CEO turnover in family firms: the corporate transparency perspective Open Access

This study examines how corporate transparency affects the relationship between family firms and Chief Executive Officer (CEO) turnover, focusing on the Type II agency problem, characterized by conflicts between controlling and minority shareholders due to the rent-extracting activities of dominant shareholders (Faccio, Lang, & Young, 2001; Anderson & Reeb, 2004; Villalonga & Amit, 2006; Anderson, Duru, & Reeb, 2009; Ding, Qu, & Zhuang, 2011; Santos, Moreira, & Vieira, 2013) ^[1]. These dominant shareholders may entrench family members or affiliated managers in key leadership positions, such as the CEO, to maintain family control and to advance their private interests (Morck & Yeung, 2003; Prencipe, Markarian, & Pozza, 2008; Young, Peng, Ahlstrom, Bruton, & Jiang, 2008; Gomez-Mejia, Cruz, Berrone, & De Castro, 2011; De Cesari, Gonenc, & Ozkan, 2016). The desire to maintain family dominance results in reluctance to dismiss CEOs who are either family members or closely connected to the family, compromising the board’s ability to effectively monitor and control the CEO (Chen, Cheng, & Dai, 2013; Cheng, 2014). Moreover, this entrenchment can weaken overall firm performance due to poor management oversight by boards affiliated with the controlling family (Anderson & Reeb, 2004; Gomez-Mejia et al., 2011). In response to rent-extracting behaviors by dominant shareholders, minority shareholders may impose significant costs on these shareholders through mechanisms such as discounting the firm’s stock price, thereby penalizing the firm for poor governance practices (Claessens, Djankov, Fan, & Lang, 2002; Chen, Chen, Cheng, & Shevlin, 2010).

Corporate transparency empowers minority shareholders to monitor dominant shareholders and detect rent-extracting activities, enhancing governance effectiveness and reducing the risk of expropriation (Ali, Chen, & Radhakrishnan, 2007; Cheng, 2014). Ali et al. (2007) argue that the severity of the Type II agency problem decreases with increased corporate transparency, which limits the ability of dominant shareholders to pursue private benefits at the expense of minority shareholders. Wang (2006) argues that the opportunity for family entrenchment often stems from information asymmetry between minority and dominant shareholders. However, the specific influence of corporate transparency on the Type II agency problem and its impact on CEO turnover remain underexplored. Although prior research (e.g. Chen, Cheng, & Dai, 2013) examines the impact of family firm ownership on CEO turnover, the role of corporate transparency in this relationship remains unexplored. Our study fills this gap by investigating whether corporate transparency moderates the impact of family control on CEO turnover.

Using 15,726 firm-year observations of Taiwan’s publicly listed firms from 2002 to 2019, we examine the association between family firms and Chief Executive Officer (CEO) turnover and the moderating role of corporate transparency in this association. Following prior studies (e.g. Chen, Dasgupta, & Yu, 2014; Lee & Bose, 2021), we measure corporate transparency using a composite index of analyst forecast error, analyst forecast dispersion, analyst following, and bid–ask spread. Our findings indicate that family control typically reduces CEO turnover, but greater corporate transparency weakens this effect, making family firms less inclined to retain CEOs, resulting in higher turnover rates. Thus, enhanced corporate transparency not only facilitates better governance but also actively influences leadership stability within family firms. To mitigate potential endogeneity concerns stemming from both observable and unobservable selection bias, we used the propensity score matching (PSM) approach and Heckman’s (1979) two-stage analysis. Additionally, we employ the simultaneous equation technique to address endogeneity issues related to corporate transparency. The findings contribute to our understanding of how increased transparency can mitigate conflicts between controlling and minority shareholders related to CEO turnover, providing insights for family firm owners, shareholders, and policymakers. To further investigate the mechanism underlying our main H1, we conducted a mediation test using tax avoidance as a proxy for rent-extracting activities (Desai & Dharmapala, 2006; Lee & Bose, 2021). By including this mediation test, we provide further evidence supporting the role of corporate transparency in reducing rent-extracting opportunities, leading to higher CEO turnover in family firms.

