Abstract
We investigate the relation between managers’ personal ideologies and financial reporting quality. We use Federal Elections Commission data to develop three proxies for managers’ personal ideologies and test their relations with two financial reporting quality metrics: discretionary accruals and financial statement restatements. We find that both the absolute value of discretionary accruals and the probability of restatement decrease in the degree to which firms’ managers have conservative ideologies. These results are robust in the post-SOX period, to controls for potential self-selection bias, to alternative measures of both ideology and financial reporting quality, and to controls for firm political sensitivity, lobbying activity, governance, operational complexity and auditor strength. Our findings contribute to a growing literature demonstrating that business outcomes are partially explained by manager-specific factors including managers’ personal views and priorities.
Similar content being viewed by others
1 Introduction
Our study extends an emerging area of research that considers the implications of managers’ individual attributes for business and financial reporting decisions. We investigate the relation between managers’ personal ideologies and financial reporting quality. We use individual political contribution histories to develop firm and manager-specific proxies for personal ideologies, and investigate the relation between managers’ ideologies and both discretionary accruals and financial statement restatements.
The assumption that managers are rational, economically self-interested decision makers underlies most accounting research examining financial reporting quality. For example, empirical evidence documents earnings management as a phenomenon that is affected by managers’ economic incentives and extant constraints on opportunistic behavior (Jones 1991; Healy 1985; Iliev 2010). The underlying assumption is that managers are a homogenous group of decision makers who will, at least on average, make predictable earnings management decisions in response to various incentives.
Yet, humans are complex and individual managers may very well have priorities other than immediate personal economic gain. Prior research has documented associations between financial reporting quality and the professional expertise and other personal demographics of board members and managers (Bedard et al. 2004; Aier et al. 2005; Dhaliwal et al. 2010; Krishnan et al. 2011). Archival research is beginning to examine the implications of managerial heterogeneity and identify personal attributes other than expertise that may be associated with managers’ business and reporting decisions (Bertrand and Schoar 2003; Bamber et al. 2010; Brochet and Welch 2011; Hutton et al. 2014, 2015; Liu et al. 2016). This recent empirical evidence shows that heterogeneity in managers’ personal characteristics helps explain operating, financing, reporting and disclosure decisions. Related research (Hutton et al. 2015) provides empirical evidence that managers’ personal ideologies affect the probability of corporate misconduct.Footnote 1
Understanding the role that managerial heterogeneity plays in decision making is important because it may identify personal attributes that help predict and explain observed business behaviors that otherwise cannot be fully understood. We contribute to this emerging line of research by examining the relation between individual managers’ personal ideologies and financial reporting quality. Like Bamber et al. (2010), we examine the role of a manager-specific characteristic. Bamber et al. (2010) use manager-specific data and document an association between voluntary corporate forecast disclosure patterns and manager-specific demographic attributes. Unlike our study, however, they do not examine financial reporting quality. In addition, our manager-specific personal ideology construct reflects a broader and more aggregate attribute than the highly specific demographic characteristics used in Bamber et al. (2010). McGuire, Omer and Sharp (2012) find a relation between the religiosity of a firm’s business environment and earnings management and reporting irregularities. The religiosity explanatory variable used in McGuire et al. (2012), however, is geographically-based and does not reflect manager-specific characteristics. Finally, Hutton et al. (2014) examine the role of managers’ personal ideologies in financing and operational decisions, but they do not consider its role in explaining financial reporting quality. Thus, we extend these related managerial heterogeneity studies by examining the relation between a manager-specific characteristic, i.e., personal ideology, and financial reporting quality.
Another recent study, Lee et al. (2014), finds evidence that political alignment between a firm’s top managers and directors increases the probability of corporate fraud. Our study differs from Lee et al. (2014) in several ways. First, Lee et al. (2014) investigate a fundamentally different research question about the governance effects of alignment in political ideology among top managers and directors. Their study focuses on explaining the relation between firm value and governance, using a metric based on political contributions data. Although Lee et al. (2014) include a test of financial reporting quality, it is a supplemental test incorporated into a broader governance study. They use a fraud metric to proxy for financial reporting quality, while we use discretionary accruals and restatements. The occurrence of financial statement fraud indicates extremely low reporting quality that reflects illegal behavior. This is appropriate for Lee et al.’s (2014) research question which investigates governance and governance failure. Discretionary accruals and restatements generally result from aggressive, but legal, accounting choices as appropriate for our research question investigating financial reporting quality. Second, Lee et al. (2014) estimate their fraud model using the CEO’s ideology as the explanatory variable. In contrast, we test the relation between financial reporting quality and ideology metrics for the top management team, the CEO, and the CFO. Our CFO ideology variable allows us to specifically examine the reporting quality implications of the ideology of the manager most directly involved in financial reporting decisions.
We choose to examine the relation between managers’ personal ideologies and financial reporting quality because considering the role of manager-specific characteristics in this context can deepen our understanding of observed variation in this economically important dimension of business. In addition, financial reporting quality is likely to be sensitive to differences in managers’ personal ideologies because ideologies are related to risk preferences. Prior studies investigating the importance of personal ideology link ideology to several risk-averse behaviors (Hutton et al. 2014, 2015). In matters such as leverage and R&D investment, outcomes consistent with more risk-averse choices are observed when managers have more conservative personal ideologies (Hutton et al. 2014). Low quality financial reporting subjects firms and their managers to potential litigation and other adverse consequences. If managers’ personal ideologies matter for firm behavior, and high quality financial reporting reduces risk, we expect firms with ideologically conservative managers to have higher financial reporting quality.
To test this proposition we compute three empirical measures of managers’ personal ideology. The first, as used in Hutton et al. (2014), is a firm-year measure based on the individual political contribution histories of the top managers employed at each sample firm. Using this measure, we empirically test the relation between the personal ideology of firms’ managers and two metrics of financial reporting quality: discretionary accruals and restatements. Our results indicate that companies with more ideologically conservative managers have higher quality financial reporting.
The Hutton et al. (2014) proxy reflects an average ideology of managers with a variety of roles in their corporations. In order to more closely examine the importance of individual manager ideology, we develop two alternative personal ideology proxies that are manager-specific. One proxy represents the ideology of the CFO, the manager most directly affecting accounting and financial reporting decisions. The other represents the ideology of the CEO, whose ideology may best reflect the tone at the top of the corporation. We find that when CFOs are more ideologically conservative, companies have higher reporting quality. Results using the CEO proxy are mixed, consistent with CFOs having more influence over accounting and financial reporting quality than CEOs.
2 Background and literature review
A well-developed body of accounting research examines cross sectional variation in financial reporting quality and assumes that managers homogeneously respond to economic incentives within the boundaries of existing constraints on opportunistic behavior. In this literature, firm-level metrics are typically used to explain reporting quality (Jones 1991; Bannister and Newman 1996; Burgstahler and Dichev 1997; Dechow et al. 1996; Jones et al. 2008). Earnings qualityFootnote 2 studies have found that earnings are managed to influence stock price (Perry and Williams 1994; Teoh et al. 1998a, 1998b; Dechow et al. 1996; Burgstahler and Dichev 1997), managers’ incentive compensation levels (Healy 1985; Holthausen et al. 1995), and regulatory intervention (Jones 1991; Cahan 1992; Li et al. 2016). Some international studies have found evidence of a relation between financial reporting quality and political connectedness of various firm stakeholder groups, which is indicative of a degree of governmental protection (Chaney et al. 2011; Batta et al. 2014). While these studies use a variety of economic incentives and constraints on opportunism to explain earnings quality, there is no attempt to identify or consider manager-specific attributes that might affect how individual managers respond to economic incentives and constraints.
McGuire et al. (2012) extends this literature by examining the impact of one nonfinancial factor, the degree of religiosity in the geographic region surrounding the firm’s headquarters. McGuire et al. (2012) observe that virtually all religious texts value honesty, and conclude that social norms in highly religious areas are more likely to include honesty than are those in less religious areas. Using geographic religiosity as a measure of the social norms of the environment in which corporate managers operate, McGuire et al. (2012) report results supporting their prediction that firms headquartered within highly religious regions manage earnings less than those headquartered in less religious areas. These findings suggest that managers’ financial reporting decisions are influenced by noneconomic reputational and social costs of deviating from the social norms of the regions where managers live and operate. However, this research continues to use firm-level metrics and does not consider whether the observed religiosity effect varies from manager to manager.
The above studies implicitly adopt an assumption of managerial homogeneity in decision-making. Managers are viewed as interchangeable, homogenous actors in the decision making process (Bertrand and Schoar 2003), and it is assumed that their business decisions are influenced only by economic incentives, constraints, and/or characteristics of their business environments. Thus, any manager-specific characteristics are ignored, and implicitly assumed to be irrelevant to decision outcomes. More recently, however, researchers have begun testing this assumption by examining the relation between manager-specific characteristics and reporting decisions. Bamber et al. (2010) test whether individual managers and their personal demographic characteristics explain their firms’ voluntary financial disclosure decisions. They find that financial disclosure styles are related to individual managers’ functional career track, age cohort, military experience, and education. Brochet and Welch (2011) also investigate individual manager attributes and identify a positive relation between goodwill impairment decisions and managers’ financial expertise. Schrand and Zechman (2012) find that executive overconfidence can result in optimistic bias in earnings, reducing financial reporting quality below minimally acceptable levels. In summary, there is a growing body of research that suggests characteristics of individual managers help explain corporate disclosure and accounting decisions. Thus, the assumption that managers are homogenous decision makers may limit empirical research.
Some corporate finance studies further extend the literature investigating manager-specific characteristics by examining implications of differences in managers’ personal ideologies (Hutton et al. 2014, 2015; Jiang et al. 2012; Hong and Kostovetsky 2012; Chin and Parawada 2010). Hutton et al. (2014) use political contributions to proxy for managers’ personal ideologies and find that firms with more ideologically conservative managers invest in less R&D, are less leveraged, and maintain higher dividend payouts. Hutton et al. (2015) find that firms with more ideologically conservative managers violate securities and property rights (labor and environmental) laws less (more). Other research using political orientation metrics finds that ideologically conservative sell-side analysts forecast earnings more conservatively than their less conservative counterparts (Jiang et al. 2012), and conservative (liberal) money managers tend to overweight (underweight) stocks in industries aligned with the Republican Party (Chin and Parawada 2010; Hong and Kostovetsky 2012).
Using political contributions data, Lee et al. (2014) measure the ideological alignment between the CEO and independent directors to relate this homophily to governance weakness. They find this homophily is positively related to many forms of weakened corporate governance, including a greater propensity for fraud litigation. Unlike the other, above cited, studies using political contributions data, Lee et al. (2014) focuses on the alignment of ideology among a firm’s managers and directors rather than the level of ideology held by management.
