Impression and earnings management : a collaborative tool to mislead analysts

(1)

Impression and earnings management

:

a collaborative tool to mislead analysts

Student: Gabriela Bocmaru Student number: 10860762 Date: 22nd_{of June, 2015} Word count: 16.486

Program: MSc Accountancy and Control, variant Control

Institution: Faculty of Economics and Business, University of Amsterdam Supervisor: prof. dr. Vincent O' Connell

(2)

2

Statement of originality

This document is written by Gabriela Bocmaru who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Abstract

Prior literature shows a consistent pattern in the use of self-serving bias in management’s explanations of corporate performance. However, there is no determined consensus whether managers use self-serving attributions as an impression management tool to mislead market users’ views of firm performance or for providing incremental, useful information to users of corporate disclosures. This study integrates elements from agency theory and social psychology to establish a connection between impression management and earnings management as elements of managerial opportunistic behavior. I expect that managers will consciously use impression management when they anticipate that earnings management will trigger undesirable consequences from market users. Using content analysis of 104 earnings press releases to detect self-serving bias in performance attributions, I found that income increasing accruals earnings management is positively related to the use of self-serving bias in attributional statements. Furthermore, I examined how analysts, as users of corporate narratives, are influenced by the presence of impression management in corporate disclosures in their forecasting activity. The results suggest that the forecast accuracy decreases for firms that use both impression management and earnings management. These findings should raise concerns among regulators and practitioners related to the susceptibility of analysts to impression management manipulation through unregulated corporate disclosures.

Keywords: impression management, self-serving bias, earnings management, managerial opportunistic behavior, performance explanations, analyst bias

(4)

4

The main purpose of corporate disclosures is to communicate private information to market users in order to reduce information asymmetry and agency conflicts between a firm’s managers and its external stakeholders. Financial disclosures use both numerical and narrative information to send credible signals of present and future firm value. The reason why corporate narratives became such a fruitful area for accounting research in the past few years is that fundamental numeric financial information explains only a small fraction of share price movements (Cenesizoglu and Timmermann, 2012). It is more and more acknowledged that the strategy used by managers to release narrative disclosures has a great impact on user’s perceptions of firm future performance. A growing body of literature is trying to integrate social theories, mainly from psychology into disclosure policy analysis, to get more insight into the strategic reasoning behind choosing a particular disclosure policy. Narrative disclosures, from the manager’s perspective, can have two purposes: either to provide incremental useful information to improve decision-making, or to spread biased information with the sole purpose of misleading investors (Merkl-Davies and Brennan, 2007). In the context of corporate reporting, impression management is the process in which managers select the information that is to be released to the public and present it in such a way that distorts the reader’s perceptions of corporate achievements (Neu et al., 1998).

The impression management hypothesis of corporate disclosures entails managers opportunistically taking advantage of information asymmetries because the market has a perceived weak efficiency. According to behavioral finance studies, market inefficiency occurs in conditions of informational uncertainty, thus users make decisions based on bounded rationality (Shefrin, 2002). This leads to heuristic-driven decisions, when people take into consideration not only the substance (financial numbers), but also the form (the way outcomes are framed or expressed), disregarding the fact whether the framing of outcomes is economically relevant or not (Merkl-Davies and Brennan, 2007). An example of impression management strategy in corporate narratives that accompany financial results is the use of self-serving attributions in performance explanations. Previous research gives almost unequivocal results regarding the presence of self-serving bias in past performance explanations (Baginski et al., 2000, 2004; Davis et al., 2012; Keusch et al., 2012, Kimbrough and Wang, 2014). Empirical evidence shows that managers tend to attribute positive organizational outcomes to permanent, internal factors and negative organizational outcomes to temporary, external

(6)

6 circumstances. In this way, they take credit for good performance and shift the blame away when the financial results are not according to prior expectations (Baginski et.al, 2000).

This paper aims to deepen and advance our understanding of the use of impression management in corporate narratives. It focuses on the association between earnings management and impression management in performance explanations in the management commentary that commonly accompany the quarterly earnings press releases. The motivation for this research question is twofold. First, there is extensive evidence of the prevalence of self-serving biased attributions in past performance explanations, however it is very difficult to determine in real market settings whether they are the result of impression management or represent useful, incremental information. Prior research suggests that social and psychological factors motivate the use of self-serving attributions. Accordingly, impression management arises from the anticipation of potential negative consequences of information release (Merkl-Davies and Brennan, 2011). Earnings management and ex-ante user uncertainty regarding earnings management represent such an “accountability predicament” (Aerts and Cheng, 2010, p.432). If detected, accruals earnings management brings considerable reputation and agency costs; this gives managers sufficient motives to engage in extensive explanations of past performance in which managers selectively manipulate the causality descriptions of performance attributions resulting in self-serving bias.

The second part of the paper addresses another important issue that arises from previous literature. The objective is to examine how analysts, as users of corporate narratives are influenced by the presence of impression management in corporate disclosures in their forecasting activity. In prior literature, there is little research on the perception of narrative disclosures by market users of corporate information (Merkl Davies and Brennan, 2011) and most of it focuses on investor response to impression management as users of corporate disclosures (Staw et al., 1983; Baginski et al., 2000, 2004; Lee et al., 2004). Analysts have very important inputs to capital allocations (Leuz and Verrecchia, 2000) and the functioning of the market, however very little is known about how analysts react to impression management in narrative disclosures, whether they are misled by manager’s’ self-serving attributions in their forecasting activity or not (Barton and Mercer, 2005; Merkl-Davies and Brennan, 2007). Previous literature has produced mixed results regarding the ability of analysts as sophisticated market users to process company released information. Therefore, this study focuses on analysts as market users of corporate information and answers the call by Ramnath et al. (2008) for more research on the reasons why analysts' forecasts reflect psychological biases.

(7)

7 In a more in depth analysis, I study the use of both impression management and earnings management as a collaborative tool to mislead analyst’s’ future earnings forecasts. Prior literature suggests that earnings management and impression management are both dimensions of opportunistic behavior. Because in theory, impression management produces fewer costs for managers and firms, it might be preferred to earnings management (Guillamon-Saorin and García Osma, 2010). On the other hand, self-serving attributions in management commentary can be used to provide a reasonable explanation for earnings results. Based on this argument, I hypothesize that used jointly, earnings management and impression management represent a more effective tool to conceal past performance and mislead analysts in their forecasting activity. To my knowledge, this is the first study that examines the effect of both types of opportunistic behavior on the analyst forecast characteristics: forecast dispersion and forecast accuracy.