The current study contributes to the literature in several ways. Our study is among the first to examine the moderating role of corporate transparency in the negative impact of family control on CEO turnover. While prior research (e.g. Chen, Cheng, & Dai, 2013) finds that family firms generally have lower CEO turnover, the authors do not explore how corporate transparency might influence this relationship. Resolving this oversight is crucial, as the effects of increased transparency, including potential reductions in the Type II agency problem and enhancements in firm value (Ali et al., 2007; Anderson et al., 2009), may not uniformly affect governance practices of all firms, such as CEO turnover. Ali et al. (2007) argue that greater corporate transparency can help to mitigate the Type II agency problem and family entrenchment. However, the effect on the relationship between family-owned firms and CEO turnover remains unexamined. Therefore, our study investigates the effect of corporate transparency on the family firm–CEO turnover association.

Our findings contribute to the family firm literature and have theoretical implications for governance effectiveness in family firms. Our study contributes to agency theory as it pertains to the Type II agency problem from a corporate transparency perspective. This is achieved by highlighting the behavior of family firms regarding CEO turnover, with our findings adding to the still scarce literature on how the dominant shareholder–minority shareholder conflict varies at different levels of corporate transparency. The findings indicate that a higher level of corporate transparency can mitigate the Type II agency problem. Therefore, policymakers should enforce corporate transparency to reduce this agency problem. Policymakers should consider mandating higher levels of corporate transparency, such as more stringent disclosure requirements and regular independent audits, to mitigate the Type II agency problem in family firms. For family firm owners, embracing transparency can lead to better governance and potentially higher firm value, as it aligns more closely with minority shareholders’ interests and reduces conflicts. Minority shareholders can advocate for increased transparency through active participation in annual general meetings, thereby enhancing their ability to monitor management and influence key decisions like CEO turnover. Our findings also have practical implications, offering insights into how employment relationships with family owners vary depending on the level of corporate transparency.

Our study enriches the family business literature by addressing a gap identified by Van den Berghe and Carchon (2003), who advocate for more research into agency relationships within family firms. As interest grows in understanding the implications of family entrenchment for the Type II agency problem (Conyon & He, 2012; Kuo & Hung, 2012; Chen, Gray, & Nowland, 2013; Prencipe, Bar-Yosef, & Dekker, 2014; Hsu, Lin, & Tsao, 2018), our findings provide valuable insights, particularly concerning CEO turnover. Previous research links family firms with reduced CEO turnover (Chen, Cheng, & Dai, 2013); poorer earnings quality (Fan & Wong, 2002; Jaggi, Leung, & Gul, 2009; Ding et al., 2011; Sue, Chin, & Chan, 2013; Achleitner, Günther, Kaserer, & Siciliano, 2014); lower levels of voluntary disclosure (Ali et al., 2007; Chen, Chen, & Cheng, 2008); family-dominated boards (Lam & Lee, 2008; Chen, Gray, & Nowland, 2013); less incentive to remediate internal control weaknesses (Chen, Feng, & Li, 2020); choice of auditor (Ho & Kang, 2013; Hsu et al., 2018); and a higher demand for audit quality (Niskanen, Karjalainen, & Niskanen, 2010).

Our research makes a unique contribution by demonstrating the moderating effect of corporate transparency on the influence of family control over CEO turnover. By doing so, we offer a more detailed explanation of the Type II agency problem, particularly in contexts such as Taiwan, where family firms are prevalent and corporate transparency varies significantly. Taiwan’s institutional environment—characterized by high family control and low investor protection—provides a distinctive backdrop for our study. Approximately 76% of publicly listed firms in Taiwan are family controlled, a rate surpassing those in Standard and Poor’s (S&P) 500 and Western European corporations. Consistent with Lee and Bose (2021), our choice of Taiwan as a study setting is strategic, allowing us to explore the dynamics of corporate transparency and family control of firms in a region where these factors have a particularly significant impact.

Our findings not only advance theoretical discussions around family firms and CEO turnover but also have practical implications for policymakers and corporate governance in regions with similar institutional features.

