Family identification and earnings management in listed firms

Abstract

In this paper, we investigate the earnings management behavior of listed family firms holding the name of the family (eponymous FF). Specifically, we use a Swiss sample of 1,544 firm-year observations from 2006 to 2018 to examine the association of eponymous FF with accrual-based earnings management in general, and identify circumstances where this association does not hold. First, we find that, on average, eponymous FF exhibit less earnings management than non-FF. Second, we exploit a Swiss-specific option to voluntarily turn away from IFRS to local GAAP. Using a difference-in-differences approach, we find that eponymous FF exhibit higher levels of earnings management immediately after the switch. Finally, we show that eponymous FF exhibit higher earnings management when the family is directly involved in the board of directors or the managing board. Our findings provide a more nuanced understanding of the effects of family identification on earnings management incentives in listed firms.

Keywords:

• Earnings management
• eponymous firms
• family firms
• family identification
• IFRS turn away

1. Introduction

Family firms (FF) are major players in the economy and a growing body of research has explored the relationships between FF and earnings quality. The extant literature shows that, on average, FF are less likely to engage in earnings management than non-FF (Ali et al., 2007; Jiraporn & DaDalt, 2009; Martin et al., 2016; Wang, 2006). However, FF appear to manipulate earnings in specific situations. For example, compared to non-FF, FF are more likely to manage accruals to decrease earnings, as doing so reduces investor pressure to declare dividends (Achleitner et al., 2014). In addition, FF with high leverage are more likely to capitalize R&D expenses on the balance sheet, possibly to avoid debt covenant violations (Prencipe et al., 2008). FF also report a larger proportion of the purchase price as goodwill compared to non-family acquirers (Frii & Hamberg, 2021), which the authors attribute to a greater opaqueness motive.

The underlying assumptions of theories developed outside the FF framework are sometimes inconsistent with the FF setting (Berrone et al., 2010, 2012; Gomez-Mejia et al., 2014; Lim et al., 2010). For example, the classical principal-agent conflict that underlies agency theory is likely to be less relevant when managers are family members themselves whose ‘wealth’ may be extended to include nonpecuniary dimensions (Stockmans et al., 2010). However, heterogeneity across FF suggests that agency theory has validity for FF that have relatively an ownership structure that is relatively more diluted and/or employ a CEO or directors outside the family (e.g. Westhead & Howorth, 2007). As a result, the specific circumstances under which FF are likely to engage in earnings management remain an open question (Eddleston & Mulki, 2021; Martin et al., 2016; Prencipe & Bar-Yosef, 2011).

Previous literature documents that family owners are concerned about both financial and socioemotional aspects of their stake in the company when making business decisions (Berrone et al., 2010, 2012; Gomez-Mejia et al., 2014). For example, FF may refrain from undertaking questionable business practices if such practices are perceived to negatively affect the family’s reputation (Martin et al., 2016). Although family identification with the firm is a key characteristic for FF in general (Arena & Michelon, 2018; Gomez-Mejia et al., 2014), there is significant inter-FF variation in the degree to which family identity and firm identity are congruent (Eddleston & Mulki, 2021; Rousseau et al., 2018; Sundaramurthy & Kreiner, 2008). In other words, FF should not be treated as a single homogenous group (Rousseau et al., 2018; Zellweger et al., 2010).

To proxy for strong identification of the family with the firm, prior research uses the concept of eponymy, i.e. firms whose name contains the family name (hereafter eponymous FF). In eponymous FF, the family’s reputation is directly tied to the firm (Arena & Michelon, 2018; Deephouse & Jaskiewicz, 2013; Minichilli et al., 2022; Zellweger et al., 2010).¹ Consistent with the argument that family reputation concerns act as a disciplining mechanism and reduce earnings management, recent research shows that private eponymous firms in Italy and Norway exhibit higher financial reporting quality than other FF (Minichilli et al., 2022; Sundkvist & Stenheim, 2022).

However, it is not obvious whether the negative association between eponymous FF and earnings management should also exist for listed firms. Given that the span of decision makers interested in a publicly-listed firm is broader than for a private firm (Bar-Yosef et al., 2019), the greater complexity of agency structures and capital market pressure faced by a listed firm increase both the firm’s incentives to manage earnings and its incentives to implement strong monitoring mechanisms (Beuselinck et al., 2023). On the one hand, the greater visibility of financial information increases the importance of providing high quality financial statements in order to preserve the family’s reputation. On the other hand, given the risk that the controlling family may extract private economic benefits at the expense of other investors (Stockmans et al., 2010), the family’s reputation can also be affected by its ability to provide superior financial performance to minority shareholders, increasing its incentives to overstate reported performance.² Due to this greater complexity, whether listed eponymous FF exhibit lower earnings management is an empirical question.

The earnings management decision can be viewed as a gamble in which the expected benefits from earnings management, such as a higher perceived firm performance, are weighted against the expected costs incurred by ‘getting caught’, such as tarnishing the firm’s reputation or lawsuits (e.g. Gomez-Mejia et al., 2014; Martin et al., 2016). It is a gamble because the benefits and costs are both uncertain at the time of the decision. In particular, the expected costs are a function of both the probability of getting caught and the actual costs incurred if the earnings management scheme is uncovered. Even if strong identification of the family with the firm reduces earnings management for listed FF on average, the family’s willingness to curtail earnings management practices could be weakened in situations in which the probability of getting caught is perceived to be lower.

Whether family involvement (i.e. the CEO and/or the board chairperson is a member of the family) impacts the association between eponymous FF and earnings management behavior is also unclear. On the one hand, family involvement may strengthen the baseline negative association between eponymous FF and earnings management because family involvement provides the family with the ability, in addition to the willingness, to manage earnings less (De Massis et al., 2014; Evert et al., 2018). On the other hand, involvement may weaken the negative association through two mechanisms likely to be exacerbated in publicly-listed eponymous FF. First, eponymous family-promoted CEOs/board chairpersons may a priori suffer from a legitimacy deficit and suspicion of underperformance (Amore et al., 2021; Bennedsen et al., 2007; Pérez-González, 2006), and consequently develop incentives to meet other shareholders’ and analysts’ expectations through increased earnings management. Second, family-promoted CEOs/board chairpersons may be prone to implicit employment commitments that reduce their ability to make performance-enhancing changes (Bach & Serrano-Velarde, 2015; Lemos & Scur, 2019) and therefore need another lever (i.e. earnings management) to manage financial performance.

In this paper, using different statistical methods (OLS, Heckman (1979) two-stage procedure and entropy balancing), we examine the association between family identification with the firm, as proxied by eponymy, and earnings management for listed FF in Switzerland, and we uncover settings in which this association does not hold. Swiss FF have several characteristics that make them particularly interesting for this investigation. First, Swiss listed FF are older than in most other countries, as the average age of Swiss listed FF is higher than in every other world region (Crédit Suisse, 2017). While the average age of Norwegian (Italian) private FF in Sundkvist and Stenheim (2022) (Minichilli et al., 2022) is 17 (29) years, the average age of FF in our sample is 79 years. Therefore, as the firms we study are comparatively more distant from the founder than those analyzed in previous studies, we argue that this is a strong setting to assess the impact of the ‘family’ compared to that of the ‘founder’ only.

Second, Swiss listed FF are also larger. According to Crédit Suisse (2017), out of 25 countries, Switzerland ranks fourth for the average market capitalization of family-owned companies. In addition, on a per capita basis, Switzerland hosts the highest concentration of large FF in Europe.³ As a result, relative to their counterparts in other countries, Swiss FF are particularly visible to the public, which makes the family more likely to be concerned by its reputation as part of its socioemotional wealth. This greater visibility makes the identity of Swiss FF salient, rendering the family more vigilant about the implications of earnings management.

Third, Swiss listed firms have the option to voluntarily turn away from IFRS to Swiss GAAP without any other important changes in listing or legal requirements (Fiechter et al., 2018). While such a change likely has permanent implications for the firm’s reporting environment,⁴ it also provides a break in the time series comparability of accounting numbers, temporarily making earnings management more difficult to detect. We exploit this distinctive feature to analyze the consequences of such a decision on earnings management for eponymous FF.

Using a sample of 1,544 firm-year observations (for 125 firms) over the 2006–2018 period, we first find that, on average, listed eponymous FF exhibit lower levels of earnings management than non-FF, while there is no difference in earnings management between other FF and non-FF. Second, we find evidence of higher earnings management for eponymous FF following a voluntary switch from IFRS to Swiss GAAP, relative to non-FF switchers. This finding only applies to the first year under Swiss GAAP. This is consistent with the argument that the expected costs in the earnings management gamble, which are generally higher for eponymous FF than for non-FF, are temporarily lower due to the break in comparability that accompanies the switch. Third, we find that the negative association between eponymous FF and earnings management disappears when family members are involved at the CEO and/or board chairperson levels, indicating that the interplay of individual and family incentives may be more complex than the desire to safeguard the family reputation at all cost.

Our paper contributes to the literature in several ways. First, we complement past literature by offering a more granular understanding of the association between the identity dimension – the most central, distinctive, and enduring organizational feature (Albert & Whetten, 1985) – and earnings management for listed firms. Although recent literature has documented that private eponymous FF are less likely to engage in accrual-based earnings management than other private firms (Minichilli et al., 2022; Sundkvist & Stenheim, 2022), the large differences in reporting incentives between listed and private FF limit their generalizability. Our results are consistent with the argument that the greater complexity of agency structures faced by listed firms should lead to a more nuanced assessment of whether the family’s incentive to protect its reputation systematically leads to lower earnings management.

