types of management earnings
forecasts. The Accounting Review 68: 896-912.
Rogers, J. L., D. J. Skinner and A. V. Buskirk. 2009. Earnings guidance and market uncertainty. Journal of Accounting
and Economics 48: 90-109.
Rogers, J. L., and P. C. Stocken. 2005. Credibility of management forecasts. The Accounting Review 80: 1233-1260.
Sengupta, P. 1998. Corporate disclosure quality and the cost of debt. The Accounting Review 73: 459-474.
Shefrin, H. 2007. Behavioral corporate finance: Decisions that create value. McGraw-Hill Irwin. New York, NY.
Shin, H-H, and R. M. Stulz. 2000a. Shareholder wealth and firm risk. Working paper, Dice center (Available at SSRN:
http://ssrn.com/abstract=254271).
Shin, H-H, and R. M. Stulz. 2000b. Firm value, risk, and growth opportunities. Working paper, Dice center (Available
at SSRN: http://ssrn.com/abstract=234344).
Skinner, D. 1994. Why firms voluntarily disclose bad news. Journal of Accounting Research 32: 38-60.
Sorescu, A. B., and J. Spanjol. 2008. Innovation’s effect on firm value and risk: insights from consumer packaged
goods. Journal of Marketing 72: 114-132.
Stocken, P. C. 2000. Credibility of voluntary disclosure. RAND Journal of Economics 31: 359-374.
Verrecchia, R. 2001. Essays on disclosure. Journal of Accounting and Economics 32: 91-180.
Wang, I. 2007. Private earnings guidance and its implications for disclosure regulation. The Accounting Review 82:
1299-1332.
Waymire, G. 1985. Earnings volatility and voluntary management forecast disclosure. Journal of Accounting Research
23: 268-295.
Welker, M. 1995. Disclosure policy, information asymmetry, and liquidity in equity markets. Contemporary
Accounting Research 11: 801-827.
Williams, P.A. 1996. The relation between a prior earnings forecast by management and analyst response to a current
management forecast. The Accounting Review 71: 103-115.
Zivot, E., R. Startz and C. R. Nelson. 1998. Valid confidence intervals and inference in the presence of weak
instruments. International Economic Review 39: 1119-1144.
34
Table 1
Management Earnings Forecast Sample Breakdown
The sample contains 52,206 firm-year observations from 1998 to 2007 for 12,610 firms with total assets greater than
$1 million. The total sample of 52,206 includes 12,815 (24.6%) firm-year observations that provided annual
management earnings forecasts, and 39,391(75.4%) observations that did not provide annual management earnings
forecasts. This table indicates the breakdown of the management earnings forecast portion. Reputational forecasters
refer to firms which have built their reputation by issuing point forecasts, quarterly forecasts, and more accurate
forecasts (i.e., if the absolute value of the difference between the forecast earnings and the actual realized earnings is
below our sample median of 0.29) in our sample period.
Management Earnings Forecast Sample
Overall
Both Annual and quarterly forecasts issued
within the fiscal year
Qualitative forecasts
Open-interval forecasts
Range forecasts
Point forecasts
Reputational forecasters
Observations
12,815
10,457
% of Total
100%
81.6%
1,088
498
8,617
2,612
871
8.5%
3.9%
67.2%
20.4%
6.8%
35
Table 2
Summary Statistics
Dependent and independent variables used in the analysis for our sample which contains 52,206 firm-year
observations from 1998 to 2007 for 12,610 firms. Idiosyncratic Risk is estimated as the standard deviation of the
residuals (slope coefficient) from the CAPM regression in equation (1) using three years of monthly data; Stock
Return Volatility is the annualized standard deviation of each firm’s return series using three years of monthly data;
Beta is estimated as the slope coefficient from the CAPM regression in equation (1) using three years of monthly data;
Average Spread is calculated as the average of the end-of-month closing bid and ask prices over the past year using
prices from CRSP; Tobin’s Q is computed as the market value of equity plus the book value of the total liabilities in
the numerator and book value of assets in the denominator; Size represents the log of the market value of common
equity, in $millions, as of the firm’s fiscal year-end; Liability is the ratio of total liabilities to total assets measured at
the fiscal year-end; Intangible is the ratio of total value of intangible assets to total assets measured at the fiscal yearend; Growth is the percentage increase in sales over the past 3 years; Analysts is the number of analysts following a
firm, as proxied by the number of forecasting estimates in a given year; Loss is an indicator variable taking on the
value of 1 if a firm reported a loss during the previous 12 months, and 0 otherwise; ROA is the ratio of net profit to
total assets; Good_News is an indicator variable taking on the value of 1 if the current-period EPS is greater than
previous-period EPS, and 0 otherwise; and Surprise is the absolute value of the difference between the current-period
EPS and the previous-period EPS, deflated by the price at the fiscal year end.