The Type II agency problem typically emerges in family firms where dominant shareholders extract private benefits at the expense of minority shareholders (Jensen & Meckling, 1976). This concept is well documented in the literature (Shleifer & Vishny, 1986; Zingales, 1994; La Porta et al., 2000, 2002; Anderson & Reeb, 2004; Hsu et al., 2018). Family owners may exert their control to influence board composition and CEO appointments, often leading to boards that feature a higher representation of family members and fewer independent directors (Mustakallio, Autio, & Zahra, 2002; Chen et al., 2008; Gomez-Mejia et al., 2011; Srinidhi, He, & Firth, 2014; Chi, Hung, Cheng, & Lieu, 2015). Although Chen, Gray, and Nowland (2013) argue that the concentrated positions and long investment horizons of founding family members can foster a focus on firm value, leading to more effective monitoring and potentially higher CEO turnover, the reality often reflects the opposite. The significant influence and active involvement of founding family members can lead to the prioritization of their personal interests over those of the firm. Consequently, family members serving as CEOs may resist leaving their positions even after poor performance due to the associated benefits—ranging from financial compensation to social status (Anderson & Reeb, 2004; Chen, Cheng, & Dai, 2013). Furthermore, family firm owners often hesitate to dismiss CEOs to maintain family control and safeguard their private interests, which contributes to lower CEO turnover compared with non-family firms (Chen, Cheng, & Dai, 2013; Li, 2018). This entrenchment can adversely affect firm value (Villalonga & Amit, 2006; Anderson et al., 2009) and earnings quality (Fan & Wong, 2002; Jaggi et al., 2009; Ding et al., 2011; Sue et al., 2013; Achleitner et al., 2014), which are critical concerns for minority shareholders.

Family entrenchment can promote CEO duality (i.e. when the CEO is also the chairperson of the board of directors), which affects the board’s ability to monitor the CEO’s performance. Specifically, CEO duality may impair the board’s capacity to oversee management (i.e. the CEO), thus exacerbating the Type II agency problem (Fama & Jensen, 1983). Lam and Lee (2008) argue that CEO duality, associated with family entrenchment, leads to expropriation by dominant shareholders in family firms to the detriment of minority shareholders. They found that CEO duality is negatively associated with the performance of family firms.

To remediate rent-extracting activities undertaken by dominant shareholders that may not enhance firm value, minority shareholders may discount the stock price of their specific firm. This action can impose significant costs on both dominant shareholders and the firm (Claessens et al., 2002; Chen et al., 2010). In response, these firms and their dominant shareholders may increase the board’s monitoring capabilities and even dismiss the CEO: as chief executives, CEOs are held accountable for firm performance, while (independent) directors primarily serve as monitors. Consequently, to mitigate the intensified Type II agency problem, dismissal of the CEO may be the decision of the family firm owners.

Prior studies indicate that family firms typically exhibit lower levels of corporate transparency than non-family firms (Ali et al., 2007; Chen et al., 2008; Chau & Gray, 2010). Ali et al. (2007) note that family firms issue fewer managerial forecasts compared with non-family firms, suggesting that enhancing transparency in corporate governance could reduce family entrenchment. Chen et al. (2008) argue that family firms are less likely to engage in voluntary financial disclosures, as these could increase scrutiny from minority shareholders. Conversely, Anderson et al. (2009) found that firms with greater transparency have higher firm value. Chau and Gray (2010) report that, with higher levels of family ownership, the demand for voluntary disclosure increases, indicating that corporate transparency can alleviate the Type II agency problem. Despite the significance of transparency for shareholders, its impact on the link between family ownership and CEO turnover in these firms has not been thoroughly investigated. Our study highlights the importance of considering the joint effects of corporate transparency and family control to better explain and predict a firm’s CEO turnover, in relation to the Type II agency problem.