Second, our study also contributes to the literature on IFRS turn away (Fiechter et al., 2018; Raffournier & Schatt, 2018). We provide empirical evidence of one of its consequences, namely higher levels of earnings management in the year of the switch for eponymous FF. We thus respond to the recent call by Gordon et al. (2019, p. 10) suggesting more analyses to study ‘[…] what happens if an IFRS adopting country allows companies to drop IFRS and revert to domestic GAAP, like what happened in Switzerland […].’

Finally, our paper adds to the body of literature examining the influence of family involvement on accounting outcomes (Ferramosca & Allegrini, 2018). Our finding that listed eponymous FF in Switzerland exhibit more earnings management when the CEO or the board chairperson are family members than when they are not suggests that socioemotional incentives such as the preservation of the family’s reputation may vary depending on the mechanisms through which the family is associated with the firm.

2. Hypotheses development

2.1. Family identification with the firm in light of the behavioral agency theory

The behavioral agency theory, originally developed by Wiseman and Gomez-Μejia (1998) and subsequently adapted to FF (Lim et al., 2010), is based on some of the shortcomings of agency theory. Under this theory, FF are characterized by the interactions of actors that are emotionally tied to the firm and simultaneously care about both financial and nonpecuniary benefits (Gomez-Mejia et al., 2011; Prencipe et al., 2014). There is a growing body of evidence suggesting that on top of protecting their economic wealth for current and future generations, family members consider noneconomic goals when making business decisions (e.g. Cennamo et al., 2012; Gomez-Mejia et al., 2000). Not necessarily in contradiction with agency theory (Martin et al., 2016; Westhead & Howorth, 2007), researchers posit that family members are not only (financial) risk averse economic agents but also loss-averse to their socioemotional wealth (Gomez-Mejia et al., 2007), which leads them to consider factors beyond economic self-interest in business decisions.

Within this framework, a widely-documented distinctive characteristic of FF is the degree of identification of the family to the firm, which inextricably ties the family’s reputation and external image to that of the firm (Deephouse & Jaskiewicz, 2013; Dyer & Whetten, 2006; Zellweger et al., 2010). In most cases, the identity characteristic is appraised through the congruence between family name and company name. According to Deephouse and Jaskiewicz (2013, p. 340), a shared family name ‘[…] enhances family members’ identification with the family firm and consequently their motivation to pursue a favorable reputation.’ For example, Belenzon et al. (2017) show that eponyms have better performance than other firms and that this effect is stronger for families with rarer names, a finding that they attribute to the greater reputational benefits/costs that stem from successful/unsuccessful business outcomes.

2.2. Association between eponymous FF and earnings management for listed firms

Earnings management is often depicted as a gamble where the objective is to generate short-term financial benefits while avoiding the costs associated with misconduct detection and a subsequent detrimental reputation loss (Gomez-Mejia et al., 2014; Martin et al., 2016). In this respect, it can be argued that FF, and especially eponymous FF, face higher costs in that tradeoff, reducing their willingness to engage in earnings management. Indeed, favorable family reputation and image maintenance are among the crucial goals FF pursue to preserve socioemotional wealth (Dyer & Whetten, 2006; Zellweger et al., 2013). As argued by Deephouse and Jaskiewicz (2013), a strong identification with the FF motivates family members to preserve the firm’s legitimacy and reputation. However, this link has bidirectional effects: any negative media coverage of the firm, such as the uncovering of improper financial reporting practices, could represent ‘indelible stains’ on family members, their company, and their family name (Adams et al., 1996; de Vries, 1993; Dyer & Whetten, 2006). For instance, Akio Toyoda’s − President of the Toyota Motor Corporation − 2010 testimony to the U.S. Congress included those terms (The Guardian, 2010): ‘as you well know, I am the grandson of the founder, and all the Toyota vehicles bear my name. For me, when the cars are damaged, it is as though I am as well. […] My name is on every car. You have my personal commitment that Toyota will work vigorously and unceasingly to restore the trust of our customers.’

Employees demonstrate personal attachment to companies named after the family (Belot & Waxin, 2017) and non-family employees can behave as stewards of the FF identity and values (Zellweger et al., 2010). As such, they contribute to achieving the family’s objectives oriented toward firm’s reputation preservation underpinned by honesty in communication. In this context, Minichilli et al. (2022) recently used a large sample of Italian private firms to show that eponymous FF have higher financial reporting quality than other FF, and Sundkvist and Stenheim (2022) showed that, in Norway, eponymous FF exhibit lower measures of accrual-based earnings management than other FF.

How a public company listing affects the private firm calculus of family reputation costs is not immediately apparent. On the one hand, for listed firms, financial statements are a more visible and therefore more important aspect of communications with stakeholders (Peek et al., 2010), potentially increasing the reputational damages sustained when the firm’s questionable business practices come to light. Indeed, Burgstahler et al. (2006) show that, in general, listed firms manage earnings less than private firms. Listed firms are also less likely to choose aggressive tax treatments (e.g. Beuselinck et al., 2015; Hoopes et al., 2018; Mills & Newberry, 2001), which has direct implications for accrual-based earnings management in settings where book-tax conformity is high (e.g. Blaylock et al., 2015).

On the other hand, the greater complexity of agency structures in listed firms requires a more nuanced characterization of incentives (Beuselinck et al., 2023). For example, listed firms have minority shareholders, whose focus on stock price generates pressure to deliver short-term performance. As a result, minority shareholders and other outsiders will scrutinize firm performance and the family’s reputation will be tied to the firm’s financial successes or failures. In this context, Vural (2018) illustrates that Swedish FF provide less disclosure than non-FF so as to avoid potential reputation damages from proprietary information disclosure. In turn, families may have greater incentive to manipulate earnings in specific circumstances where poor financial performance could cause them to lose control of the firm (Gomez-Mejia et al., 2014). Although focused on all firms rather than only FF, Hope et al. (2013) show that the superior accounting quality of listed firms decreases or disappears in settings where firms face higher potential benefits or lesser potential costs from earnings management, such as when the demand for financial information is lower.

This discussion implies that there is a potentially greater set of factors affecting family reputation when the family is associated with a listed firm than with a private firm. Nevertheless, because the avoidance of business scandals is a permanent incentive while the incentive to provide high financial performance to minority shareholders is context-specific, we argue that, on average, identity FF should exhibit lower levels of earnings management than other firms, leading to the following hypothesis:

Hypothesis 1 (H1): Listed eponymous FF are negatively associated with earnings management.

2.3. Association between eponymous FF and earnings management – specific contexts

The behavioral agency theory allows for varying risk preferences under different decision conditions (Kumeto, 2015). It thus makes it possible to investigate specific contexts where the baseline effect (H1) may not hold. To do so, we study distinct settings, by first exploring a temporary decline in the time series comparability of financial reports – IFRS turn away – and then a more permanent structural context – family involvement.

2.3.1 IFRS turn away

Firms quoted on the SIX Swiss Exchange can choose between IFRS, US GAAP and local (Swiss) GAAP, without any substantive differences in other listing or legal requirements (Fiechter et al., 2018). Firms can also voluntarily switch from one set of standards to the other. Although the basic principles underlying Swiss GAAP and IFRS are similar, important differences exist regarding the accounting treatment of goodwill, pension and hidden reserves, as well as segment disclosure requirements (Fiechter et al., 2018; Raffournier, 2017).

Fiechter et al. (2018) investigate the voluntary decision to switch from IFRS to Swiss GAAP and find that smaller firms with higher ownership concentration and fewer foreign investors have a greater propensity to switch. In addition, they find that IFRS complexity and implementation costs are the two most frequent reasons cited by switching firms in the press release announcing the move. On average, switching firms’ notes to financial statements are 11 pages shorter after the switch than before the switch. Despite this substantial reduction in the quantity of audited disclosures, Raffournier and Schatt (2018) find that switching back to Swiss GAAP does not result in a decrease in audit fees, suggesting that at least some of the expected benefits from switching back to Swiss GAAP may not come to fruition.⁵

Given the greater ownership concentration and smaller size of many listed FF, one could expect FF to be more likely to switch back to Swiss GAAP. However, an empirical question is whether eponymous FF, for whom reputation costs are a key issue, follow the same pattern. On the one hand, if the owning family wants the firm to maintain a reputation of high earnings quality, it could be argued that continued IFRS reporting would be preferable. On the other hand, given the absence of a negative market reaction to the IFRS turn away (Fiechter et al., 2018), one could argue that there may be no difference in perceived financial reporting quality between IFRS and Swiss GAAP and that eponymous FF may not suffer from reputation loss due to the switch.

In the earnings management decision ‘gamble’ (e.g. Martin et al., 2016), firms trade off the (ex ante) benefits and costs of earnings management. The expected costs depend on the probability that outsiders will learn about the attempt, which is generally high (Das et al., 2011; Zang, 2012), and on the reputational or legal costs incurred if the earnings management scheme is uncovered, which are generally higher for FF than non-FF due to families’ reputation concerns (Gomez-Mejia et al., 2014). Whether firms engaging in earnings management face a greater probability of ‘getting caught’ under Swiss GAAP than under IFRS is unclear. The absence of a post-switch decline in liquidity (Fiechter et al., 2018) suggests that the market does not perceive Swiss GAAP to provide more extensive earnings management opportunities than IFRS. However, the switch itself may provide firms with a short time period during which the probability of detection is temporarily lower. Indeed, a change in accounting standards requires an adaptation period, for both the firm and its auditor, which may give the firm a time window during which more flexibility is tolerated. In addition, the switch can be expected to generate a break in the time series comparability of accounting numbers, making detection of earnings management more difficult for outsiders. In turn, this lower probability of detection temporarily reduces the expected cost of managing earnings. This reduction applies to all switching firms but should be most pronounced for firms that have the most to lose if they get caught (i.e. eponymous FF).⁶ As a result, we posit that the voluntary decision to switch from IFRS to Swiss GAAP is a unique opportunity to manage earnings for eponymous FF and propose the following hypothesis:

Hypothesis 2 (H2): Listed eponymous FF are positively associated with earnings management around the voluntary decision to switch from IFRS to Swiss GAAP.