25th
Percentile
0.490
0.536
0.504
0.047
75th
Percentile
1.151
1.229
Idiosyncratic Risk
Stock Return Volatility
Beta
Average Spread
Mean
0.908
0.969
1.213
0.231
Median
0.757
0.818
1.031
0.117
Standard
Deviation
0.610
0.627
1.113
2.896
Tobin’s Q
2.072
1.346
3.050
1.050
2.135
Size (log value)
Size (raw value)
Liability
Intangible
Growth
5.647
283
0.565
0.123
0.392
5.585
266
0.542
0.039
0.298
2.087
8.061
0.429
0.173
0.993
4.148
63
0.325
0.000
0.033
7.029
1129
0.764
0.185
0.653
Analysts
Loss
ROA
Good_News
Surprise
5.702
0.316
-0.063
0.588
0.768
3.000
0.000
0.017
1.000
0.029
6.725
0.465
0.582
0.492
49.796
0.000
0.000
-0.031
0.000
0.010
8.000
1.000
0.064
1.000
0.100
36
1.712
0.250
Table 3
Relationship Between Risk Measures and Management Earnings Forecasts
This table presents the regression results of the effects of the occurrence of management earnings forecasts (MEF) on firm risks. Panel A presents the results on
mixed-effect models. Panels B and C summarize the results on two-stage Heckman treatment models. The first stage model (Panel B) is described in equation
(4a), which is a probit model with MEF (i.e., taking on a value of 1 when a firm releases a management earnings forecasts, and 0 otherwise) as the dependent
variable. The independent variables in equation (4a) are Size, Loss, Analysts, Good_News, and Surprise. These variables are described in Table 2. In the second
stage (Panel C), the Inverse Mills ratio is inserted into the regression models listed below. The dependent variables are Idiosyncratic Risk, Stock Return
Volatility, Beta, and Average Spread as described in Table 2. The independent variables include MEF, Size, Liability, Intangible, Growth, as described in Table
2, and Industry and Year Dummies. We use a series of ten industry classifications based on the standard break-down according to 2-digit SIC codes.
Specifically, we define the industries as agriculture/forestry/fishing, mining, construction, manufacturing, transportation, wholesale trade, retail trade,
finance/insurance/real estate, services and other. The coefficients on Industry and Year dummies are not reported in this table. For each regression the first row is
the coefficient estimates, the second row is the standard errors, and the third row is the t-statistics. The R2 is the MacFadden’s (pseudo) R2 which indicates the
degree to which the model parameters improve upon the prediction of the null model, i.e., a model predicting the dependent variable without any independent
variables.