Furthermore, prior research suggests that increased corporate transparency may mitigate the Type II agency problem. The reason is that greater transparency allows minority shareholders to more easily detect and curb rent-extracting activities by dominant shareholders, with the resultant actions possibly leading to significant costs for the firm through stock price discounts (Ma, Ma, & Tian, 2015). We argue that increased corporate transparency reduces costs associated with minority shareholders’ monitoring efforts, enabling them to take corrective actions against dominant shareholders and even to dismiss the CEO. Consequently, enhanced transparency plays a crucial role in alleviating the Type II agency problem.

Conversely, a low level of corporate transparency makes it difficult for minority shareholders to detect and control rent-extracting activities by dominant shareholders, potentially allowing opportunism and rent-extracting activities (Ma et al., 2015). This view is supported by Kerr (2019) and Lee and Bose (2021), who suggest that, in less transparent firms, minority shareholders struggle to detect and control dominant shareholders’ rent-extracting activities. When corporate transparency is low, minority shareholders may not find it easy to discipline dominant shareholders. For example, private firms are unlisted and may have low corporate transparency, as well as being less subject to public scrutiny and capital market discipline (Chen et al., 2010). Minority shareholders may not be able to discount the stock of firms nor take corrective actions against dominant shareholders. In these environments, minority shareholders may fail to detect rent-extracting activities, thereby diminishing their influence over dominant shareholders who may wish to retain incumbent CEOs to sustain family control and entrenchment. This situation can lead to lower CEO turnover. Thus, the influence of family control on CEO turnover does not necessarily diminish with decreased corporate transparency. We hypothesize that corporate transparency moderates the impact of family control on CEO turnover, as stated in the following hypothesis:

Our sample consists of all firms listed on the Taiwan Stock Exchange (TSE) from 2002 to 2019. We selected 2002 as the starting year as family firm data first became available that year, and we continued until 2019, marking the end of our data collection period. Family firm and CEO turnover data were sourced from the Taiwan Economic Journal (TEJ) database, extensively used in previous studies (e.g. Tsai, Hung, Kuo, & Kuo, 2006; Kuo & Hung, 2012; Lin & Shen, 2015; Lien, Teng, & Li, 2016) to measure variables related to family firms and CEO turnover. In Taiwan, it is challenging to distinguish between forced CEO dismissals and voluntary resignations (Li, 2018) ^[2]. Financial data were obtained from Standard and Poor’s (S&P) Compustat Global database, while analysts’ earnings forecasts were sourced from the Institutional Brokers’ Estimate System (I/B/E/S) database. We merged firm-year observations from all three databases for the years 2002 through 2019. After excluding observations with incomplete financial or analysts’ earnings forecast data, our initial sample consisted of 1,785 unique firms, resulting in 15,726 firm-year observations covering the period 2002–2019.

Table 1, Panel A provides the sample distribution using industry classifications developed by Fama and French (1997). Our sample is dominated by firms operating in the electronic equipment industry (29.85%), followed by those operating in the computer industry (8.88%), while the shipbuilding and railroad equipment industry has the lowest proportion of firms (0.04%). Furthermore, the highest number of firm observations (10.07%) was in 2018, followed by 2017 (9.95%) and 2016 (9.43%), while 2002 had the lowest number of observations (0.48%). Overall, an upward trend is observed in the number of observations per year, although this trend does not extend to 2019 ^[3].

Following Chen et al. (2014), and Lee and Bose (2021), we developed a composite index of corporate transparency (TRANS) based on the factor score obtained from a principal component analysis (PCA) of four variables: analyst following (ANALYST), analyst forecast dispersion (AFDISP); analyst forecast error (AFERROR), and bid–ask spread (SPREAD). Specifically, ANALYST is the total number of analysts covering a firm (Chen et al., 2014). AFDISP is the 12-month average of the standard deviations of analyst earnings forecasts divided by the stock price at the beginning of the fiscal year for each year (Ali et al., 2007). AFERROR is the absolute value of the 12-month average of analyst forecast errors, calculated as the difference between actual and median forecast earnings divided by the stock price at the beginning of the fiscal year (Ali et al., 2007). SPREAD is the annual average of the daily closing bid–ask spread divided by the daily closing price. Higher ANALYST values indicate higher corporate transparency. Higher AFERROR, AFDISP and SPREAD values indicate lower corporate transparency (Kerr, 2019). Therefore, AFDISP, AFERROR and SPREAD are multiplied by minus one (−1), with higher values interpreted as indicating a higher level of corporate transparency (Chen et al., 2014). A higher value of TRANS indicates higher corporate transparency, whereas a lower value indicates lower corporate transparency. This index provides a comprehensive measure of corporate transparency by integrating analyst behavior with stock market information, thereby offering a detailed view of corporate transparency dynamics (Chen et al., 2014) ^[4].

