1. Introduction

1The debate around CEO compensation is pervasive in many developed countries. In France, since the 2000s, the increasing level of executive compensation has attracted the attention of investors, regulators, academics, the media, and the public at large. The public debate is mainly focused on the justification of the level of CEO pay and the likelihood and efficiency of improving corporate performance (e.g., Air France-KLM, Carrefour, Alcatel-Lucent, and Renault). For example, in their 2015 annual meeting, 54% of Renault investors voted against the CEO’s (Carlos Ghosn) payout, including 1.78 million in variable pay and a further 4.18 million in deferred bonuses and stock. However, as the vote was non-binding, the Renault board of directors endorsed Ghosn’s payout.

3The academic debate on CEO compensation distinguishes between two main theoretical approaches. The optimal contracting (agency theory) approach (Jensen and Meckling, 1976) views executive pay as a mechanism to reconcile the interests of executives with that of shareholders to reduce agency problems and costs. The agency model assumes that agents are self-interested and risk-averse, and that the principal has incomplete information about the agent’s actions. The principal has two options to overcome this problem. Either deploying more resources in monitoring activities to obtain more information about the agent’s actions, or shifting the risk to the agent through outcome-based contracts. According to Bebchuk et al. (2002), there is no contract that perfectly aligns the interests of managers and shareholders. The optimal contract is therefore one that minimizes agency costs. Under the optimal contracting approach, this is exactly what executive compensation packages are designed to do. The board, attempting to maximize shareholder wealth, seeks to establish optimal incentives for the executives. By contrast, the managerial power approach considers executive compensation as part of the agency problem. Introduced by Bebchuk and Fried (2004), this approach considers an alternative explanation for many executive compensation practices considered as puzzles of sorts under the optimal contracting approach. The tendency to strengthen the relationship between performance and compensation may motivate executives to make decisions that maximize their own interests, as in the case of earnings management (Almadi and Lazic, 2016; Harris and Bromley, 2007; Harris et al., 2019). Therefore, the level of CEO incentive compensation depends on performance criteria, including earnings targets (Dechow et al., 2010; Ibrahim and Lloyd, 2011). For example, in France, the AMF (Autorité des Marchés Financiers) annual report (2017, p. 61) shows that for a sample of 58 firms, 97% condition bonus attribution to achieving quantifiable objectives. Among these quantifiable objectives, there are numerous operational criteria, such as sales and net earnings. Thus, CEOs’ incentives could motivate them to increase the earnings reported in two ways: accounting choices in the form of accrual-based earnings management (AEM) or operational decisions. The literature pays a great deal of attention to CEO’s unusual operational decisions that impact earnings reporting. These decisions include abnormal reductions of discretionary expenditure, decreasing research and development (R&D) and advertising expenditure, employee training, maintenance budgets, and overproduction. These practices are known as components of real earnings management (Almadi and Lazic, 2016; Harris and Bromley, 2007; Harris et al., 2019). Many previous studies document empirical evidence on real earnings management practices (e.g., Cohen et al., 2008; Graham et al., 2005; Enomoto et al., 2015; Ferentinou and Anagnostopoulou, 2016; Healy and Wahlen, 1999; Roychowdhury, 2006). For example, the survey among 400 executives that Graham et al. (2005) conducted reveals that the vast majority of US CEOs admit to using real earnings management to secure their jobs, despite being aware of the negative impact on their firm’s future performance. Their findings show that 80% of survey participants preferred making operational decisions (such as cutting R&D, maintenance and advertising expenditure), notwithstanding the negative long-term effects, rather than using accrual-based methods (AEM) to meet their earnings reporting targets. Indeed, recent studies show that REM has a negative effect on future performance (e.g., Cohen and Zarowin, 2010; Cupertino et al., 2016; Leggett et al., 2016; Moradi et al., 2015; Tabassum et al., 2015).

4The literature review on the relation between incentive pay and earnings management reveals two main gaps. First, the majority of studies have been conducted in the US context (Almadi and Lazic, 2016; Croci et al., 2012; Ronen and Yaari, 2008). Other recent studies show the relevance of the institutional context in influencing the relationship between CEO pay and earnings management (Almadi and Lazic, 2016; Cuomo et al., 2015; Ding et al. 2018; Jouber and Fakhfakh, 2014; Leuz et al., 2003). For example, Ding et al. (2018) demonstrate that politically affiliated privately-held firms in China engage more in real activities to manipulate earnings than non-affiliated firms. Furthermore, in a comparative study between a European continental model (German and Austrian) and an Anglo-Saxon model (British and Australian), Almadi and Lazic (2016) found that institutional factors influence the relationship between CEO incentive-based compensation and AEM. Second, almost all of these studies focus on the relationship between CEO incentives and accrual-based earnings management (AEM) (Almadi and Lazic, 2016; Bergstresser and Philippon, 2006; Burns and Kedia, 2008; Cheng and Warfield, 2005; Larcker et al., 2007). However, recent studies show executives increasingly use REM (Graham et al., 2005; Cohen and Zarowin, 2010; Jungeun et al., 2012). Considering these two gaps issues and the ongoing debate on CEO compensation in France, our study aims to bridge the gap in the literature by raising the question about the relation between CEO pay and REM. The French context has several specific characteristics that could influence this relation. First, ownership tends to be concentrated with a high presence of the State, family, and institutional investors in the ownership structure of many firms (Dardour and Boussaada, 2015; Mard and Marsat, 2012; Broye et al., 2018). For example, from a sample of 89 companies listed on the SBF 120, Labelle and Schatt (2005) observe that the first shareholder owns on average 29% of outstanding shares. More recently, Dardour and Boussaada (2015) reported that the average voting rights of the first owner are about 36.53%. Indeed, with a large share of voting rights, the investor has the ability and resources to monitor and advise executives. Corporate monitoring by larger investors can reduce the opportunistic behaviors of CEOs (Cornett et al., 2008). Furthermore, larger shareholder investments are often motivated by long-term value rather than myopically focusing on short-term profits (Mard and Marsat, 2012). On the other side, concentrated ownership may lead to the expropriation of minority shareholder interests, leading to an increase in earnings management (EM) practices (Liu and Lu, 2007), especially in institutional contexts with lower protection of small shareholders (La Porta et al., 1999). Additionally, countries with a civil law system, such as France, will be more exposed to earnings management than common law countries, such as the US and the UK (Almadi and Lazic, 2016; Leuz et al., 2003; Jouber and Fakhfakh, 2014). In France, organizations representing private firms, such as AFEP and MEDEF, have a more dominant role in producing the recommendations included in governance codes and in controlling the application of these codes compared to other European countries. This may call into question the effectiveness of these codes in limiting the discretionary power of managers (Boncori and Cadet, 2013). A comparative study published by the French financial markets authority (AMF) in 2016 shows that the AFEP-MEDEF French corporate governance code is the only code drawn up exclusively by private organizations representing the issuers. In contrast, the drafting of codes in other European countries involves multiple stakeholders (market authorities, issuing associations, unions, etc.). In line with these specific institutional characteristics, some studies have analyzed AEM in French firms. First, based on a sample of SBF 250 listed firms between 2004 and 2008, Mard and Marsat (2012) show a curvilinear relationship between the percentage of capital held by the main shareholder and the level of AEM. In addition, family and institutional shareholding limits the intensity of earnings management. Second, Jouber and Fakhfakh (2014) carried out a comparative study over the same period (2004–2008) between four countries (US, Canada, UK, and France) on the impact of incentive compensation on AEM, identifying a positive and significant relation. Moreover, the study shows higher earnings management intensity in Euro-Continental countries such as France. The authors correlate these results with certain institutional factors, such as the quality of governance, the degree of protection of the rights of minority shareholders, and the level of laws strengthening or limiting earnings management incentives in Anglo-Saxon countries.