2.3.2 Family involvement in governance structures

Decisions on earnings levels involve both the executive board and the board of directors (Achleitner et al., 2014), making them relevant candidates to investigate interactions affecting our baseline hypothesis (H1). Favoring the selection of family members to occupy positions such as CEO and/or chairperson occurs frequently as families attempt to maintain transgenerational control and subsequently enjoy non-financial private benefits associated with this kind of involvement (He & Yu, 2019; Prencipe & Bar-Yosef, 2011). However, the effect of family involvement on earnings management in eponymous FF is unclear. Some arguments predict an additional decrease while opposite ones predict an increase in earnings management.

On the one hand, family involvement may be expected to enhance the negative association between eponymous FF and the propensity to manage earnings. First, family CEOs are more likely to develop strong organizational identity and to internalize the family’s non-financial motives (Cui et al., 2018; Martin et al., 2016), such as good reputation preservation – a particularly salient objective within eponymous FF. The alignment of these goals between family CEO and/or family chairperson and the family as a whole is favored by a long process of socialization, reduced information asymmetry and higher trust and intimate knowledge of the firm (Visintin et al., 2017). As a corollary, the family is able to closely monitor one of its members. Martin et al. (2016, p. 456) state that, in this configuration, family principals ‘[…] may analyze major decisions and financial statements more directly.’ In fine, in addition to voting rights, family involvement in key leadership positions represents an increase in effective control, improving the family’s ability to advance its agenda (De Massis et al., 2014; Evert et al., 2018). In other words, given the stronger incentive to preserve the firm’s reputation in an identity FF, one could argue that involvement at the CEO/chairperson level provides the family with an additional lever to curb earnings management practices within the firm. This leads us to formulate the following directional hypothesis:

Hypothesis 3a (H3a): Family involvement at the CEO/board chairperson levels strengthens the negative association between listed eponymous FF and earnings management.

Alternatively, family involvement in listed firms could weaken the negative association between eponymous FF and earnings management, or even reverse the association, for at least two reasons. First, family-promoted CEOs/board chairpersons a priori suffer from a legitimacy deficit as FF run by family CEOs underperform comparable FF led by nonfamily CEOs (Bennedsen et al., 2007; Pérez-González, 2006). Amore et al. (2021, p. 1434) summarize the leading explanations for this phenomenon: family-drawn CEOs may be prone to dysfunctional nepotism, nomination is based on kinship rather than merit and selection is made from a small family talent pool. As a reaction, involved members in listed identity FF may develop their own set of incentives, such as the need to show their ability to generate performance for minority shareholders and financial analysts (i.e. build legitimacy). Facing this pressure, involved family members may increase earnings management, thereby adopting a behavior inconsistent with the whole family’s incentive to avoid reputation loss. Second, family-promoted CEOs/board chairpersons may be prone to implicit employment commitments, according to which current family leadership is unwilling to question or undo previous choices, which hinders innovation and the remediation of poor decisions (Bach & Serrano-Velarde, 2015; Lemos & Scur, 2019). Consequently, involved family members may resort to earnings management to boost financial performance. If either of these arguments – legitimacy deficit or implicit employment commitments – were to be true, it would weaken the listed eponymous FF’s ability to reflect its founding family’s reputation-preserving objective through the avoidance of earnings management.⁷ This leads us to formulate the following directional hypothesis, which is the opposite of H3a:

Hypothesis 3b (H3b): Family involvement at the CEO/board chairperson levels weakens the negative association between listed eponymous FF and earnings management.

3. Research design

3.1. Sample selection

We start with all Swiss firms that are publicly-listed on the SIX Exchange from 2006 to 2018 on Refinitiv Datastream and remove financial firms, duplicates, firms with an ISIN not starting with CH, and firms with missing data. Then, using annual reports and company publications, we hand-collect the name and percentage of voting rights held by the main shareholders, whether the CEO and/or the chairperson are family members for FF (we provide details on how FF are defined in section 3.2.), the accounting standards followed, and corporate governance information. We retrieve other variables from Refinitiv Datastream. The final sample is composed of 1,544 firm-year observations for 125 different companies. Table 1 Panel A depicts the sample selection process.

Table 1. Sample.

Steps				Firm-year observations	Firms
Panel A. Sample selection.
All firms listed on SIX between 2006 and 2018 (primary quote, active or dead, incorporated in Switzerland)				8,606	662
Minus firms for which Datastream reports an error				−3,250
Minus financials, duplicates, firms with an ISIN not starting with CH, and firms with missing data on Datastream				−3,679
Minus firms with missing data in annual reports				−133
Final sample				1,544	125
ICB industry name	All non-FF	All FF	Eponymous FF	Non-eponymous FF	Total
Panel B. Firm-year observations by firm type and industry.
Basic materials	73	52	0	52	125
Consumer goods	100	113	0	113	213
Consumer services	97	32	0	32	129
Health care	167	60	13	47	227
Industrials	348	292	83	209	640
Technology	94	42	12	30	136
Telecommunication	13	0	0	0	13
Utilities	61	0	0	0	61
Total	953	591	108	483	1,544

3.2. Measurement of family-related variables

In this study, we follow the now ‘standard practice’ (Eugster & Isakov, 2019, p. 4) and use a 20% threshold of voting rights owned by a family to define a FF. As the classification between FF and non-FF is sometimes difficult due to the complexity of the ownership structure (e.g. a shareholder holding shares in his own name but also through a holding company that he controls), we supplement the analysis of the annual reports with internet searches to trace the history of the company’s shareholding structure. FAM is a dummy variable equal to one if the percentage of voting rights owned by the family is equal or above 20%, and zero otherwise.⁸ Among FF, FAM_EPONYM is a dummy variable equal to one if the name of the FF contains the family name, and zero otherwise (Arena & Michelon, 2018; Deephouse & Jaskiewicz, 2013). Finally, FAM_OTHER is a dummy variable equal to one if one or more families own more than 20% of the voting rights but do not name the company after the family, and zero otherwise. As shown in Table 1 Panel B, out of the 1,544 firm-year observations in the full sample, 591 observations are classified as FF. Among these, 108 observations are classified as identity FF and 483 observations relate to non-identity ones. Regarding the distribution of the sample across industries, most FF belong to the Industrials (N = 292) or the Consumer goods (N = 113) industries, while most other firms belong to the Industrials (N = 348) or the Health care (N = 167) industries.

3.3. Earnings management proxy

Our earnings management proxy is accruals-based and follows the Francis et al. (2013) total accruals model. Abnormal accruals are defined as the firm-year-specific residuals of a cross-sectional modified Jones model that controls for contemporaneous firm performance (Dechow et al., 1995; Jones, 1991; Kothari et al., 2005) and industry and year fixed effects (FE): (1) $\mathrm{T}\mathrm{O}\mathrm{T}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\mu }_1{\displaystyle \frac{1}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_2{\displaystyle \frac{\left(\mathrm{\Delta }{\mathrm{S}}_{\mathrm{i}\mathrm{t}}\text{-}\mathrm{\Delta }\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\right)}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_3{\displaystyle \frac{\mathrm{P}\mathrm{P}{\mathrm{E}}_{\mathrm{i}\mathrm{t}}}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_4\mathrm{R}\mathrm{O}{\mathrm{A}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}}}}$(1) In Equation (1(1) $\mathrm{T}\mathrm{O}\mathrm{T}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\mu }_1{\displaystyle \frac{1}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_2{\displaystyle \frac{\left(\mathrm{\Delta }{\mathrm{S}}_{\mathrm{i}\mathrm{t}}\text{-}\mathrm{\Delta }\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\right)}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_3{\displaystyle \frac{\mathrm{P}\mathrm{P}{\mathrm{E}}_{\mathrm{i}\mathrm{t}}}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_4\mathrm{R}\mathrm{O}{\mathrm{A}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}}}}$(1) ), TOTACC is total accruals (net income before extraordinary items, less cash flows from operations, scaled by total assets), TA is the firm’s total assets, ΔS is the variation in sales, ΔAR is the variation in accounts receivable, PPE is gross property, plant, and equipment, and ROA is the return on assets. Our proxy for earnings management is absolute abnormal accruals (ABSABNACC), defined as the absolute value of the residual${ϵ}_{it}$. A lower value of ABSABNACC indicates less earnings management.⁹

