Asimultaneous examination of the effectes of reporting incentives of earning properties

A Simultaneous Examination of the Effects of Reporting Incentives on Earnings Properties Jana Raedy The University of North Carolina at Chapel Hill Wendy Wilson Southern Methodist Univer

A Simultaneous Examination of the Effects of Reporting Incentives on Earnings Properties

Jana Raedy The University of North Carolina at Chapel Hill

Wendy Wilson Southern Methodist University

March 2012

Abstract

This paper examines how multiple reporting incentives affect the ways that managers utilize discretion when reporting earnings In particular, this study directly addresses two financial reporting trade-offs that have received limited attention from the accounting literature: multiple reporting incentives and the endogenous determination of earnings characteristics Using a two-stage least squares regression model, we provide a comprehensive analysis of the effects of reporting incentives arising from equity-based compensation, debtholders, equity stakeholders, tax, and labor unions on three different earnings characteristics in order to understand the implications of modeling determinants of accounting choice in a multi-incentive context We find that after considering the influence of multiple reporting incentives and the endogeneity inherent in the earnings process, the importance of individual reporting incentives varies

considerably across different earnings characteristics, and is not always consistent with findings

in prior literature

JEL classification: M41

Keywords: reporting incentives; earnings management; discretionary accounting choices

We thank Doug Hanna and seminar participants at Southern Methodist University for useful comments We also thank Mark Lang, Zining Li, Darius Miller, Karen Nelson, Tom Omer, Suzanne Paquette, Wayne Shaw, Greg Sommers, Johan Sulaeman, Michael Van Breda, Ram Venkataraman, and seminar participants at the 2009 Lone Star Accounting Conference for helpful comments on a previous version of this paper We thank Justin Hopkins for excellent

research assistance

interesting question of the interactions between, and trade-offs among, incentives (Fields et al

2001, 290) In this study, we draw on evidence from the literature on accounting choice that suggests that numerous reporting incentives affect characteristics of reported earnings

simultaneously to provide a comprehensive analysis of how multiple incentives impact

characteristics of reported earnings

In general, papers analyzing the determinants of accounting choice focus on the relation between one incentive and one variable that measures a form of accounting discretion Under this approach, a conclusion is drawn that the incentive has an influence on the particular earnings characteristic if statistical significance between the two variables is found However, this

empirical design can be problematic for two reasons First, it focuses on the effect of a single reporting incentive on reported earnings, which is not reasonable in the context of multiple reporting incentives Some studies (e.g., Beatty et al 2006) enhance the analysis of the primary incentive of interest by including control variables to account for one (or more) other plausible reporting incentives (i.e., those that could influence reported earnings in addition to the primary incentive) However, thorough consideration of the numerous incentives documented to affect characteristics of reported earnings has not been consistently applied in this area of research

Consider research on incentives to report earnings that meet or beat analyst forecast targets as one example Evidence from this research shows that equity-based compensation incentives provide motivation for managers to report earnings that achieve forecast targets (Cheng and Warfield 2005, McVay et al 2006) In addition, evidence from studies in this area shows that managers have incentives to beat forecasts as a result of pressure from equity

shareholders (Graham et al 2005), and that managers face disincentives to report earnings that achieve forecast targets when labor union demands are strong (Bova 2012) If the three

incentives for reporting earnings that beat (or do not beat) earnings forecasts are not modeled jointly, it is not clear which incentives dominate or whether a significant relation can be properly accredited to the modeled incentive(s) rather than to one(s) that is excluded from the model As Schipper and Vincent (2003) note, the existence of multiple, and often incongruent, incentives complicates interpretation of the results of research on accounting choice because it is not clear which incentives dominate