5Our paper aims to complement this literature by studying the relationship between REM and CEO compensation in French firms listed on the SBF 120 Index over the more recent 2006–2017 period marked by a significant increase in the level of incentive compensation and the need to disclose the quantitative and qualitative performance criteria for the attribution of executive pay components in the annual report. Overall, we find that CEO incentive pay is positively associated with the extent of real earnings management. This first result is consistent with previous empirical studies in different developed countries (e.g., Cohen et al., 2008; Ferentinou and Anagnostopoulou, 2016). In addition, the CEOs of firms controlled by families or institutions engage less in REM when they have greater incentive compensation. By contrast, State ownership does not have any moderating effect on this relationship. Thus, the effect of increasing CEO incentives on REM is determined by the major shareholder ownership type. Our study significantly contributes to the literature on earnings management and CEO pay by examining how CEO incentive pay affects the extent of REM, shedding light on how ownership type moderates this relationship in the French context. Our results extend the debate on the efficiency of CEO pay design and could help improve the regulation and/or corporate governance codes in France and other European countries with similar institutional settings. Furthermore, our findings highlight serious concerns when the incentive ratio is high. The results suggest a need to enhance incentive compensation policies with more efficient monitoring mechanisms to constrain the opportunistic operational decisions of CEOs.

6The remainder of this article proceeds as follows. Section 2 provides a summary of prior research and develops the hypotheses. Section 3 describes the sample and data, and presents the research design. Section 4 reports the empirical results and discusses their implications. We perform some robustness checks in Section 4, while Section 5 concludes and presents our research limitations and avenues for future research.

2. Literature review and hypotheses development

2.1. Real earnings management and CEO compensation

7In view of the information asymmetry between managers and various stakeholders, publishing financial information is a crucial means to promote and control the firm’s progress. The earnings reporting derived from financial statements is part of the mandatory information that firms disclose, and is closely followed by investors and financial analysts. In addition, earnings are broadly used by stakeholders to evaluate executive performance and firm investment decisions. Due to its relevance to stakeholders, earnings are widely used to set performance criteria in CEO compensation contracts (Dechow et al., 2010; Ibrahim and Lloyd, 2011). Consequently, using earnings in compensation contracts may incentivize potential earnings management practices (Dechow et al., 2010). Indeed, the latitude that managers have in their decisions allows them to shape accounting information in compliance with the legal framework (Schatt et al., 2018), known as earnings management. According to Schipper (1989), earnings management is a “purposeful intervention in the external financial reporting process with the intent to obtaining one private gain” (p. 92). In addition, Healy and Wahlen (1999) consider that “earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers” (p. 368). Furthermore, Ronen and Yaari (2008, p. 27) define earnings management as “a collection of managerial decisions that result in not reporting the true short-term, value-maximizing earnings as known to management”. These authors distinguish between three categories of actors on which earnings management research can focus: managers, stakeholders, and gatekeepers. Previous studies distinguish between two types of earnings management (Bartov and Cohen, 2009; Cohen et al., 2008; Zang, 2012): accounting or accrual-based earnings management (AEM) and real earnings management (REM). AEM is defined as accounting choices made to influence earnings reporting (Beneish; 2001; Dechow and Skinner; 2000; Degeorge et al., 1999). According to Gunny (2010), REM occurs when executives take actions that change the timing of structuring an operation, investment, and/or financing transaction in an effort to influence the output of the accounting system. Numerous studies confirm the widespread use of REM by executives (Cohen et al., 2008; Graham et al., 2005; Enomoto et al., 2015; Ferentinou and Anagnostopoulou, 2016; Healy and Wahlen, 1999; Roychowdhury, 2006), identifying several real manipulation activities, such as abnormal cash from operations, discretionary expenditures, and overproduction (Cohen et al., 2008; Harris et al., 2019; Roychowdhury, 2006; Zang, 2012). Recent studies find that executives increasingly substitute AEM with REM (Graham et al., 2005; Cohen and Zarowin, 2010; Jungeun et al., 2012), showing that REM is more harmful to firm value than AEM (Cohen et al., 2008; Cohen and Zarowin, 2010; Dechow et al., 2010; Eisele, 2012; Roychowdhury, 2006).