3.4. Main models

To test H1 on the association between eponymous FF and earnings management, we use the following model: (2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) In Equation (2(2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) ), FAMVAR is alternatively equal to FAM, FAM_EPONYM or FAM_OTHER. Consistent with H1, we expect a negative coefficient on FAM_EPONYM, which would indicate that eponymous FF exhibit lower levels of earnings management than a control group of non-FF. CONTROLS is a vector of control variables (Dechow et al., 2010). Among these, FAM_INVOLVEMENT is a dummy variable equal to one if the CEO and/or the board chairperson is a family member, and zero otherwise. Growth opportunities are measured with the market-to-book ratio (MTB), ownership structure is assessed via the percentage of closely held shares (OWN),¹⁰ financial leverage is measured as the debt-to-assets ratio (LEVERAGE), firm size is measured as the natural logarithm of total assets (SIZE), and the age of the firm (AGE) is measured as the current year minus the incorporation year, ranked into quartiles. We also include dummy variables taking the value of one if the firm has an audit committee that is only composed of independent members (AC_INDEP), applies IFRS (IFRS), incurs a loss (DLOSS), or is audited by a Big 4 company (BIG4), along with industry and year fixed effects. In line with Chen et al. (2018), CONTROLS includes all of the first-step regressors that were used to calculate abnormal accruals in Equation (1(1) $\mathrm{T}\mathrm{O}\mathrm{T}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\mu }_1{\displaystyle \frac{1}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_2{\displaystyle \frac{\left(\mathrm{\Delta }{\mathrm{S}}_{\mathrm{i}\mathrm{t}}\text{-}\mathrm{\Delta }\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\right)}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_3{\displaystyle \frac{\mathrm{P}\mathrm{P}{\mathrm{E}}_{\mathrm{i}\mathrm{t}}}{\mathrm{T}{\mathrm{A}}_{\mathrm{i}\mathrm{t}\text{-}1}}+{\mu }_4\mathrm{R}\mathrm{O}{\mathrm{A}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}}}}$(1) ).¹¹

In order to investigate whether the voluntary decision to turn away from IFRS to Swiss GAAP is related to a change in earnings management for eponymous FF (H2), we use a staggered difference-in-differences model to account for the fact that the switch occurs in different firms at different points in time: $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+{\text{β}}_2\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{1}_{\mathrm{i}\mathrm{t}}+{\text{β}}_3\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{2}_{\mathrm{i}\mathrm{t}}+{\text{β}}_4\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{3}_{\mathrm{i}\mathrm{t}}$ $+{\text{β}}_5\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{1}_{\mathrm{i}\mathrm{t}}+{\text{β}}_6\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{2}_{\mathrm{i}\mathrm{t}}$ (3) $+{\text{β}}_7\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{3}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{m}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(3)

Given that the decision to switch back to Swiss GAAP is not random, we use a Heckman (1979) two-stage model in which we model the propensity to switch in the first stage, and then include the inverse Mills ratio as a regressor in the second stage. In line with common usage of staggered difference-in-differences models (Dou et al., 2016; Fauver et al., 2017; Giroud, 2013; Jiang et al., 2019; Kausar et al., 2016), the model in Equation (3(3) $+{\text{β}}_7\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{3}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{m}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(3) ) includes firm and year fixed effects. All firms that apply the same accounting standards throughout the sample period (non-switchers) are considered to be the control group. The treated group is composed of firms that turned away from IFRS to Swiss GAAP between 2006 and 2018 (switchers). Our test uses four years of data for switchers: the last year under IFRS and the first three years under Swiss GAAP. In Equation (3(3) $+{\text{β}}_7\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{3}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{m}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(3) ), POST1, POST2 and POST3 are binary post-treatment variables for the treated group equal to one for the first, second and third year under Swiss GAAP, respectively. Because switching firms include FF and non-FF, we include interaction terms between FAMVAR and these three variables. As a result, the coefficient on POST1 (POST2, POST3) measures the average difference in ABSABNACC between the first (second, third) year after the switch and the pre-switch period for non-FF switchers. In turn, the coefficients on the interaction terms represent the incremental difference in ABSABNACC for FF switchers relative to non-FF switchers. Consistent with H2 and the argument that reputation costs are temporarily lower immediately after the switch for listed identity FF, we expect a positive coefficient on FAMVAR*POST1.¹² However, we make no predictions regarding the coefficients on FAMVAR*POST2 and FAMVAR*POST3 because we expect reputation costs to return to pre-switch levels. CONTROLS includes the same control variables as in Equation (2(2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) ) but also the SALES_GROWTH variable to make sure that our results are not due to a growth differential between eponymous and other FF.¹³

Finally, to test H3a/H3b on the interaction effect of eponymous FF and family involvement on earnings management, we use the following model: $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}\mathrm{R}\mathrm{\_}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}+{\text{β}}_2\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}\mathrm{R}\mathrm{\_}\mathrm{C}{}_{\mathrm{i}\mathrm{t}}\ast \mathrm{F}\mathrm{A}\mathrm{M}\mathrm{\_}\mathrm{I}\mathrm{N}\mathrm{V}\mathrm{O}\mathrm{L}\mathrm{V}\mathrm{E}\mathrm{M}\mathrm{E}\mathrm{N}{\mathrm{T}}_{\mathrm{i}\mathrm{t}}$ (4) $+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(4) Equation (4(4) $+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(4) ) is identical to Equation (2(2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) ), except that FAMVAR_C is a continuous variable to allow for the inclusion of the additional interaction term FAMVAR_C*FAM_INVOLVEMENT. Indeed, we include the percentage of voting rights owned by the family to form the set of FAMVAR_C continuous variables (FAM_C, FAM_EPONYM_C or FAM_OTHER_C). A negative (positive) and significant coefficient on FAM_EPONYM_C*FAM_INVOLVEMENT would be consistent with H3a (H3b).

4. Results

4.1. Descriptive statistics and correlation matrix

Table 2 reports descriptive statistics about firm-specific characteristics for the full sample (N = 1,544) in Panel A, and for FF (N = 591) and non-FF (N = 953) in Panel B. FF observations represent 38%¹⁴ of the full sample (42 firms in 2006 up to 48 in 2018). Eponymous FF represent 18% of the FF subsample. Among FF, family involvement (FAM_INVOLVEMENT) occurs in 54% of observations. Average absolute abnormal accruals (ABSABNACC) are equal to 0.081 for the full sample.¹⁵ It is significantly smaller for FF (0.065) compared to non-FF (0.091).¹⁶ FF differ from non-FF on a number of other aspects. FF have fewer fully independent audit committees (AC_INDEP) and tend to apply IFRS less frequently than non-FF (56.0% versus 66.6%). Non-FF report a loss (DLOSS) more frequently than FF (19.5% versus 12.4%). Growth opportunities measured as the market-to-book ratio (MTB) tend to be lower for FF (2.653 versus 2.783), although the difference is not statistically significant. Ownership concentration (OWN) is higher for FF than non-FF (47.8% versus 33.4%). Mean leverage (LEVERAGE) measured by the debt-to-asset ratio is around 17% for both subsamples. Finally, FF tend to be smaller (SIZE) and older (AGE) than non-FF and less likely to mandate a Big 4 as an external auditor (BIG4).

Table 2. Descriptive statistics.

	Mean	Std. dev.	Q1	Median	Q3
Panel A. Full sample (N = 1,544).
ABSABNACC	0.081	0.076	0.029	0.059	0.106
FAM	0.380	0.486	0.000	0.000	1.000
FAM_EPONYM	0.069	0.254	0.000	0.000	0.000
FAM_INVOLVEMENT	0.205	0.404	0.000	0.000	0.000
AC_INDEP	0.647	0.478	0.000	1.000	1.000
IFRS	0.626	0.484	0.000	1.000	1.000
DLOSS	0.168	0.374	0.000	0.000	0.000
MTB	2.734	3.030	1.073	1.835	3.231
OWN	0.389	0.257	0.169	0.360	0.591
LEVERAGE	0.171	0.155	0.034	0.141	0.271
SIZE	13.566	1.846	12.348	13.432	14.691
BIG4	0.884	0.321	1.000	1.000	1.000
AGE	0.666	0.285	0.5000	0.750	1.000
	Family firms		Non-family firms
	Mean	Median	Mean	Median
Panel B. Family firms (N = 591) and non-family firms (N = 953).
ABSABNACC	0.065	0.050	0.091***	0.066***
FAM_EPONYM	0.183	0.000	0.000***	0.000***
FAM_INVOLVEMENT	0.540	1.000	0.000***	0.000***
AC_INDEP	0.614	1.000	0.667**	1.000**
IFRS	0.560	1.000	0.666***	1.000***
DLOSS	0.124	0.000	0.195***	0.000***
MTB	2.653	1.781	2.783	1.890
OWN	0.478	0.516	0.334***	0.294***
LEVERAGE	0.170	0.126	0.171	0.154
SIZE	13.458	13.053	13.632*	13.689**
BIG4	0.863	1.000	0.896**	1.000**
AGE	0.689	0.750	0.636***	0.500***
	Eponymous FF		Other FF
	Mean	Median	Mean	Median
Panel C. Eponymous family firms (N = 108) and other family firms (N = 483).
ABSABNACC	0.065	0.054	0.065	0.049
FAM_INVOLVEMENT	0.815	1.000	0.478***	0.000***
AC_INDEP	0.639	1.000	0.609	1.000
IFRS	0.500	0.500	0.573	1.000
DLOSS	0.056	0.000	0.139**	0.000**
MTB	2.865	2.061	2.685	1.829*
OWN	0.459	0.487	0.482	0.517
LEVERAGE	0.176	0.116	0.169	0.127
SIZE	13.543	13.228	13.439	12.963
BIG4	0.861	1.000	0.863	1.000
AGE	0.826	1.000	0.658***	0.750***

Notes: This table presents descriptive statistics for dependent, independent and control variables. All variables are defined in the Appendix. In Panels B and C, ***, ** and * indicate statistical significance of a t-test of difference in means, or Mann-Whitney test of difference in medians (FF versus non-FF in Panel B and identity FF versus other FF in Panel C) at the 1%, 5% and 10% levels, respectively.

Panel C provides descriptive statistics for eponymous FF (N = 108) and other FF (N = 483). Overall, only three variables appear to have a significantly different mean. Greater family involvement is found in eponymous FF, while these firms also disclose losses less frequently and are older than their other FF counterparts.