Using a sample of earnings press releases for 52 randomly selected U.S. firms from 2013 and 2014, I use content analysis to detect self-serving attributions in the management commentary of past performance. I find that managers use impression management in their past performance explanations in association with income-increasing earnings management. The joint use of both types of opportunistic behavior makes it possible to report good performance by managing earnings and to avoid being detected by analysts by using impression management in order to provide a reasonable explanation for past performance. Furthermore, the results show that for firms that use both impression management and earnings management the forecast accuracy decreases. This means that analysts are susceptible to impression management manipulation in narrative disclosures.

My work makes several contributions to the literature. First, thiss study complements Aerts and Cheng (2011) and Aerts and Zhang (2013) by providing incremental evidence of the use of impression management in past performance explanations that help to mitigate user’s ex-ante uncertainty about earnings management. Specifically, the findings suggest that only upwards earnings management represent an accountability predicament for managers to use self-serving attributions in narratives that accompany quarterly earnings results. More important, I extend the work of Aerts and Cheng (2011); Aerts and Zhang (2013) and Keusch et al. (2012) by increasing our understanding of how market users react to these disclosures. This is of particular interest to both preparers and users of corporate narrative disclosures, because it shows that in conditions of information uncertainty, managers can use impression management that successfully accomplishes the goal of hiding the

(8)

8 inconsistency between the managerial private information regarding performance and the publicly disclosed information.

Second, this study adds to the impression management literature by taking a social psychology perspective together with the classical agency theory, thus responding to the call of Merkl-Davies and Brennan (2011) for more research on alternative theoretical perspectives of managerial self-serving bias. The results show that firms that engage in accruals earnings management have incentives to use impression management in a remedial mode, because of the anticipation of negative user reactions. Self-serving biased attributions forestall the negative impact of financial results disclosures by creating sound explanations of managers’ past actions and decisions.

Third, the study adds to earnings management literature by providing evidence that accruals earnings management needs a corroborating tool such as impression management to justify performance changes. The findings show that in a strong scrutiny environment such as the US, when the expected costs of income increasing earnings management are perceived to be high (Aerts et al., 2013), managers have strong incentives to use self-serving attributions in performance explanations to justify the reported results.

The remainder of this paper is organized as follows. First, the literature review and hypothesis development part explains the theoretical framework used to draw the connection between earnings management and impression management and the incentives that managers have to behave opportunistically. I present prior evidence of the importance of analysts as users of corporate disclosures and the effect of impression management and earnings management on forecast accuracy and forecast dispersion. This is succeeded by the description of the sample and the method used for content analysis of earnings press releases in section III and the empirical models to be tested. Section IV presents the results of the hypotheses testing, followed by additional analysis and a sensitivity test. All the regressions’ results are depicted in table 14. The final section summarizes and discusses the findings, presents the limitations of the paper and directions for future research.

2. Literature review and hypotheses development 2.1 Agency and attribution theory

Previous research employed diverse theoretical lenses to explain the reasons why managers provide performance explanations for their past results in corporate disclosures. I will use the conceptual

(9)

9 framework developed by Merkl-Davies and Brennan (2011) and build my hypotheses on the theories derived from economics (agency theory) and social and cognitive psychology (attribution theory).

Agency theory involves delegation of decision-making authority based on a contract between the two parties: owner – principal and manager – agent. Agency theory predicts that the relationship between the principal and the agent is always affected by conflicting interests. Jensen and Meckling (1986) argue that parties are utility maximizers, therefore the agent will not always act in the best interest of the principal and both will bear agency costs. In this context, the manager and the shareholder are regarded as rational and self-interested decision-makers. By using the utilitarian behavior concept, corporate reporting and investment decisions will depend on the cost-benefit tradeoff (Merkl-Davies and Brennan, 2011). Managers act in an environment in which remuneration and wealth depends on the good financial performance of the companies they manage (Rutherford, 2003), therefore they will engage in impression management strategies in corporate narratives to “disguise” the true performance. Similarly, investors’ act on the information provided by the managers, motivated by the expectations of future returns in the form of cash flows.

2.2 The relationship between earnings and impression management

Social psychology literature predicts the occurrence of self-serving bias when managers publicly report on their past performance as a response to the inquiry of the external users of information (Keusch et al., 2012). Four potential explanations for the presence of self-serving bias emerged: the impression management explanation, the cognitive bias explanation (hubris), the incremental information explanation, and retrospective sense-making (Merkl-Davies and Brennan, 2011).

The impression management strategy occurs in the presence of certain socio-psychological factors. Biased performance attributions are impression management techniques that arise from the anticipation of potential negative consequences of information disclosures. Managers selectively manipulate the descriptions of causality of performance attributions, by obfuscating failures (blame on external factors) and emphasizing successes (a result of internal efforts). The second explanation interprets biased performance attributions as egocentric bias or hubris, meaning that a cognitive dissonance exists between self-image and the evidence of performance (Merkl-Davies and Brennan, 2011). Overconfident or optimist managers have an unconscious desire to protect their work and self-image, they expect success rather than failure, therefore will make biased self-assessments (Miller and Ross, 1975). Keusch et al. (2012) argue that the attributional process is neither purely driven by

(10)

10 cognitive bias nor solely by impression management intentions, but rather contains substantial elements from both.

Baginski et al. (2000, 2004) adopt the incremental information explanation, they argue that if security price reactions are affected by the presence of attributions this is evidence of the usefulness of attributional statements for investors. Finally, attributions may not be biased if they represent retrospective sense making ex-post rationalization of decisions in order to give the impression of rational decision-making (Aerts, 2005).

In a real life setting, it is difficult to determine the reasons behind presenting self-serving attributions in corporate discretionary disclosures. My research will focus on one circumstance in which managers are believed to engage in making self-serving biased attributions that reflect opportunistic behavior. I expect that managers will consciously use impression management when they anticipate that their actions will trigger undesirable punishments from market users. Aerts and Zhang (2014) argue that earnings management represents an “accountability predicament”, which motivates managers to offer more causal performance explanations in order to signal a rational and reasoned management.

Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company, or to influence contractual outcomes that depend on reported accounting numbers (Healy and Wahlen, 1999, p.368). Earnings management is similar to impression management because it represents an inconsistency between the managerial private information regarding performance and the publicly disclosed information.