The TEJ database defines family firms as follows: (1) family members hold positions on the board of directors or in top management roles such as General Manager or CEO; (2) family members hold more than 50% of the internal directorships on the board, with combine external and affiliated directorships accounting for less than 33%; (3) at least three family members serve either on the board of directors or in top management, with their combined directorships exceeding 33% of the board; and (4) the family’s controlling share percentage exceeds the critical control level, granting family members overwhelming voting rights and full control over key firm decisions. Following Lee and Bose (2021), we code family firms as an indicator variable that equals 1 if the TEJ database classifies a firm as a family firm, and 0 otherwise.

Following prior studies (Lee, Matsunaga, & Park, 2012; Chen, Cheng, & Dai, 2013; Fiordelisi & Ricci, 2014; Gao, Harford, & Li, 2017), we control for firm size (SIZE) as larger firms tend to have lower CEO turnover rates than smaller firms (Chen, Cheng, & Dai, 2013). We also control for financial leverage (LEV) as firms with higher leverage are subject to increased monitoring by creditors, which may influence CEO turnover (Fiordelisi & Ricci, 2014). Prior studies found that firms with poor performance have higher CEO turnover (Denis & Denis, 1995; Dikolli, Mayew, & Nanda, 2014). Therefore, we control for financial performance measured by return on assets (ROA), stock performance measured by annualized stock returns (RET), and operating cash flow performance (CFO). Recognizing the relationship between a firm’s growth opportunities and CEO turnover (Dikolli et al., 2014), we control for book-to-market ratio (BM) and sales growth (SGROWTH). Additionally, we control for cash flow volatility (STD_CFO) to control for risk (Lee et al., 2012). We also control for employee turnover (EMPTURN) as firms with higher employee turnover may have higher CEO turnover. Furthermore, older firms generally experience lower CEO turnover (Gao et al., 2017); thus, we control for firm age (FAGE). Given that CEO turnover is a critical corporate decision, prior studies argue that corporate governance effectiveness and CEO power have a significant impact on this decision (Chen, Cheng, & Dai, 2013). Therefore, we control for block ownership (BLOCK), board size (BSIZE) and board independence (BIND) as proxies for corporate governance effectiveness, as well as controlling for CEO duality (CEO_DUAL), CEO gender (CEO_GENDER) and CEO ownership (CEO_OWN) as proxies for CEO power.

Following prior studies (Kim, Li, & Zhang, 2011; Kim, Li, & Li, 2014; DeFond, Mingyi, Siqi, & Yinghua, 2015), we use the following model to test our H1:

where CEO_TURN_i,t is the CEO turnover indicator provided by the TEJ database, which is coded 1 if a firm experiences CEO turnover, and 0 otherwise. The variable FF denotes a family firm, as defined in Section 3.3. The control variables are firm size (SIZE); leverage (LEV); profitability (ROA); book-to-market ratio (BM); stock return (RET); sales growth (SGROWTH); operating cash flow performance (CFO); cash flow volatility (STD_CFO); firm age (FAGE); employee turnover (EMPTURN); block ownership (BLOCK); board size (BSIZE); board independence (BIND); CEO duality (CEO_DUAL); CEO gender (CEO_GENDER); and CEO ownership (CEO_OWN).