Intercept
1.3303
MEF
-0.0947
Size
-0.0838
Liability
-0.1038
Intangible
-0.0116
Growth
0.0501
Standard Error
0.0513
0.0066
0.0015
0.0101
0.0171
0.0029
T-statistic
25.94
-14.27
-56.34
-10.30
-0.68
17.07
Stock Return Volatility
1.349
-0.1092
-0.0772
-0.1107
-0.0090
0.0517
Standard Error
0.0531
0.0069
0.0015
0.0105
0.0177
0.0030
T-statistic
25.39
-15.89
-50.13
-10.59
-0.51
16.98
Beta
1.6387
-0.2107
0.0402
-0.1813
0.0273
0.0662
Standard Error
0.1047
0.01355
0.0030
0.0206
0.0349
0.0060
T-statistic
15.65
-15.56
13.26
-8.81
0.78
11.04
Average Spread
14.9627
-0.2019
0.0814
0.1087
-0.0529
0.0168
Standard Error
0.3037
0.0393
0.0088
0.0597
0.1014
0.0174
T-statistic
49.27
-5.14
9.25
1.82
-0.52
0.97
Panel A
Idiosyncratic Risk
37
R2
0.32
0.31
0.20
0.15
Table 3 (continued)
Relationship Between Risk Measures and Management Earnings Forecasts
Panel B
Intercept
Size
Loss
Analysts
Good_News
Surprise
R2
MEF Dummy
-2.2378
0.2149
-0.4314
0.0099
-0.0509
-0.0000
0.83
Standard Error
0.1436
0.0060
0.0185
0.0015
0.0152
0.0004
T-statistic
-15.59
35.95
-23.30
6.43
-3.34
-0.08
Idiosyncratic Risk
0.8690
Inverse
Mills
-1.0975
Standard Error
0.0536
0.0263
0.0037
0.0105
0.0174
0.0035
T-statistic
16.22
-41.78
13.12
-9.81
-3.80
16.61
Stock Return Volatility
0.8452
-1.2008
0.0683
-0.1118
-0.0701
0.0584
Standard Error
0.0557
0.0273
0.0039
0.0109
0.0181
0.0036
T-statistic
15.18
-43.97
17.56
-10.28
-3.88
16.04
Beta
0.8672
-1.9792
0.2813
-0.2029
-0.0771
0.0592
Standard Error
0.1163
0.0570
0.0081
0.0227
0.0377
0.0076
7.46
-34.71
34.65
-8.94
-2.04
7.79
Average Spread
15.4745
-0.0014
0.0654
0.0934
-0.1169
0.0347
Standard Error
0.3443
0.1689
0.0240
0.0672
0.1117
0.0225
T-statistic
44.94
-0.01
2.72
1.39
-1.05
1.54
Panel C
T-statistic
Intercept
Size
Liability
Intangible
Growth
R2
0.0491
-0.1026
-0.0660
0.0582
0.47
38
0.46
0.32
0.28
Table 4
Relationship Between Firm Value and Management Earnings Forecasts
This table reports the regression results of the effects of the occurrence of management earnings forecasts on firm valuation. Panel A presents the
results on mixed-effect models. Panel B summarizes the results on two-stage Heckman treatment models. The first stage model (see Panel B of
Table 3) is described in equation (4a), which is a probit model with MEF (i.e., taking on a value of 1 when a firm releases a management earnings
forecast, and 0 otherwise) as the dependent variable. The independent variables in equation (4a) are Size, Loss, Analysts, Good_News, and
Surprise, which are described in Table 2. The dependent variable for the second-stage models, reported below, is Tobin’s Q as described in Table
2. In the second stage, the Inverse Mills ratio is inserted into the regression models listed below. Other independent variables include Size,
Liability, Growth, ROA, one risk measure (i.e., Idiosyncratic Risk, Stock Return Volatility, Beta, or Average Spread) and Industry and Year
Dummies. All these variables are described in Table 2. The coefficients on Industry and Year dummies are not reported in this table. For each
regression the first row is the coefficient estimates, the second row is the standard errors, and the third row is the t-statistics. The R2 is the
MacFadden’s (pseudo) R2 which indicates the degree to which the model parameters improve upon the prediction of the null model, i.e., a model
predicting the dependent variable without any independent variables.