Table 3 presents the Pearson correlation matrix for the dependent, independent and control variables, for the full sample. The correlation coefficients between the dependent variable ABSABNACC and FAM, FAM_OTHER, FAM_INVOLVEMENT, AC_INDEP, IFRS, SIZE, BIG4 and AGE are negative and significant (at a 1% level), while it is positive and significant with DLOSS. Family variables (variables 2–5) are significantly and positively correlated with one another.

Table 3. Correlation matrix.

	1	2	3	4	5	6
1: ABSABNACC	1
2: FAM	−0.1365*	1
3: FAM_EPONYM	−0.059	0.3529*	1
4: FAM_OTHER	−0.0931*	0.5436*	−0.1144*	1
5: FAM_INVOLVEMENT	−0.106*	0.649*	0.4091*	0.2018*	1
6: AC_INDEP	−0.1665*	−0.0450	−0.0030	0.0840*	−0.0990*	1
7: IFRS	−0.1041*	−0.0871*	−0.0624*	−0.0350	−0.0677*	0.3110*
8: DLOSS	0.1838*	−0.0870*	−0.0781*	−0.0300	−0.0210	−0.0340
9: MTB	0.0612	−0.023	0.0154	−0.0160	0.0060	0.1030*
10: OWN	−0.0270	0.2456*	0.0702*	0.0719*	0.2045*	−0.1029*
11: LEVERAGE	0.0641	0.0212	0.0196	0.0137	0.0083	0.0389
12: SIZE	−0.2459*	−0.0300	−0.0030	−0.0480	−0.0955*	0.2588*
13: BIG4	−0.1487*	−0.0480	0.0005	−0.0160	−0.0310	0.0753*
14: AGE	−0.0653*	0.0939*	0.1563*	−0.0025	−0.0416	0.0304
Table 3. Correlation matrix (continued).
	7	8	9	10	11	12	13	14
7: IFRS	1
8: DLOSS	0.0879*	1
9: MTB	0.0761*	0.0407	1
10: OWN	−0.0714*	0.0504	−0.094*	1
11: LEVERAGE	−0.0200	0.1277*	0.0962*	−0.1043*	1
12: SIZE	0.2162*	−0.2866*	−0.0300	−0.2168*	0.1291*	1
13: BIG4	0.1606*	−0.0619*	0.0676*	−0.1697*	−0.0715*	0.1692*	1
14: AGE	−0.1104*	−0.1533*	−0.2028*	0.0133	−0.0627*	0.1296*	−0.0599	1

Notes: This table presents Pearson correlation coefficients for dependent, independent and control variables. All variables are defined in the Appendix.

* indicates statistical significance at the 1% level.

Although the Pearson correlation coefficients displayed in Table 3 do not suggest serious multicollinearity problems, we compute (untabulated) variance inflation factors (VIF) to further assess potential multicollinearity problems. VIF values¹⁷ range from 1.15–1.86 and therefore do not show problematic multicollinearity in our data.

4.2. Eponymous FF and earnings management

Table 4 presents regression results from Ordinary Least Squares (OLS) as well as entropy balancing estimations of Equation (2(2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) ). Entropy balancing is a multivariate matching approach that has the advantage, compared to propensity score matching, of not requiring to adjust a propensity model, but also of ensuring that ‘[…] higher-order moments of covariate distributions are nearly identical across treated and control samples […]’ (McMullin & Schonberger, 2020, p. 93). Indeed, in line with prior studies (e.g. Bhandari et al., 2020; Burke, 2022; Hainmueller, 2012; McMullin & Schonberger, 2020), entropy balancing can be used to identify a control group that is similar to the treated group in terms of observable covariates without losing observations as it might be the case with propensity score matching. In the present study, we apply this statistical method by balancing on all control variables, which enables us to identify a control sample of non-FF that is nearly identical to the treated sample of FF.

Table 4. Influence of eponymous FF on accrual-based earnings management.

	Absolute abnormal accruals (ABSABNACC)
	OLS model			Entropy balancing
	FF	Eponymous	Non-eponymous	FF	Eponymous	Non-eponymous
		FF	FF		FF	FF
	versus non-FF			versus non-FF
	1	2	3	4	5	6
FAM	−0.02**			−0.01**
	(−2.10)			(−2.44)
FAM_EPONYM		−0.04***			−0.04***
		(−2.65)			(−2.83)
FAM_OTHER			−0.01			−0.01
			(−1.42)			(−1.18)
FAM_INVOLVEMENT	−0.00	0.02	−0.01	0.00	0.03**	−0.00
	(−0.50)	(1.41)	(−1.15)	(0.13)	(2.03)	(−0.61)
AC_INDEP	−0.01	−0.02***	−0.01	0.00	−0.00	0.00
	(−1.57)	(−3.04)	(−1.53)	(0.44)	(−0.03)	(0.24)
IFRS	−0.01	−0.01	−0.01*	−0.01*	−0.01	−0.01*
	(−0.96)	(−1.29)	(−1.72)	(−1.79)	(−1.34)	(−1.86)
DLOSS	0.02***	0.02***	0.02**	0.02**	0.04***	0.02**
	(3.38)	(2.94)	(2.47)	(2.40)	(2.86)	(2.00)
MTB	0.00	0.00*	0.00	0.00	0.00	−0.00
	(1.48)	(1.92)	(0.93)	(0.11)	(0.74)	(−0.63)
OWN	−0.01	0.02	−0.02*	−0.02**	0.02	−0.03***
	(−0.59)	(1.07)	(−1.90)	(−2.27)	(1.02)	(−3.77)
LEVERAGE	0.00	0.00	0.01	0.03	−0.04*	0.04
	(0.12)	(0.15)	(0.25)	(1.30)	(−1.79)	(1.51)
SIZE	−0.00	−0.00	−0.00**	−0.00	−0.00	−0.00*
	(−1.05)	(−0.41)	(−2.43)	(−0.85)	(−0.33)	(−1.72)
BIG4	−0.02**	−0.01	−0.01	−0.01*	−0.04***	−0.01
	(−2.42)	(−1.30)	(−1.55)	(−1.95)	(−3.38)	(−0.78)
AGE	−0.01	−0.01	−0.02	0.01	0.05***	0.01
	(−0.84)	(−0.72)	(−1.44)	(1.01)	(2.91)	(0.76)
Constant	0.07*	0.03	0.16***	0.06**	0.04	0.10***
	(1.87)	(0.62)	(5.84)	(2.03)	(0.58)	(3.48)
Year and industry FE	Incl.	Incl.	Incl.	Incl.	Incl.	Incl.
First-step regressors	Incl.	Incl.	Incl.	Incl.	Incl.	Incl.
Observations	1,544	1,061	1,436	1,544	1,061	1,436
Adj. R-squared	0.29	0.33	0.29	0.29	0.33	0.29
F-statistic	23.57***	28.11***	10.19***	23.57***	28.11***	10.19***

Notes: This table reports regression results (using both OLS and entropy balancing models) on the association between FF and earnings management and is based on Equation (2(2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) ). In columns 1 and 4, we compare all FF versus a control group of non-FF. In columns 2 and 5, all eponymous FF are compared to the non-FF control group. In columns 3 and 6, all FF except eponymous FF (FAM_OTHER) are compared to the non-FF control group. All variables are defined in the Appendix. Standard errors are clustered at the firm level and are corrected for heteroscedasticity. t-statistics are reported in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.

In columns 1–3 (OLS model), the generic FAMVAR term is successively replaced by FAM, FAM_EPONYM and FAM_OTHER. The sample size on which equation (2(2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) ) is estimated varies across columns: while the control group remains the same (non-FF), the FF group of interest varies across columns. In column 2, only eponymous FF (108 observations) are compared to non-FF (953 observations), providing a total sample of 1,061 observations. In column 3, non-eponymous FF (483 observations) are compared to non-FF (953 observations), for a total of 1,436 observations.

In column 1, as a preliminary finding and in accordance with previous studies (Ali et al., 2007; Cascino et al., 2010; Jiraporn & DaDalt, 2009; Martin et al., 2016; Prencipe et al., 2011; Wang, 2006), the negative and significant coefficient on FAM (−0.02, p < 0.05) shows that FF are associated with lower levels of absolute abnormal accruals (ABSABNACC) compared to non-FF. In column 2, the negative and significant coefficient on FAM_EPONYM (−0.04, p < 0.01) suggests that eponymous FF exhibit less earnings management than non-FF, which is consistent with H1. In contrast, in column 3, the results show that other FF do not manage earnings differently compared to their non-FF counterparts, as documented by the non-significant coefficient on FAM_OTHER. Although most control variables are not significant, each model as a whole is globally significant as shown by the F-statistic and an adjusted-R² of around 30%.¹⁸

Next, we replicate the tests of columns 1–3 using entropy balancing in columns 4–6. The results are similar, which attests to the robustness of the findings. Overall, the results of Table 4 provide evidence that not all FF exhibit lower levels of earnings management. Indeed, only eponymous FF manage earnings less than non-FF. For other FF, this finding does not hold. Taken together, these results are consistent with the argument that the family’s incentive to maintain its reputation, which is stronger when the firm bears the family name, is a significant deterrent to earnings management in listed eponymous FF.

4.3. Association between eponymous FF and earnings management – specific contexts

4.3.1 IFRS turn away

Table 5 reports the distribution of voluntary switches from IFRS to Swiss GAAP throughout the sample period, for all firms and for FF only. In total, we document 44 switches. FF are over-represented: although they represent only 38.3% of the total sample, they represent 50% of all switchers (22 out of 44). From 2008 to 2018, for the full sample, between one (in 2008) and seven firms (in 2010) per year decided to turn away from IFRS. Most FF that decided to make the change did it during the first half of the sample period (before 2014).