The second reason the “one incentive to one earnings characteristic” research design is problematic is because it assumes that when managers have incentives to report earnings with a particular attribute other impending goals with respect to reported earnings are fixed However, this is not accurate when reporting one earnings characteristic affects the likelihood that a

secondary earnings characteristic will be realized For example, evidence from the literature suggests that equity-based compensation incentives affect reporting earnings that achieve

forecast targets (Cheng and Warfield 2005, McVay et al 2006) and reporting smooth earnings (Cheng and Warfield 2005) However, managers’ use of discretion to report earnings that beat the forecast target will also affect whether reported earnings are smooth Empirical models of the determinants of earnings characteristics that treat earnings properties as exogenous are

misspecified if decisions with respect to accounting discretion contemporaneously affect

reported earnings

For the most part, accounting research has not made widespread attempts to address how firms use accounting discretion in the presence of multiple objectives and how endogenous earnings characteristics collectively affect properties of reported earnings To this point, Dechow

et al (2010, 390) highlight the need for accounting research to consider these reporting trade-offs faced by managers by making a call for future research to (1) examine how managers choose between competing objectives and (2) examine choices about the portfolio of accounting

policies, specifically within the context of meeting multiple objectives The most thorough studies of the impact of several reporting incentives on earnings have been conducted in

restricted settings with specialized incentive structures.1 Beatty et al (1995) use data from

commercial banks to investigate how firms manipulate various bank loan-related accounts in order to achieve regulatory capital, tax, and earnings objectives, and Hunt et al (1996) examine how firms use several inventory-related accruals to manage accounting under LIFO to meet conflicting tax and financial reporting objectives.2 While these studies reinforce the importance

of considering the multiple reporting trade-offs faced by managers, inferences from these studies are not applicable to the broader settings most commonly examined by research on the effects of reporting incentives due to the lack of generalizability

Based on the framework for the usefulness of accounting information provided by Fields

et al (2001), we analyze the effects of five specific stakeholders that have vested interests in

1

The joint effect of multiple incentives on managers’ decisions has also been the subject of research that is

indirectly related to reported earnings For example, Collins et al (1997) examine the joint effects of tax, regulatory, and other incentives on capital gain realizations by life insurers

2

Qiang (2007) examines the significance of multiple explanations for accounting conservatism, which is similar to the approach in our paper of considering the existence of multiple reporting incentives simultaneously However, Qiang (2007) does not incorporate the endogenous nature of the earnings process or other accounting properties beyond evidence of conservatism

reported earnings: managers with equity-based compensation and debtholders (for contracting purposes), equity shareholders (for asset pricing purposes), and tax authorities and labor unions (to influence information prepared for external parties) We utilize a panel dataset of over 3,000 firm-year observations comprised of data from the years 2000 to 2008 to analyze the relation between variables representing these reporting incentives and three different earnings properties that exhibit evidence of discretionary reporting: (1) meeting or beating analysts’ forecasts, (2) reporting smooth earnings, and (3) timely recognition of losses Our analysis incorporates

multiple conjectures with respect to how specific reporting incentives influence characteristics of reported earnings Empirically, we use two-stage least squares (2SLS) regression models to take into account the fact that accounting decisions interactively affect earnings characteristics

We find that after considering the influence of multiple simultaneous reporting incentives and the endogeneity inherent in the earnings process, the importance of individual incentives varies considerably across different earnings characteristics The results indicate dominance of one reporting incentive over others with respect to achieving forecast targets and reporting smooth earnings In particular, the evidence shows that incentives from equity-based

compensation are stronger than incentives from equity shareholders and labor effects when discretion is used to achieve analyst forecast targets In addition, the evidence shows that

reporting incentives from equity shareholders have a significant effect on reporting smooth earnings relative to insignificant effects from equity-based compensation and labor union

incentives Evidence from our test of the determinants of timely loss recognition reveals a

different pattern in that we find that multiple stakeholder incentives have a significant influence

In particular, we find that timely loss recognition is increasing in the importance of debtholders

and labor unions, which is consistent with the contracting explanation for conservatism (Watts 2003)

We conduct additional tests to gain an understanding of whether consideration of

multiple incentives or endogeneity affects the inferences from our analyses When the

endogenous nature of the earnings process is relaxed and the earnings characteristics of

achieving forecast targets and reporting smooth earnings are modeled in an OLS (or logit, in the case of forecast targets) regression on multiple incentives, we find significance for additional reporting incentives These results are notable because individual incentives that have been found to be significant determinants of earnings characteristics in one-to-one analyses (e.g., labor union incentives affect the propensity of managers to report earnings that achieve forecast

targets) are not found to have a significant effect once we control for the endogeneity of the earnings characteristics