8First, REM and AEM have different impacts on cash flows. For example, in the case of REM, the decision to reduce prices to accelerate sales and increase turnover has a direct and negative effect on cash flows. In contrast, AEM only indirectly affects cash flows, as in the case of using accruals-based earnings management to avoid breaching the terms of debt contracts (Cohen and Zarowin, 2010; Burgstahler and Eames, 2003; Franz et al., 2014). Consequently, REM activities produce higher costs than AEM. Mizik (2010) and Roychowdhury (2006) show that REM relates to management decisions that are not optimal, which can lead to weakening firm performance. In this vein, Zang (2012) finds that a reduction in advertising or R&D expenses leads to a decrease in future cash flows, as in the case of unjustifiable increases in these expenses.

9Second, REM could be more difficult to control through different governance mechanisms, such as mandatory audits, as it concerns managerial and not accounting decisions (Schatt et al., 2018). Thus, REM is more challenging for investors and regulators to detect (Zang, 2012; Schatt et al., 2018). Finally, as Zang (2012) argues, accounting choices are made by executives when preparing financial statements at the end of the financial year, unlike managerial decisions related to REM that take place during the financial year. An example is the decision to delay advertising expenditure during the accounting year.

10Consistent with agency theory, studies have focused on earnings management incentives, such as CEO compensation (Healy and Wahlen, 1999; Ibrahim and Lloyd, 2001). Balsam (1998) analyzed the way in which earnings components affect CEO pay, showing that discretionary accruals are associated with CEO cash compensation (base salary and bonuses). More recently, Carter et al. (2009) found a positive relationship between earnings management and executive bonuses in US firms. Other studies demonstrate the greater use of discretionary accruals in firms with higher equity incentive pay (e.g., Bergstresser and Philippon, 2006; Cheng and Warfield, 2005). In Japan, Shuto (2007) found that discretionary accruals are positively and significantly associated with executive compensation, suggesting that executives use discretionary accruals to increase their pay.

11Hence, one fundamental driver of earnings management is the pressure on executives to deliver on short-term performance criteria. In the French context, the executive incentive pay level has continued to increase since the 2000s, but also the requirements to link executive compensation to firm performance (Broye et al., 2018). Indeed, the AMF annual report (2017, p. 61) shows that for a sample of 58 firms, 97% condition bonus attribution to achieving quantifiable objectives. According to agency theory and the arguments presented, we test the relationship between REM and CEO incentive pay ratio, positing the following hypothesis:

H1. Real earnings management is positively associated with CEO incentive pay ratio.

2.2. Ownership structure, CEO incentives, and REM

12The relationship between CEO incentive pay and REM might be affected by the ownership type and structure of the major shareholder. Consistent with agency theory, ownership concentration can have two opposite effects in resolving agency conflicts. On the one hand, the presence of a controlling shareholder acts as a governance mechanism to monitor executives and limit their discretionary power (Demsetz and Lehn, 1985; Shleifer and Vishny, 1986). Thus, corporate monitoring by larger investors can reduce CEOs’ opportunistic behavior (Cornett et al., 2008). On the other hand, the controlling shareholder may be tempted to expropriate minority shareholders’ interests, leading to an increase in earnings management practices, especially in institutional contexts with lower protection of small shareholders (La Porta et al., 1999; Fama and Jensen, 1983; Liu and Lu, 2007). In this case, ownership concentration leads to the entrenchment of the controlling shareholder (Shleifer and Vishny, 1997). Furthermore, the type of controlling ownership can also influence the resolution of agency conflicts. Accordingly, different motivations might affect the relationship between incentive pay and earnings management. Moreover, the involvement of one or more controlling shareholders might affect the attribution of CEO incentive pay (David et al., 1998; Tosi and Gómez-Mejía, 1989). Therefore, concentrated ownership may strengthen or weaken the expected positive direct effect between CEO incentive pay and REM. To better determine the potential moderating effect, we test the impact of ownership concentration on the CEO incentive pay ratio-REM relation.

H2: Ownership concentration affects the CEO incentive pay ratio-REM relation

13Furthermore, to better determine the possible moderating effect, we test the impact of controlling shareholder ownership type as moderator on the CEO incentive pay ratio-REM relation. Following previous studies, we distinguish between three main types of controlling shareholders: institutional, family, and the French State (Mard and Marsat, 2012; Dardour and Boussaada, 2015).

2.2.1. Institutional ownership and REM-incentive pay ratio

14Institutional shareholders can nominate their own candidates as members of the board of directors. As a result, CEOs are restricted in the pursuit of personal goals (Dardour and Boussaada, 2017). In addition, they are more active in monitoring executive compensation policy and its impact on performance targets. Indeed, in this case, agency costs can be reduced by using tools other than incentive plans (Zattoni, 2007). However, the presence of a dominant shareholder may lead to the expropriation of minority shareholders’ interests by satisfying their private interests, which could increase earnings management activities (Liu and Lu, 2007). Furthermore, the likelihood of minority shareholder expropriation can be amplified in the case of a weak legal environment and lack of investor protection (La Porta et al., 1999). Studies in the US context indicate that REM is negatively related to the level of institutional ownership (Roychowdhury, 2006; Zang, 2012). Roychowdhury (2006) also found a negative association between institutional ownership and real activities manipulation using abnormal production costs and abnormal discretionary expenses. This is consistent with the findings of Mard and Marsat (2012) on the relation between AEM and institutional shareholders in the French context. Based on these arguments, we expect a moderating effect of institutional ownership on the CEO incentive pay-REM relation, without predicting the sign of this effect. Thus, we posit:

H3a: Institutional ownership affects the CEO incentive pay ratio-REM relation

2.2.2. Family ownership and REM-incentives pay ratio

15Prior studies on family ownership and REM find inconclusive results. Ali et al. (2007), Veider et al. (2016), and Wang (2006) show a negative relationship between family firms and earnings management explained by better monitoring, long-term orientation, and reputation concerns. In addition, some studies comparing family and non-family firms find that family firms have a lower level of earnings management (e.g., Wang, 2006). However, in Taiwan, Chi et al. (2011) document that family ownership is positively related to earnings management. Adıgüzel (2013) argues that managers that have close family or personal relationships with the controlling family members can manipulate earnings in the interests of the controlling family. The family might seek to minimize taxes or political costs, inducing earnings management (Ali et al., 2007). The conflicts between family and non-family shareholders are more pronounced in civil law countries characterized by lower governance quality. In such environments, these conflicts are more likely to lead to the expropriation of minority shareholder interests. Furthermore, in family-controlled firms, incentive-based compensation can be a means for majority shareholders to expropriate minority shareholders (Bertrand and Schoar, 2006; Zattoni and Minichilli, 2009). The strong commitment and firm-specific knowledge of family members make them better monitors (Bertrand and Schoar, 2006; Martínez-Ferrero et al., 2016), tending to maximize the firm’s value in the long run (Gonzalez et al., 2014). Therefore, based on the specific family ownership type and the above theoretical arguments, we posit:

H3b: Family ownership affects the CEO incentive pay ratio-REM relation

2.2.3. State ownership and REM-incentives pay ratio

16Studies on the influence of State ownership on the relationship between incentive compensation and earnings management are scarce (Mard and Marsat, 2012). By holding part of the capital of several large listed firms, the State plays an important role in France. State ownership may be motivated by non-financial considerations, such as maintaining employment, revitalizing territories, or even controlling strategic industries (Clarke, 2003). Nevertheless, the State also pursues the objective of creating economic value and can link executive compensation to firm performance (Dardour and Boussaada, 2017). The empirical results are thus far inconclusive. In China, the study of Ding et al. (2007) reveals that State-owned firms tend to have superior accounting performance, but are also more likely to engage in real activities to manipulate earnings than privately-held firms. On the contrary, Ben-Nasr et al. (2009), based on an international sample, reveal that State-ownership is associated with lower accounting results quality and higher discretionary accruals. Finally, Mard and Marsat (2012) do not find a significant effect of State-ownership on accruals earnings management. Based on the above arguments, we posit:

H3c: State ownership affects the CEO incentives pay ratio-REM relation

3. Research method

3.1. Sample and study period

17Our study of the relationship between REM and CEO compensation covers the period 2006–2017. We chose 2006 as the starting point to avoid the effect of France adopting International Financial Reporting Standards (IFRS) (Elage and Mard, 2018) and introducing the granting of performance shares to executives (Article 83 of the Finance Act) in 2005. Furthermore, we control for the potential impact of the 2008 financial crisis on our results. Our sample derives from the SBF 120 Index where we identified all firms included at least once between 2006 and 2017. The initial sample consisted of 146 firms. Due to their particular accounting methodologies, we removed banking and insurance institutions from the sample (19 firms), and excluded a further 11 firms due to lack of financial or governance data. Specifically, we eliminated all firms with more than four missing values over the 2006–2017 period. Thus, the final sample comprises 108 firms (see Appendix 1 for the sample composition). For the accounting and financial data, we used the Thomson Reuters Datastream database, for CEO compensation data, we used the IODS and Bloomberg databases, in addition to hand-collecting data from the firms’ annual reports.

3.2. Variables measurement

3.2.1. Dependent variables

19Following Cohen et al. (2008), Farooqi et al. (2014), and Zang (2012), we compute the measure of REM as the sum of the three variables as follow:

20Following previous studies (e.g., Roychowdhury, 2006; Martinez and Serve, 2011; Mard and Marsat, 2012; and Elage and Mard, 2018), we regrouped the firms by industry to obtain a sufficient number of firm/industry-year observations to estimate the parameters of the models. The regroupings are based on the Industry Classification Benchmark (ICB). We estimated the earnings management models for each industry-year pair, which resulted in 72 regressions per earnings management model (6 industries ^* 12 years).

3.2.2. Independent variables

21A CEO’s total compensation includes different components based on salary, bonuses, and equity-based incentive rewards (stock options plus performance shares). We use the CEO incentive pay ratio, which is equal to the sum of bonuses and equity-based compensation divided by CEO total compensation (Dardour et al., 2015). CEO salary and bonuses are awarded during a given year, while long-term incentives are primarily composed of the potential value of stock options and performance shares valued at the grant date. We estimate the value of CEO stock option awards using the Black-Scholes (1973) valuation model for the period 2006–2007, as since 2008, companies publish the value of long-term incentives. The accumulation of these two components is justified by the executive’s incentive effect from the short-term component in the form of an annual bonus and the long-term component (stock options and performance shares). The common point between these components is the need to achieve performance objectives that are predefined by the board of directors as proposed by the compensation committee. Compared to the annual bonus, equity-based compensation is not attributed to executives each year. It is therefore difficult to distinguish the incentive effect of one pay component from others on CEO behavior. In addition, in the case of poor performance, the company may decide to reduce the CEO bonus while increasing the level of the long-term pay components. Thus, our overall incentive ratio takes into account the potential substitution effect between these pay components. With regard to ownership structure, prior studies highlight the coexistence of two types in the main industrialized countries: concentrated ownership and widely-held ownership. However, concentrated ownership is predominant in continental Europe (La Porta et al. 1999; Faccio and Lang, 2002). For example, the share of voting rights of the largest shareholders of listed companies in the SBF 120 was on average 36.53% for the period 2003–2012 (Dardour and Boussaada, 2017). In contrast, La Porta et al. (1999) demonstrate that at the 20% threshold, 80% of US firms are widely-held, and this rises to 100% for UK listed companies. At the 10% threshold, the vast majority of UK companies (90%) have widely-held ownership, while at this threshold, only 15% of French companies do (Dardour and Boussaada, 2015). This explain why arbitrary 5% or 10% thresholds are often used in the US context (Dardour et al., 2015; Broye et al., 2018). Therefore, in our study, we first consider the threshold of 20% of voting rights to distinguish between widely-held and concentrated ownership. Due to the existence of double voting rights in certain firms in our sample, we considered it more appropriate to retain voting rights instead of the percentage of capital (Mard and Marsat, 2012). We distinguish between four ownership categories: widely-held ownership (without major shareholder), ownership controlled by the family or founder (Family_20%), ownership controlled by the French State (State_20%), and ownership controlled by institutional investors other than the State (Instit_20%). We then use the 10% threshold in our additional analysis.

