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

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.