Furthermore, we create an interaction term between family firm (FF) and corporate transparency (TRANS) and include it, along with TRANS, in Equation (1) to test our H1. The model is expressed as follows:

where TRANS represents corporate transparency, a composite index derived from the factor score obtained through a principal component analysis of four variables: analyst following (ANALYST), analyst forecast dispersion (AFDISP), analyst forecast error (AFERROR), and bid–ask spread (SPREAD), as discussed previously. The positive coefficient of the interaction term FF×TRANS in Equation (2) is consistent with our H1 that corporate transparency moderates the impact of family control on CEO turnover. We control for industry and year fixed effects in all our research models. We apply robust standard errors clustered by firm and year to control for heteroscedasticity and serial correlation in our regression models.  Appendix defines all variables used in Equations (1) and (2).

Table 2, Panel A provides the descriptive statistics of the variables used in our study. As shown in the table, about 21.80% of firms in our sample experience CEO turnover (CEO_TURN) during the sample period. Additionally, about 62.70% of firms in our sample are family firms, which is higher than the 55% reported by Lee and Bose (2021). This difference can be attributed to the specific sample used by Lee and Bose (2021), covering 1998–2014. The average (median) value of corporate transparency (TRANS) is −0.225 (0.418). The mean (median) value of firm size (SIZE), measured as the natural logarithm of total market capitalization, is 8.095 (7.930). The average book-to-market ratio (BM) is 0.513, indicating that firms in our sample have high growth opportunities. Moreover, firms in our sample have leverage (LEV), profitability (ROA) and cash flow performance (CFO) of about 7.80, 3.40 and 5.10%, respectively, of total assets. The average value of the stock market return (RET) and sales growth (SGROWTH) is 12.00 and 5.50%, respectively, while the cash flow volatility (STD_CFO) is 0.070. The mean value of firm age (FAGE), measured as the natural logarithm of firm age, is 3.258, implying that the average age of firms in our sample is 26.15 years. The average employee turnover (EMPTURN) is 13.70%.

Furthermore, the average block ownership (BLOCK) for observations in our sample was 48.40%. Firms in our sample, on average, have 7.15 board members (BSIZE) ^[5], while approximately 22.30% of board members are independent (BIND). Regarding CEO characteristics, approximately 34% of observations in our sample have the same person as the chairperson and CEO (CEO_DUAL), while approximately 94.70% of CEOs are male (CEO_GENDER). The average CEO ownership (CEO_OWN) is 0.06%.

Table 2, Panel B provides the results of the mean and median tests of the variables used in Equation (1), differentiating between family firms and non-family firms. The results indicate that family firms (FF) have lower CEO turnover (CEO_TURN) compared with non-family firms. Additionally, family firms are smaller (SIZE), have higher leverage (LEV) and book-to-market ratio (BM), and are of longer standing in the market (FAGE). Family firms (FF), in contrast with non-family firms, have lower operating cash flow performance (CFO) and cash flow volatility (STD_CFO), a smaller board size (BSIZE), less board independence (BIND), and are managed by more female CEOs (CEO_GENDER). The median test also suggests comparable results, except for profitability (ROA), when comparing family firms with non-family firms.

Table 2, Panel C reports the results of the mean and median tests of the variables used in Equation (1), based on high versus low corporate transparency (HIGH_TRANS vs LOW_TRANS), which is computed based on the industry–year adjusted median of corporate transparency. The results report that firms with higher corporate transparency (HIGH_TRANS) have higher CEO turnover (CEO_TURN) than those with lower corporate transparency. Additionally, high-transparency firms are more likely to be family firms (FF) and to demonstrate higher levels of leverage (LEV), stock returns (RET), sales growth (SGROWTH), cash flow volatility (STD_CFO), employee turnover (EMPTURN), block ownership (BLOCK), board independence (BIND), and instances of CEO duality (CEO_DUAL). Firms characterized by higher levels of corporate transparency, compared to those with lower corporate transparency levels, tend to be smaller in size (SIZE) with lower profitability (ROA). They also have lower book-to-market ratios (BM) and operating cash flow performance (CFO), are of shorter standing in the market (FAGE) and have a smaller board size (BSIZE). These results provide insights into the differences in firm characteristics between high and low corporate transparency groups, shedding light on the potential implications of these characteristics for CEO turnover and corporate governance practices.