3.2.3. Control variables

22Following prior studies (e.g., Cohen and Zarowin, 2010; Hope et al., 2013; Roychowdhury, 2006; Zang, 2012), we include numerous control variables shown to be associated with REM in general.

23We control for the monitoring effect with internal governance variables: CEO-chairman duality (CEO duality) and board structure (board duality). In addition, we control for CEO change (CEO turnover) and audit quality (Big Four firm). We use leverage as a proxy for debt-covenant violations. Previous studies also suggest that firms manipulate earnings to conceal poor performance (Dechow et al., 2010; Zang, 2012). Following Gunny (2010) and Healy and Wahlen (1999), we use industry-adjusted return on assets (adjusted ROA) to control for the differences in industry that may affect the performance measure. Farooqi et al. (2014) and Bartov (1993) found that large firms engage less in earnings management than small firms. As a result, we also control for firm size using the natural logarithm of total assets as a proxy. We include the dummy variable loss because companies with negative net income might have greater incentives to manage earnings (Mard and Marsat, 2012). We also include two dummy variables (industry and year), as executive compensation and REM may vary with industry and economic conjuncture.

3.3. Empirical model

24After estimating the REM measures, we used the following model to test the impact of CEO pay and corporate governance mechanisms on earnings management.

25To mitigate the effect of extreme values, all continuous variables are winsorized at the 1% and 99% level to ensure that the results are not sensitive to extreme values. Firm performance (adjusted ROA), firm size, financial leverage, CEO turnover, CEO duality, board duality, Big Four audit, and major shareholder voting rights are included as controls for other firm-specific characteristics, such as ownership structure, that likely affect REM (Cohen and Zarowin, 2010; Roychowdhury, 2006). We employ the fixed effects specification to remove unobserved time and industry heterogeneities that may be associated with earnings management and executive compensation. Thus, we estimate the models with industry and year-fixed effects (Industry_FE and Year_FE). The variable definitions are provided in Appendix 2. Finally, to control for heteroscedasticity, given the nature of the data (panel data), we applied the regressions through clustering by firms (Petersen, 2009; Dimitropoulos and Asteriou, 2010; Elage and Mard, 2018).

4. Results

4.1. Descriptive statistics

26Regarding CEO pay, CEOs received an average total compensation of €2.13 million, and CEO incentives represent €1.4 million on average. We note a change in total compensation mainly due to the variation in incentive compensation (bonus and equity-based pay). Indeed, Figure 1 shows a 42% drop in the level of average incentive compensation between 2007 and 2009 from €2.577 to €1.678 million. This sharp drop is explained by the 2008 financial crisis. The average level of compensation was stable over the 2010–2014 period and increased again from 2015.

Fig 1. The evolution of executive compensation components of SBF 120 listed firms between 2006 and 2017 (excluding financial companies).

Source: authors’ own.

27With regard to CEO pay structure, incentive pay ratio accounts for on average 48% of CEO total compensation. Figure 2 shows that the lowest level of median incentive compensation was observed in 2009 and amounted to 43.5% of CEO total pay. From 2010, this median level was above 50%. However, 17% of firm year-observations (198/1160) do not attribute any of the three incentive components to CEOs (bonuses, stock options, performance shares).

Fig 2. The evolution of CEO incentive pay ratio of SBF 120 listed firms between 2006 and 2017 (excluding financial companies).

Source: authors’ own.

28REM has a mean value of 0.010 and a median of 0.042 of total assets with a standard deviation of 0.305. We observe a high difference between the mean and standard deviation values of this variable. This difference can be explained by the wide dispersion of the value ​​of the elements constituting the estimation models (R&D expenditure, advertising exemptions, etc.). This result is consistent with other empirical studies (e.g., Zang, 2012; Cohen et al., 2008; Harris et al., 2019), and we deal with this issue in our robustness tests. Concerning ownership, on average, the majority shareholder holds 37.04% of voting rights, similarly to Dardour and Boussaada (2017). Family shareholders as controlling investors represent approximately 39% of firms in the sample, institutional shareholders 18%, and nearly 7% of firms are primarily controlled by the State. The average firm in our sample has an adjusted ROA of approximately -0.003%. Firm size measured by the natural logarithm of total assets is on average around 22.36. Our proxy for financial leverage, the debt-to-asset ratio, has a mean value of approximately 25.04%. Table 1 provides the descriptive statistics.

Table 1. Descriptive statistics

Variables	Mean	St. Dev	25th	Median	75th
CEO total compensation (€)	2,126,505	2,018,866	680,330	1,587,526	2,831,159
CEO incentives (€)	1,389,845	1,772,916	200,000	806,338	1,909,262
CEO_incentive pay ratio (%)	47.95	28.17	30.38	52.92	70.05
REM	.010	.305	-.153	.042	.209
Abn.CFO	-.0002	.0513	-.030	-.0038	.025
Abn.Prod	.0001	.1580	-.0785	.0152	.1014
Abn.Dex	-.0003	.1535	-.0899	-.0148	.0685
Adjusted ROA	-.0030	.0500	-.024	-.0005	.022
Firm size (log)	22.36	1.786	21.076	22.35	23.870
Financial leverage (%)	25.04	15.93	14.35	23.35	33.37
Major Shareholder VR (%)	37.04	24.85	14.120	32.49	59.145
Dummy variables	1	0		1	0
Audit quality (Big Four%)	44	56	Incentives= 0 (%)	17	83
CEO duality (%)	57	43	Family ownership%	39	61
Board duality (%)	20	80	Institutional ownership (%)	18	82
CEO Turnover (%)	8	92	State ownership (%)	7	93
Loss (%)	13	87	Widely-held ownership (%)	36	64

Notes: for the description of the variables, please see Appendix 2

29Table 2 reports the Pearson correlation of the variables in our main analysis. The aggregate REM measure is not correlated with CEO incentive pay ratio. This result deserves more attention. Indeed, when we exclude from our sample the observations of CEO incentives equal to 0 (198 observations over the study period), the association between these two variables becomes positive and significant at the 1% level (the unreported coefficient is 0.083^***). As these observations concern the compensation practices of French firms, we have added a binary variable (incentives= 0) in our model that equals 1 when CEO incentives are positive, 0 otherwise. Furthermore, there is a negative and significant relation between firm performance and REM, and a positive and significant association between firm performance and CEO compensation at the 1% level. The correlation between REM and major shareholder voting rights is negative, as is the correlation between CEO incentive pay ratio and major shareholder with a relatively high coefficient (-.254^***). This result suggests that ownership structure, as an internal governance mechanism, influences CEO compensation.