Panel A
Dependent variable: Tobin’s Q;
―Risk‖ independent variable:
No Risk Measure
Standard Error
T-statistic
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
Intercept
MEF
Size
Liability
Growth
ROA
0.2073
0.2140
0.97
-1.1368
0.2290
-4.96
-1.0687
0.2293
-4.66
-0.1207
0.2323
-0.52
0.1636
0.2389
0.68
-0.4525
0.0318
-14.24
-0.3151
0.0294
-10.72
-0.3096
0.0294
-10.52
-0.3760
0.0300
-12.55
-0.3940
0.0299
-13.17
0.2134
0.0068
31.33
0.3337
0.0068
48.88
0.3227
0.0068
47.54
0.2631
0.0068
38.70
-0.5580
0.0465
-12.00
0.5560
0.0352
15.78
-0.3621
0.0458
-7.90
-0.3642
0.0459
-7.94
-0.5310
0.0466
-11.40
0.1373
0.0131
10.47
0.1477
0.0133
11.08
0.0939
0.0129
7.28
0.0953
0.0129
7.38
0.1310
0.0131
9.98
-0.1119
0.0382
-2.93
0.4046
0.0273
14.82
0.2787
0.0387
7.21
0.2759
0.0388
7.11
-0.0642
0.0386
-1.66
-0.0106
0.0042
-2.55
39
Risk
R2
0.08
0.9565
0.0238
40.12
0.8923
0.0231
38.60
0.0933
0.0118
7.88
0.3019
0.3284
0.92
0.27
0.27
0.27
0.27
Table 4 (continued)
Relationship Between Firm Value and Management Earnings Forecasts
Panel B
Dependent variable: Tobin’s Q;
―Risk‖ independent variable:
No Risk Measure
Standard Error
T-statistic
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
Intercept
Inverse
Mills
-0.1801
0.1887
-0.95
-1.1228
0.2263
-4.96
-1.0385
0.2266
-4.58
-0.3629
0.2303
-1.58
-0.1970
0.2373
-0.83
-0.8502
0.1041
-8.17
-0.2913
0.1145
-2.54
-0.2846
0.1149
-2.48
-1.0064
0.1164
-8.65
-1.0951
0.1150
-9.53
Size
Liability
Growth
ROA
0.2990
0.0142
21.07
0.3300
0.0157
21.04
0.3176
0.0157
20.18
0.3462
0.0163
21.30
0.3605
0.0160
22.52
0.2823
0.0307
9.19
-0.2865
0.04471
-6.41
-0.2894
0.0448
-6.46
-0.4479
0.0456
-9.83
-0.4633
0.0454
-10.20
0.2359
0.0132
17.82
0.1929
0.0144
13.36
0.1966
0.0145
13.59
0.2384
0.0147
16.21
0.2419
0.0147
16.46
-0.3649
0.0283
-12.91
0.1342
0.0391
3.43
0.1279
0.0392
3.26
-0.1808
0.0392
-4.61
-0.2059
0.0389
-5.29
40
Risk
R2
0.30
0.9281
0.0244
38.01
0.8542
0.0236
36.21
0.0554
0.0114
4.85
-0.0080
0.0039
-2.03
0.43
0.43
0.42
0.42
Table 5
Relationship Between Management Earnings Forecasts Frequency and Firm Risks and Valuation
This table reports the regression results of the effects of the frequency of management earnings forecasts on firm risks and firm valuation. Panel A presents
results on firm risks, while Panel B presents results on firm valuation. Heckman treatment models are used to control for self-selection bias. The first stage model
is described in equation (4a), which is a probit model with D_Frequency (i.e., taking on a value of 1 if a firm also issued quarterly forecasts, and 0 otherwise) as
the dependent variable. The independent variables in equation (4a) are Size, Loss, Analysts, Good_News, and Surprise, which are described in Table 2. In Panel
A, the dependent variables are Idiosyncratic Risk, Stock Return Volatility, Beta, and Average Spread as described in Table 2. The independent variables include
Inverse Mills Ratio for D_Frequency, Size, Liability, Intangible, Growth, as described in Table 2, and Industry and Year Dummies. In Panel B, the dependent
variable for the second-stage models, reported below, is Tobin’s Q as described in Table 2. Other independent variables include Size, Liability, Growth, ROA,
one risk measure (i.e., Idiosyncratic Risk, Stock Return Volatility, Beta, or Average Spread) and Industry and Year Dummies. All these variables are described in
Table 2. The coefficients on Industry and Year dummies are not reported in this table. For each regression the first row is the coefficient estimates, the second
row is the standard errors, and the third row is the t-statistics. The R2 is the MacFadden’s (pseudo) R2 which indicates the degree to which the model parameters
improve upon the prediction of the null model, i.e., a model predicting the dependent variable without any independent variables.