Prior literature suggests numerous reasons for managers to use earnings management, such as to meet or beat earnings benchmarks (Cheng and Warfield, 2005), to avoid debt-covenant violation (Dichev and Skinner, 2002) or for income smoothing purposes (Grant et al., 2009) to effortlessly meet future targets. However, earnings management does have negative consequences for managers and the firm, such as litigation costs (Jones and Wu, 2010), increased cost of debt and/or capital (Dechow et al., 2010) and reputational costs (Desai et al., 2006). Previous literature shows that managers can use both accrual and/or real earnings management to manipulate performance perceptions (Badertscher, 2011). While accrual-based earnings management makes use of the flexibility in accounting methods or estimates within the boundaries of generally accepted accounting

(11)

11 principles (Dechow and Skinner, 2000), real earnings management strategies involves interfering in the execution of certain business transactions (Roychowdhury, 2006). Real earnings management is considered more costly for the firm because of the potential long-term economic consequences, but it has the advantage of being more difficult to detect by outsiders compared to accrual earnings-management (Graham et al., 2005; Gunny, 2010; Badertscher, 2011).

Managers anticipate that if detected, earnings management will negatively affect both their reputation and the stock price (Hunton et al., 2006; Libby and Seybert, 2009). They will try to decrease informational ambiguity by either switching to real earnings management activities with lower detection risk (Cohen et al., 2008), by temporary increasing the volume of disclosures (Jo and Kim, 2007) or by using graphical impression management (Godfrey et al., 2003). Moreover, Louis and Robinson (2005) argue that when managers use accruals to signal private information, they need a second backing up signal to support the credibility of their earnings management message. In this respect, managers can make use of performance explanations that mitigate earnings manipulation concerns. Davis and Tama-Sweet (2012) give empirical evidence that high levels of accruals are associated with additional performance explanations in earnings press releases and MD&A and with lower level of pessimistic tone in these disclosure outlets. Aerts and Cheng (2011) show that in an IPO setting, accruals earnings management is associated with more intense assertive causal disclosure and managers tend to avoid using explicit defensive causal disclosure tactics. In a more generalized setting, Aerts and Zhang (2014) give empirical evidence that firms make strategic use of causal reasoning on past performance in order to create appropriateness and cognitive legitimacy, thus increasing the credibility of reported performance and mitigating earnings management-related investor concerns.

I hypothesize that earnings management and impression management will be jointly used by managers with the purpose of presenting his/her performance in a more favorable light, in order to avoid punishments and/or increase market rewards.

H1: Firms that engage in earnings management also make use of impression management techniques in the form of self-serving attributions in earnings press releases.

(12)

12

2.3 Financial analysts as users of corporate disclosures

Analysts have important input to capital allocations, they are information intermediaries that affect the collective perceptions of investors and contribute to the functioning of the market (Graham et al., 2005). Previous literature shows that analysts add value through the discovery of private information (before earnings announcements) or through their superior ability to process public information (after earnings announcements) (Frankel et al., 2006; Chen et al., 2010). Analysts have been perceived as sophisticated processors of financial information who are less likely to misunderstand the implications of financial information compared to naïve investors. In general financial analysts are believed to possess superior information processing abilities because they are better trained, more experienced and have more knowledge about the firm and industry (Chen et al., 2010). This means that if analysts fail to process information efficiently this is strong evidence of overall inefficiency by market participants (Ramnath et al., 2008).

With the passing of Regulation Fair Disclosure Act of 2000, the reliance on public disclosure in the formation of their forecasts has increased due to the restriction of private communication channels between managers and analysts. Kross and Suk (2012) studied the impact of Regulation FD on analysts’ use of public disclosures as information sources in their forecasting activity. Their empirical results show that Regulation FD increased the reliance of analysts on management public disclosures (earnings announcements, management forecasts or conference calls). While research that focuses on the effects of Regulation FD show an increase in the quantity of public voluntary disclosure (Bailey et al, 2003), there is an ongoing debate of whether Regulation FD impaired the information environment by impeding analysts to obtain and interpret private and valuable information which contribute to forming an opinion and market consensus. Bailey et al. (2003) and Agrawal et al. (2006) found that consensus forecasts are not less accurate following the implementation of Regulation FD but the forecast dispersion increases. One explanation for these results is that analysts’ beliefs change positions relative to each other around the interim earnings announcements, indicating differential interpretation of interim earnings. Given the fact that analysts cannot rely anymore on their private information to help them predict firm future performance, it is safe to assume that the forecast dispersion will increase if managers use self-serving explanations in public disclosures for impression management purposes.

(13)

13 H2a: Analyst forecast dispersion will increase for firms that use impression management in the form of self-serving attributions in past performance explanations.

A growing body of research studied the relationship between managers and analysts. Analysts have significant influence on manager behavior and company disclosure strategy. By surveying high-level managers, Graham et al. (2005) found that managers view analysts as one of the most important users of financial information that affect the share price of the firms they manage. There is no conclusive evidence, however, regarding the effect of analysts on performance manipulations, such as earnings management. Analysts can act as external monitors given their superior technical financial ability to detect and report earnings management (Jensen and Meckling, 1976; Healy and Palepu, 2001), but they also exert pressure on management to meet or beat analysts’ expectations by managing earnings (Yu, 2008).

Considering prior evidence, I hypothesize that management will be concerned about analysts detecting earnings manipulation, which negatively affects firm value. This might create the need to provide more disclosure in order to avoid ex-ante uncertainty with regard to earnings management motives. Managers may try to mitigate reputation and other agency costs by providing self-serving biased explanations of past performance to shift the blame away from them or to take credit for good results. Barton and Mercer (2005) argue that it is possible that analysts will perceive the performance explanations as “cheap talk” (Bhattacharya and Krishnan, 1999), because they are voluntary and non-binding, thus difficult to verify. Previous studies show nevertheless that analysts use management’s explanations when assessing firms’ prospects (Rogers and Grant, 1997; Clarksonet al., 1999). Potential reasons for this are given by Abarbanell and Lehavy (2003), who suggest that analysts are aware of the use of some sorts of earnings management but are unable to detect it in quarterly earnings announcements (“unexpected accruals”) in the absence of publicly observable signals of the likelihood of earnings management. Moreover, evidence shows that analysts cannot identify the specific firms that engage in earnings management to avoid specific losses (Burgstahler and Eames, 2003). This gives sufficient evidence to presume that they will read the manager’s explanation for reported performance to assess the appropriateness and plausibility of the financial results.

It is therefore expected that performance explanations will be taken into consideration in case of ex-ante uncertainty regarding earnings management, but their impact on analyst perceptions will depend

(14)

14 on the explanation’s credibility. It is possible that other factors besides managers explanations will be taken into consideration in an additional cognitive processing to decide if the information found in the disclosures is to be relied on (Hutton et al., 2003) or to ignore it. On the other hand, assessing the plausibility of seemingly self-serving attributions in actual market setting might prove to be a difficult task (Kimbrough and Wang, 2014). These researchers argue that managers do not make attributions that are easily detected and refuted, and the causes cited, especially the internal factors are not verifiable or observable, thus analysts are susceptible to impression management manipulation.