Our paper makes several contributions to the literature on incentives for accounting choice First, our study makes a comprehensive attempt to understand how managers balance financial reporting trade-offs in the presence of multiple incentives We find that taking account

of multiple reporting motivations, as well as the endogenous nature of the earnings process, leads

to different inferences than have been obtained in prior research when the reporting trade-offs faced by managers have been not been modeled comprehensively Understanding management’s incentives is important in understanding the ways in which managers choose to engage in

earnings management (Dechow and Skinner, 2000), and our consideration of the effects of multiple motivations in a simultaneous framework fills a noted gap in the literature (Fields et al

2001, Dechow et al 2010)

Second, results from our study provide implications for related research on incentives for earnings management Evidence from this study demonstrates that consideration of multiple incentives in a simultaneous framework has implications for understanding the significance of determinants of earnings characteristics that are indicative of accounting discretion Collectively, the results from this study show that without studying the comprehensive effects of multiple reporting incentives shown to be significant by existing research, our understanding of the

significance of incentives for reporting earnings with particular earnings properties, and thus the potentially conflicting trade-offs involved in discussions related to earnings management,

II Reporting Incentives and Earnings Properties

Our study is based on the premise that multiple reporting incentives influence

management’s use of discretionary accounting choices Therefore, we identify variables that represent incentives that are likely to influence managers in their financial reporting choices as well as variables that represent various earnings properties Because a large number of variables fall within the broad categories of ‘reporting incentives’ and ‘earnings properties’, we reference comprehensive review papers on accounting choice and earnings quality for assistance in

identifying the particular measures that are best-suited for our analysis

We derive our categorization of reporting incentives from Fields et al (2001), where the determinants of accounting choice arise from the effects of reported earnings on (1) contracts, (2)

asset prices, and (3) information prepared for external parties As described in their review paper, Fields et al (2001) note that in the absence of complete and perfect markets (Modigliani and Miller 1958) accounting choices have decision-making usefulness for information users, which results in accounting information being useful for contracting purposes (to overcome agency costs), for asset pricing (to overcome information asymmetries), and in order to influence

external parties (by imposing positive externalities)

Using these three determinants of accounting choice as a framework, Fields et al (2001) further identify specific stakeholders within each area to whom reporting incentives should matter Fields et al (2001) suggest that the reporting preferences of the following parties affect accounting choice: managers with equity-based compensation and debtholders (for contracting purposes), equity shareholders (for asset pricing purposes), and tax authorities and labor unions (to influence external parties) We incorporate these five stakeholder interests in our paper and examine whether their reporting incentives are determinants of the ways in which managers utilize discretion to report earnings.3

We derive our categorization of earnings properties from the proxies for earnings quality identified by Dechow et al (2010) in their review of the earnings quality literature Dechow et al (2010) summarize the large number of variables used in accounting research to represent the notion of earnings quality, which they admit is a somewhat elusively-defined construct that is not consistently applied in the literature As we are not interested in ‘earnings quality’ per se, but rather are focused on earnings outcomes that are more indicative of discretionary accounting

3 Accounting information is also useful for regulatory purposes and evidence from accounting research shows that regulatory incentives influence discretionary accounting choice in specific contexts (e.g., Beatty et al 1995) We do not incorporate regulatory incentives in our framework because the types of firms subject to regulatory scrutiny (e.g., financial services firms, banks, utilities) have been intentionally omitted from our sample because of their unique reporting incentives In addition, as noted by Armstrong et al (2010), it is likely that other groups have an interest in a firm’s financial reports beyond the five stakeholders that we consider in our paper (e.g., customers, suppliers, etc.) We incorporate these five stakeholders in order to follow the theoretical development of Fields et al (2001), for comparison with inferences from prior research, and because of data availability

choices, we narrow the attributes identified by Dechow et al (2010) to three earnings properties that suggest the use of accounting discretion: meeting analysts’ forecasts, smooth earnings, and timely loss recognition.4 While no empirical proxy for discretion is perfect because