Panel A: The Effect of Management Earnings Forecasts Frequency on Firm Risks
Inverse
Intercept
Size
Liability
Mills
1.1538
-0.3419
-0.0437
-0.0879
Idiosyncratic Risk
0.0538
0.0137
0.0025
0.0106
Standard Error
21.43
-24.98
-17.26
-8.26
T-statistic
1.1518
-0.3820
-0.0320
-0.0955
Stock Return Volatility
0.0561
0.0143
0.0026
0.0111
Standard Error
20.55
-26.81
-12.16
-8.62
T-statistic
1.2835
-0.7696
0.1371
-0.1751
Beta
0.1158
0.0294
0.0054
0.0229
Standard Error
11.09
-26.15
25.22
-7.65
T-statistic
15.5484
0.1150
0.0480
0.0925
Average Spread
0.34
0.0864
0.0160
0.0672
Standard Error
45.73
1.33
3.00
1.38
T-statistic
41
Intangible
Growth
R2
-0.0648
0.0177
-3.66
-0.0688
0.0184
-3.73
-0.0738
0.0380
-1.94
-0.1178
0.1117
-1.05
0.0585
0.0036
16.40
0.0587
0.0037
15.80
0.0578
0.0077
7.54
0.0362
0.0225
1.61
0.46
0.45
0.32
0.28
Panel B: The Effect of Management Earnings Forecasts Frequency on Firm Valuation
Dependent variable: Tobin’s Q;
―Risk‖ independent variable:
No Risk Measure
Standard Error
T-statistic
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
Intercept
0.1152
0.1864
0.62
-0.9529
0.2241
-4.25
-0.8666
0.2244
-3.86
-0.0131
0.2280
-0.06
0.1825
0.2350
0.78
Inverse
Mills
-0.0895
0.0532
-1.68
0.1055
0.0568
1.86
0.1151
0.0569
2.02
-0.1087
0.0581
-1.87
-0.1523
0.0577
-2.64
Size
Liability
Growth
ROA
0.2048
0.0093
22.01
0.2784
0.0103
26.96
0.2653
0.0103
25.70
0.2325
0.0106
21.85
0.2424
0.0105
23.04
0.2713
0.0308
8.82
-0.2873
0.0447
-6.43
-0.2901
0.0448
-6.47
-0.4473
0.0456
-9.81
-0.4660
0.0455
-10.24
0.2388
0.0133
18.01
0.1944
0.0144
13.45
0.1981
0.0145
13.68
0.2400
0.0147
16.28
0.2441
0.0147
16.57
-0.4015
0.0279
-14.37
0.1114
0.0386
2.89
0.1058
0.0387
2.73
-0.2437
0.0386
-6.32
-0.2800
0.0381
-7.35
42
Risk
R2
0.30
0.9448
0.0242
39.10
0.8715
0.0233
37.36
0.0681
0.0114
5.99
-0.0078
0.0039
-2.00
0.43
0.43
0.42
0.42
Table 6
Relationship Between Management Earnings Forecasts Precision and Firm Risks and Valuation
This table reports the regression results of the effects of the precision of management earnings forecasts on firm risks and firm valuation. Panel A presents results
on firm risks, while Panel B presents results on firm valuation. Heckman treatment models are used to control for self-selection bias. The first stage model is
described in equation (4a), which is a probit model with D_precision (i.e., taking on a value of 1 for point forecasts, and 0 otherwise) as the dependent variable.
The independent variables in equation (4a) are Size, Loss, Analysts, Good_News, and Surprise, which are described in Table 2. In Panel A, the dependent
variables are Idiosyncratic Risk, Stock Return Volatility, Beta, and Average Spread as described in Table 2. The independent variables include Inverse Mills
Ratio for D_Precision, Size, Liability, Intangible, Growth, as described in Table 2, and Industry and Year Dummies. In Panel B, the dependent variable for the
second-stage models, reported below, is Tobin’s Q as described in Table 2. Other independent variables include Size, Liability, Growth, ROA, one risk measure
(i.e., Idiosyncratic Risk, Stock Return Volatility, Beta, or Average Spread) and Industry and Year Dummies. All these variables are described in Table 2. The
coefficients on Industry and Year dummies are not reported in this table. For each regression the first row is the coefficient estimates, the second row is the
standard errors, and the third row is the t-statistics. The R2 is the MacFadden’s (pseudo) R2 which indicates the degree to which the model parameters improve
upon the prediction of the null model, i.e., a model predicting the dependent variable without any independent variables.