4.2. Multivariate analysis

30Table 3 presents the pooled panel regression. For each regression model, we examined the variance inflation factors (VIF) to check for multicollinearity issues. The maximum VIF values are far below 10, the conventional threshold for excessive multicollinearity. Indeed, the values ​​observed are between 1.09 and 6, confirming the absence of serious multicollinearity problems among the variables. As Column 1 in Table 3 shows, the coefficient of CEO_incentive pay ratio (0.189) is positive and significant at 1%, indicating that a 10% increase in a firm’s average incentive pay ratio increases the level of earnings management by 1.89% of total assets. This result confirms H1 stating a positive association between CEO incentive ratio and REM. Consistent with the managerial power perspective, this suggests that CEOs with a higher incentive pay ratio are associated with a higher use of REM. Previous empirical studies show similar results (e.g., Bergstresser and Philippon, 2006; Almadi and Lazic, 2016; Ding et al., 2018; Jouber and Fakhfakh, 2014; Harris et al. 2019). We note however that most of these studies focus on AEM rather than REM.

Table 3. Regression results of CEO incentive pay ratio on REM

	Dependent variable: REM OLS regression
	Column 1		Column 2
	Coef.	t-stat	Coef.	t-stat
Intercept	-.751***	-5.593	-0.802***	-5.530
CEO_incentive pay ratio	.189***	3.629	0.354***	4.790
Adjusted ROA	-1.971***	-9.296	-2.010***	-9.501
Firm size (log)	.033***	5.611	0.031***	5.286
CEO turnover	-.041	-1.274	-0.040	-1.231
Financial leverage%	-.009	-.168	0.006	0.103
Major shareholder VR%	.001	1.368	0.003	1.394
Major shareholder VR%* CEO_incentive pay ratio			-0.004	-1.490
CEO duality	-.064***	-2.800	-0.061***	-2.681
Audit quality	-.066	-1.497	-0.066	-1.482
Board duality	.033	1.169	0.039	1.381
Loss	-.060*	-.191	-0.059 *	-1.774
Incentives=0	.107***	3.006	0.091***	2.535
Industry and year fixed effects	Yes		Yes
Number of firms	108		108
Number of obs.	1 160		1160
R2 adjusted	0.118		0.125
F-statistic	6.737***		6.899***

Notes: CEO_incentive pay ratio is the sum of bonuses and equity-based compensation attributed to the CEO at the end of the fiscal year divided by CEO total pay. ^***, ^**, and ^* indicate statistical significance at the 1%, 5%, and 10% level, respectively.

31With regard to the control variables, firm performance (adjusted ROA) is negatively and significantly associated with REM. Thus, when a firm performs poorly compared to its competitors, the CEO is more incentivized to engage in earnings management practices, which include decisions such as abnormal cutting of discretionary expenditure, reducing R&D and advertising expenditure, employee training, maintenance budget, overproduction, and reducing prices to accelerate sales.

32In addition, firm size has a significant positive effect on REM. This result might be explained by the fact that larger firms with relatively complex activities have more R&D expenses, higher sales and discretionary expenses than smaller firms (Cheng et al., 2016). For example, the correlation coefficient between total discretionary expenses and firm size (log) in our sample is high (0.67) and significant at the 1% level. Concerning CEO turnover, financial leverage, major shareholder voting rights, audit quality, and board duality, the coefficients are not significant. CEO duality shows a negative relation with REM. In other words, when the CEO and chairman are the same person, REM practices are reduced. This cautious behavior could be explained by the CEO’s interest in preserving the sustainability of the firm’s performance. Our result confirms the findings of Cornett, McNutt, and Tehran (2009) reporting a significant and negative association between CEO duality and EM, consistent with stewardship theory. The control variable loss is negative and significant, also consistent with prior empirical studies (e.g., Mard and Marsat, 2012; Wang, 2006). Furthermore, we note a positive and significant effect of the variable Incentives=0 on the REM. This result could be explained by the pressure exerted on the CEOs whose companies achieve poor financial performance. This can motivate them to engage in real earnings management practices to achieve those performance goals. Moreover, the impact of the CEO_Incentive pay ratio variable is greater than the impact of the of Incentives= 0 variable. Column 2 of Table 3 shows that the coefficient of the interaction variable Major shareholder VR^*CEO_incentive pay ratio is not significant, thus rejecting H2. This result might be explained by the fact that this variable cannot distinguish between different types of shareholders. Taking into account the heterogeneity of shareholder characteristics could be relevant to explain the impact of major shareholder on the CEO incentive pay-REM relation.