In an international context, Gul et al. (2013) use Chinese data to examine the audit quality effects of several individual characteristics of the signing auditor. One such characteristic is the signing auditor’s membership in the Chinese Communist Party (CCP). Gul et al. (2013) use membership in the CCP as a proxy for the individual auditor’s desire for lax regulatory oversight, and find that when a signing auditor is a member of the CCP audits are of lower quality than those with where signing auditors are not CCP members.
In sum, there is growing evidence that manager-specific characteristics affect business decisions, and differences in managers’ personal ideologies can help explain these decisions. We extend this literature by examining the relation between managers’ personal ideologies and financial reporting quality.
3 Hypothesis development
We adopt Jost’s (2006, p 653) definition of ideology as “an interrelated set of moral and political attitudes that possesses cognitive, affective, and motivational components.” This broad definition allows ideology to encompass both moral and political beliefs and implies that ideology relates to several aspects of an individual’s personality. An individual’s ideology captures core personality traits and is stable over time (Gerring 1997).
Ideological dichotomies are common, and the vast majority of individuals relate to one ideology or the other as opposed to being in the middle or undecided (Jost 2006). An individual’s identification with an ideology results from the realization that one particular ideology is more consistent with characteristics of his or her personality (McClosky 1958). Research has identified several characteristics that differ across individuals with conservative versus liberal ideologies. Compared to their counterparts, individuals with conservative ideologies tend to be more risk averse (Kam and Simas 2010; Joe 1974; Peterson and Lane 2001), less open to new experiences (Joe et al. 1977), and value security more than excitement (Joe 1974; Feather 1979). Individuals with conservative ideologies also place more value on strict adherence to rules (Joe 1974; Feather 1979; Atieh et al. 1987). Lastly, conservatives are more likely than those with liberal ideologies to value hard work and delayed gratification (Joe 1974; Atieh et al. 1987). These factors, among others, affect individuals’ self-identification with a conservative or liberal ideology.
The theory of behavioral consistency suggests that personality “traits” are persistent, and this causes individuals to behave similarly across different situations (Burton 1963; Alker 1972; Epstein 1979; Sherman et al. 2010).Footnote 3 An individual’s personal ideology is one of these persistent traits, and therefore may affect an individual’s decisions in a consistent way across a variety of areas (McClosky 1958). The theory of behavioral consistency motivates the predictions of many studies examining the relation between personal ideology and decisions (Cronqvist et al. 2010; Hutton et al. 2015; Jiang et al. 2012). Behavioral consistency suggests that regardless of the context, individuals tend to make decisions that are consistent with their personal ideologies. Thus, when exercising discretion over financial reporting, managers will likely make reporting decisions that are consistent with their personal ideologies.
Regardless of personal ideology, managers face economic incentives that can motivate financial reporting choices, as well as consequences that may constrain opportunistic behavior. When facing similar incentives and constraints however, the above-cited research suggests that, all else being equal, managers with different personal ideologies may make different decisions. Those with conservative ideologies will be more risk averse, place more value on security and rule adherence, and be more accepting of delayed gratification (Kam and Simas 2010; Joe 1974; Feather 1979; Atieh et al. 1987; Peterson and Lane 2001). In the context of financial reporting choices, each of these characteristics of individuals with conservative ideologies is consistent with decisions leading to higher (rather than lower) financial reporting quality.
Low reporting quality is risky for both companies and their senior managers. DuCharme et al. (2004) find that both the likelihood of shareholder litigation and settlement amounts increase with levels of earnings management around new issues of equity. Jones et al. (2008) reports that some measures of earnings management are positively related to the incidence of financial statement restatements and financial statement fraud charges. Firms with fraudulent financial statements also experience increased costs of capital and reduced analyst following (Dechow et al. 1996). For CEOs and CFOs, mandated certification of financial statements (Sarbanes-Oxley Act, Section 302, 2002) has likely increased the personal litigation risk related to low quality financial reporting. All managers, regardless of their personal ideologies, will face these risks. Because of greater risk aversion and a preference for security over excitement, however, high quality reporting choices are consistent with the characteristics of managers with more conservative personal ideologies.
Additionally, financial reporting quality is ultimately defined by accounting standards and low quality reporting may also result from a violation of accounting rules. Accounting standards set forth parameters establishing a boundary between acceptable accounting practices and financial reporting violations. Undoubtedly, this boundary is not a bright line. Both U.S. GAAP and IFRS allow considerable reporting discretion and a spectrum of quality exists within the bounds of accounting standards. Managers with conservative ideologies, however, value strict conformity to rules and are thus likely to be more concerned about the possibility of crossing this line. A preference for strict rule adherence, along with the value placed on security versus excitement, suggests that high reporting quality choices are consistent with the characteristics of mangers with conservative ideologies.
Finally, short-term economic self-interest is perhaps the most well established motivation for lower reporting quality. For example, earnings management has been identified in contexts where managers are motivated by bonus plan parameters (Healy 1985). Managers who are more willing to delay their own economic gratification are likely less motivated to manage earnings in order to inflate short-term bonus payments. Thus, a characteristic of mangers with conservative ideologies, a high value placed on hard work and delayed gratification, is again consistent with high reporting quality.
In summary, risk aversion, delayed gratification, a preference for security, and strict adherence to rules may motivate managers with conservative personal ideologies to avoid earnings management and make financial reporting choices that result in higher reporting quality. This suggests the following hypothesis, stated in the alternative form:
H 1A
Firms whose managers have more conservative personal ideologies manage earnings less.
As noted above, low financial reporting quality can be associated with restatements (Jones et al. 2008). Financial reporting quality can be viewed as a continuum, where restatements lie at one extreme. Like low reporting quality more generally, a restatement imposes risk and potentially adverse consequences for both companies and their senior managers. Restatement announcements are associated with negative stock market reactions and the strength of this reaction is affected by a number of factors including the magnitude of the restatement and the presence of fraud (Palmrose et al. 2004; Palmrose and Scholz 2004). In addition, managers of firms making restatements are more likely to lose their jobs and have less desirable subsequent employment opportunities (Desai et al. 2006).
Unlike the degree of earnings management, restatement is, by nature, a binary metric. In general, a restatement indicates that the entity’s financial statements, as originally reported, did not fully comply with accounting standards. While there can be variation in reporting quality within the boundaries of accepted accounting standards, a restatement provides something akin to a bright line test for reporting violations. Thus, it is possible that conservative managers may consider some decisions that slightly reduce financial reporting quality to be acceptable as long as those reporting decisions do not violate accounting standards. Strict rule adherence, however, would be at odds with financial reporting decisions that violate accounting standards and trigger restatements.
Thus, risk aversion, a preference for security, and in particular, strict adherence to rules, may motivate managers with conservative personal ideologies to make financial reporting choices that reduce the probability that their companies’ financial statements will require restatement. This suggests the following hypothesis, stated in the alternative form:
H 2A
Firms whose managers have more conservative personal ideologies are less likely to restate their financial statements.
It is possible that a variety of mechanisms may reduce or eliminate the ability of ideology to influence financial reporting quality. These mechanisms may include financial statement audits, internal controls over financial reporting, oversight by audit committees and boards of directors, and reputational concerns. If these mechanisms are sufficiently effective, there may be no difference in reporting quality across ideology. Thus, the relation between managers’ personal ideology and financial reporting quality remains an empirical question and requires model specifications that control for governance and alternative mechanisms.
4 Methods
We empirically investigate the relation between managers’ personal ideologies and financial reporting quality. We construct three measures, each based on individual managers’ political contributions, as proxies for the conservatism of managers’ personal ideologies. We use two measures of financial reporting quality: the absolute value of discretionary accruals and the probability of a firm restating their financial statements. As described below, data were obtained from COMPUSTAT (financial statement data), CRSP (price and returns data), EXECUCOMP (manager and compensation data), Audit Analytics (restatements data), the Federal Election Commission of the United States (FEC) contributions files (political contribution data), IRRC (governance data), and the American Religion Data Archive’s Churches and Church Membership survey (religiosity data).
4.1 Personal conservative ideology
To develop our measures of personal ideology, we use managers’ individual political contributions as an indicator of their ideologies. Based on the theory of behavioral consistency, we use a metric that directly reflects political conservatism as a proxy for personal ideological conservatism. The underlying assumption is that managers who make more contributions to the Republican Party are more politically conservative, and more likely to follow conservative personal ideologies (Hutton et al. 2014). Gallup polls suggest that political party affiliation can be good indicator of conservatism. An average of 20 Gallup and USA Today/Gallup polls conducted in 2011 shows that 71 percent of Republicans describe their political views as conservative or very conservative, while only 20 percent of Democrats self-identify as having conservative views (http://www.gallup.com/poll/152021/conservatives-remain-largest-ideological-group.aspx). Extant empirical results are also consistent with this interpretation; companies whose managers are identified with more conservative political ideologies (based on contributions to the Republican Party) make more financially conservative decisions (Hutton et al. 2014).Footnote 4 Similarly, a recent study finds that firms with more Republican-leaning top managers have higher effective tax rates, which is consistent with a personally conservative ideology despite being at odds with the Republican Party’s commitment to lower taxes (Christensen et al. 2015).
We first construct a firm-level variable, CONSRVTOP5, measuring the ideological conservatism of the firm’s top managers following Hutton et al. (2014). This variable is constructed using data available in the FEC individual contribution files for the period 1980–2010 (Federal Elections Commission of the United States website, http://www.fec.gov). The FEC contribution files contain detailed information on contributions to every federal election candidate, arranged by election cycle, beginning with the 1979–1980 election cycle. Data in these files include the contributor’s name, employer, the amount of the contribution, and the recipient candidate and his or her party affiliation.
We use FEC contribution files to determine each individual’s political contributions during the period 1980–2010. For each individual,Footnote 5 we calculate an election-cycle conservatism score. Individuals receive a conservatism score of one if they make contributions during the election cycle to Republican candidates only, and zero otherwise.Footnote 6 Since individuals’ conservatism characteristics are best revealed over their entire lifetimes (Green et al. 2002), we combine election-cycle-level conservatism scores into a lifetime conservatism metric for each individual by calculating the equal-weighted average of each individual’s election-cycle-level conservatism measures. The lifetime conservatism metric is the percentage of times that an individual’s political contributions are made to Republican candidates only. Thus, the lifetime conservatism measure for each individual ranges from 0 to 1. It takes the value of 1 if the individual has made contributions exclusively to Republican candidates and 0 if, in every election cycle in which the individual has contributed, the individual made some contribution to Democrats. This approach sets a rigorous standard for classification as ideologically conservative, since only individuals making political contributions exclusively to Republican candidates will be coded as non-zero in any election cycle. The use of a lifetime metric also enhances the strength of the proxy for ideological conservatism. The lifetime score is based on long-term contribution histories and is more likely to reflect a persistent core characteristic of individual managers. Thus it is appropriate as a proxy for ideology and consistent with the concept of behavioral consistency.