management’s choices are not directly observed, each of these earnings characteristics represents

a reporting outcome that incorporates management’s use of accounting judgment

Our analysis incorporates the fact that managers are faced with two distinct trade-offs when reporting earnings First, trade-offs exist to the extent that multiple, and potentially

conflicting, incentives simultaneously dictate preferences for different earnings characteristics Overall, managers must evaluate the importance of reporting to provide value to traditional investor stakeholders (e.g., equity investors, debtholders) relative to other investor stakeholders (e.g., managers with equity-based compensation) versus reporting to satisfy noninvesting

stakeholders (e.g., labor union members, tax authorities) Considering the simple example of the incentives of two stakeholder parties (equity shareholders and labor union members), Tirole (2001) describes that when investors have greater influence managers place higher priority on emphasizing capital value, and when labor has greater influence managers place higher priority over employee welfare relative to equity value enhancement To the extent that reporting

preferences of the two parties are different, managers face trade-offs in considering the

dominance of one versus the other Incorporating the effects of more than two stakeholder

groups further complicates consideration of trade-offs arising from incentives for financial reporting

Notwithstanding potentially conflicting incentives, trade-offs also exist when managers must evaluate how to use discretion from a technical perspective since achieving one earnings objective affects the ability to report earnings with an alternative desirable characteristic For example, due to the endogenous nature of the earnings process, it would likely be impossible for

a manager to report smooth earnings, earnings which beat analysts’ forecasts, and earnings that incorporate losses on a timely basis at the same time When faced with the option of how to utilize accounting discretion, managers must prioritize the effect of such discretion on desirable earnings characteristics, knowing that the decision affects other earnings characteristics as well

We describe the incentives for financial reporting that are expected to affect each

earnings characteristic below In particular, we rely on predicted relations between incentives and earnings characteristics as prescribed by existing theory and/or evidence from empirical research A summary of the predictions is reported in Table 1.5

Forecast targets

Accounting research interprets earnings that either just meet or exceed forecasted

earnings by a small amount as potentially having been manipulated (Degeorge et al 1999) Matsumoto (2002), Ayers et al (2006) and Burgstahler and Eames (2006) provide evidence that firms utilize either earnings management, expectations management, or a combination of the two

to meet or beat forecast targets While this evidence supports the idea that achieving forecast targets is a documented goal of financial reporting, several different motivations have been brought forth to explain target-achieving behavior

5

We do not make conjectures about the association between certain pairs of reporting incentives and earnings characteristics when there is a lack of support for a systematic relation from either theoretical or empirical research Thus, we directly examine fewer relations than the number of possible combinations of reporting incentives/earnings properties within our research framework

Based on evidence that equity markets reward firms that report earnings which meet or beat current period expectations (Bartov et al 2002, Kasznik and McNichols 2002), motivations for earnings management to achieve forecasts arise for parties that will benefit from the wealth effect of rising equity prices Consistent with equity-based compensation providing such an incentive, empirical evidence from Cheng and Warfield (2005) and McVay et al (2006) shows that management’s equity-based compensation incentives are positively associated with just meeting or beating analyst forecasts.6 In addition, survey evidence reported by Graham et al (2005, 27) found that an overwhelming majority of CFOs claimed capital market pressures from equity shareholders were the number one reason to achieve earnings benchmarks On the other

hand, labor unions provide an incentive for managers not to beat forecast targets Bova (2012)

finds that unionized firms have a tendency to barely miss analysts’ forecasts, sending a negative signal of the firm’s performance which is useful in lowering the firm’s cost of labor

6 Compensation-related penalties to missing forecasts also serve as an incentive to meet or beat targets

Mergenthaler et al (2011) find that executives receive lower equity compensation and face a greater likelihood of forced turnover after missing multiple quarterly earnings benchmarks