Table 5. IFRS turn aways by year.

Year	Number of switches from IFRS to Swiss GAAP		Percentage of FF among switching firms
	All firms	FF
2008	1	0	0%
2009	6	5	83%
2010	7	6	86%
2011	4	2	50%
2012	3	3	100%
2013	6	2	33%
2014	4	0	0%
2015	3	0	0%
2016	2	1	50%
2017	5	1	20%
2018	3	2	67%
Total	44	22	50%

Notes: This table reports the distribution of voluntary switches from IFRS to Swiss GAAP by year over the 2008–2018 period.

Table 6 reports the analysis of the impact of the voluntary IFRS turn away on earnings management for FF and non-FF. Using a Heckman (1979) two-stage model, we first analyze the determinants of the switch in column 1. Then, in columns 2–4, we examine the impact of the switch on earnings management including the inverse Mills ratio (IMR) derived from the first-stage regression as an independent variable. Note that to fulfill the exclusion restriction, MTB is included in the first stage (with a coefficient that is strongly significant), while it is excluded from the second one (where its coefficient would not be significantly different from 0) (Certo et al., 2016). In column 1, the results indicate that FF have a higher propensity to switch back to Swiss GAAP than their non-FF counterparts, as documented by the positive and significant coefficient on FAM (0.38, p < 0.01). Firms with higher growth rates (MTB), larger firms (SIZE), and firms incurring an accounting loss (DLOSS) have a lower propensity to switch. In contrast with Fiechter et al. (2018), who find that the proportion of closely held shares (OWN) is positively associated with switching back to Swiss GAAP, the coefficient on OWN is not significant in this model. Results from an additional test (untabulated) indicate that when the FAM indicator is excluded, the coefficient on OWN is indeed positive and significant. Taken together, these results indicate that FF are responsible for the positive association between insider ownership and the IFRS turn away documented by Fiechter et al. (2018).¹⁹

Table 6. Specific settings influencing earnings management in FF – IFRS turn away.

	Absolute abnormal accruals (ABSABNACC) {−1; + 2}
	First stage	Second stage
	The determinants of the switch	Switching FF or switching non-FF versus non-switching firms	Switching eponymous FF or switching non-FF versus non-switching firms	Switching other FF or switching non-FF versus non-switching firms
	1	2	3	4
POST1		−0.02	−0.03*	−0.02
		(−1.25)	(−1.71)	(−1.26)
POST2		0.02	0.01	0.02
		(0.56)	(0.32)	(0.56)
POST3		0.02	0.01	0.03
		(1.27)	(0.72)	(1.27)
POST1*FAM		0.03
		(1.51)
POST2*FAM		−0.02
		(−0.65)
POST3*FAM		−0.01
		(−0.47)
POST1*FAM_EPONYM			0.11**
			(2.13)
POST2*FAM_EPONYM			−0.04
			(−0.76)
POST3*FAM_EPONYM			0.05
			(1.44)
POST1*FAM_OTHER				0.02
				(0.93)
POST2*FAM_OTHER				−0.02
				(−0.58)
POST3*FAM_OTHER				−0.03
				(−0.87)
FAM	0.38***
	(4.80)
MTB	−0.05***
	(−3.74)
AC_INDEP	0.03	−0.00	−0.00	−0.01
	(0.38)	(−0.54)	(−0.58)	(−0.93)
DLOSS	−0.29***	0.01	−0.00	0.01
	(−2.85)	(0.52)	(−0.06)	(0.45)
OWN	−0.04	−0.02*	−0.04*	−0.03*
	(−0.22)	(−1.72)	(−1.74)	(−1.86)
LEVERAGE	0.23	0.06	0.08	0.05
	(1.00)	(1.47)	(1.52)	(1.34)
SIZE	−0.29***	−0.04	−0.07***	−0.04
	(−11.65)	(−1.34)	(−2.96)	(−1.29)
BIG4	−0.16	−0.00	−0.00	−0.00
	(−1.44)	(−0.30)	(−0.00)	(−0.29)
SALES_GROWTH	0.00	0.00*	0.00*	0.00*
	(0.89)	(1.86)	(1.75)	(1.67)
IMR		0.02	0.05	0.02
		(0.42)	(0.98)	(0.43)
Constant	4.03***	0.60	1.00***	0.58
	(8.28)	(1.59)	(3.36)	(1.52)
Fixed effects	Year	Firm and year	Firm and year	Firm and year
First-step regressors	No	Incl.	Incl.	Incl.
Observations	1,720	1,123	791	1,047
Pseudo/Adj. R-squared	0.24	0.16	0.25	0.16
LR Chi-squared/F-statistic	440.2***	2.2***	31.4***	32.6***

Notes: This table reports the results of a Heckman (1979) two-stage model (probit in column 1 and OLS in columns 2–4) estimating the association between FF and earnings management around the IFRS turn away and is based on Equation (3(3) $+{\text{β}}_7\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{O}\mathrm{S}\mathrm{T}{3}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{m}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(3) ). IMR is the inverse Mills ratio calculated in the first stage. All variables are defined in the Appendix. Standard errors are clustered at the firm level and are corrected for heteroscedasticity. z-statistics in column 1 and t-statistics in columns 2–4 are reported in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.

In the second stage of the model, the coefficients on POST1, POST2 and POST3 measure the change in earnings management for FF and non-FF switchers, while the coefficients on the interaction terms are difference-in-differences coefficients capturing the incremental effect of FAM, FAM_EPONYM or FAM_OTHER.²⁰ Overall, the years following a switch are not associated with different levels of earnings management as the coefficients on POST1, POST2 and POST3 (and their interactions with FAM) are not significant in column 2. Estimations in column 3 indicate that eponymous FF manage earnings more only in the year of the switch, as the positive and significant coefficient (0.11, p < 0.05) on POST1*FAM_EPONYM documents. This suggests that some eponymous FF use the temporary break in the comparability of accounting numbers caused by the switch to behave opportunistically. This highlights a context in which identity FF deviate from their generally more ‘virtuous’ behavior (i.e. Table 4). As expected, this deviation appears to be time-limited as the coefficients on POST2*FAM_EPONYM and POST3*FAM_EPONYM are not significant. In contrast, the switch event does not modify the inclination of other FF to manage earnings as shown in column 4. Overall, consistent with H2, identification of the family with the firm appears to increase the propensity to manage earnings when turning away from IFRS. However, this effect does not persist over time.²¹

These results are consistent with the argument that eponymous FF do not play the risky game of earnings management unless they face an exceptional situation in which the detection of earnings management is more difficult, such as the decision to switch from international accounting standards to local ones.

4.3.2 Family involvement in governance structures

In Table 7, using both OLS and entropy balancing, we investigate whether the propensity of eponymous FF to manage earnings is affected by family involvement at the CEO and/or board chairperson levels. In columns 1 and 4, the coefficients on FAM_C are negative and significant (−0.04, p < 0.01), consistent with prior results (e.g. Table 4). However, the positive and significant coefficients on the interaction terms (0.12 and 0.14, p < 0.01) suggest that the involvement of a family member at the CEO or board chairperson level is associated with higher earnings management.²² We observe similar results in columns 2 and 5 (eponymous FF) as well as 3 and 6 (non-eponymous FF).

Table 7. Specific settings influencing earnings management in FF – Family involvement.

		Absolute abnormal accruals (ABSABNACC)
		OLS model			Entropy balancing
		FF	Eponymous FF	Non-eponymous FF	FF	Eponymous FF	Non-eponymous FF
		versus non-FF			versus non-FF
		1	2	3	4	5	6
FAM_C	${\text{β}}_1$	−0.04***			−0.04***
		(−3.25)			(−4.29)
FAM_EPONYM_C	${\gamma }_1$		−0.07***			−0.07***
			(−3.14)			(−3.11)
FAM_OTHER_C	${\delta }_1$			−0.04***			−0.03***
				(−2.80)			(−3.52)
FAM_C*FAM_INVOLVEMENT	${\text{β}}_2$	0.12***			0.14***
		(3.16)			(5.52)
FAM_EPONYM_C*FAM_INVOLVEMENT	${\gamma }_2$		0.36***			0.28***
			(4.65)			(3.55)
FAM_OTHER_C*FAM_INVOLVEMENT	${\delta }_2$			0.09**			0.10***
				(2.27)			(3.77)
FAM_INVOLVEMENT	${\theta }_1$	−0.06***	−0.18***	−0.05**	−0.07***	−0.13***	−0.06***
		(−3.23)	(−4.14)	(−2.49)	(−5.24)	(−3.13)	(−3.93)
Control variables		Incl.	Incl.	Incl.	Incl.	Incl.	Incl.
Year and industry FE		Incl.	Incl.	Incl.	Incl.	Incl.	Incl.
First-step regressors		Incl.	Incl.	Incl.	Incl.	Incl.	Incl.
Observations		1,544	1,061	1,436	1,544	1,061	1,436
Adj. R-squared		0.3	0.34	0.31	0.24	0.3	0.25
F-statistic		23.16***	30.47***	22.86***	10.10***	9.93***	9.44***
Test on linear combinations of coefficients:
${\text{β}}_1+{\text{β}}_2+{\theta }_1=0$
		0.013			0.027**
		(0.68)			(2.25)
${\gamma }_1+{\gamma }_2+{\theta }_1=0$
			0.105***			0.077***
			(2.94)			(2.16)
${\delta }_1+{\delta }_2+{\theta }_1=0$
				−0.003			0.013
				(−0.13)			(0.96)

Notes: This table reports OLS and entropy balancing estimation results on the association between FF and earnings management and is based on Equation (4(4) $+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(4) ). In columns 1 and 4, we compare all FF versus a control group of non-FF. In columns 2 and 5, all eponymous FF are compared to the non-FF control group. In columns 3 and 6, all non-eponymous FF are compared to the non-FF control group. Control variables include AC_INDEP, IFRS, DLOSS, MTB, OWN, LEVERAGE, SIZE, BIG4, and AGE. All variables are defined in the Appendix. Standard errors are clustered at the firm level and are corrected for heteroscedasticity. t-statistics are reported in parentheses. *** and ** indicate statistical significance at the 1% and 5% levels, respectively.