As an example, assume that FEC data identifies George Smith, employed by Acme Corporation, as making contributions of $750 to Democrat candidates in the 1999–2000 election cycle and $2000 to Republican candidates during the 2009–2010 election cycle. The FEC data includes no other contributions made by George Smith during the period 1980–2010. George Smith’s conservatism score for the 1999–2000 election cycle would be zero since George made no contributions to Republican candidates in this election cycle.Footnote 7 For the 2009–2010 election cycle, George Smith’s conservatism would be scored as one since all of his contributions in this election cycle were made to Republican candidates. We would construct the lifetime score for George Smith as the average of 0 and 1, or 0.5. Therefore, the interpretation of the lifetime conservatism metric is that George Smith made political contributions exclusively to Republican candidates in 50 percent of the election cycles during which he made contributions to political candidates.
We construct the top 5 manager measure of conservatism, CONSRVTOP5, by merging the dataset of individual lifetime conservatism values computed as described above, with data from EXECUCOMP. EXECUCOMP contains data for a firm’s highest paid managers, so we are limited to matching data for these managers. In most cases, EXECUCOMP includes data for only the top five managers, however in a small number of cases data is included for more than five managers (for example, when there is turnover within the top 5 managers during the fiscal year). We include all managers available in EXECUCOMP in our computation of CONSRVTOP5. Following Hong and Kostovetsky (2012) and Hutton et al. (2014), if we fail to match a manager in EXECUCOMP to an individual in the FEC contributions file, we assign that manager a lifetime conservatism value of zero.Footnote 8 The firm-year value of CONSRVTOP5 is then computed as the weighted average of the top managers’ individual conservatism measures, with weight given in descending order of total compensation.Footnote 9 Therefore, CONSRVTOP5 is an index that is bounded by 0 and 1. A CONSRVTOP5 value of 1 means each of the firm’s top managers were identified as contributors in the FEC dataset, and each made contributions to Republican candidates only. A value of zero means that none of the managers were identified as having contributed exclusively to Republicans within any election cycle. We use the CONSRVTOP5 variable as a proxy for the firm’s top managers’ personal ideologies, where a higher value indicates a more conservative ideology.
We also consider two alternative personal ideology proxies for individual managers. These proxies are the lifetime conservatism scores (computed as above described) for the CEO and the CFO. For the measure CONSRVCEO (CONSRVCFO), we assign the individual conservatism score of the CEO (CFO) to the firm. For investigating our research question, we believe that the individual manager metrics are better proxies for personal ideology than the top five metric for two key reasons. First, many of the top managers included in the CONSRVTOP5 measure are not directly involved in accounting decisions. This has the potential to create noise in the measurement of the impact of manager ideology on financial reporting quality. Second, in many cases the CFO or CEO may not be among those listed in EXECUCOMP. In EXECUCOMP, the CEO is included in 72 percent of firm years in our sample, while the CFO is included in only 36 percent. Thus, separate ideology variables that are explicitly linked to the managers with the greatest influence over accounting may be more appropriate when investigating financial reporting quality.
4.2 Financial reporting quality measures
Estimates of discretionary accruals are among the most commonly used measures of reporting quality in both the accounting and finance literatures. Discretionary accruals serve as an inverse indicator of financial reporting quality. Our measure of discretionary accruals, or DA, is the performance adjusted discretionary accruals model as described in Gong et al. (2008).Footnote 10 Kothari et al. (2005) present simulation evidence on the properties of alternative measures of discretionary accruals and conclude that the discretionary accruals adjusted for performance are both well specified and powerful. To develop our measure of discretionary accruals, we first estimate the following cross-sectional model by quarter and two-digit SIC code:
TA is total accruals, Q is one of four indicator variables corresponding to the fiscal quarters, ΔSALE is the change in sales, PPE is beginning property, plant and equipment, LTA is beginning total accruals, and ASSET is beginning total assets. All variables in the model are scaled by beginning total assets. To mitigate the effect of outliers, we delete the top and bottom one percentiles of the scaled TA, ΔSALE, PPE, and LTA variables. We require at least 20 observations for each estimation. The performance matched discretionary accrual for the quarter, DA, is defined as residuals from the model (unadjusted abnormal accruals) performance-adjusted by subtracting the median abnormal accrual of the matching ROA quintile of the same quarter and two-digit SIC code. The annual performance matched discretionary accrual is then the sum of the four quarterly DA. Because earnings can be managed either upward or downward, we use the absolute value of the performance matched discretionary accrual, |DA| as our proxy for earnings quality.
Some research criticizes the use of discretionary accruals in earnings quality studies (Dechow et al. 1995; McNichols 2000; Nagy 2010; Paterson and Valencia 2011). McNichols (2000) and Dechow et al. (1995) suggest that researchers’ lack of knowledge about the behavior of unmanaged discretionary accruals justifies questions about the validity of results obtained when using discretionary accruals in reporting quality studies. Additionally, Patterson and Valencia (2011, p 1514) state that “Restatements strike at the heart of financial reporting quality,” suggesting that a restatement is more logically associated with deficient quality than is discretionary accruals. Restatements have also been used as a measure of deficient financial reporting quality in international contexts (Sue et al. 2013; Young et al. 2008). In addition, the line between earnings management and reporting violations necessitating restatement may be especially important for rule adhering, personally conservative managers. For these reasons we also investigate the relation between managers’ ideological conservatism and the probability of a restatement.Footnote 11
We identify restating firms using the restatements dataset available from Audit Analytics. We follow Paterson and Valencia (2011) by including only restatements coded as “accounting rule application failures” or “financial fraud, irregularities, and misrepresentation” and eliminating all restatements with a positive effect on earnings. Lastly, we eliminate non-restating years for restating firms.Footnote 12 For a control sample, we include only firms without restatements during our entire sample period. This creates clean restatement and control samples. We code a restatement indicator variable, RESTATE, with a value of one if the firm-year is in our restatement sample or zero if the firm-year is in our control sample. We use RESTATE as the dependent variable in logistic regressions to determine the relation between manager ideology and the probability of restating the firm’s financial statements.
We hypothesize that managers with conservative ideologies make decisions that increase financial reporting quality, i.e., they manage earnings and violate accounting standards less than managers whose personal ideologies are more liberal. Lower discretionary accruals (|DA|) and lower probabilities of restatement (RESTATE = 1) reflect higher financial reporting quality. Thus, we expect an inverse relation between CONSRV proxies and each of our financial reporting quality variables.
4.3 Control variables
To robustly test the relation between CONSRV and our financial reporting quality variables, we include multiple control variables. Based on prior research, we use different sets of control variables in the discretionary accruals and restatement models.
In our discretionary accrual regressions we include controls for factors that are documented in prior literature as related to reporting quality. Control variables expected to be negatively associated with the absolute value of discretionary accruals include cash flows from operations (Dechow et al. 1995), bankruptcy risk (Petroni 1992)Footnote 13 measured by Altman’s (1983) Z-score, and firm size (Becker et al. 1998). Control variables expected to be positively associated with the absolute value of discretionary accruals include sales growth (Menon and Williams 2004), leverage (DeFond and Jiambalvo 1994), and an indicator variable for reported losses (Francis and Yu 2009). We also incorporate controls based on volatility measures. We expect both the volatility in cash flows and the volatility in sales to be positively related to the absolute value of discretionary accruals (Hribar and Nichols 2007). Because earnings management is more likely in riskier firms, we include a control for risk measured as the standard deviation of the last 12 monthly stock returns, as well as the ratio of the market value of equity to the book value of equity as a control for capital market incentives to manage earnings (Francis and Yu 2009). We expect both risk controls to be positively related to our discretionary accruals dependent variable.
Control variables included in our restatements regressions are also identified based on prior research (Aier et al. 2005; Nagy 2010). We include controls for additional external capital demands (Dechow et al. 1996), leverage, (Aier et al. 2005), and loss avoidance incentives (Nagy 2010). Consistent with prior literature, we predict that each of these variables will be positively related to RESTATE. We also control for factors that prior literature demonstrates are negatively associated with restatements. These include size and free cash flow (Aier et al. 2005; Nagy 2010). Detailed variable definitions for each control are provided in Tables 1 and 2 and in the text that follows.
Additionally, McGuire et al. (2012) finds a positive relation between earnings quality and the religiosity of the geographical area where a firm is headquartered. They suggest that religious social norms influence managers’ financial reporting choices. Our variable of interest is the personal ideology of top managers. It is possible that our proxy for managerial ideology could capture or otherwise proxy for social norms related to geographical religiosity. For this reason, we include a geographical religiosity control variable in our regression models.Footnote 14 We measure religiosity, REL, as the percentage of the population of the firm’s headquartering county who self-identify as religious adherents as reported on the American Religion Data Archive’s (ARDA) Churches and Church Membership survey divided by the U.S. Census population for the county (as reported in the ARDA survey).Footnote 15 Since the survey is only administered once a decade, we linearly interpolate values following Hillary and Hui (2009).
Lastly, prior literature finds that aggressive financial reporting is both limited by the strength of a firm’s governance and incentivized by equity-related pay (Ghosh et al. 2010; Lee et al. 2014; Bergstresser and Philippon 2006; Cheng and Warfield 2005; Jiang et al. 2010; Chao and Horng 2013). It is particularly important that we control for the reporting quality effects of these factors because of the possibility that more conservative managers may self-select into firms with more conservative governance and compensation policies. Therefore, we control for both governance strength and managerial incentive pay. We follow prior literature by proxying for governance strength using the size of the board of directors and for managerial equity pay with the ratio of equity pay to total compensation for the CEO (Ghosh et al. 2010; Bebchuck et al. 2011).Footnote 16 We expect that financial reporting quality is positively (negatively) related to governance strength (executive equity pay).
4.4 Controls for manager self-selection
Hutton et al. (2014) describe potential problems associated with statistical inference inherent when trying to infer that managers’ characteristics drive a firm-level effect. A question arises about whether manager characteristics really explain the observed firm-level effect. An alternative explanation for an observed relation between manager characteristics and a firm-level effect is that managers simply self-select into firms with the effect. In this case, personal characteristics might explain a manager’s selection of a firm as an employer, but not necessarily the observed behavior of that firm. As in Hutton et al. (2014), this issue is relevant for our model specifications and interpretation of results.
As discussed above, the most obvious self-selection issue involves managers self-selecting into firms with governance and compensation policies that are aligned with their personal ideologies. To address the potential self-selection issue, we include governance and compensation controls that directly address this concern. In addition, we also include firm-level fixed effects indicators when estimating our discretionary accruals models (Lennox et al. 2012).Footnote 17, Footnote 18 Our governance and compensation controls directly address this concern. Hutton et al. (2014) and Chen et al. (2015) suggest using fixed effects to control for other firm-level factors that could be driving their results. Introducing firm-level fixed effects indicators has its own problems including a reduction in statistical power, making it more difficult for a relation to achieve significance. Hutton et al. (2014) reported diminished significance of their results when a firm-level fixed effects specification is used. Nevertheless, we use a firm-level fixed effects approach in our discretionary accruals specifications in order to address the potential for manager self-selection.