Earnings management and impression management jointly used becomes a powerful tool for managers (Guillamon-Saorin and García Osma, 2010) that increases the information asymmetry and uncertainty. Prior literature also suggests that analysts possess different interpretational skills of public information (Kim and Verrecchia, 1994) and they have restricted use of private information after the adoption of Regulation FD to compensate for the information uncertainty. Research found that Regulation FD provoked disagreement and differential informed judgment about future annual earnings (Bailey et al., 2003). Land and Lundghom (1996) argue that the effect of increased disclosure on the analyst forecast dispersion depends on whether the differences in the forecasts are due to differences in information or in forecasting models. If analysts use a common forecasting model and have the same firm provided information, it will result in a decrease of forecast dispersion and increase in consensus. On the other hand, if analysts have the same information but value differently the firm provided disclosures, additional disclosure might actually increase the dispersion of analyst forecasts. Prior literature studies that examined this behavior on investors conclude that due to the differences in interpretation of firm disclosures, additional disclosure can result in divergence of beliefs (Harris and Raviv, 1993; Kandel and Pearson, 1995). Obviously if the earnings release contains uninformative information, analysts have no opportunity to make use of their superior information processing ability (Chen et al., 2010) and the forecast dispersion will increase. To summarize, prior evidence suggests that managers that use earnings management to conceal their true performance anticipate that if detected, analysts will punish them. They will provide an increased volume of causal performance explanations (Aerts and Cheng, 2011; Aerts and Zhang, 2014), which signals rational behavior in order to shift blame away for poor results and get rewards for good performance. Analysts do not disregard these performance explanations because they cannot directly distinguish if they are truly informative or are driven by managers’ psychological biases or

(15)

15 opportunism (Kimbrough and Wang, 2014). They will take other factors into consideration to assess the plausibility of performance explanations. External factors are relatively easier to verify in causal explanations (such as firm environment, economic conditions, industry and peer performance, manager reputation, etc.) compared to internal, unobservable and unverifiable factors in causal explanations (management plans, actions, strategies). In cases when no other useful information is publicly available, I argue that at least some financial analysts are prone to impression management in earnings releases and the disagreement among analysts regarding future performance will increase. H2b: Analyst forecast dispersion is positively associated with the use of earnings management and impression management in the form of self-serving attributions in past performance explanations. Another major characteristic of analyst forecasts is accuracy. It is expected that forecast accuracy of future earnings will increase with the informativeness of firm public disclosures (Lang and Lundholm, 1996). According to the impression management hypothesis, the information disclosure might not be purposefully informative, but rather has the ultimate goal of deceiving analyst perceptions and influencing their future earnings forecast.

H3a: Analyst forecast accuracy will decrease for firms that use impression management in the form of self-serving attributions in past performance explanations.

Prior literature suggests that analysts do not do not initially anticipate the implications of the most recent year’s accruals for the subsequent year’s earnings (Bradshaw et al., 2001). This happens because analysts gradually appear to realize the implications of accruals for subsequent earnings as the year progresses and firms release more information through quarterly earnings releases. As a result, future analysts’ forecasts are usually optimistically biased because they do not fully incorporate the predictable earnings reversals associated with extreme levels of accruals. Moreover, most of firms do not provide management earnings forecasts in their earnings press releases to guide analysts’ future forecasts, and rely on their past performance explanations that might send signals of rational decision-making and future profitability. Because of the reasons cited above, I hypothesize that the use of impression management and earnings management will have an impact on the forecast error. H3b: Analyst forecast error is positively associated with the use earnings management and impression management in the form of self-serving attributions in past performance explanations.

(16)

16

3. Data and Methodology 3.1 Data Sources

I chose to analyze performance explanations from quarterly earnings press releases because of their multiple advantages compared to other types of narrative disclosures. Earnings press releases comprise an important element of a firm’s overall disclosure strategy. It communicates consolidated information to market users in both numerical and narrative forms and often contains management commentary on past performance. They have the advantage of being released after each financial quarter compared with the annual report, which is a once in a year event. Previous research suggests that earnings announcements are a superior information source in the competitive corporate information market, in the sense that they convey more information than other information sources individually, such as dividend announcements, management forecasts, preannouncements, and 10-K and 10-Q filings (Basu et al., 2013; Levi 2008).

Earnings press releases are the major news event of the season for many companies as well as investors, analysts, financial media, and other market users (Davis et al., 2012). Recent changes in regulation, as well as advances in technology, have increasingly made press releases directly accessible to market users. On October 23, 2000, the Securities and Exchange Commission (SEC) implemented ‘‘Regulation Fair Disclosure” (FD). Regulation FD provides that when an issuer discloses material nonpublic information to certain individuals or entities (securities market professionals, such as stock analysts), the issuer must make public disclosure of that information. In this way, Regulation FD aims to promote the full and fair disclosure, and earnings press releases became an important source of public company information. Moreover, following the Sarbanes-Oxley Act of 2002, from March 2003 the SEC requires that all earnings press releases be filed, thus ensuring that stakeholders have direct access to the actual text of firms’ earnings releases (Form 8-K).

Francis et al. (2002) characterized earnings press releases as being a ‘‘package of information’’ to shareholders. Researchers found evidence that the information content of earnings press releases has increased significantly over time as witnessed by the increase in length of these disclosures (Francis et al., 2002; Davis et al., 2012). Because of limited regulation of their content, it is interesting to investigate whether they convey useful information for decision-making purposes or merely serve to manipulate external perceptions of company performance with the use of self-serving performance explanations.

(17)

17 The analyzed earnings press releases were retrieved from company websites. Because the sample chosen reflects highly visible firms, the investor relations internet pages were usually very interactive and user-friendly making earnings press releases easy to find. For every earnings press release, I captured the publication date, saved a copy to my computer, read the full text and manually coded the performance explanations identified.

3.2 Sample selection

I analyzed the incidence of impression management in earnings press releases focusing on US firms. Aerts et al. (2013) and Aerts and Tarca (2010), show that the institutional setting influences the disclosure behavior of firms. Although earnings press releases are voluntary disclosures, the fact that USA is a country with high regulatory and litigation costs (La Porta et al., 2006), I expect we expect that such an environment will lead to more detailed explanations of performance. For instance, Skinner (1997) argues that voluntary disclosures occur more frequently in quarters that result in litigation than in quarters that do not, this is evidence that the institutional environment plays an important factor in the decision of providing voluntary performance explanations in corporate narratives.