4.2.1. Major shareholder types and REM

33Column 1 of Table 4 presents the effect of major shareholder ownership type on the CEO incentives-REM relation. We find no direct relationship between REM and institutional ownership or family ownership. By contrast, the coefficient of the variable State_20 is negative and significant at the 1% level. State ownership exerts less pressure on firms to achieve financial performance criteria that lead to constraining real activities manipulation. In State-controlled firms, value maximization may only be of secondary importance due to political or social reasons (Liang et al., 2015). In addition, managerial power and entrenchment may be restricted (Dardour and Boussaada, 2017), and the State might be more involved in closely monitoring management decisions. Overall, these first results suggest that only State ownership mitigates REM. We previously confirmed that REM is correlated with CEO incentive pay as stimuli, and test the moderating effect of ownership type on the REM-CEO incentives relation. To do so, we use three interaction variables: State_20%^*CEO_incentive pay ratio, Instit_20%^*CEO_incentive pay ratio, and Family_20%^*CEO_incentive pay ratio. Column 2 of Table 4 shows that the coefficients of these interaction variables are negative and significant, except for the variable State^*CEO incentive pay ratio. Our findings suggest that the CEOs of family and institutional investor controlled firms engage less in REM when they have a higher incentive pay ratio. The incentive pay ratio level increase leads to a change in CEO behavior. Indeed, this result can be explained by the controlling shareholder’s better ability to access information, monitor the effects of the manager’s choices, and estimate the impact of operational decisions on the firm’s future performance, thus limiting executive discretion to manage earnings. In addition, in the case of family businesses, conflicts of interest between the executive and the majority shareholder are less frequent, especially when the CEO is a family member. This might also explain the negative and significant effect of the interaction variable CEO incentive pay ratio^*Family ownership. Conversely, we find that the incentives levels in State-controlled firms has no effect on REM activities. This result could be explained by at least two reasons. First, the French State is more involved in operational decisions and the CEO nomination process, which can reduce CEO discretion (latitude). Second, CEOs of State-controlled companies may feel less pressure to achieve earnings targets since the State might be motivated by non-financial considerations, such as maintaining employment or controlling strategic industries (Dardour and Boussaada, 2017). This result is consistent with Mard and Marsat (2012) who do not find a significant effect of State ownership on accruals earnings management.

Table 4. Regression results of ownership type on REM, and the moderation effect of ownership type on the CEO incentive pay ratio-REM relation

	Column 1			Column 2
	Coef.	t-stat		Coef.	t-stat
Intercept	-.843***	-5.700		-.888***	-5.953
CEO_incentive pay ratio	.132**	2.532		.227***	3.269
Adjusted ROA	-1.970***	-9,347		-2.011***	-9,494
Family_20%	.023	1.049		.109**	2.463
Instit_20%	-.019	-.766		.085	1.36
State_20%	-.132***	-3.642		-.134*	-1.66
Family_20%* CEO_incentive pay ratio	-	-		-.164**	-2.196
Instit_20%*CEO_incentive pay ratio	-	-		-.085*	-1.762
State_20%*CEO_ incentive pay ratio	-	-		.031	.202
Firm size	.04***	6.418		.04***	6.31
CEO turnover	-.039	-1.229		-.040	-1.251
Financial leverage	.012	-.213		-.013**	-.228
CEO duality	-.057**	-2.528		-.059***	-2.614
Audit quality	-.064	-1.438		-.059	-1.308
Board duality	.036	1.292		.038	1.388
Loss	-.059*	-1.812		-.06*	-1.842
Incentives=0	.077**	2.139		-.061*	-1.676
Industry and year fixed effects	Yes			Yes
Number of firms	108			108
Number of observations	1,160			1,160
R2	.150			.155
Adjusted R2	.128			.131
F-statistic	6.89***			6.47***

Notes: Family ownership, institutional ownership, and state ownership are three dummy variables that indicate the firm’s major shareholder ownership type when holding at least 20% of VR. ^***, ^**, and ^* indicate statistical significance at the 1%, 5%, and 10% level, respectively.

34In the case of increased CEO incentive pay, we show that families and institutional investors are more involved in monitoring CEO activities and evaluating the impact of operational decisions. Incentive pay is used as a control mechanism and improves the risk that managers take, which is consistent with the optimal contract approach. In other word, an increased incentive pay ratio and better monitoring by active investors could lead to reducing non-optimal operational decisions. These investors are more involved in determining CEO pay designed to protect the firm’s long term value.

4.3. Robustness check

35Given the characteristics of the French institutional context, we conducted the following additional tests. First, we use the threshold of 10% (instead of 20%) of majority shareholder voting rights, as the classification is significantly affected by the choice of threshold. Indeed, at 10% of voting rights, we obtain the following distribution: family 45%, State 11%, institutional 25%, and widely-held companies 34%, against respectively 39%, 7%, 18% and 17% for the 20% threshold. At the 10% threshold, the number of widely-held firms is only half of the first classification (from 34% to 17%). Columns 1 and 2 of Table 5 show that there is no significant relation between the shareholder type variables and the interaction variables with REM. In comparison to the results reported in Table 4 (Column 2), we find that in the French context marked by highly concentrated ownership, a threshold of 10% is not sufficient to measure the impact of shareholder ownership type on REM, especially as our descriptive statistics reveal that the first owner holds on average 37.04% of voting rights.

Table 5. Regression results of ownership type at the 10% threshold on REM

	Column 1			Column 2
	Coef.	t-stat		Coef.	t-stat
Intercept	-.812***	-5.401		-.894***	-5.730
CEO_incentive ratio	.159**	2.994		.224**	2.451
Adjusted ROA	-1.940***	-9.194		-1.930***	-9,116
Family_10%	.029	1.144		.099	1.668
Instit_10%	.031	1.102		.080	1.131
State_10%	-.029	-.854		.085	1.005
Family10%*CEO_incentive pay ratio	-	-		-.127	-1.361
Institutional10%*CEO_incentive pay ratio	-	-		-.081	-.697
State10%*CEO_ incentive pay ratio	-	-		.130	.930
Firm size	0.037***	5.812		.039***	6.030
CEO turnover	-0.042	-1.316		-0.044	-1.354
Financial leverage	-0.015	-.267		-0.015	-0.267
CEO duality	-0.061**	-2.678		-0.062***	-2.705
Audit quality	-.065	-1.420		-0.060	-1.309
Board duality	.033	1.180		0.036	1.287
Loss	-.058*	-1.747		-0.055	-1.637
Incentives=0	.098***	2.687		0.087**	2.347
Industry and year fixed effects	Yes			Yes
Number of firms	108			108
Number of observations	1,160			1,160
R2	.140			0.145
Adjusted R2	.118			0.120
F-statistic	6.355***			5.949***

Family ownership, institutional ownership, and state ownership are three dummy variables that indicate the firm’s major shareholder ownership type when holding at least 10% of VR. ^***, ^**, and ^* indicate statistical significance at the 1%, 5%, and 10% level, respectively.

37Following Bergstresser and Philippon (2006), we also control for the impact of the extreme values of incentive pay ratio on our results by regressing a model on the highest decile of the CEO incentive pay ratio variable (see Table 6, column 2). For this decile, the incentive ratio is greater than 79.44%. Again, our results remain robust.