Panel A: The Effect of Management Earnings Forecasts Precision on Firm Risks
Inverse
Intercept
Size
Liability
Mills
1.3420
-0.3428
-0.0901
-0.0894
Idiosyncratic Risk
0.0542
0.0853
0.0019
0.0108
Standard Error
24.78
-4.02
-46.39
-8.32
T-statistic
Intangible
Growth
R2
-0.0708
0.0179
-3.96
0.0625
0.0036
17.37
0.45
Stock Return Volatility
Standard Error
T-statistic
1.3560
0.0565
24.02
-0.4532
0.0890
-5.10
-0.0828
0.0020
-40.91
-0.0970
0.0112
-8.66
-0.0761
0.0186
-4.08
0.0631
0.0038
16.81
0.44
Beta
Standard Error
T-statistic
1.6034
0.1164
13.78
-1.9717
0.1833
-10.75
0.0503
0.0042
12.06
-0.1748
0.0231
-7.57
-0.0988
0.0384
-2.57
0.0652
0.0077
8.43
0.31
Average Spread
Standard Error
T-statistic
15.4609
0.3387
45.65
-0.1649
0.5336
-0.31
0.0677
0.0121
5.57
0.0939
0.0672
1.40
-0.1185
0.1118
-1.06
0.0344
0.0225
1.53
0.27
43
Panel B: The Effect of Management Earnings Forecasts Precision on Firm Valuation
Dependent variable: Tobin’s Q;
―Risk‖ independent variable:
No Risk Measure
Standard Error
T-statistic
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
Intercept
0.4712
0.1901
2.48
-0.5304
0.2288
-2.32
-0.4461
0.2291
-1.95
0.3298
0.2332
1.41
0.5225
0.2400
2.18
Inverse
Mills
2.1329
0.3339
6.39
3.0752
0.3615
8.51
3.0993
0.3625
8.55
1.7501
0.3697
4.73
1.4941
0.3680
4.06
Size
Liability
Growth
ROA
0.1229
0.0122
10.10
0.1891
0.0138
13.67
0.1765
0.0139
12.75
0.1557
0.0143
10.92
0.1684
0.0142
11.90
0.2533
0.0308
8.23
-0.2901
0.0447
-6.49
-0.2925
0.0448
-6.53
-0.4522
0.0456
-9.92
-0.4737
0.0455
-10.41
0.2420
0.0132
18.29
0.1945
0.0144
13.48
0.1981
0.0145
13.71
0.2422
0.0147
16.45
0.2471
0.0147
16.79
-0.4261
0.0281
-15.18
0.0717
0.0388
1.85
0.0668
0.0389
1.72
-0.2777
0.0389
-7.14
-0.3173
0.0385
-8.25
44
Risk
R2
0.30
0.9620
0.0241
39.89
0.8886
0.0233
38.18
0.0763
0.0114
6.73
-0.0078
0.0039
-2.00
0.43
0.43
0.42
0.42
Table 7
Relationship Between Management Earnings Forecasts Accuracy and Firm Risks and Valuation
This table reports the regression results of the effects of the accuracy of management earnings forecasts on firm risks and firm valuation. Panel A presents results
on firm risks, while Panel B presents results on firm valuation. Heckman treatment models are used to control for self-selection bias. The first stage model is
described in equation (4a), which is a probit model with D_Accuracy as the dependent variable. Specifically, D_Accuracy takes on a value of 1 if the absolute
value of the difference between the forecast earnings and the actual earnings is below the median 0.29, and 0 otherwise. For a range forecast, the median is
chosen as the forecast earning; for an open-ended forecast, the upper or lower bound is chosen as the forecast earning. The independent variables in equation (4a)
are Size, Loss, Analysts, Good_News, and Surprise, which are described in Table 2. In Panel A, the dependent variables are Idiosyncratic Risk, Stock Return
Volatility, Beta, and Average Spread as described in Table 2. The independent variables include the Inverse Mills Ratio for D_Accuracy, Size, Liability,
Intangible, Growth, as described in Table 2, and Industry and Year Dummies. In Panel B, the dependent variable for the second-stage models, reported below, is
Tobin’s Q as described in Table 2. Other independent variables include Size, Liability, Growth, ROA, one risk measure (i.e., Idiosyncratic Risk, Stock Return
Volatility, Beta, or Average Spread) and Industry and Year Dummies. All these variables are described in Table 2. The coefficients on Industry and Year
dummies are not reported in this table. For each regression the first row is the coefficient estimates, the second row is the standard errors, and the third row is the
t-statistics. The R2 is the MacFadden’s (pseudo) R2 which indicates the degree to which the model parameters improve upon the prediction of the null model, i.e.,
a model predicting the dependent variable without any independent variables.