Previous literature (Merkl-Davies and Brennan, 2007) suggests that impression management tends to be more pronounced in highly visible firms, therefore the sample will consist of firms from the S&P 500 Index. Given their market capitalization, it can be expected that these firms are closely followed by analysts, investors and regulators, and management is incentivized with keeping them in the public attention (Keusch et al., 2012). Manual content analysis of corporate disclosures creates generalizability concerns because of the small sample. In order to make the task more practical and the findings generalizable, I drew a random sample of firms from the full population. Using a small random sample allows accurate hand-collection on a representative sample at a reasonable cost and is common in disclosure research (Hutton et al., 2003; Kimbrough and Wang, 2014).

My final sample consists of 52 unique firms for which I have analyzed the earnings press releases for two quarters, which results in 104 earnings press releases. Firms are classified using the Standard Industry Classification system (SIC). I did not purposely exclude any industries, however, there are no firms from the Agriculture, Forestry and Fishing or Public Administration sectors in the sample. The most highly represented industry is Manufacturing, with 23 firms (44.23%), followed by Transportation and Public Utilities with 8 firms (15.38%).

(18)

18 TABLE 1

Distribution of firm observations by industry

SIC n %

1000-1400 Mining 5 9,62

1500-1700 Construction 1 1,92

2000-3900 Manufacturing 23 44,23

4000-4900 Transportation and Public Utilities 8 15,38

5000-5100 Wholesale Trade 1 1,92

5200-5900 Retail Trade 6 11,54

6000-6700 Finance, Insurance, Real Estate 2 3,85

7000-8900 Services 6 11,54

Total 52 100,00

I have selected the years 2014 (quarter 1) and 2013 (quarter 4) for analysis because they are the most recent with all the available data and are different in terms of general economic environment. The first half of 2014 was characterized by a 2.1% decline in Gross Domestic Product (“GDP”), weather being the main responsible factor for the decline. This came as a surprise after economists predicted that economic growth would be higher in the first quarter of 2014 when investments and purchases would be made that could not be made in late 2013 because of the 16 days federal government shut down. It is interesting to see the pattern of the use of attributional statements in performance explanations in “”good” and “bad” quarters. Previous research shows that both types of self-serving attributions will be significantly present in both quarters and particularly in the “bad” quarter. Due to the visible weak performance of the economy in the first quarter of 2014, the role and credibility of performance explanations could change. Keusch et al. (2012) argue that in a crisis situation there is less need for impression management so the role for acclaiming attributions is weaker. At the same time, defensive attributions are less relevant due to management’s desire to appear in control of the business in unstable times.

3.3 Content analysis

Content analysis of archival data was used to investigate corporate attributional behavior in earnings press releases. In the related literature several studies used content analysis to detect self-serving reporting bias in attributional statements (Aerts, 2005; Baginski et al., 2004; Keusch et al., 2012; Kimbrough and Wang 2014). They all try to establish an explicit cause-and-effect relationship, where the link between action and outcome can be determined easily. I read each earnings press

(19)

19 release to determine whether an attribution is present in the explanation of past quarter performance. An attributional statement is defined as a “phrase or a sentence in which a corporate event or performance outcome is linked with a reason or a cause for the event or outcome” (Aerts, 2005). In order to increase the reliability of the results, I included those cause-and-effect relationships where the link between antecedent and outcome is explicit or implicit, and cause and effect could be confidently related. An explicit link might be preceded a conjunction, such as ‘as a result of’, ‘because of’, ‘due to’ or a verb such as ‘leading to’, ‘driven by’ or ‘generated’ (Keusch et al., 2012). When I could not establish an explicit or implicit link directly or could not determine if the valence of the outcome or cause was positive or negative, I did not code the statement. This reduces the possible subjectivity bias in the coding process. Another solution was to code the cause or outcome as undetermined, but such observations did not add value to the results.

3.4 Coding procedure

I developed a coding procedure to guide me in the process. For each attribution present, I first identified the explained content or the outcomes linked to a company’s income statement such as revenues, earnings, income or profitability ratios. Next, I coded the way of expressing the performance change, namely, if the change in the outcome is qualitatively or quantitatively expressed. The following step was to capture the valence of the outcome. Outcomes were identified as positive if they show an improvement over past years, a higher level in absolute or relative terms or were framed with a positive connotation. Outcomes are coded as negative if the opposite occurs. The same rationale was applied to code the valence of the cause. A cause had a positive valence if it explicitly or implicitly mentioned that the cited factor positively impacted the outcome, in numerical expressions (quantitatively) or verbally (qualitatively). In the cases in which I was unable to confidently determine the valence of the explained outcome or the cause, I did not code it. Causes were explicitly (preceded by a conjunction or verb as explained above) or implicitly related to the outcome. Moreover, I coded the direction of the causal link. A direct influence was suggested by phrases such as ‘due to’, ‘because of’ or ‘leading to’ whereas an opposite influence was indicated by expressions such as ‘despite’, ‘even though’, or ‘in spite of’. The cause was attributed to internal or external factors. Internal attributions refer to causal factors internal to the organization, such as strategy or management skill. External attributions refer to events not controllable by the company, such as competition and certain economic factors or natural phenomenon (Baginski, Hassell, Kimbrough, 2004). In most of press releases, I identified a large number of potential causes in

(20)

20 attributional statements; this is why I used the Classification Scheme of Attributions developed by Baginski et al. (2004). This framework was extensively used in prior literature on this topic (Kimbrough and Wang, 2014; Keusch et al., 2012).

TABLE 2

Classification Scheme of Attributions EXTERNAL CAUSES

General Economic/Environmental Issues Governmental/Third Party Issues Recession/inflation

Dollar weakness/strength Foreign currency fluctuation

Input cost changes—increasing/decreasing costs Change in market for product

General loss/gain of customers

Weather/catastrophe/uncontrollable factors Order backlogs

Geopolitical events

Tax law/other law changes SEC actions/regulatory actions

Expropriation by foreign governments Lawsuits/legal actions

Competition action/issues Involuntary accounting changes

INTERNAL CAUSES

Product/Services Issues/Actions Organizational Issues/Actions

Changes in product prices Changes in product mix New products/processes/production/partnerships Advertising/marketing Management techniques/strategies/plans/repositioning Cost cutting/savings Cost increase

Changes in management personnel Asset write downs

Going public

Selling/buying stocks

Merger/ acquisition/ disposal of a business segment

Investment in plant assets Voluntary accounting changes *developed by Baginski et al. (2004)

One issue that I encountered is that often companies explained one single performance outcome with several causes or, conversely, one antecedent was presented as the cause for several outcomes. In these situations, each of the cause-outcome relationships was coded separately with the condition that they referred to different financial line items or financial performance measures.