7

We are not interested in commenting on debates in existing research as to whether certain reporting characteristics enhance or diminish the informativeness of earnings (e.g., the informativeness of reporting smooth earnings as examined by Dichev and Tang (2009) and Leuz et al (2003)) The purpose of our paper is to examine the relation between particular characteristics of earnings and reporting incentives provided by a firm’s stakeholders, regardless

of whether a managers use discretion opportunistically or in an informative manner Inferences from our analysis regarding the trade-offs arising from multiple incentives are valid irrespective of the opportunistic or informative intent of the manager’s use of discretion

demonstrates that smoothness is a desirable earnings property Empirical evidence that smooth earnings are associated with a lower cost of equity (Francis et al 2004) also supports a

preference by equity shareholders for smooth earnings

In addition to incentives from equity shareholders, labor unions and equity-based

compensation should also motivate the use of discretion to report smooth earnings With respect

to labor unions, Moses (1987) hypothesizes that financial reports that contain patterns of smooth earnings circumvent attention from uneven and extraordinarily high profits and allow firms to avoid greater compensation demands from union members In addition, based on the idea that managers have compensation-based motivations to reserve excess earnings for future periods (through the effect of reported earnings on stock price), Cheng and Warfield (2005) find that managers with higher levels of equity-based compensation report smoother earnings

Timely loss recognition

Another attribute of financial reporting is the extent to which earnings are reported conservatively, where more timely recognition of losses relative to gains is interpreted as a characteristic of more conservative and higher-quality earnings (Basu 1997, Pope and Walker 1999) While some degree of timely loss recognition results from the application of U.S GAAP, evidence also suggests that management can utilize discretion to incorporate conservative

accounting decisions to facilitate the manipulation of earnings (Jackson and Liu 2010)

Several motivations for conservative reporting have been suggested by the literature, most notably the contracting explanation (Watts 2003) that describes the role of conservative accounting to address moral hazard problems due to asymmetric information or asymmetric payoffs in the firm’s contracts Accordingly, conservatism should be increasing in debt

contracting since more stringent verification of gains relative to losses provides important

information for debtholders in light of the their asymmetric payoff function with respect to the firm’s net assets (Watts 2003, Beatty et al 2006) With respect to contracts with the firm’s employees, Leung et al (2009) hypothesize that labor union contracts generate conservatism on both the demand and supply sides since asymmetric recognition of losses protects union

members from downside risk and also protects management from labor unions’ rent-seeking demands Leung et al (2009) find that more heavily-unionized firms report losses more timely, suggesting a positive relation

In addition to the contracting explanation, the link between accounting earnings and taxable earnings suggests that firms have an incentive to accelerate the reporting of losses to minimize the present value of tax obligations (Shackelford and Shevlin 2001, Watts 2003) Furthermore, evidence from several studies demonstrates that public equity firms report more conservatively relative to firms with private equity (Ball and Shivakumar 2005, Givoly et al 2010), suggesting a positive relation between the equity shareholder reporting preferences and more timely loss recognition.8

<< Insert Table 1 here >>

III Research design

The objective of our paper is to examine the relation between multiple reporting

incentives and earnings characteristics that are indicative of discretionary accounting Because of the interactive effect of accounting discretion on multiple characteristics of earnings (i.e., when managers utilize discretion to achieve one reporting goal other properties of earnings are also

8 LaFond and Watts (2008) find that timely loss recognition is increasing in a firm’s level of information asymmetry, where information asymmetry is measured by the probability of informed trading (PIN) and bid/ask spread While these measures of information asymmetry between firm outsiders and insiders are different from our equity

construct that is based on the level of widespread common share ownership, to the extent that PIN and bid/ask spread and diffusion in equity ownership are correlated as measures of information asymmetry, the evidence from LaFond and Watts (2008) also supports a positive association with timely loss recognition

affected), we utilize a two-stage least squares (2SLS) approach that incorporates the endogenous nature of the earnings characteristics (Maddala and Lahiri, 2009).9 Specifically, we identify the three earnings characteristics as jointly endogenous variables and the five reporting incentives as exogenous variables The three first-stage regressions take a similar form in that each of the earnings measures is regressed on all of the reporting incentives and relevant control variables The first-stage regressions are as follows:10

EARN_MEASURE = a1 + a2INCENT_RATIO + a3DEBT + a4EQUITY + a5TAX +

a6LABOR + ∑ aj CONTROLS + e (1)