Panel A: The Effect of Management Earnings Forecasts Accuracy on Firm Risks
Intercept
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
1.4706
0.0631
23.32
1.5096
0.0662
22.79
2.1346
0.1506
14.17
29.1036
0.5132
56.71
Inverse
Mills
-1.0402
0.0276
-37.76
-1.1664
0.0289
-40.31
-1.9974
0.0658
-30.35
-0.3512
0.2242
-1.57
Size
Liability
Intangible
Growth
R2
-0.0705
0.0018
-38.35
-0.0635
0.0019
-32.90
0.0247
0.0044
5.63
0.0925
0.0150
6.18
-0.1699
0.0113
-15.10
-0.1824
0.0118
-15.43
-0.2578
0.0269
-9.59
0.0487
0.0916
0.53
-0.1528
0.0169
-9.05
-0.1603
0.0177
-9.05
-0.2085
0.0403
-5.17
-0.1220
0.1373
-0.89
0.0798
0.0036
21.91
0.0791
0.0038
20.69
0.0566
0.0087
6.51
0.0577
0.0296
1.95
0.68
45
0.66
0.50
0.43
Panel B: The Effect of Management Earnings Forecast Accuracy on Firm Valuation
Dependent variable: Tobin’s Q;
―Risk‖ independent variable:
No Risk Measure
Standard Error
T-statistic
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
Intercept
0.0468
0.2500
0.19
-1.6123
0.2898
-5.56
-1.4876
0.2905
-5.12
-0.2083
0.2952
-0.71
0.1443
0.3116
0.46
Inverse
Mills
-0.5625
0.1234
-4.56
0.4450
0.1350
3.30
0.4591
0.1358
3.38
-0.4047
0.1377
-2.94
-0.5135
0.1358
-3.78
Size
Liability
Growth
ROA
0.2528
0.0079
31.96
0.3507
0.0087
40.26
0.3362
0.0087
38.75
0.2784
0.0087
31.92
0.2813
0.0087
32.27
-0.7133
0.0467
-15.28
-0.7235
0.0535
-13.54
-0.7358
0.0536
-13.72
-0.9780
0.0544
-17.97
-0.9971
0.0543
-18.37
0.2991
0.0155
19.36
0.2219
0.0168
13.21
0.2308
0.0168
13.71
0.2987
0.0171
17.47
0.3022
0.0171
17.69
-0.3486
0.0384
-9.08
-0.0580
0.0451
-1.29
-0.0699
0.0453
-1.54
-0.3836
0.0454
-8.43
-0.4087
0.0451
-9.05
46
Risk
R2
0.50
1.1192
0.0311
36.02
1.0069
0.0298
33.82
0.0615
0.0133
4.61
-0.0089
0.0040
-2.26
0.57
0.57
0.57
0.56
Table 8
Relationship Between Management Earnings Forecasts Reputation and Firm Risks and Valuation
This table reports the regression results of the effects of the reputation of management earnings forecasts on firm risks and firm valuation. Heckman treatment
models are used to control for self-selection bias. The first stage model is described in equation (4a), which is a probit model with D_Reputation as the dependent
variable. Specifically, D_Reputation takes on the value of 1 if a firm issued point forecasts (D_Precision=1) and more accurate forecasts (D_Accuracy=1) and
more frequent forecasts (D_Frequency=1) in a given year. The independent variables in equation (4a) are Size, Loss, Analysts, Good_News, and Surprise, which
are described in Table 2. In Panel A, the dependent variables are Idiosyncratic Risk, Stock Return Volatility, Beta, and Average Spread as described in Table 2.