I used the framework presented by Keusch et al. (2012) for determining self-serving bias in attributional explanations. ‘Acclaiming attributions’ are used to take credit for positive outcomes. They are of two types: performance explanations that stress management’s responsibility for positive

(21)

21 outcomes (‘entitlement’) and performance explanations that emphasize that favorable outcomes were achieved in spite of adverse external forces (‘enhancement’). ‘Defensive attributions’ are used to deny responsibility for bad performance. Defensive attributions help managers to dissociate oneself from negative outcomes, either by emphasizing the role of adverse external forces (‘excuse’), or by accentuating that the right managerial action was taken even though the overall performance is bad (‘causality denial’). Previous empirical literature on management’s discussion of performance in the annual report narrative has produced almost consistent results. Positive outcomes are more often related to management actions (‘entitlements’), while unfavorable outcomes are related to adverse external forces (‘excuses’) (Keusch et al., 2012). I classified an attributional statement as self-serving if the valence of the outcome was positive (negative) and the cited cause was internally (externally) attributed. An offsetting attribution names a factor that negatively (positively) affects results, but the news (outcome) has a positive (negative) valence. The following scheme, derived from Keusch et al. (2012) summarizes the process for identifying self-serving bias.

Self-serving bias Two explanations: • Intentional (impression management) or

• Unintentional (cognitive bias)

Acclaiming attributions

• Seeking credit for favorable outcomes

Defensive attributions

• Dissociating oneself from unfavorable outcomes

Enhancements

• Favorable outcomes were achieved despite of adverse external forces

Entitlements

• Favorable outcomes were achieved because of one’s actions

Excuses

• Unfavorable outcomes occurred due to adverse external forces

Causality denials

• Unfavorable outcomes occurred despite of one’s actions

(22)

22 Besides self-serving attributions, a number of companies had cause-and-effect performance explanations that could not be perceived as impression management. Following Kimbrough and Wang (2014), I coded this attributions as incentive incompatible. An incentive incompatible attribution names a factor that positively (negatively) affects results but attributes the factor to an external (internal) cause. The attributions that were only incentive-incompatible constitute the

non-self-serving sample. 3.5 Descriptive statistics for content analysis

Table 3 contains descriptive statistics regarding the frequency in both absolute and relative numbers of the characteristics of attributional statements. From the full sample of 283 attributions, 71.38% were coded as self-serving and 28.62% were interpreted as non self-serving. From the results, I can conclude that firms tend to make more self-serving attributions, and their number increases in the fiscal year 2014 compared to fiscal 2013. From the self-serving sample, almost 80% of attributions are acclaiming, meaning that managers take credit for good firm performance, and the majority of acclaiming attributions are entitlements (managers explain favorable outcomes by relating them to one’s actions). From the defensive sample, most of the attributions are of the causal denial type, managers defend themselves by claiming that unfavorable outcomes occurred despite of one’s actions. The non self-serving sample contains only incentive incompatible attributions, this is the case when a positive (negative) outcome is explained by an external (internal) cause. Some of the incentive incompatible attributional statements were also offsetting. Attributions are classified as offsetting if the valence of the outcome is positive (negative), whereas the valence of the cause is negative (positive). Offsetting attributions can be both enhancements, causality denials, or, in some cases, they incentive incompatible attributions.

Managers use more often internal causes to explain their results compared to external causes. Internal and external attributions have different degrees of credibility. Baginski et al. (2004) argue that external attributions are more easily verifiable by analysts. The causes are observable because they relate to economic conditions and/or government actions, and the attribution’s credibility, then depends on analyst’s sophistication and market/industry knowledge. Internal attributions, on the other hand are unobservable to the external stakeholder and it is more difficult for the analyst to assess their plausibility. Dontoh (1989) show that internal attributions may contain proprietary information because they are related to management’s strategic plans and actions. If this is the case, managers will try to avoid providing the subset of internal attributions that convey proprietary

(23)

23 information or, in case managers do disclose it, the internal attributions are perceived as highly credible. Of course, there is the possibility that both internal and external attributions are viewed as cheap talk and the past performance explanations are not taken into consideration by analysts in the earnings forecasting process.

From the statistics in table 3, we can see that managers avoid reporting bad news, or in cases in which the performance has worsened, they employ different strategies to make their performance look good. For example, in many of the press releases, the negative results are adjusted for better “comparisons” with past quarters results. In this way, the performance is “framed” as improved, and the manager can make a good impression. Managers can use their discretion in the presentation of earnings results because they are allowed to disclose non-GAAP and non-audited financial numbers. The most often explained outcomes are earnings and revenues. The link between the cause and outcome is usually explicit, and managers express the impact of the cause on the outcome in qualitative terms in 75% of the cases.

Table 4 provides a first insight into the extent of the usage of attributions in quarterly earnings press releases by showing means and other descriptive statistics for each type of attributional statements. On average, I have identified 2.72 attributions in each earnings press release. From the total number of attributions, on average 1.94 were perceived as self-serving and 0.77 as non self-serving (incentive incompatible). The acclaiming type was used more often than the other four types of self-serving attributions. Moreover, each earnings press release had a mean of one entitlement statement and on average 0.49 enhancement statements. Barton and Mercer (2004) argue that managers have incentives to take credit for good performance because it usually brings positive effects on both firm’s cost of capital and manager’s evaluation of performance and his compensation. Managers were more reluctant in providing defensive attributions, they provide less than half defensive attributions compared to acclaiming ones. I coded an average number of 0.23 causality denials and 0.19 excuses in each earnings release.