The dependent variable in each of the first-stage regressions (EARN_MEASURE) takes the value of one of the earnings characteristics that are indicative of discretionary reporting Consistent with prior research (Cheng and Warfield 2005), achieving analyst earnings targets by

a small amount (POSFE), is an indicator variable equal to one if the difference between actual earnings and the median analyst forecast, on a per share basis, is between 0 and 0.01, and zero otherwise.11 The second construct (CF_ACC) represents smooth earnings, measured as the Pearson correlation between operating cash flows and operating accruals over a five-year rolling window, where cash flows and accruals are scaled by total assets (Leuz et al 2003) Finally, the third earnings property (C_SCORE) measures the extent of timely loss recognition inherent in earnings Following the methodology of Khan and Watts (2009), C_SCORE is a firm-year

timely loss coefficient based on a modified version of the Basu (1997) model measured on a

9 Other contexts in which 2SLS regression models have been used to incorporate joint decision effects include analyst coverage and investor decisions to invest in firms (O’Brien and Bhushan 1990), the relationship between institutional ownership and a firm’s auditor choice (Velury et al 2003), analyst following and stock return

synchronicity (Chan and Hameed 2006), and loan covenants and accounting conservatism (Beatty et al 2006)

10 Each observation is a firm-year measure We omit the firm and year subscripts for simplicity

11 Actual earnings and the median forecast are collected from the I/B/E/S adjusted annual summary file

cross-sectional basis by year All three of the earnings properties are calculated such that a higher value of the variable is indicative of a greater degree of discretionary reporting

Reporting incentives are represented by INCENT_RATIO, DEBT, EQUITY, TAX, and LABOR in equation 1 Equity-related compensation incentives are measured by INCENT_RATIO, which represents the share of a CEO’s total compensation that would result from a one percentage point increase in the equity value of the firm (Core and Guay 2002), and is commonly used as a measure of equity-based compensation incentives (Bergstresser and Philippon 2006) Debtholder incentives (DEBT) are defined as the ratio of the sum of notes and bonds payable to the sum of the book value of liabilities and the market value of equity Equity shareholders are represented by EQUITY, defined as the ratio of the market value of common shares which are not closely held to the sum of the book value of liabilities and the market value of equity Tax incentives (TAX) are measured by the simulated marginal tax rate as estimated by Graham (1996), where the importance

of tax-related incentives is increasing in the marginal tax rate Finally, we follow Hilary (2006) and construct a variable to measure the power of labor unions (LABOR), which represents the product

of the number of employees (a firm-level measure of labor intensity) and the industry-level

unionization rate.12

We include a comprehensive list of control variables in equation 1 to account for specific characteristics that are likely to affect earnings properties such as size, growth (Cheng and Warfield 2005), and litigation risk (Frankel et al 2002) We include control variables related

firm-to operational performance, such as return-on-assets (DeFond and Park 1997), sales variability (Berstresser and Phillipon 2006), and an indicator variable for net losses (Brown 2001) We also

12 We use a combination of firm and industry-level data to calculate the labor incentives variable because we do not have access to a firm-specific representation of labor union intensity for our sample As noted by related research on the effects of labor unionization (e.g., Klasa et al 2009), it is difficult to collect reliable data on firm-level

unionization rates because firms are not required to provide this information in SEC filings Therefore, industry unionization rates are accepted by the literature as a reasonable proxy for firm-level unionization

incorporate control variables related to the firm’s information environment (Frankel et al 2002), such as the level of analyst coverage, and analyst forecast dispersion Finally, we include the percentage of outside board members as a corporate governance control variable (Klein 2002, Larker et al 2007, Bowen et al 2008) The appendix provides a detailed description of each control variable used in this analysis

In the second stage, we incorporate the fitted values of the earnings properties from the first-stage regressions as instrumental variables Each of the second-stage regression models takes the following general form, where each of the non-fitted earnings measures is regressed on the relevant reporting incentives, fitted earnings properties, and control variables:

EARN_MEASURE = a1 + ∑ aj INCENTIVES + ∑ ak FITTED_EARN_MEASURE +

The variables of interest from equation 2 are the coefficients on the reporting incentives (i.e., INCENTIVES) and are represented by the following: INCENT_RATIO, DEBT, EQUITY, TAX, and LABOR These variables are described above The reporting incentives and control variables are customized for each of the second-stage regressions such that only the variables predicted to be associated with the particular earnings property are included The independent variables used in each individual regression are discussed in more detail in section V and

complete specifications of the second-stage regressions are presented in tables 4 through 6 The fitted values of each earnings characteristic (FITTED_EARN_MEASURE) from the first-stage regression are included in the model along with variables used to control for other determinants

of earnings characteristics The control variables are listed in the appendix

We note the potential existence of multiple incentives for each firm/year observation For example, it is possible that firm characteristics might indicate a strong level of management

equity-based compensation incentives and also a high level of labor union importance In such a case, results from prior research predict the firm would have increasing incentives to use

abnormal accruals arising from equity-based compensation (Cheng and Warfield 2005,

Bergstresser and Phillipon 2006), and also decreasing incentives to use abnormal accruals based

on labor union incentives (Liberty and Zimmerman 1986, Bowen et al 1995) Due to competing predictions from conflicting reporting preferences, it is an empirical question as to how managers utilize discretion to meet reporting objectives

After taking into consideration the simultaneity of the decision-making process

underlying reported earnings, if the reporting incentives of the firm’s stakeholder groups

influence management’s use of accounting discretion then the sign and significance levels of the coefficients on stakeholder variables should be consistent with our expectations However, if the trade-offs inherent in applying accounting discretion prohibit firms from effectively achieving specific earnings properties, then we may fail to detect significant relations for some or all of our predictions

IV Data and Sample

We begin collecting our sample with the broad dataset of firms that are covered by

Compustat between 2000 and 2008 We then restrict the usable observations to those with

necessary data for computation of the earnings characteristics, reporting incentives, and control variables We intersect our base Compustat dataset with Execucomp (data to measure equity-based compensation incentives), the Corporate Library (data to measure equity shareholder interests and the corporate governance control variable), and from the Union Membership and Coverage Database (data to compute labor union intensity) maintained by Hirsch and

Macpherson (Hirsch and Macpherson 2003).13 We also collect analyst forecasts, reported

earnings, and analyst coverage from I/B/E/S, and we calculate the marginal tax rate according to the methodology in Graham (1996).14 Data necessary for measurement of the regression

variables not mentioned above are collected from Compustat and CRSP After restricting our sample to firm-year observations with complete data for all variables, we have a panel dataset with 3,359 observations Table 2, panel A provides details regarding sample attrition

<< Insert Table 2 here >>

The number of observations by year (panel B) and industry (panel C) are reported in table

2 The number of sample observations is steadily increasing over time, which is reflective of growing coverage of firms by several of our data sources, most notably the Corporate Library The industry categorization in panel C follows Hirsch and Macpherson (2003), where the largest percentage of sample firms are represented by the durable and nondurable goods manufacturing, retail trade, and information sectors The industries with the highest average labor unionization rate are healthcare and social assistance, educational services, and transportation and

warehousing

Descriptive statistics on the variables used in our empirical analyses are provided in panel

A of table 3 The mean value of POSFE indicates that approximately 12.6 percent of annual observations show evidence of earnings beating analysts’ forecasts by a small amount The distribution of CF_ACC is skewed to the right in that the median value (0.655) is greater than the mean (0.468), and the descriptive statistics for C_SCORE are lower than those reported by Khan

13 The Union Membership and Coverage Database is available from http://unionstats.gsu.edu/ This data is widely used in labor economics, accounting, and finance research involving labor unions (e.g., Chen et al 2012, Leung et

al 2009, Klasa et al 2009)

14 Marginal tax rates are available for download at http://faculty.fuqua.duke.edu/~jgraham/taxform.html We thank John Graham for supplying this data

and Watts (2009) for their full sample.15 Analysis of the incentive measures reveals that the mean value of INCENT_RATIO (0.259) is similar to that reported by Berstresser and Phillipon (2006) and that the average ratio of notes and bonds payable to firm value is approximately 6.9% (mean DEBT = 0.069) Approximately 59% of common shares are widely held (mean value of