4.5 Empirical models and estimation
Equation (2) describes the model we use for testing the relation between managerial personal ideology and discretionary accruals:
where
- \(\left| {DA_{jt} } \right|\) :
-
The absolute value of firm j’s performance matched discretionary accruals in year t, estimated cross-sectionally as described in Gong et al. (2008).
- \(CONSRV_{jt}\) :
-
One of three measures of the conservatism of top managers’ ideologies at firm j during fiscal year t. These variables include CONSRVTOP5, CONSRVCEO, and CONSRVCFO measured as described in the text.
and controls are
- CFO jt :
-
Firm j’s Cash Flows from Operations (COMPUSTAT mnemonic OANCF) divided by total assets (AT), for fiscal year t
- GRO jt :
-
Firm j’s sales growth in fiscal year t, calculated as the percentage change in sales
- SIZE jt :
-
The natural log of firm j’s total assets at the end of fiscal year t
- DEBT jt :
-
The natural log of the ratio of the sum of firm j’s debt in current liabilities (DLC) and long-term debt (DLTT) to total assets at the end of fiscal year t
- D LOSSjt :
-
An indicator variable that takes the value of one if firm j’s income before extraordinary items (IB) is less than zero for fiscal year t, and zero otherwise
- Z jt :
-
Altman’s Z-score for firm j at the end of fiscal year t
- VOL CFjt :
-
Firm j’s volatility in Cash Flows from Operations in fiscal year t. This variable is measured as the standard deviation of the previous 3 years cash flows from operations
- VOL SALjt :
-
Firm j’s sales volatility in fiscal year t, calculated as the standard deviation of the previous 3 years sales values
- VOL RETjt :
-
The standard deviation of firm j’s monthly stock returns over fiscal year t
- MB jt :
-
The market-to-book ratio, which is the natural log of the market value of equity (CRSP mnemonic PRC times the number of shares outstanding SHROUT) divided by the book value of equity (SEQ), for firm j at the end of fiscal year t
- BrdSize jt :
-
The number of directors serving on firm j’s board during fiscal year t
- EBC jt :
-
The ratio of the sum of the fair value of restricted stock granted (STOCK_AWARDS_FV) + fair value of options granted (OPTION_AWARDS_FV) to total compensation (TDC1) during year t for the CEO of firm j
- REL jt :
-
the percentage of the population in firm j’s headquarters’ county identifying themselves as members of a religion during fiscal year t
- Firm Fixed Effects :
-
firm indicator variables for each unique GVKEY in our sample
We test for a relation between managerial ideology and the probability of a financial statement restatement by estimating a logistic regression of RESTATE on CONSRV and control variables described earlier. This regression model is described in Eq. 3:
where
- RESTATE jt :
-
An indicator variable that takes the value of one if firm j is a restatement firm that restated financial statements during fiscal year t
- CONSRV jt :
-
One of three measures of the conservatism of top managers’ ideologies at firm j during fiscal year t. These variables include CONSRVTOP5, CONSRVCEO, and CONSRVCFO measured as described in the text
and controls are
- FREEC jt :
-
Free Cash Flow, measured as firm j’s Cash Flows from Operating Activities less average capital expenditures, divided by lagged total assets, over fiscal year t
- FINRAISED jt :
-
The sum of proceeds from new debt issued and proceeds from new equity issued divided by lagged total assets for firm j over fiscal year t
- LAGD LOSSjt :
-
An indicator variable that takes the value of one if firm j’s lagged income before extraordinary items (IB) is less than zero, or else it is equal to zero for fiscal year t
- SIZE jt :
-
The natural log of firm j’s total assets at the end of fiscal year t
- LEV jt :
-
Firm j’s total liabilities divided by lagged total assets at the end of fiscal year t
- BrdSize jt :
-
The number of directors serving on firm j’s board during fiscal year t
- EBC jt :
-
The ratio of the sum of the fair value of restricted stock granted (STOCK_AWARDS_FV) + fair value of options granted (OPTION_AWARDS_FV) to total compensation (TDC1) during year t for the CEO of firm j
- REL jt :
-
The percentage of the population in firm j’s county identifying themselves as members of a religion during fiscal year t
Our formal test statistic of our hypotheses is the test statistic for the coefficient estimates of \(\partial_{1}\) and \(\Delta_{1}\). We predict \(\partial_{1} < 0 \,{\text{and }}\,\Delta_{1 } < 0\), so we interpret the statistical significance of the results using one-tailed tests.
5 Results
5.1 Sample description, summary statistics, and correlation
Our samples are comprised of all observations with all data available to meet several data requirement screens. Since EXECUCOMP data was unavailable prior to 1992, we limit our sample to observations with fiscal years occurring after 1992. Election contribution data from the FEC and geographical religiosity data from the ARDA were not available after 2010, so this limits our sample to observations with fiscal years ending no later than 2010. Consistent with prior literature, we exclude all utilities (SIC between 4900 and 4999) and financial companies (SIC between 6000 and 6999) because managers have less discretion over firm policies in these heavily regulated industries (Francis and Yu 2009; Hutton et al. 2014). All observations must have the required data available to calculate all variables in the estimated models. All continuous variables are winsorized at the top and bottom one percent levels.
Using the matching procedure described above, we are able to successfully match managers in both the FEC and EXECUCOMP databases for 35,948 firm-years. We deleted 16,737 firm years that lacked COMPUSTAT, CRSP, or ARDA data needed to compute discretionary accruals and control variables or operated in either the financial or utilities industries. Additionally, requiring BRDSIZE and EBC reduces our sample size by an additional 7981 observations that lack IRRC data. Using the personal ideology proxy, CONSRVTOP5, our final sample includes 11,230 firm-year observations representing 2154 unique firms, spanning fiscal years from 1992 to 2010. When using CONSRVCEO (CONSRVCFO) as the personal ideology proxy, the same sample attrition process yields a final sample of 11,230 (9335) firm-year observations. The sample for CONSRVCFO is smaller because we exclude firms where the CFO is not available in EXECUCOMP.
Our restatements analysis sample is comprised of a sample of restating firm-years and a control sample of non-restating firm-years. The initial sample was obtained from the Audit Analytics-Restatements and Audit Analytics-Audit Opinions datasets. We started with 1656 non-positive effect restatement firm-years and 8721 control firm-years that have non-missing data in FEC and EXECUCOMP datasets, spanning fiscal years 2003–2010.Footnote 19 We eliminated 551 restatement observations and 1099 control sample observations that lacked data necessary to compute the REL variable. Then, we deleted 62 restatement firm-years that were unrelated to accounting judgments.Footnote 20 Lastly, we eliminated 388 restatement observations and 2269 control sample observations that lacked data necessary to compute BRDSIZE and EBC. Our final sample is comprised of 655 restatement observations and 5353 control observations (6008 in total) when estimating the model using the CONSRVTOP5 proxy. The same sample attrition process yields a final sample of 6008 (5494) observations for the model using the CONSRVCEO (CONSRVCFO) proxy.
Table 1 reports descriptive statistics for the discretionary accruals samples. Each panel includes the mean, standard deviation, quartile 1 (25th percentile), median, and quartile 3 (75th percentile) for each of our three variables of interest (ideological conservatism CONSRVi, the absolute value of discretionary accruals |DA|) and control variables. All values reported are winsorized at the top and bottom one percent levels.
Descriptive statistics for the discretionary accruals sample using CONSRVTOP5 are reported in Panel A. The sample mean (median) value for CONSRVTOP5 is 0.24 (0.15). These values are very close to those reported in prior research (Hutton et al. 2014). The observed mean indicates that the typical firm in our sample has managers with an overall ideology that is somewhat conservative, but our sample shows substantial variation in the conservatism of managers’ personal ideologies. The 75th percentile (Q3) is 0.44, the 25th percentile (Q1) is 0.00, and the standard deviation is large (0.26). This suggests that the sample is not pooled within either ideology, enabling us to use our variable to effectively separate companies led by conservative managers from those led by non-conservatives. Reported descriptive statistics for other variables are generally consistent with those reported in recent studies (Francis and Yu 2009; Krishnan et al. 2011; McGuire et al. 2012; Hilary and Hui 2009; Ghosh et al. 2010; Bebchuck et al. 2011).Footnote 21
Panels B and C report statistics for the CONSRVCFO and CONSRVCEO samples, respectively. The sample mean and median for CONSRVCFO (CONSRVCEO) are 0.15 and 0.00 (0.34 and 0.00), respectively. Thus, the scores for CFOs (CEOs) are slightly less (more) conservative than the weighted average score for the top managers. While the average CFO and CEO has a slightly conservative ideology, the standard deviation remains high (0.33 and 0.42 for the two samples, respectively), suggesting that these CONSERV variables continue to effectively separate the sample based on the extent to which the manager’s personal ideology is conservative.
Descriptive statistics for the samples used in the restatement analyses are reported in Table 2, with Panels A, B, and C presenting statistics for the CONSRVTOP5, CONSRVCFO, CONSRVCEO samples, respectively. Descriptive statistics for CONSRVTOP5, CONSRVCFO, CONSRVCEO, REL, and SIZE are very close to those for the discretionary accruals model reported in Table 1. Descriptive statistics for other variables of interest are generally consistent with those reported in other recent studies (Aier et al. 2005; Francis and Yu 2009; Ghosh et al. 2010; Bebchuck et al. 2011; Nagy 2010).
Table 3 presents the correlation matrices of variables used in our discretionary accrual analyses, including control variables. Panels A, B, and C present the correlation matrices for the variables in the CONSRVTOP5, CONSRVCFO, and CONSRVCEO samples, respectively.
Correlations reported in Table 3 provide some bivariate support for our hypothesis with a significantly negative correlation between all three of our conservative ideology proxies and our discretionary accruals dependent variable reported in Panel A. The correlations between REL and |DA| is significantly negative in each of our samples, suggesting that Hillary and Hui’s (2009) measure of religiosity likely captures a similar construct to the McGuire et al. (2012) measure.
We include a control for geographical religiosity in our models because of the potential overlap it might have with our measure of management’s ideology. Correlations reported in Table 3 shed some light on the relation between these variables. There is a small but statistically significant negative correlation between the geographical religiosity variable, REL, and both CONSRVTOP5 (Panel A) and CONSRVCEO (Panel C), but an insignificant relation with CONSRVCFO (Panel B). This suggests that these variables, while conceptually related, may be capturing different constructs.