(24)

24 TABLE 3

Attribution incidence and type

Frequency _attributions% of total 2013 2014 _totals% of

Total Attributions: 283 100% 132 151 Self-serving sample: 202 71,38% 87 115 Total Acclaiming: 158 69 89 100% Entitlements 107 49 58 67,72% Enhancements 51 20 31 32,28% Total Defensive: 44 18 26 100% Excuses 20 9 11 45,45% Causality denials 24 9 15 54,55%

Non self-serving sample: 81 28,62% 45 36

Incentive incompatible 81 45 36

Total Offsetting: 106 37,46% 46 60 100%

Incentive incompatible and Offsetting 31 17 14 29,25%

Enhancements and Offsetting 51 20 31 48,11%

Causality denials and Offsetting 24 9 15 22,64%

External Causes: 98 34,63% 42 56 100%

General Economic/Environmental Issues 78 34 44 79,59%

Governmental/Third Party Issues 20 8 12 20,41%

Internal Causes: 185 65,37% 90 95 100% Product/Services Issues/Actions 72 32 40 38,92% Organizational Issues/Actions 113 58 55 61,08% News: Negative 73 25,80% 40 33 Positive 210 74,20% 92 118 Outcome explained: 283 132 151 Revenues 96 33,92% 38 58 Earnings 107 37,81% 54 53 Income 54 19,08% 28 26 Profitability Ratio 26 9,19% 12 14

Terms for expressing the relationship: 283 132 151

Qualitative 211 74,56% 89 122

Quantitative 72 25,44% 43 29

Explicitness of the relationship: 283 132 151

Implicit 39 13,78% 15 24

(25)

25 Next, I analyzed whether there are any differences in the use of attributions in the two time periods selected. Because in the first quarter of 2014 the general economic environment has worsened compared to the last quarter of 2013, I expect the attributional pattern to change. Keusch et al. (2012) show that in a crisis situation, managers will make more use of self-serving bias, because the external economic conditions allow managers to be portrayed in the best possible light. In table 5, I present the results for independent samples t-test on the amount of attributional statements found in earnings press releases. We can see that the mean number of attributions made by a firm in an earnings release has slightly increased in 2014 compared to 2013, although the difference is not statistically significant. The number of self-serving attributions was, on average 0.53 higher in 2014 compared to 2013. The p-value for the two tailed test and for the one tailed test is statistically significant (<0.05) and this is evidence that in a worse general economic environment, or a “crisis” period, managers will engage in making more self-serving attributions in past performance explanations. Furthermore, managers used significantly more acclaiming attributions in the first quarter of 2014 compared to the last quarter of 2013. One reason can be that managers want to take more credit for positive outcomes in a bad quarter compared to a good quarter. The use of enhancements has significantly increased from a mean of 0.38 in 2013 to 0.59 in 2014 (one tailed p-value<0.1). This finding supports the impression management opportunistic behavior; when firms are exposed to adverse external forces, positive outcomes will be emphasized more. I found no

TABLE 4

Descriptive statistics of the number of attributional statements at firm level Quartile

Variables n Mean Std. Dev. Min 0,25 Median 0,75 Max

Total attributions 104 2,7211 1,5976 1 1 2 4 8 Self-serving attributions: 104 1,9423 1,3353 0 1 2 3 6 Total acclaiming: 104 1,5192 1,307 0 0,5 1 2 5 Entitlements 104 1,0288 0,9497 0 0 1 2 4 Enhancements 104 0,4903 0,7757 0 0 0 1 4 Total defensive: 104 0,423 0,8888 0 0 0 0 4 Excuses 104 0,1923 0,5039 0 0 0 0 2 Causality denials 104 0,2307 0,5613 0 0 0 0 2

Non self-serving attributions: 104 0,7788 0,9446 0 0 0 1 3

Incentive incompatible 104 0,7788 0,9446 0 0 0 1 3

(26)

26 significant difference in the use of defensive attributions in the two quarters, thus no support for the impression management explanation of voluntary disclosures that managers will blame more the bad outcomes on the negative external factors.

TABLE 5

Independent samples t-test on the amount of attributional statements in earnings press releases

Variables Mean _DifferenceMean t Two-tailed

p-value One-tailed p-value 2013 2014 Total attributions 2,5384 2,9038 -0,3653 -1,1682 0,2454 0,1227 Self-serving attributions: 1,673 2,2115 -0,5384 -2,0895 0,0392 0,0196 Total acclaiming: 1,3269 1,7115 -0,3846 -1,5097 0,1342 0,0671 Entitlements 0,9423 1,1153 -0,1730 -0,9286 0,3553 0,1777 Enhancements 0,3846 0,5961 -0,2115 -1 0,1655 0,0828 Total defensive: 0,3461 0,5 -0,1538 -0,8816 0,3801 0,19 Excuses 0,173 0,2115 -0,0384 -0,3876 0,6991 0,3496 Causality denials 0,173 0,2884 -0,1153 -1,0485 0,2969 0,1484

Non self-serving attributions: 0,8653 0,6923 0,173 0,9337 0,3527 0,1763

Incentive incompatible 0,8653 0,6923 0,173 0,9337 0,3527 0,1763

3.6 Regression Models and variable measurement

3.6.1 Impression management and earnings management

Hypotheses 1 states that firms that engage in impression management are motivated by the use of earnings management. To test the hypothesis, I estimate the following empirical model:

The dependent variable IM_Dummy equals 1 when a self-serving attribution is present in a coded statement, and 0 otherwise. Each variable and its measurement is described below, starting with the independent variable, discretionary accruals which is the proxy for earnings management and ending with the control variables.

(27)

27 As a proxy for earnings management I used discretionary accruals. Accruals are prone to management manipulation because they require management estimation and judgment. However not all accruals are the result of earnings management, some accrual adjustments are necessary because they reflect the operational conditions of particular industries. Total accruals can be therefore, decomposed into two parts, nondiscretionary accruals (NDAs) and discretionary accruals (DAs). Previous accounting literature makes extensive use of discretionary accruals regression models, such as the modified Jones model or a performance-adjusted modified Jones model (Dechow et al., 2010). I used a performance-matched version of the modified-Jones model following Kothari et al. (2005), because other models might be misspecified due to the correlation between financial performance and accruals. Aerts and Zhang (2014) found evidence of a positive relationship between accruals earnings management and opportunistic use of causal reasoning on past performance using the model cited above. I computed total accruals as the difference between earnings before extraordinary items and cash flows from operations. The model is specified as follows:

Where:

= Income from operations before extraordinary items and discontinued operations; = Cash-flows from operations before extraordinary items and discontinued operations

The non-discretionary variables are the expected accruals and the discretionary variables are the residuals. The expected accruals are determined using the following cross-sectional regression for the specific quarter using a sample of all listed firms in the same industry categories. The model is estimated quarterly at the 3-digit industry level with at least 20 quarterly observations available, otherwise the 2 and 1-digit SIC codes are used.