EQUITY = 0.591) and the average simulated marginal tax rate is 20.4% Panel B provides a correlation matrix of the incentive and earnings characteristic variables While the majority of correlations are significant, for the most part the correlations are not particularly large in

of the coefficients for the three stakeholder variables is significant in the predicted direction Consistent with our predictions, the propensity of firms to report earnings which meet or beat analysts’ forecasts by a small amount is increasing in management’s equity compensation

incentives (coefficient on INCENT_RATIO = 0.137) The importance of equity compensation incentives for achieving target benchmarks is consistent with findings by Cheng and Warfield

15 We calculate C_SCORE as described and then multiply the calculated value by negative one so that a larger value

of this variable is indicative of more use of discretion in reporting earnings While the C_SCORE values for our sample are on the lower end of the range reported by Khan and Watts (2009), the mean value of C_SCORE for our sample is consistent with that reported for larger and older firms (see table 7 of Khan and Watts 2009), which are the types of firms included in our sample

(2005) and McVay et al (2006), but the lack of significance on coefficients for the equity

shareholder and labor union reporting incentives are in contrast with expectations

With respect to equity importance, our expectations were based on indirect evidence included in Graham et al (2005) and Bartov et al (2002) rather than prior empirical literature that directly tested the association between equity importance and meeting analysts’ forecasts Thus, we find that although this conjecture seems reasonable, it does not hold when empirically tested in our analysis that incorporates multiple incentives and multiple earnings goals in a simultaneous model.16 The lack of significance for the coefficient on LABOR is in contrast with the findings in Bova (2012) In supplemental tests reported later in the paper, we investigate whether incorporation of simultaneous reporting incentives or the endogenous nature of the earnings process in our research design is affecting our results Results are mixed with respect to the coefficients on the control variables in the forecast target analysis reported in table 4

Consistent with prior research (Matsumoto 2002, Frankel et al 2002, Cheng and Warfield 2005),

we find that the coefficient on growth is significantly positive and the coefficient on forecast dispersion is significantly negative We also include control variables to account for the effects

of size and litigation risk (Matsumoto 2002), firm performance (Frankel et al 2002), negative earnings (Brown 2001), and analyst coverage (Cheng and Warfield 2005), though the

coefficients on these variables are not significant in explaining target-achieving behavior

<< Insert Table 4 here >>

Table 5 reports results for the determinants of smooth earnings The evidence reported in table 5 indicates that the coefficient for one of the three stakeholder preferences is significant in explaining the propensity to report smooth earnings In particular, the results show a positive relation between equity shareholder importance and smooth earnings (coefficient on EQUITY = 0.415), which is consistent with the idea that managers respond to pressure from equity

shareholders when reporting earnings (Graham et al 2005) Contrary to expectations, the results

do not indicate that equity-based compensation and labor incentives have a significant influence

on reporting smooth earnings In supplemental tests reported later in the paper, we investigate whether incorporation of simultaneous reporting incentives or the endogenous nature of the earnings process in our research design is affecting our results Several of the coefficients on the control variables are significant in the predicted direction, in particular that larger firms with greater operating performance report smoother earnings (Cheng and Warfield 2005) and firms with greater operating variability report less smooth earnings

<< Insert Table 5 here >>

Results for tests of the reporting incentives for timely recognition of losses are reported in table 6.17 The evidence shows that two of the four reporting incentives, namely debtholders and labor unions, are significant in explaining management’s propensity to record losses in a less-timely manner In particular, the results show that timely loss recognition is increasing in

debtholder incentives (coefficient on DEBT = -0.044) and is increasing in labor importance (coefficient on LABOR = -0.092) The positive relation between debt and labor is consistent with

17

The dependent variable, C_SCORE, is calculated such that a higher value indicates greater discretion over

reported earnings and thus represents less timely recognition of losses Therefore, inferences with respect to timely

loss recognition should be interpreted with the opposite sign relative to what is reported in table 6 For example, the negative coefficient on debt is indicative of timely loss recognition increasing in debt reporting incentives