38Moreover, we conducted a logistic regression test where the dependent variable represents the highest decile of our REM measure. This variable REM_decile is a dummy that takes the value 1 if the REM value is included in the highest decile, 0 otherwise. The relation between CEO_incentive pay ratio and REM_decile remains positive and significant at the 5% level (the results of this regression are available upon request).

Table 6. Regression of the CEO incentive pay ratio decile on REM

	Column 1 (2010-2017)			Column 2 (decile)
	Coef.	t-stat		Coef.	t-stat
Intercept	-0.918***	-5.336		-0.716***	-4.920
CEO_incentive pay ratio	0.222***	3.502		-	-
CEO_incentive pay ratio_decile	-	-		0.057*	1.861
Adjusted ROA	-2.140***	-8.065		-1.889***	-8.932
Firm size	0.037***	5.397		0.036***	6.189
CEO turnover	-0.001	-0.012		-0.041	-1.257
Financial leverage	-0.277	-0.019		-0.017	-0.304
Major shareholder VR%	0.001	0.453		0.001	0.630
CEO duality	-0.083***	-3.087		-0.064***	-2.798
Audit quality	-0.045	-1.037		-0.062	-1.390
Board duality	0.067**	1.969		0.023	0.837
Loss	-.090***	-2.302		-0.063*	-1.897
Incentives=0	0.153***	3.390		0.018	0.732
Industry and year fixed effects	Yes			Yes
Number of firms	108			108
Number of observations	758			1,160
R2	.173			0.131
Adjusted R2	.147			0.110
F-statistic	6.67***			6.325***

Notes: ^***, ^**, and ^* indicate statistical significance at the 1%, 5%, and 10% level, respectively.

39To control for causality in our main model, we conducted a robustness test including the variables lagged by one year. Column 1 of Table 7 shows that our results remain consistent.

Table 7. Causality test of CEO incentive pay ratio and REM

	Column 1 (lag -1)
	Coef.	t-stat
Intercept	-0.758***	-4.923
CEO_Incentives%(-1)	0.168***	3.201
Adjusted ROA(-1)	-1.838***	-8.560
Firm size (-1)	0.035***	5.684
CEO turnover(-1)	-0.087***	-2.793
Financial leverage (-1)	-0.014	-0.243
Major shareholder VR%(-1)	0.000	1.263
CEO duality(-1)	-0.077***	-3.288
Audit quality(-1)	-0.058	-1.286
Board duality(-1)	0.009	0.317
Loss(-1)	-0.051	-1.524
Incentives=0(-1)	0.109***	3.014
Industry and year fixed effects	Yes
Number of firms	108
Number of observations	1,116
R2	0.134
Adjusted R2	0.112
F-statistic	6.216***

Notes: All independent variables are lagged by one year. ^***, ^**, and ^* indicate significance at 1%, 5%, and 10% level, respectively.

40Second, consistent with Zang (2012) and Cohen and Zarowin (2010) who propose five alternative combinations for measuring REM, we test the impact of CEO incentives on these five combinations based on the same three REM estimation models used in our main model:

REM1= (-Abn.CFO) + Abn.Prod

REM2= (-Abn.CFO) + (-Abn.Dex)

REM3= Abn.Prod

REM4= (-Abn.CFO)

REM5= (-Abn.Dex)

41The results show that the coefficients of the CEO incentive pay ratio variable in each model are positive and significant, except for (-Abn.CFO), which can be explained by the fact that executives combine the overall extent of REM. This is consistent with Cohen and Zarowin (2010) who argue that REM activities are conducted as strategic decisions by CEOs, making these activities more difficult to detect. Finally, we reran our models by including other measures of firm performance, such as return on equity and market-to-book ratio, instead of adjusted ROA. Again our results remain consistent (although not reported, all the robustness tests for alternative measures and statistical methods are available upon request).

5. Conclusion and future research directions

42This study provides insights into the effect of CEO incentive pay ratio on REM in the context of SBF 120 listed companies. We find that a higher incentive pay ratio encourages CEOs to engage more in REM activities (H1). This result is consistent with many empirical studies in different developed countries (e.g., Cohen et al., 2008; Graham et al., 2005; Ferentinou and Anagnostopoulou, 2016; Roychowdhury, 2006). In the French context, a high level of incentive pay ratio in the compensation design may exert more pressure on executives to achieve short-term performance goals. This finding highlights some serious concerns about CEO pay practices that might encourage opportunistic behavior. REM decisions may prove harmful to the firm’s future performance, in line with the managerial power approach. Nevertheless, taking into account the specific ownership structure in France, we find that the controlling shareholder ownership type can mitigate the undesired effect of incentive compensation on REM practices. Our findings point to the diverse motivations of different ownership types. We show that firms controlled by families or institutional investors are more active in monitoring CEO activities and evaluate the impact of their operational decisions. Therefore, in these firms, the attribution of incentive pay could play an important role as a governance mechanism to help align the interests of executives with those of shareholders, and hence reduce agency problems, consistent with the optimal contract approach. Our study also sheds light on the limitations of using one theoretical approach to evaluate the efficiency of incentive pay as a governance mechanism. Our results demonstrate the importance of considering the interaction between incentive pay and other governance mechanisms. We hence extend the debate on the efficiency of CEO pay design, which could help improve the regulation and/or corporate governance codes in France and other European countries with similar institutional contexts.

43Last, our study has some limitations. First, we only study REM as a measure of earnings management practices, which is less studied than AEM. Second, we do not consider the impact of risk-taking on CEO decisions concerning discretionary expenses. In fact, incentive compensation can lead COE to make “risky” decisions, especially with regard to R&D expenses. These limitations could provide promising future research avenues. Thus, we suggest extending our research to examine the effect of CEO incentive pay on other types of earnings practices, such as classification shifting, another earnings management method less studied in the literature, which CEOs might use to influence earnings reporting (Haw et al., 2011; Poonawala and Nagar, 2019). Furthermore, as an additional factor, it could be relevant in future research to consider the impact of risk-taking on CEO decisions concerning discretionary expenses on the relation between CEO pay and REM.