The following model estimates total accruals: ( ) ( ) ( ) Where:

(28)

28 = total accruals = income before extraordinary items - cash from operations for a firm i in quarter t;

= lagged total assets for a firm i in quarter t;

= change in net sales revenues in quarter t from quarter t-1; = change in net receivables in quarter t from quarter t-1; = gross property, plant and equipment for a firm i in quarter t; = return on assets for a firm i in quarter t; equal with the

The estimated coefficients are then used to compute the asset-scaled nondiscretionary accruals for each firm i in each quarter t.

̂ ( ) ̂ ( ) ̂ ( )

The residual discretionary accruals are the asset scaled excess accruals for firm I in quarter t, which is determined as follows:

(

)

The model uses lagged total assets as deflator for control of heteroskedasticity. I used the residual from the model as the proxy for discretionary accruals.

I used absolute values for discretionary accruals, because extant literature on earnings management suggests that market users of corporate information respond differently to upwards and downwards earnings management (Dechow et al., 2010). For this reasons signed discretionary accruals are more appropriate to use in studying the relationship between earnings management and opportunistic behavior (impression management) (Aerts and Zhang, 2014). Despite the fact that most earnings management research uses annual data, I used quarterly data to detect discretionary accruals. Quarterly accruals have been used in several studies because they have the advantage of detecting earnings management around specific corporate events (Jeter and Shivakumar, 1999). Moreover, quarterly results are often unaudited, allowing greater managerial discretion and higher probability of earnings management.

(29)

29 I control for company and industry related variables that proxy for the demand and supply of corporate disclosures and are found to affect the characteristics of attributional statements. Firm size is generally used to measure the amount of information that a company discloses. Lang and Lundlom (1996) argue that larger firm will provide more disclosures; reasons for this might be the lower information production costs and lower costs of competitive disadvantage associated with disclosures. Baginski et al. (2004) give empirical evidence that the use of causal explanations is positively associated with firm size. Aerts (2005) finds that larger firms use more enhancements and entitlements and less defensive attributions. Moreover, previous studies show that size is also related to analyst forecast bias (e.g., Wu, 1999). Analysts can obtain more information about larger firms, therefore the accuracy of their forecasts can increase. I included a control for firm size in my model of analyst forecast dispersion and error. I will use the natural logarithm of market capitalization in order to avoid skewness in data distribution.

Prior literature shows that litigation affects the quality of disclosed information. Firms with high litigation risk are found to provide more quantitative and qualitative information accompanying their earnings forecasts (Johnson et al., 2001). On the other hand, Rogers and Van Buskirk (2009) argue that firms reduce the amount of disclosures after litigation. Following prior research (Aerts and Cheng (2011); Kim and Skinner, 2011), I will include a litigation dummy equal to 1 for industries that with the following SIC codes: 2833–2838 (Biotech firms), 3570–3577 (Computer firms), 3600–3674 (Electronics firms), 5200–5961 (Retail firms), 7370–7374 (Computer firms), 8731–8734 (Biotech firms), and 0 otherwise. Prior studies identified firms from the above mentioned industries to be prone to higher litigation risk than other firms (Field et al., 2005).

Jensen and Meckling (1976) show that financial leverage is positively related to disclosure levels, because leveraged firms incur higher monitoring costs and are incentivized to disclose more to reduce these costs. Aerts (2005) uses leverage as a primary financial risk indicator and argues that financial leverage could be related to impression management techniques and the use of more verbal coping tactics. Leverage is proxied by the debt-to-equity ratio or total liabilities divided by total equity.

I also include a profitability measure since prior research shows a relationship between disclosure levels and profitability (Aerts, 2001; 2005). Changes in firm performance should result in changes in the attributional pattern, but, in past research this relationship is mostly weak, which supports the

(30)

30 impression management hypothesis (Keusch et al., 2012). Profitability is determined as income before extraordinary activities on total assets.

Lastly, I include a dummy variable to control for year effects. The first quarter of 2014 was characterized by a change in general economic environment so it is expected that the attributional behavior to change as explained in the section 3.2. The year dummy takes the value of 1 for the year 2014, and 0 for 2013. I present the description of all the variables used is presented in Table 6.

3.6.2 Impression management, earnings management and analyst forecast characteristics

To test the second and third hypothesis which state that forecast dispersion and forecast error are associated with the use of impression management and earnings management I use the following empirical models:

The dependent variable forecast characteristic is alternatively measured as the forecast standard deviation and the forecast error. The description of all the variables is provided in Table 6. I use the independent variable LagROA to control for possible over or under reaction to prior operating performance (Abarbanell and Bernard, 1992). I also control for growth opportunities, inversely captured by BTM as growing firms are likely to suffer from greater information asymmetries and to experience greater forecasts errors (Wilson and Wu, 2011). The loss dummy is used because Gu and Wu (2003) give evidence that systematic forecast optimism is likely to be related with firms that report losses. There is evidence that analyst forecasts are biased depending on firm size. Prior literature shows that forecasts are more optimistic for small firms compared to larger firms (Gu and Wu, 2003). I also include leverage as indebted firms are more likely to experience future financial problems which increases the uncertainty regarding future performance.

(31)

31 TABLE 6

Variables description

Variable Definition

IM_Dummy 1 if a firm makes a self-serving attribution in its

earnings press release, and 0 otherwise;

Discretionary_Accruals The residual from the performance-matched

version of the modified-Jones model;

Forecast_Dispersion The standard deviation of analyst forecasts for

the t+4 EPS;

Forecast_Error The difference between t+4 (one year later)

actual EPS and the most recent mean EPS forecast, after the earnings results release, deflated by the stock price at the end of year (t+4);

Size Log of Market Capitalization at the end of

quarter;

BTM Ratio of book value of equity to market value of

equity for firm i, in quarter t;

Leverage Long term debt to equity ratio for firm i, in

quarter t;

LagROA Income before extraordinary items and

discontinued operations on total assets in quarter t-1;

ΔProfitability 1 if the change in ROA from previous quarter is

negative, 0 otherwise;

Year_Dummy 1 if the fiscal year is 2014, 0 if it is 2013;

Loss_Dummy 1 if the income before extraordinary items and

discontinued operations ≤0, 0 otherwise;

Litigation_Dummy 1 if the firm’s industry is from the following SIC

codes: 2833–2838 (Biotech firms), 3570–3577 (Computer firms), 3600–3674 (Electronics firms), 5200–5961 (Retail firms), 7370–7374 (Computer firms), 8731–8734 (Biotech firms), and 0 otherwise.

Analyst following Numbers of estimates for the consensus EPS

forecast

Firm related information necessary for determining discretionary accruals, the characteristics of analyst forecasts and for the control variables are taken from Compustat and I/B/E/S from Thomson Reuters.

Impression and earnings management : a collaborative tool to mislead analysts