Table 3 provides the Pearson bivariate correlation matrix for the variables used in Equation (1), indicating that the variable FF (family firm) is negatively correlated with CEO turnover (r = –0.049); firm size (r = −0.029); operating cash flow performance (r = −0.024); cash flow volatility (r = −0.054); board size (r = –0.082); board independence (r = –0.119); and CEO gender (r = −0.015); whereas the variable FF is positively correlated with transparency (r = 0.058); leverage (r = 0.074); book-to-market ratio (r = 0.027); and firm age (r = 0.239). These coefficients are significant at least at the 1% level, and are in line with the suggestion by Gujarati and Porter (2009) that bivariate correlations below 0.80 do not create multicollinearity problems. Furthermore, we assess the multicollinearity of our models using variance inflation factor (VIF) values. The unreported mean VIF value is 1.22, with the lowest VIF value of 1.03 and the highest VIF value of 1.66. All VIF values are less than the threshold of 10 suggested by Greene (2008), indicating that multicollinearity is unlikely to affect our results.

Although our study’s H1 predicts that corporate transparency moderates the impact of family firms on CEO turnover, we first examine the impact of family firms on CEO turnover. Table 4 provides the logistic regression results. Model (1) reports the association between family firms and CEO turnover, whereas Model (2) reports the moderating role of corporate transparency in the association between family firms and CEO turnover. The results from Model (1) indicate that the coefficient for the family firm variable (FF) is significantly negative (β = −0.262, p < 0.01), suggesting that family firms experience a lower rate of CEO turnover compared to non-family firms. This finding implies that family firms have a 26.20% lower CEO turnover rate than non-family firms.

Furthermore, to test our H1, the variable of interest in Equation (2) is the interaction term between FF and TRANS (FF×TRANS). Model (2) reports the regression results. The coefficient of the interaction term between FF and TRANS suggests that corporate transparency has different impacts on CEO turnover in family firms compared with non-family firms. In our analysis, family firms (FF) are represented by a dummy variable. Consequently, the coefficient of TRANS specifically captures the effect of corporate transparency on CEO turnover in non-family firms. Conversely, the coefficient of FF reflects the impact of family ownership on CEO turnover, with this effect considered under the conditions set up by corporate transparency. This set-up allows us to isolate and understand how corporate transparency influences CEO turnover differently in family firms versus their non-family firm counterparts.

The coefficient of TRANS is significantly negative (β = −0.052, p < 0.01), indicating that higher corporate transparency is associated with lower CEO turnover in non-family firms. The significantly positive coefficient (β = 0.050, p < 0.05) of the interaction term FF×TRANS suggests that increased corporate transparency in family firms leads to higher CEO turnover rates, in contrast to the effect observed in non-family firms. This result implies that, within family firms, heightened corporate transparency may prompt greater scrutiny and faster changes in leadership. An F-test result confirms the significance of the combined effects of TRANS and FF×TRANS, validating our H1 that corporate transparency significantly moderates the impact of family firm status on CEO turnover. The analysis demonstrates that, with a higher level of corporate transparency, family firms are more likely to change CEOs compared to their non-family firm counterparts, thus supporting our H1 ^[6].

Regarding the control variables in our analysis, the coefficients of RET, STD_CFO, EMPTURN and BSIZE are significantly positive, suggesting that firms experiencing higher stock returns, cash flow volatility and employee turnover, with a larger board size have higher CEO turnover. In contrast, the coefficients of ROA, CEO_DUAL and CEO_GENDER are significantly negative, indicating that higher profitability, CEO duality and having a male CEO are associated with lower CEO turnover.

In this study, endogeneity may arise from both observable and unobservable selection biases. Additionally, our moderator variable, corporate transparency, may be affected by endogeneity. To address observable and unobservable selection bias, we employ propensity score matching (PSM) and Heckman’s (1979) two-stage analysis, respectively. We also employ a simultaneous equation framework to address the endogeneity associated with corporate transparency.