Table 4 presents correlation matrices for variables used in our restatement analyses. Panels A, B, and C present the correlation matrices for the variables in the CONSRVTOP5, CONSRVCFO, and CONSRVCEO samples, respectively. Here we find bivariate support for our hypothesis with a significantly negative correlation between each of the three CONSRV ideology proxies and our dependent variable, RESTATE. The geographical religiosity variable, REL, is negatively correlated with RESTATE in all three (significant in two) of the panels. As in our discretionary accrual sample, there is a significantly negative correlation between REL and two of our CONSRV proxies, CONSRVTOP5 and CONSRVCEO (Panels A and C).
5.2 Multivariate regression results
In Table 5 we report the results of our multivariate tests of the relation between earnings quality and management’s ideology using our discretionary accruals model. We estimate multiple regressions of |DA| on CONSRVTOP5 and a variety of controls. Model 1 is a regression of |DA| on CONSRVTOP5 and our primary set of control variables, but excluding controls for religiosity, governance strength, and equity incentives. Model 2 is a regression of |DA| on CONSRVTOP5 and all non-religiosity control variables. Model 3 (Model 4) is a replication of Model 1 (Model 2), but includes the geographic religiosity control variable, REL. To control for the potential self-selection bias we include firm level fixed effects in each regression.
Consistent with our first hypothesis, in each of the four models estimated we find significantly negative coefficients on our independent variable of interest, CONSRVTOP5. Additionally, control variables are generally significant, with signs that are consistent with predictions.Footnote 22 The control variables CFO and Z had, in some specifications, significant coefficients with signs that were inconsistent with those predicted. We note that other similar studies report coefficient signs that are insignificant or inconsistent with predictions for some control variables (Francis and Yu 2009; Boone et al. 2010). The reported results support our hypothesis, H1A, that greater ideological conservatism of management is associated with higher financial reporting quality (lower absolute magnitude of discretionary accruals).
The previously discussed conceptual relation between political conservatism and religiosity warrants careful consideration. We incorporate the control for geographic religiosity in Models 3 and 4. REL is significantly related to |DA| only in model 3, suggesting that once controlling for governance and equity-based pay there is no relation between REL and earnings quality. The controls for governance and equity-based pay, however, do not diminish the significance of the coefficient on our proxy for personal ideology, CONSRV.
Table 6 presents the results of our full sample logistic regressions of RESTATE on CONSRVTOP5 and the control variables. Consistent with our second hypothesis, in each regression specification we find negative coefficients on CONSRV that are significant at the one percent level. Additionally, coefficients on REL are significantly negative, consistent with the findings of McGuire et al. (2012). All control variables have the predicted signs or are insignificant.
The results presented in Tables 5 and 6 provide strong and consistent support for our hypotheses. The proxy for management’s ideology, CONSRVTOP5, however, is a firm-level metric that represents an average across the most highly compensated managers. Since not all of the managers included in this metric are directly involved in financial reporting decisions, we develop a manager-specific proxy for the personal ideology of the CFO, CONSRVCFO. This allows us to examine the relation between the personal ideology of the manager most directly responsible for financial reporting decisions and financial reporting quality. We thus re-estimate all the discretionary accrual and restatement models using CONSRVCFO instead of CONSRVTOP5.
Results for the discretionary accruals models are presented in Table 7 and are similar to those reported in the CONSRVTOP5 models. It is notable that despite the reduced sample size, the coefficients on CONSRVCFO remain significantly negative in all four models estimated. One control variable’s coefficient, CFO, has a coefficient that is significant with a sign that is inconsistent with the predicted direction in two of four specifications, but all remaining control variables have predicted signs or are insignificant.
Table 8 presents results of the restatement models estimated with the CONSRVCFO proxy. These results are quite similar to those reported in Table 6. The coefficients on the CONSRVCFO variables are significantly negative at conventional levels in each of estimated models. The coefficients on REL are significantly negative. All control variables either have the predicted signs or are insignificant.
Finally, we report results using a second manager-specific proxy for conservative personal ideology, CONSRVCEO, in Tables 9 and 10. CEOs are not necessarily directly involved in financial accounting and reporting decisions. Nevertheless, the CEO can set a tone at the top that may be important for those directly making accounting and reporting decisions. We find significantly negative relations between CONSRVCEO and |DA| in only our two most limited specifications (Table 9), but we continue to find a significantly negative relation with restatements (Table 10). This suggests that CEOs and their ideologies may not influence accounting choices affecting financial reporting quality within the boundaries of accounting standards, as captured by the absolute value of discretionary accruals. Restatement is a more rigorous measure of financial reporting quality than discretionary accruals. In a spectrum of financial reporting quality, restatement lies further toward the lower end of the spectrum. Our results suggest that CEOs’ personal ideologies help to explain financial reporting choices associated with the most extreme deviations from expected thresholds of reporting. These results further suggest that the CEO’s ideology may not be driving the results reported for CONSRVTOP5.
In sum, we find evidence that companies with CFOs and top management teams having more conservative personal ideologies are less likely to aggressively manage earnings and violate GAAP and mixed evidence about impacts of the CEO’s ideology. It’s also important to note that these relations are economically significant. The coefficient of interest in the full discretionary accruals model when using CONSRVTOP5 (CONSRVCFO) is estimated to be − 0.77 (− .23), suggesting that a one unit shift in the conservatism level (moving from non-conservative to fully conservative) for the top five management team (CFO) is estimated to reduce discretionary accruals by 0.77% (0.23%) of lagged total assets. Considering that absolute discretionary accruals are 6% of lagged total assets, our results suggest that a one unit increase in the conservatism level of the top five managers (CFO) is associated with a 12.8% (3.8%) reduction in discretionary accruals. Also, the estimated marginal effects of our logistic regression coefficients on CONSRVTOP5 (CONSRVCFO) [CONSRVCEO] evidence that a one unit shift in the conservatism level for the top 5 management team (CFO) [CEO] reduces the probability of restatement by approximately 9.53% (3.86%) [3.71%].
6 Additional analysis
In our primary analyses, we find support for both of our hypotheses. With the exception of CONSRVCEO in the discretionary accruals model, our regression results consistently report a significantly positive relation between our measures of managers’ ideological conservatism and financial reporting quality. In this section we describe tests for the robustness of our results.
6.1 Effects of Sarbanes–Oxley
We analyze the robustness of our results to changes in opportunities for earnings management as a result of the Sarbanes–Oxley Act (SOX). Prior literature documents stronger internal control systems and less opportunity for earnings management after implementation of Sarbanes–Oxley (Iliev 2010; Lobo and Zhou 2006; Cohen et al. 2008; Ge and McVay 2005). We control for the potential effects of Sarbanes–Oxley on our results by replicating the full model analyses reported in Tables 5 through 10 using only the subsample of observations from the post-SOX period. To accomplish this, we limit observations to those for years 2004 through 2010. For comparative purposes we present results of our full models using each of our three conservatism proxies (CONSRVTOP5, CONSRVCFO, and CONSRVCEO) side by side.
Table 11 reports the post-SOX sample period regression results for the discretionary accruals models. Consistent with results of our primary analyses, we find significantly negative relations between |DA| and CONSRVCFO, but insignificant relations between |DA| and both CONSRVTOP5 and CONSRVCEO. Thus, Sarbanes–Oxley appears to have limited the impact of the ideology of some managers on the quality of earnings, but the CFO’s ideology still matters.
Table 12 presents the results of our restatement regressions estimated in the post-SOX period. Consistent with our second hypothesis and results reported in Tables 6, 8, and 10 for the full sample period, the coefficient on each measure of CONSRV is significantly negative in all models. These results suggest that even in the post-SOX period, firms with more conservative managers report higher quality earnings than those with less conservative managers.
6.2 Controls for manager self-selection
We interpret our results as evidence that managers’ personal ideologies explain financial reporting quality. An alternative explanation is that managers with conservative ideologies may seek out companies that have governance policies consistent with these ideologies. If managers with conservative ideologies self-select into firms with policies that reduce incentives for and/or increase constraints on low reporting quality, then we might observe results consistent with our hypotheses because the firms where conservative managers choose to work share characteristics that affect observed reporting quality. Some of these firm characteristics might be consistent with conservative ideologies that value rule adherence, risk reduction, and security. As previously described, we use governance control variables and a firm fixed effects approach in our |DA| regressions to control for these factors, but concerns about a potential self-selection bias may remain.Footnote 23
To further address the potential self-selection issue, we re-estimate our discretionary accruals analyses using a changes specification.Footnote 24 This approach addresses self-selection by controlling for the level of any firm-specific factors that could be driving our reported personal ideology results. We measure changes in discretionary accruals as the year-over-year change in the absolute value of performance-matched discretionary accruals. We regress this variable on the changes in CONSRV proxies and changes in control variables. Results of estimating these changes models are reported in Table 13. Once again, the coefficient on CONSRVCFO, in this case specified as ∆CONSRVCFO, is significantly negative at the five percent level. Additionally, the conservatism of the top five managers, ∆CONSRVTOP5, is significantly negative.Footnote 25 As in other specifications, the proxy for the CEO’s ideology is insignificant. Thus, even after introducing rigorous controls for potential self-selection bias, we find evidence that managers’ personal ideologies explain their firms’ reporting quality decisions, but the personal ideologies of individual managers within a firm may differentially influence decisions including those related to financial reporting quality.
6.3 Alternative measure of management’s personal ideology
In our study we proxy for each manager’s personal ideology using the ideology measure from Hutton et al. (2014). This measure implicitly assumes a relatively high hurdle for a manager to be considered conservative; a manager is considered conservative in a period only if 100 percent of his or her political contributions are made to Republican candidates. Otherwise the manager is considered non-conservative for that period. Some readers may question this measure, leading to concerns about whether or not our results would hold under an alternative specification, particularly one that assumes ideology is more continuous. Accordingly, we re-estimate our main analyses using an alternative ideology proxy used by Hong and Kostovetsky (2012).
Hong and Kostovetsky (2012) measure the ideology of investment fund managers using the natural log of the absolute value of the difference between each manager’s lifetime contributions to Democrats and the manager’s lifetime contributions to Republicans, and the value is multiplied by negative one if the manager’s Republican contributions are greater than the manager’s Democrat contributions. As in the previous analyses, we examine the ideology of the CEO, the CFO, and a compensation rank weighted average of the ideologies of the top executives. Note however that this proxy for ideology, LN(NETDEM) has the opposite sign as the CONSRV proxies used in our main analyses, and results consistent with our expectations will have a positive coefficient. The results of these analyses are reported in Tables 14 and 15, and are consistent with those reported for our primary analyses with the exception that the coefficient on CONSRVCEO is insignificant in the restatements specification reported in Table 15.
6.4 Signed discretionary accruals
In our primary analyses we use unsigned discretionary accruals in each of our discretionary accruals models to capture the management of earnings both upwards and downwards. However, an argument could be made that conservative managers may be less prone to manage earnings upward while also being more prone to manage earnings downward than their less conservative counterparts.Footnote 26 To test this possibility, we re-estimate our primary model using signed discretionary accruals, only positive discretionary accruals, and only negative discretionary accruals using Gong et al. (2008) discretionary accruals as the dependent variable. The results are tabulated in Table 16 where, for brevity, we include only the results of estimating the CONSRVCFO specification. For signed, positive only, and negative only discretionary accruals models the coefficient of interest is significant and in the predicted direction. We find similar results when estimating the model using CONSRVTOP5. However, consistent with most of our analyses, the coefficient of interest was insignificant when using the CONSRVCEO specification.