To test for the total effect of family involvement on absolute abnormal accruals, we estimate linear combinations of coefficients (FAMVAR + FAM_INVOLVEMENT + FAMVAR*FAM_INVOLVEMENT = 0). At the bottom of Table 7, in columns 1–3 (OLS estimations), the results indicate that family involvement in FF is positively associated with earnings management for eponymous FF only. When using entropy balancing, the sum of coefficients is different from 0 for FF in general (column 4) and for eponymous FF specifically (column 5), but not for other FF (column 6). In other words, the result on FF in general is driven by eponymous FF. This striking result suggests that, in our sample of listed firms, family incentives and individual (family member) incentives are sufficiently different to reverse the baseline result that eponymous FF exhibit lower levels of earnings management in order to uphold the family’s reputation. This result is consistent with the argument that involved family members suffer from a legitimacy deficit, which may lead them to use tools such as earnings management in order to report higher financial performance to minority shareholders. At any rate, this suggests that, in specific contexts, eponymous FF tend to manage earnings more than their non-FF counterparts.

4.3.3 Total accruals and signed abnormal accruals

Our main tests use absolute abnormal accruals as the dependent variable because we have no basis on which to expect that FF are more or less likely to systematically manage earnings in a particular direction in the absence of a specific motivating event (e.g. Beuselinck et al., 2015). In addition, given that many accruals reverse in a subsequent period (e.g. Baber et al., 2011; DeFond & Park, 2001), a regression of signed accruals on a constant firm characteristic such as FAM_EPONYM would not succeed in picking up earnings management related to income smoothing or other period-specific reasons. Besides, given the high level of book-tax conformity in Switzerland (Blaylock et al., 2015), it is not straightforward to interpret positive (negative) abnormal accruals as ‘aggressive’ (‘conservative’).

Nevertheless, in Table 8, we replicate the OLS tests of Table 4 using total accruals (TOTACC) and signed abnormal accruals (ABNACC) as dependent variables. This tests whether the accruals of FF and eponymous FF are systematically biased relative to other firms. For example, one could argue that FF may be more likely to manage earnings downwards for tax purposes while other firms may be more likely to manage earnings upwards to meet market expectations. The coefficients on FAM, FAM_EPONYM and FAM_OTHER are never significant. Hence, even though FF in general and eponymous FF in particular exhibit lower levels of earnings management when the absolute value of abnormal accruals is used as a dependent variable (Table 4), these firms are not systematically more or less likely to manage earnings upwards or downwards (Table 8).

Table 8. Influence of eponymous FF on total accruals and signed abnormal accruals.

	OLS model
	Total accruals (TOTACC)			Signed abnormal accruals (ABNACC)
	FF	Eponymous	Non-eponymous	FF	Eponymous	Non-eponymous
		FF	FF		FF	FF
	versus non-FF			versus non-FF
	1	2	3	4	5	6
FAM	−0.00			−0.00
	(−0.38)			(−0.63)
FAM_EPONYM		0.01			0.00
		(0.68)			(0.32)
FAM_OTHER			−0.01			−0.00
			(−1.35)			(−0.41)
FAM_INVOLVEMENT	0.01*	0.00	0.01	0.00	0.00	0.00
	(1.84)	(0.49)	(1.46)	(0.53)	(0.20)	(0.31)
AC_INDEP	0.00	−0.00	−0.00	−0.02**	−0.03**	−0.02*
	(0.19)	(−0.02)	(−0.59)	(−2.03)	(−2.10)	(−1.82)
IFRS	0.00	−0.00	0.00	0.01	0.01	0.00
	(0.33)	(−0.04)	(0.95)	(0.93)	(1.25)	(0.58)
DLOSS	−0.07***	−0.07***	−0.06***	−0.05***	−0.06***	−0.05***
	(−8.68)	(−7.50)	(−7.87)	(−5.36)	(−4.69)	(−5.27)
MTB	−0.00	−0.00	−0.00	0.00	0.00	0.00
	(−0.54)	(−0.33)	(−0.70)	(0.24)	(0.21)	(0.02)
OWN	−0.00	−0.00	0.01	−0.00	−0.00	−0.01
	(−0.12)	(−0.12)	(1.19)	(−0.01)	(−0.05)	(−0.88)
LEVERAGE	0.00	−0.00	0.00	0.02	0.00	0.01
	(0.34)	(−0.19)	(0.07)	(0.92)	(0.07)	(0.53)
SIZE	−0.00***	−0.00***	−0.00**	−0.02***	−0.02***	−0.02***
	(−3.39)	(−3.04)	(−2.29)	(−4.98)	(−4.80)	(−5.90)
BIG4	0.01	0.00	0.01	−0.00	0.00	−0.00
	(1.45)	(0.36)	(1.47)	(−0.08)	(0.08)	(−0.19)
AGE	−0.01	−0.01	0.01	−0.01	−0.02	−0.02*
	(−1.03)	(−1.13)	(0.89)	(−1.31)	(−1.50)	(−1.95)
Constant	0.08***	0.09***	0.00	0.13**	0.18***	0.19***
	(2.98)	(2.63)	(0.04)	(2.40)	(2.68)	(4.48)
Year and industry FE	Incl.	Incl.	Incl.	Incl.	Incl.	Incl.
First-step regressors	Incl.	Incl.	Incl.	Incl.	Incl.	Incl.
Observations	1,544	1,061	1,436	1,544	1,061	1,436
R-squared	0.23	0.25	0.22	0.42	0.40	0.42
F-statistic	103.6***	97.6***	32.9***	21.3***	19.5***	20.4***

Notes: This table reports regression results (using OLS estimations) on the association between FF and earnings management and is based on Equation (2(2) $\mathrm{A}\mathrm{B}\mathrm{S}\mathrm{A}\mathrm{B}\mathrm{N}\mathrm{A}\mathrm{C}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{F}\mathrm{A}\mathrm{M}\mathrm{V}\mathrm{A}{\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}\mathrm{L}{\mathrm{S}}_{\mathrm{i}\mathrm{t}}+\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{F}\mathrm{E}+{ϵ}_{\mathrm{i}\mathrm{t}}$(2) ). In columns 1 and 4, we compare all FF versus a control group of non-FF. In columns 2 and 5, all eponymous FF are compared to the non-FF control group. In column 3 and 6, all FF except eponymous FF (FAM_OTHER) are compared to the non-FF control group. All variables are defined in the Appendix. Standard errors are clustered at the firm level and are corrected for heteroscedasticity. t-statistics are reported in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.

5. Conclusion

In this study, we examine whether listed FF whose family owner is closely identified with the firm (as proxied by the inclusion of the family name in the firm’s name) exhibit different earnings management patterns versus non-FF. Our study is motivated by calls for investigation of nonpecuniary decision-making drivers in FF (Stockmans et al., 2010) and for a detailed understanding of family-specific features affecting accounting phenomena (Salvato & Moores, 2010). Although recent studies have examined the effect of family identification on earnings management for private firms, we argue that the greater complexity of agency structures and stakeholder interactions experienced by publicly-listed firms warrant a separate investigation.

First, consistent with the argument that the incentive to avoid questionable business practices, in order to preserve the family’s reputation, is a key characteristic of eponymous FF, we find that, on average, listed eponymous FF exhibit less earnings management than non-FF. In contrast, we find no difference in the level of earnings management between other listed FF and non-FF in our Swiss sample.

Second, we exploit a feature unique to Switzerland, whereby listed firms are allowed to voluntarily switch from IFRS to Swiss GAAP and remain listed. Using a difference-in-differences approach, we find that listed eponymous FF exhibit higher levels of earnings management than other switching firms immediately after the switch, although this effect does not persist over time. This result is consistent with the argument that eponymous FF avoid earnings management in settings in which the likelihood of detection is high, but that they may be inclined to manage earnings during the switching event as the change in accounting standards triggers a break in the time series comparability of accounting numbers.

Finally, we show that listed eponymous FF exhibit higher earnings management when the family is directly involved in the board of directors or the managing board. This finding is consistent with the argument that family (shareholding principal) and individual (CEO/board chairperson agent) incentives can be significantly misaligned. In particular, the legitimacy deficit caused by the historical underperformance of family versus non-family CEOs in FF (e.g. Pérez-González, 2006) may lead involved family members to manage earnings as a means to deliver superior financial performance to minority shareholders. Taken together, our findings provide a more nuanced understanding of the effects of the family identification dimension on earnings management incentives in listed FF.