The independent variables include Inverse Mills Ratio for D_Reputation, Size, Liability, Intangible, Growth, as described in Table 2, and Industry and Year
Dummies. In Panel B, the dependent variable for the second-stage models, reported below, is Tobin’s Q as described in Table 2. Other independent variables
include Size, Liability, Growth, ROA, one risk measure (i.e., Idiosyncratic Risk, Stock Return Volatility, Beta, or Average Spread) and Industry and Year
Dummies. All these variables are described in Table 2. The coefficients on Industry and Year dummies are not reported in this table. For each regression the first
row is the coefficient estimates, the second row is the standard errors, and the third row is the t-statistics. The R2 is the MacFadden’s (pseudo) R2 which indicates
the degree to which the model parameters improve upon the prediction of the null model, i.e., a model predicting the dependent variable without any independent
variables.
Panel A: The Effect of Management Earnings Forecasts Reputation on Firm Risks
Intercept
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
1.2551
0.0537
23.36
1.2678
0.0560
22.64
1.5591
0.1158
13.46
15.3480
0.3378
45.44
Inverse
Mills
-4.0099
0.2149
-18.66
-4.3804
0.2240
-19.56
-7.3853
0.4633
-15.94
-4.3767
1.3512
-3.24
Size
Liability
Intangible
Growth
R2
-0.0512
0.0028
-18.45
-0.0416
0.0029
-14.36
0.1021
0.0060
17.07
0.1132
0.0175
6.49
-0.1155
0.0108
-10.72
-0.1258
0.0112
-11.20
-0.2270
0.0232
-9.78
0.0661
0.0677
0.98
-0.0883
0.0178
-4.96
-0.0945
0.0186
-5.09
-0.1182
0.0384
-3.08
-0.1396
0.1119
-1.25
0.0628
0.0036
17.56
0.0635
0.0037
17.03
0.0676
0.0077
8.76
0.0345
0.0225
1.53
0.46
47
0.44
0.32
0.28
Panel B: The Effect of Management Earnings Forecasts Reputation on Firm Valuation
Dependent variable: Tobin’s Q;
―Risk‖ independent variable:
No Risk Measure
Standard Error
T-statistic
Idiosyncratic Risk
Standard Error
T-statistic
Stock Return Volatility
Standard Error
T-statistic
Beta
Standard Error
T-statistic
Average Spread
Standard Error
T-statistic
Intercept
0.2929
0.1851
1.58
-0.8574
0.2229
-3.85
-0.7758
0.2232
-3.48
0.1606
0.2267
0.71
0.3779
0.2334
1.62
Inverse
Mills
5.1764
0.8285
6.25
6.1590
0.8838
6.97
6.1816
0.8857
6.98
4.3048
0.9038
4.76
3.9313
0.9030
4.35
Size
Liability
Growth
ROA
0.1381
0.0103
13.43
0.2285
0.0113
20.25
0.2166
0.0113
19.19
0.1702
0.0115
14.78
0.1780
0.0115
15.50
0.2753
0.0307
8.97
-0.2549
0.0449
-5.68
-0.2573
0.0450
-5.72
-0.4278
0.0458
-9.34
-0.4511
0.0457
-9.87
0.2410
0.0132
18.21
0.1933
0.0144
13.39
0.1969
0.0145
13.62
0.2413
0.0147
16.39
0.2462
0.0147
16.74
-0.4279
0.0281
-15.22
0.0748
0.0389
1.92
0.0696
0.0391
1.78
-0.2804
0.0390
-7.19
-0.3204
0.0385
-8.32
48
Risk
R2
0.30
0.9500
0.0240
39.54
0.0863
0.0232
37.81
0.0739
0.0113
6.54
-0.0076
0.0039
-1.93
0.43
0.43
0.42
0.42