6.5 Additional controls to address potential alternative explanations for our results
Our primary variable of interest, CONSRV, is measured using the political contributions of firms’ top managers. However, it is possible that the political contributions of managers could be correlated with other related factors that drive our results. These potential factors include the level of a firm’s political sensitivity, the level of a firm’s lobbying efforts, the strength of the firm’s governance systems, the complexity of the firm’s operating activities, the strength of the firm’s internal control system, and the sensitivity of executive options based pay to company performance.Footnote 27
Mills et al. (2013) suggest that more politically sensitive firms receive greater scrutiny from outside groups, and this scrutiny acts as an additional layer of governance. They find that increased scrutiny results in politically sensitive firms paying higher taxes. In the context of our study, extra scrutiny may also result in politically sensitive firms having less incentive to manage earnings. Mills et al. (2013) proxy for political sensitivity using the proportion of revenue from government contracts. However, companies that are managed by loyal political donors may more easily secure government contracts. Therefore, the political sensitivity of our sample firms may provide a plausible alternative explanation for our results.
Additionally, Skaife et al. (2013) test the relation between a firm’s lobbying activities and CEO pay. They hypothesize that corporate lobbying increases the probability of policies that stimulate excess CEO pay, but at the shareholder’s cost (both the expense of lobbying and the excess pay extract wealth from the firm). Thus, corporate lobbying may evidence an agency problem at the firm, and this agency cost may take forms other than excess CEO pay. Similarly, excessive earnings management and financial statement misstatements may also indicate an agency problem. Additionally, managers who allocate corporate resources to lobbying are likely to also use their personal resources to lobby political candidates. Therefore, another potential correlated variable, not included in our primary models, is companies’ lobbying efforts.
As discussed earlier, conservative managers may be attracted to companies with stronger governance. So it is important that we control for various dimensions of a firm’s governance system. Several studies find that insiders on the board, including the CEO as chairman of the board, weaken governance and financial reporting quality (Beasley 1996; Klien 2002; Dechow et al. 1996). Other literature finds that financial statement audits performed by smaller audit firms are also related to lower financial reporting quality (Becker et al. 1998; Lennox and Pittman 2010). Although our primary models incorporate some governance controls, one might argue that another dimension of governance or audit quality could be driving our results.
Further, financial reporting may be more complex for some firms than for others, and this reporting complexity may result in higher levels of discretionary accruals and probability of restatement. Complex financial reporting likely stems from complex operating activities since complex operating activity frequently requires complex accounting. Several studies suggest a connection between complexity in the financial reporting process and operational complexity (Kinney et al. 2004; Ge and McVay 2005; Ashbaugh-Skaife et al. 2007). These studies investigate the impact of operational complexity factors including the number of operating segments, the level of foreign sales reported, and the level of merger and acquisition activity. If operating complexity impacts financial reporting quality and is correlated with the level of managerial conservatism, then one might argue that operating complexity might explain our results.
Moreover, prior research finds that financial reporting quality is impacted by internal control strength (Doyle et al. 2007). These researchers suggest that weak internal controls provide an opportunity for accounting errors and irregularities to impact the financial statements, reducing financial reporting quality. If internal control system weaknesses are risky, it may be that more conservative managers self-select into firms with stronger internal control systems. One might question whether internal control strength is the correlated omitted variable that explains our results.
Lastly, although we control for executive compensation effects on incentives to report aggressively, our control variable, EBC, may not fully capture every relevant aspect of executive compensation systems. It may be possible that some executive compensation system characteristics, not fully captured in our EBC control variable, might incentivize executives to aggressively report (Burns and Kedia 2006). This is likely a greater issue when the executive’s compensation is significantly composed of stock options with values that are highly sensitive to the level and volatility of the company’s stock. If conservative managers are less attracted to firms offering these types of compensation, then the performance sensitivity of options based compensation may provide an alternative explanation for our results.
To control for the effects of political sensitivity, lobbying, governance and auditor strength, internal control strength, sensitivities of options based executive compensation, and operating complexity we re-estimate our primary regression models with controls for each of these factors. We follow prior research in formulating our measures of firm political sensitivity, POLSENS (Mills et al. 2013), and firm lobbying activities, LOBBY (Skaife et al. 2013). We compute POLSENS as the product of each firm’s percentage of total revenues from federal contracts over a 3-year period and an indicator variable that is equal to one if the firm’s federal contract revenues are in the top 10 percent of federal contract dollar recipients in a given Federal fiscal year. We compute LOBBY as an indicator variable that is equal to one if the firm was a client of a registered lobbying firm during a given fiscal year.Footnote 28 We also include a commonly used measure of auditor size, BIGN (Becker et al. 1998), as a proxy for auditor quality. BIGN is an indicator variable that is equal to one only if the company used a Big 4, Big 5, or Big 8 auditor. Additionally, we control for the effects of insiders being on the board of directors by including the percentage of independent board members, %BOARD, and an indicator variable that is equal to one if the CEO is also the chairman of the board, CEOCHAIR (Dechow et al. 1996). We also include an internal control weakness indicator, MW, that is equal to one if the firm reported a material weakness in internal controls over financial reporting in either management’s assessment or the auditor’s report on internal controls, or zero otherwise (Doyle et al. 2007). Further, we control for the effect of option-based compensation on management’s incentives for aggressive reporting by including the delta and vega of managements’ option-based compensation in our regressions. Delta (vega) is the change in the value of the manager’s options based compensation portfolio due to changes in the value (volatility) of the company’s stock. We follow the method developed by Core and Guay (2002) and Burns and Kedia (2006) for computing the inputs for DELTA and VEGA from data publicly reported in firms’ year-end financial statements.Footnote 29 For each year, we compute the DELTA and VEGA for the CEO’s portfolio, the CFO’s portfolio, and the compensation-rank weighted average of the delta and vega of the top management team. Lastly, we include the number of distinct business segments (SEG), the level of foreign income reported as a percentage of lagged total assets (FORINC), and the number of merger and acquisition deals announced within the current fiscal year (M&A).
Results of estimating our main regression models augmented to include the added control variables are reported in Tables 17 and 18. These results provide evidence that our previously documented relations persist even after including controls for a firm’s political sensitivity, lobbying activity, governance, internal control strength, executive compensation sensitivities, audit quality, and operational complexity. In fact, when we include controls for these factors we find that each measure of managerial ideology is significantly related to both absolute discretionary accruals and the probability of restatement. This suggests that the reported positive relation between financial reporting quality and the conservative ideologies of top managers is not likely driven by these factors.
6.6 Alternative measures of financial reporting quality
In each of our earlier analyses we use one of two measures of financial reporting quality that are common in the accounting and finance literatures. Those measures include estimates of discretionary accruals and financial statement restatements. Our results are robust to using either measure and under many different model specifications. However, one might question if the tested relation between financial reporting quality and management’s personal ideology holds when using other measures of financial reporting quality.
We address this question by re-estimating our full model (including all added control variables) on two additional measures of financial reporting quality. The first additional measure of financial reporting quality is the accruals quality measure suggested in Dechow and Dichev (2002). Dechow and Dichev (2002) regress current period, one period lagged and one period leading cash flows from operating activities on the current period change in working capital accruals. We do the same, estimating the model annually, by industry (2 digit SIC), over the 7 year period from 1996 to 2002. The measure used is the standard deviation of the absolute value the model’s residuals.
The second additional measure of financial reporting quality is a variant of the previously described measure. We follow Doyle et al. (2007) in using the variant of Dechow and Dichev’s accruals quality measure as modified by McNichols (2002) and Francis et al. (2005). This measure is also the standard deviation of the absolute residuals from a Dechow and Dichev (2002) style regression. But in this case we add to their model both the current period change in sales and the current period level of property, plant and equipment.
We re-estimate the models from Table 17, replacing the dependent variable with either Dechow and Dichev’s (2002) measure of accruals quality or the modified measure of accruals quality described above. For each model (Top5, CFO, and CEO) and when using either variant of the accruals quality measure the coefficient of interest is significantly negative at the 5% level (untabulated).
7 Conclusions
We investigate whether managers’ personal ideologies help explain financial reporting quality. We find that when a firm’s managers have more conservative personal ideologies, financial reporting quality is higher. We use manager-specific political contribution data to develop three proxies for the conservatism of managers’ personal ideologies. We examine the relation between these proxies and two metrics of financial reporting quality; discretionary accruals and the probability of financial statement restatements. Interpretation of results is consistent with our hypotheses and similar across model specifications with the exception of generally insignificant relations between CEO ideology and discretionary accruals. CEOs have less direct involvement in day to day accounting decisions than CFOs, so it is not surprising to find insignificant results in the CEO specifications. Results remain consistent in analyses examining the post-SOX period when opportunities for earnings management are reduced, when rigorous controls for the potential effects of self-selection bias are incorporated, in analyses using alternative measures of ideology and financial reporting quality, after controlling for several additional components of the firm’s governance; complexity; internal control strength; executive compensation sensitivity to performance; political sensitivity and lobbying activity, and when using signed discretionary accruals in place of unsigned discretionary accruals.
Our results contribute to an emerging literature that reconsiders assumptions of managerial homogeneity. Results of our analyses support the idea that individual managers, and their personal characteristics, can make a difference in their firms’ behaviors and outcomes. While individual demographic characteristics may help explain outcomes, it becomes difficult to aggregate across multiple dimensions of individuals. We use a more aggregate measure of manager-specific characteristics, i.e., personal ideology, to make predictions. Managers with conservative ideologies are more risk averse, place more value on delayed gratification and hard work, and are more focused on strict rule adherence than their more liberal counterparts. We provide evidence that ideology, an individual trait that subsumes these characteristics and values, partially explains financial reporting quality. While it is challenging to develop a proxy for an unobservable and aggregate personal characteristic such as ideology, our approach has several advantages. We impose a rigorous standard for identifying managers with conservative ideologies that requires behavior (i.e., political contributions) that is both unambiguous and consistent over a long horizon. This reduces the risk of misclassification and incorporates evidence about the persistence of behavior that is an important dimension of ideology.
The reported relation between managers’ personal ideologies and financial reporting quality has several important implications. Our results suggest that further investigation of manager-specific factors has the potential to better explain financial reporting quality. Financial reporting quality is related to investment risk, so parameters related to reporting quality have important economic implications for investors. In addition, when considering candidates for management positions, boards of directors may be interested in understanding the implications that managers’ personal characteristics may have for firms and shareholders. Auditors conduct risk assessments that include consideration of subjective factors such as the tone set by managers. Auditors may be able to improve their risk assessments by incorporating publicly available information capturing managers’ personal ideologies that is related to financial reporting quality.