Our findings offer policy implications for initiating and/or improving a ‘family business-friendly environment’, a field in which institutional and academic actors have advocated for more research (e.g. Mandle, 2008). Indeed, according to Villalonga and Amit (2020, p. 242), ‘understanding the unique dimensions of family-owned firms […] serves as vital input to formulating public policy in such areas as wages, taxation, and more’. In particular, we show that listed eponymous FF manage earnings to a lesser extent than non-FF in general, but that this is not the case when the CEO is a family member. These nuanced results suggest that governance guidance and codes that explicitly address family businesses could, for example, refine their guidelines to recognize the heterogeneity of FF. These results are also of interest to regulators, whose identification of the conditions under which earnings quality may be impaired could consider the complex interplay between controlling families and minority shareholders when family members are directly involved at the upper management echelons of publicly-listed firms. This could result, for example, in the modification or identification of additional control/audit measures. Our results therefore generally argue in favor of targeted support and control mechanisms for each type of FF by both practitioners and policy-makers.

As is common in FF research, our study has various limitations. The main limitation stems from our use of eponyms as a binary indicator of family identification with the firm in our large sample of listed firms. While eponyms have been used for this purpose in the past (e.g. Belenzon et al., 2017; Minichilli et al., 2022), family identification with the firm is undoubtedly a more complex construct and it is difficult to ascertain the effect of this measurement error on our results. In addition, family and business identities are dynamic constructs that evolve as firms grow and as multiple generations enter the FF (Sundaramurthy & Kreiner, 2008). As a result, relevant variables likely remain omitted. For example, we do not control for variables that may affect earnings management such as whether the family founded or acquired the firm (Pazzaglia et al., 2013) or whether the founder is still involved with the firm (Martin et al., 2016). Several questions remain unexplored, including how environmental and institutional factors affect a FF’s ability and incentives to preserve its reputation, and how important the avoidance of earnings management truly is for the family’s reputation. Future research could extend the analysis to countries with different ownership concentrations, investor protection levels and other institutions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 After the 2013 European horse meat scandal involving the family business founded in 1970, members of the Spanghero family were forced to publicly announce that they were no longer running the eponymous company (having sold it in 2009) to restore their honor and reputation. In the end, they asked – in public – the rhetorical question: ‘the children, our grandchildren, what will people say tomorrow when they pass in the street? People will say: “he/she is a Spanghero”’ (France Info, 2019).

2 By contrast, in the samples used by Minichilli et al. (2022) and Sundkvist and Stenheim (2022), average family ownership is above 90% and the median firm has no non-family shareholders at all, making this issue a much lesser concern for most private firms.

3 Own computation equal to the number of large FF in 2022 per country (Center for Family Business & EY, 2022) divided by the number of inhabitants in 2022 in that country (Statista, 2023).

4 Several studies document that the benefits derived from IFRS are conditional on, among other aspects, firm-specific characteristics such as size, ownership structure, and reporting incentives (Burgstahler et al., 2006; Christensen et al., 2015; Daske et al., 2013; Fiechter et al., 2018; Hail et al., 2010).

5 In terms of capital market consequences, Fiechter et al. (2018) find that switchers do not experience negative abnormal returns around the switching announcement or a decline in liquidity following the switch back to Swiss GAAP. This contrasts with Leuz et al. (2008) who analyze the consequences of the decision to ‘go dark’ (cease SEC reporting) and find that such a decision generates large negative abnormal returns. However, the two settings are not directly comparable, as Swiss firms that turn away from IFRS still need to comply with SIX Exchange disclosure requirements following the change.

6 Analytically, if a firm’s (ex ante) expected costs are equal to the probability of detection multiplied by the actual (‘if detected’) costs then, for a given decline in probability, the largest absolute decline in expected costs is found in firms with the highest actual costs.

7 The arguments developed in this section are distinct from Gomez-Mejia et al.’s (2014) proposition that families are more likely to have the firm engage in earnings management when their decisions are motivated by the desire to ensure family control of the firm. In that framework, the main driver of earnings management is not family control itself but rather perceived threats to family control. For example, firms could manage earnings to avoid a debt covenant violation if such a violation would give the debtholder the power to veto all important operating decisions. Any empirical assessment of this proposition would require identification of specific instances in which family control is threatened, a difficult feat to achieve in large samples.

8 All variables are defined in the Appendix. On average, in our sample, families own 58% of voting rights in eponymous FF and 49% of voting rights in other FF.

9 Given the limited number of observations in our sample, we estimate abnormal accruals using a pooled sample with industry and year fixed effects, following Francis et al. (2013). However, in untabulated analyses, in line with Ecker et al. (2013), we use size- (i.e., quartiles of lagged total assets) and year-based estimation samples to estimate abnormal accruals. We then replicate our tests using the absolute value as a dependent variable and the results are qualitatively similar. We also replicate Table 4 using alternative earnings management models (Dechow & Dichev, 2002; Jones, 1991; Kothari et al., 2005) and the (untabulated) results hold.

10 Closely-held shares are shares held by insiders (e.g., cross holdings, corporations, holding company, government, employees, pension plans, blockholders). As a result, this variable includes shares held by the owning family but is not restricted to it. Kim et al. (2017) show that ownership concentration is negatively associated with earnings management. Because Eq. (2) includes OWN as a control variable, the coefficient on FAMVAR can be interpreted as its association with earnings management that is incremental to the effect of ownership concentration.

11 In all our tests, standard errors are adjusted for heteroscedasticity and firm-level clustering, consistent with Petersen (2009). Continuous variables are winsorized at 1% and 99%.

12 Because we measure accruals using data from cash flow statements, the effect of the switch on accruals is not spurious even though accounting standards obviously affect the accrual generation process. As they did when they first adopted IFRS, switching firms restated their balance sheets on the switch date through a one-time adjustment to shareholders’ equity that did not affect earnings or cash flows, and therefore did not affect accruals.

13 We thank an anonymous reviewer for this recommendation.

14 This value is in line with what Zellweger et al. (2007) and Isakov and Weisskopf (2014) find using the same definition as ours in the Swiss context.

15 Average absolute abnormal accruals vary widely across studies and models. Our sample average is in the mid-range of recent studies that use Swiss or European data (e.g., Achleitner et al., 2014; Beuselinck et al., 2019; Docimo et al., 2021; Windisch, 2021).

16 Both a t-test of difference in means and a Mann-Whitney test of difference in medians show that the difference is significant at the 1% threshold.

17 VIF values are computed on the regression estimated in column 1 of Table 4.

18 The lack of significance of control variables in earnings management models is in line with some previous studies (e.g., Achleitner et al., 2014; Qi et al., 2021).

19 In another untabulated test, we include FAM_EPONYM as an additional independent variable in the first stage model. In that specification, the coefficient on FAM is positive and significant, the coefficient on FAM_EPONYM is negative and significant, and the sum of the coefficients on FAM and FAM_EPONYM is positive and significant. This indicates that among all groups, non-eponymous FF are the most likely to switch, followed by eponymous FF, and finally non-FF.

20 In Table 6, the control group is composed of both non-switching FF and non-switching non-FF.

21 Note that the coefficients on the inverse Mills ratio (IMR) in columns 2–4 of Table 6 are not statistically different from zero. Untabulated tests document that the results excluding IMR are similar to those reported.

22 In an additional test, we separately examine family involvement at the CEO and at the Board chairperson levels. At both levels, the (untabulated) results are similar to those reported in Table 7.

Appendix

Table A1. Definition and sources of variables.

Variable	Definition	Source
ABSABNACC	The absolute value of abnormal (discretionary) accruals following Francis et al. (2013).	Datastream
TOTACC	Total accruals calculated following Francis et al. (2013).	Datastream
ABNACC	Signed value of abnormal accruals following Francis et al. (2013).	Datastream
FAM	Dummy variable equal to 1 if the family owns 20% or more of voting rights, and 0 otherwise.	Annual reports
FAM_EPONYM	Dummy variable equal to 1 if the FF name includes the family name, and 0 otherwise.	Annual reports
FAM_OTHER	Dummy variable equal to 1 if the FF name does not include the family name, and 0 otherwise.	Annual reports
FAM_INVOLVEMENT	Dummy variable equal to 1 if the chairperson of the board of directors is a member of the owning family and/or if the CEO is a member of the owning family, and 0 otherwise.	Annual reports
FAM_C	Percentage of voting rights owned by the family for FF.	Annual reports
FAM_EPONYM_C	Percentage of voting rights owned by the family for eponymous FF.	Annual reports
FAM_OTHER_C	Percentage of voting rights owned by the family for other FF.	Annual reports
AC_INDEP	Dummy variable equal to 1 if the audit committee is only composed of independent members, and 0 otherwise.	Annual reports
SIZE	Natural logarithm of total assets.	Datastream
LEVERAGE	Debt-to-total assets ratio.	Datastream
MTB	Market-to-book ratio.	Datastream
DLOSS	Dummy variable taking the value of 1 if the firm is incurring a loss, and 0 otherwise.	Datastream
IFRS	Dummy variable taking the value of 1 if the firm applies IFRS, and 0 otherwise.	Annual reports
OWN	Percentage of closely held shares.	Datastream
BIG4	Dummy variable equal to 1 if the firm mandates a Big 4 firm as external auditor, and 0 otherwise.	Annual reports
AGE	Age of the firm calculated as the current year minus the incorporation year, ranked into quartiles.	Annual reports
SALES_GROWTH	Growth rate of total revenue.	Datastream
POST1	Dummy variable equal to 1 if the firm-year observation is the year of a switch from IFRS to Swiss GAAP for a given firm, and 0 otherwise.	Annual reports / corporate websites / Swiss press
POST2	Dummy variable equal to 1 if the firm-year observation is one year after a switch from IFRS to Swiss GAAP for a given firm, and 0 otherwise.	Annual reports / corporate websites / Swiss press
POST3	Dummy variable equal to 1 if the firm-year observation is two years after a switch from IFRS to Swiss GAAP for a given firm, and 0 otherwise.	Annual reports / corporate websites / Swiss press