Although our results appear robust to many controls and alternative specifications, we acknowledge limitations. Ideology is a personal characteristic and reliance on public information may introduce important limitations in developing a proxy. Additionally, as is typical in empirical research, severe sample attrition may limit the generalizability of our findings, as our final sample may not fully represent the population of publicly traded companies or of companies in general. We caution readers of these limitations when interpreting our findings.
Notes
We use the term ideology to refer to “an interrelated set of moral and political attitudes that possesses cognitive, affective, and motivational components (Jost 2006).” This definition makes clear that the concept of ideology is much broader than an individual’s political views. However, consistent with prior work in finance (Hutton et al. 2014; Jiang et al. 2012; and Hutton et al. 2015), social psychology (Jost 2006), and political science (Gerring 1997) we simplify this multidimensional concept by assuming that individual ideologies all lie somewhere on a single ideological continuum running between conservative and liberal. Therefore, our use of the terms “conservative” and “liberal” refers to the direction or position of an individual’s personal ideology on this continuum.
We use the term “earnings quality” to refer to the subset of “financial reporting quality” that is related to amounts included in the financial statements and specifically in the determination of net income. We use the term “financial reporting quality” to refer to the overall quality of the financial statements and the related disclosures.
Although Bamber et al. (2010) cite research in psychology, they motivate their investigation, at least in part, by the finding that humans develop long-lived, individual communication styles (e.g., Pennebaker and King 1999). Results of Bamber et al. (2010) are consistent with persistent differences in manager-specific disclosure styles. This finding is consistent with the theory of behavioral consistency.
It is possible that some managers’ social views dominate their fiscal views, allowing fiscally liberal (conservative) managers who vote and contribute as social conservatives (liberals) to appear in our dataset differently than their fiscal views would dictate. However, this should only bias against finding results consistent with our hypotheses.
To aggregate contribution data by individual donor, we construct an individual identifier based on first name, last name and employer name. We use these three variables because they are the only common variables across the FEC contributions files and the EXECUCOMP database used to link the contributions data to COMPUSTAT financial data. Because of the size of the databases, we use a matching algorithm created in SAS. Due to the particulars of SAS, we have to thoroughly clean the data for non-letter characters, titles, etc. For firm name, we choose the longest name of the employer after removing several common terms (e.g., CORP, CO., etc.).
Since it is not straightforward to identify and classify the conservatism level of minor or third parties, we exclude all contributions that are not to candidates identified as Republican and Democrat.
If George Smith had made contributions to both Republican and Democrat candidates in the same election cycle (1999–2000 in this case), the conservatism score would have been coded zero since not all contributions were to Republican candidates.
It is essential to code managers in EXECUCOMP but not matched to FEC contributions data as zero rather than missing when constructing the CONSRVTOP5 proxy since this metric is an average of the top managers’ individual scores. If missing values were not coded to zero an extreme constraint would be placed on sample size (i.e., only cases where all listed managers were matched with FEC contributions data would be included).
Following Hutton et al. (2014), the weight assigned to the ith (of 5) highest paid manager’s political conservatism wi, is: \(w_{i} = \frac{{i^{ - 1} }}{{\sum\nolimits_{i = 1}^{5} {i^{ - 1} } }}\)
We also estimated our discretionary accruals models using discretionary accruals from (a) a cross-sectional modified Jones model, as recommended by Keung and Shih (2014) and (b) a cross-sectional performance adjusted modified Jones model, as recommended by Kothari et al. (2005). Our results were qualitatively unaffected by the use of these alternative discretionary accruals models.
We also considered the relation between financial reporting quality and conditional conservatism. However, upon further investigation we discovered that other researchers question the connection between financial reporting quality and conditional conservatism. Ruch and Taylor (2015) said, “In general, it is unlikely that conditional conservatism would have a significant relationship with earnings management.” Therefore, we do not investigate the relation between management’s ideology and conditional conservatism.
As a specification check, we also performed analyses using an alternative restatement sample that excludes restatements related to “Accounting Rule Application Failures”. This is a slightly more conservative definition of restatement than that used in our main analyses, and is consistent with that used by Hennes et al. (2008) and in the dataset available on Andrew Leone’s website at https://sbaleone.bus.miami.edu/. We find significantly negative coefficients on the CONSRV measures for the CFO and the top five managers. Results for the CEO CONSRV variable are insignificant, but this may be explained by the restricted sample size that this definition of restatement imposes.
Altman’s (1983) Z score is equal to 0.717 * working capital/total assets + 0.847 * retained earnings/total assets + 3.107 * EBIT/total assets + 0.42*Book Value of Equity/total liabilities + 0.998 * Sales/total assets. A lower Z-score indicates a higher risk of bankruptcy.
McGuire et al. (2012) use responses to a Gallup survey about religiosity. We use the American Religion Data Archive Churches and Church Membership surveys following Hilary and Hui (2009). This allows us to also analyze the robustness of the results of McGuire et al. (2012) using an alternative measure of geographical religiosity.
We thank the American Religion Data Archive (www.theARDA.com) for making this data available for download. We use the Churches and Church Membership in the United States, 1990 (principal investigators are the Association of Statisticians of American Religious Bodies and the Lily Foundation), Churches and Church Membership in the United States, 2000 (principal investigators are the Association of Statisticians of American Religious Bodies), and the U.S. Religion Census: Religious Congregations and Membership Study, 2010 (County file) (principal investigators are Clifford Grammich, Kirk Hadaway, Richard Houseal, Dale E. Jones, Alexei Krindatch, Richie Stanley, Richard H. Taylor).
We considered using CEO equity pay in the CONSRVCEO regressions, CFO equity pay in the CONSRVCFO regressions and the proportion of equity pay for the top 5 executives in the CONSRVTOP5 regressions. However, data losses due to missing compensation data for CFO and others led us to use only CEO equity pay in all specifications. We believe that the CEO equity pay ratio sufficiently measures the relevant construct, which is the firm’s philosophy on managerial pay.
We could not perform firm-level fixed effects analysis for our restatement sample because there is no firm level variation in the value of RESTATE. Due to sample selection criteria all firms are either RESTATE = 1 or RESTATE = 0. Therefore, controlling for firm level effects removes all variation in RESTATE.
Lennox et al. (2012) caution researchers about using an Instrumental Variables approach when controlling for self-selection bias. These researchers suggest that selection models frequently suffer from multicolinearity and generally fail to base 2nd stage regression exclusions on theory. These issues can result in spurious findings.
We follow Paterson and Valencia (2011) elimination of restatements not categorized within Audit Analytics as either “Financial Fraud, Irregularities, or Misrepresentations” or “Accounting Rule-Application Failures”. This data screen is designed to remove less severe restatements.
The one exception is our measure of the market to book ratio, MB, computed as the natural log of the market value of equity divided by the book value of equity. Our sample mean (median) was 0.83 (0.82), but the sample mean (median) for Francis and Yu (2009) was 1.367 (0.943). We believe this difference can be explained by the difference in economic conditions during Francis and Yu’s sample period (2003–2005) and ours (1992–2010).
We tested for multicolinearity using VIFs. All were below 2, suggesting that multicolinearity is not a problem.
Chen et al. (2015) suggest that firm level fixed effects may be insufficient controls when estimating the effect of political contributions variables since political contributions rise and fall with election cycles.
We do not use a changes specification for the restatement model, since there is no firm-level variation in the dependent variable.
Coefficients on control variables are generally consistent with predictions or insignificant. The only exceptions are SIZE, DEBT, VOLRet and CFO, where some specifications include significant coefficient estimates with signs inconsistent with predictions for these variables.
We thank an anonymous referee for making this point.
We sincerely thank anonymous referees for identifying these issues and suggesting the related analyses.
We acquire this data from the Lobbying Disclosure database that is maintained by the U.S. Senate’s Office of Public Records. This database covers all federal lobbyists, not just those that lobby the U.S. Senate. This database is available at http://www.senate.gov/legislative/Public_Disclosure/database_download.htm.
For pre-2009 fiscal years, these inputs are all reported in EXECUCOMP. For fiscal years 2009 and after, EXECUCOMP changed some of the option-related variables it reported. From data available in CRSP, we obtained or computed stock price, dividend yield, and stock return volatility following the method EXECUCOMP used for the pre-2009 fiscal years.
References
Aier J, Comprix J, Gunlock J, Lee D (2005) The financial expertise of CFOs and accounting restatements. Acc Horiz 119:123–135
Alker H (1972) Is personality situationally specific or intrapsychically consistent? J Pers 40:1–16
Altman E (1983) Corporate financial distress: a complete guide to predicting, avoiding, and dealing with bankruptcy. Wiley, New York
Ashbaugh-Skaife H, Collins D, Kinney W (2007) The discovery and reporting of internal control deficiencies prior to sox-mandated audits. J Acc Econ 44:166–192
Atieh J, Brief A, Vollrath D (1987) The protestant work ethic conservatism paradox: beliefs and values about work and life. Pers Indiv Differ 8:577–580
Bamber L, Jiang J, Wang I (2010) What’s my style? The influence of top managers on voluntary corporate financial disclosure. Acc Rev 85:1131–1162
Bannister J, Newman H (1996) Accrual usage to manage earnings toward financial analysts’ forecasts. Rev Quant Finan and Acc 7:259–278
Batta G, Heredia R, Weidenmier M (2014) Political connections and accounting quality under high expropriation risk. Eur Acc Rev 23:485–517
Beasley M (1996) An empirical analysis of the relation between the board of director composition and financial statement fraud. Acc Rev 71:443–465
Bebchuck LA, Cremers KJM, Peyer UC (2011) The CEO pay slice. J Financ Econ 102:199–221
Becker C, Defond M, Jiambalvo J, Subramanyam KR (1998) The effect of audit quality on earnings management. Contemp Acc Res 5:1–24
Bedard J, Chtourou SM, Courteau L (2004) The effect of audit committee expertise, independence, and activity on aggressive earnings management. Auditing 23:15–37
Bergstresser B, Philippon T (2006) CEO incentives and earnings management. J Financ Econ 180:511–529
Bertrand M, Schoar A (2003) Managing with style: the effects of managers on firm policies. Q J Econ 118:1169–1208
Boone JP, Khurana I, Raman KK (2010) Do the big 4 and the second-tier firms provide audits of similar quality? J Acc Public Pol 29:330–352
Brochet F, Welch K (2011) Top executive background and financial reporting choice: The case of goodwill impairment. Working paper, Harvard Business School
Burgstahler D, Dichev I (1997) Earnings management to avoid earnings decreases and losses. J Acc Econ 24:99–126
Burns N, Kedia S (2006) The Impact of Performance-Based Compensation on Misreporting. J Fin Econ 79:35–67
Burton R (1963) Generality of honesty reconsidered. Psychol Rev 70:481–499
Cahan S (1992) The effect of antitrust investigations on discretionary accruals: a refined test of the political cost hypothesis. Acc Rev 67:77–95
Chaney P, Faccio M, Parsley D (2011) The quality of accounting information in politically connected firms. J Acc Econ 51:58–76
Chao C-L, Horng S-M (2013) Asset write-offs discretion and accruals management in Taiwan: the role of corporate governance. Rev Quant Finan Acc 40:41–74
Chen H, Parsley D, Yang Y (2015) Corporate lobbying and firm performance. J Bus Financ Acc 42:444–481
Cheng Q, Warfield T (2005) Equity incentives and earnings management. Acc Rev 80:441–476
Chin A, Parawada J (2010) Red-blooded Republican or true-blue Democrat? The influences of political preferences on money managers’ decisions. Working paper, University of New South Wales
Christensen DM, Dhaliwal DS, Boivie S, Graffin SD (2015) Top management conservatism and corporate risk strategies: evidence from managers’ personal political orientation and corporate tax avoidance. Strateg Manag J 36:1918–1938
Cohen DA, Dey A, Lys TZ (2008) Real and accrual-based earnings management in the pre- and post-Sarbanes–Oxley periods. Acc Rev 83:757–787
Core J, Guay W (2002) Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. J Acc Res 40:613–630
Cronqvist H, Makhija A, Yonker S (2010) Behavioral consistency in corporate finance: CEO personal and corporate leverage. Working paper, The Ohio State University
Dechow P, Dichev I (2002) The quality of accruals and earnings. Acc Rev 77:35–59
Dechow P, Sloan R, Sweeney A (1995) Detecting earnings management. Acc Rev 70:193–225
Dechow P, Sloan R, Sweeney A (1996) Causes and consequences of earnings manipulation: an analysis of firms subject to enforcement actions by the SEC. Contemp Acc Res 13:1–36
DeFond R, Jiambalvo J (1994) Debt covenant violations and manipulation of accruals. J Acc Econ 17:145–176
Desai H, Hogan C, Wilkins M (2006) The reputational penalty for aggressive accounting: earnings restatements and management turnover. Acc Rev 81:83–112
Dhaliwal DS, Naiker V, Navissi F (2010) The association between accruals quality and the characteristics of accounting experts and mix of expertise on audit committees. Contemp Acc Res 27:787–827
Doyle J, Ge W, McVay S (2007) Accruals quality and internal control over financial reporting. Acc Rev 82:1141–1170
DuCharme L, Malatesta P, Sefcik S (2004) Earnings management, stock issues, and shareholder lawsuits. J Financ Econ 17:27–49
Epstein S (1979) The stability of behavior: I. On predicting most of the people much of the time. J Pers Soc Psychol 37:1097–1126
Feather N (1979) Value correlates of conservatism. J Pers Soc Psychol 37:1617–1630
Francis J, Yu D (2009) Big 4 office size and audit quality. Acc Rev 84:1521–1552
Francis J, LaFond R, Olsson P, Schipper K (2005) The market pricing of accruals quality. J Acc Econ 39:295–327
Ge W, McVay S (2005) The disclosure of material weaknesses in internal control after the Sarbanes–Oxley act. Acc Horiz 19:137–158
Gerring J (1997) Ideology: a definitional analysis. Polit Res Q 50:957–994
Ghosh A, Marra A, Moon D (2010) Corporate boards, audit committees, and earnings management: pre- and post-Sox evidence. J Bus Financ Acc 37:1146–1176
Gong G, Louis H, Sun X (2008) Earnings management and firm performance following open-market repurchases. J Financ 63:947–986
Green D, Palmquist B, Schickler E (2002) Partisan hearts and minds: political parties and the social identities of voters. Yale University Press, New Haven
Gul F, Wu D, Yang Z (2013) Do individual auditors affect audit quality? Evidence from archival data. Acc Rev 88:1993–2023
Healy P (1985) The effect of bonus schemes on accounting decisions. J Acc Econ 7:85–107
Hennes KM, Leone AJ, Miller BP (2008) The importance of distinguishing errors from irregularities in restatement research: the case of restatements and CEO/CFO turnover. Acc Rev 83:1487–1519
Hilary G, Hui KW (2009) Does religion matter in corporate decision making in America? J Financ Econ 93:455–473
Holthausen R, Larcker D, Sloan R (1995) Annual bonus schemes and the manipulation of earnings. J Acc Econ 19:29–74
Hong H, Kostovetsky L (2012) Red and blue investing: values and finance. J Financ Econ 103:1–19
Hribar P, Nichols D (2007) The use of unsigned earnings quality measures in tests of earnings management. J Acc Res 45:1017–1053
Hutton I, Jiang D, Kumar A (2014) Republican managers and corporate policies. J Financ Quant Analysis 49:1279–1310
Hutton I, Jiang D, Kumar A (2015) Political values, culture, and corporate litigation. Manag Sci 61:2905–2925
Iliev P (2010) The effect of SOX section 404: costs, earnings quality, and stock prices. J Financ 65:1163–1196
Jiang J, Petroni K, Wang I (2010) CFOs and CEOs: who have the most influence on earnings management? J Financ Econ 96:513–526
Jiang D, Kumar A, Law K (2012) Republican equity analysts. Working Paper, University of Texas
Joe VC (1974) Personality correlates of conservatism. J Soc Psychol 93:309–310
Joe VC, Jones RN, Ryder S (1977) Conservatism, openness to experience, and sample bias. J Pers As 41:527–531
Jones JJ (1991) Earnings management during import relief investigations. J Acc Res 29:193–228
Jones KL, Krishnan GV, Melendrez KD (2008) Do models of discretionary accruals detect actual cases of fraudulent and restated earnings? An empirical analysis. Contemp Acc Res 25:499–531
Jost JT (2006) The end of the end of ideology. Am Psychol 61:651–670
Kam CD, Simas EN (2010) Risk orientations and policy frames. J Polit 72:381–396
Keung E, Shih M (2014) Measuring discretionary accruals: are ROA-matched models better than the original Jones-type models? Rev Acc Stud 19:736–768
Kinney WR, Palmrose Z-V, Scholz S (2004) Auditor independence, non-audit services, and restatements: was the US government right? J Acc Res 42:561–588
Klien A (2002) Audit committee, board of director characteristics, and earnings management. J Acc Econ 33:375–400
Kothari SP, Leone AJ, Wasley CE (2005) Performance matched discretionary accruals. J Acc Econ 39:163–197
Krishnan J, Wen Y, Zhao W (2011) Legal expertise on corporate audit committees and financial reporting quality. Acc Rev 86:2099–2130
Lee J, Lee KJ, Nagarajan NJ (2014) Does political alignment between top management and directors impair board independence? J Financ Econ 112:232–250
Lennox P, Pittman J (2010) Big five audits and accounting fraud. Contemp Acc Res 27:209–247
Lennox C, Francis J, Wang Z (2012) Selection models in accounting research. Acc Rev 87:589–616
Li S, Nwaeze E, Yin J (2016) Earnings management in the electric utility industry: profit incentives. Rev Quant Finan Acc 46:633–660
Liu Y, Wei Z, Xie F (2016) CFO gender and earnings management: evidence from China. Rev Quant Finan Acc 46:881–905
Lobo G, Zhou J (2006) Did conservatism in financial reporting increase after the Sarbanes–Oxley act? Initial evidence. Acc Horiz 20:57–73
McClosky H (1958) Conservatism and personality. Am Polit Sci Rev 52:27–45
McGuire S, Omer T, Sharp N (2012) The impact of religion on financial reporting irregularities. Acc Rev 87:645–673
McNichols M (2000) Research design issues in earnings management studies. J Acc Public Pol 19:313–345
McNichols M (2002) Discussion of the quality of accruals and earnings: the role of accrual estimation errors. Acc Rev 77:61–69
Menon K, Williams D (2004) Former audit partners and abnormal accruals. Acc Rev 79:1095–1118
Mills L, Nutter S, Schwab C (2013) The effect of political sensitivity and bargaining power on taxes: evidence from federal contractors. Acc Rev 88:977–1005
Nagy A (2010) Section 404 compliance and financial reporting quality. Acc Horiz 24:441–454
Palmrose Z-V, Scholz S (2004) The circumstances and legal consequences of non-GAAP reporting: evidence from restatements. Contemp Acc Res 21:139–180
Palmrose Z-V, Richardson VJ, Scholz S (2004) Determinants of market reactions to restatement announcements. J Acc Econ 37:59–89
Paterson J, Valencia A (2011) The effects of recurring and nonrecurring tax, audit-related, and other nonaudit services on auditor independence. Contemp Acc Res 28:1510–1536
Pennebaker J, King L (1999) Linguistic styles: language use as an individual difference. Pers Soc Psychol 77:1296–1312
Perry S, Williams T (1994) Earnings management preceding managerial buyout offers. J Acc Econ 18:157–179
Peterson BE, Lane MD (2001) Implications of authoritarianism for young adulthood: longitudinal analysis of college experiences and future goals. Pers Soc Psychol Bull 27:678–690
Petroni KR (1992) Optimistic reporting in the property-casualty insurance industry. J Acc Econ 15:485–508
Raghunandan K, Read WJ, Whisenant J (2003) Initial evidence on the association between nonaudit fees and restated financial statements. Acc Horiz 17:223–234
Ruch G, Taylor G (2015) Accounting conservatism: a review of the literature. J Acc Lit 34:17–38
Schrand C, Zechman S (2012) Executive overconfidence and the slippery slope to financial misreporting. J Acc Econ 53:311–329
Sherman RA, Nave CS, Funder DC (2010) Situational similarity and personality predict behavioral consistency. J Pers Soc Psychol 99:330–343
Skaife H, Veenman D, Werner T (2013) Corporate lobbying and CEO pay. Working Paper, University of Wisconsin—Madison
Sue S, Chin C, Chan L (2013) Exploring the causes of restatements by family firms. J Bus Financ Acc 40:1068–1094
Teoh S, Welch I, Wong T (1998a) Earnings management and the post-issue performance of seasoned equity offerings. J Financ Econ 50:63–99
Teoh S, Welch I, Wong T (1998b) Earnings management and the long-term market performance of initial public offerings. J Financ 53:1935–1974
Young C, Tsai L, Hsu H (2008) The effect of controlling shareholders excess board seats control on financial restatements: evidence from Taiwan. Rev Quant Financ Acc 30:297–314
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Notbohm, M., Campbell, K., Smedema, A.R. et al. Management’s personal ideology and financial reporting quality. Rev Quant Finan Acc 52, 521–571 (2019). https://doi.org/10.1007/s11156-018-0718-5
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11156-018-0718-5
Keywords
- Earnings quality
- Management ideology
- Political contributions
- Restatements
- Earnings management
- Financial reporting quality