The results in Table 2 present systematic differences in firm characteristics between family and non-family firms. These variations could lead to observable self-selection bias, as firms may not be randomly distributed across these categories (Chen et al., 2014). Additionally, functional form misspecification, a specific type of endogeneity, could bias our results (Shipman, Swanquist, & Whited, 2017). To address these issues, we apply propensity score matching (PSM) following Shipman et al. (2017).

In the first PSM stage, we conducted a logistic regression using an indicator-dependent variable to identify the propensity of being a family firm (FF = 1) or a non-family firm (FF = 0). This analysis involves matching, without replacement, each family firm observation with a non-family firm observation using a 1% caliper in the caliper matching method. We ensure an accurate balance between the treatment group (family firms) and control group (non-family firms) by using the same set of control variables in both stages of the PSM (Shipman et al., 2017). In total, 10,152 firm-year observations are matched and used in the second-stage model.

Table 5 presents the regression results. Panel A reports the logistic regression results from the first stage, where leverage (LEV), profitability (ROA), firm age (FAGE), employee turnover (EMPTURN), and block ownership (BLOCK) are positively associated with the propensity to select a family firm. Conversely, firm size (SIZE), operating cash flow performance (CFO), board size (BSIZE); board independence (BIND), CEO gender (CEO_GENDER); and CEO ownership (CEO_OWN) are negatively associated with the propensity to select a family firm. Table 5, Panel B indicates the comparison of firm characteristics between the family and non-family firms used in the first stage. The results report that some firm characteristics significantly differ between family and non-family firms; therefore, we run the ordinary least squares (OLS) regression in the second stage to reduce the bias. Table 5, Panel C presents the second-stage regression results using propensity score-matched samples.

The coefficient of FF is significantly negative (β = −0.272, p < 0.01) in Model (1), while the coefficient of FF×TRANS is significantly positive (β = 0.042, p < 0.10) in Model (2). Moreover, the F-test results indicate that the coefficients of TRANS and the interaction term FF×TRANS are statistically significant in Model (2). These results demonstrate the robustness of our findings, consistent with the prediction that corporate transparency attenuates the negative impact of family firms on CEO turnover.

While PSM is used to address observable differences between family and non-family firms, other unobservable factors may differ between them (Chen et al., 2014). Therefore, we cannot completely rule out the possibility that self-selection bias is associated with being either a family firm or a non-family firm. To address unobservable selection bias, we use Heckman’s (1979) two-stage model. We employ the following first-stage model, following Chen et al. (2014):

In Equation (3), the dependent variable is FF, as defined in Section 3.3. Following prior research (e.g. Chen et al., 2014), we introduce two key variables in Equation (3) to meet the exclusion restriction criteria: industry-level competition (HHI) and changes in industry volatility (ΔIND_VOL). These variables, along with several other factors such as return volatility (VOLAT); share turnover (TURN); business segments (SEGMENT), are included to explore their influence on whether a firm is identified as a family firm (FF).

The rationale behind their inclusion is that founding family members may not hold control if their firms operate in industries characterized by high competition and volatility (Chen et al., 2014). These factors are used to satisfy the exclusion restrictions criteria because although they influence the likelihood of a firm being family controlled, they are unlikely to directly affect CEO turnover. This assumption holds as these industry characteristics impact all firms within that industry, not only individual firms. Industry-level competition is measured using the Herfindahl–Hirschman Index (HHI), which is based on market shares (Chen et al., 2014). Changes in industry volatility (ΔIND_VOL) are captured by the average change in stock return volatility for each industry. We expect a positive sign for HHI, indicating that less competition is positively associated with family control, while the expected negative sign for ΔIND_VOL suggests that a higher level of volatility reduces family control. All other variables are defined in  Appendix.

Table 6, Panel A presents Heckman’s (1979) first-stage regression results, indicating that the coefficients of HHI and ΔIND_VOL are significantly positive and negative, respectively (β = 0.432, p < 0.01; β = −0.182, p < 0.01). The model reports a pseudo-R-squared (R²) value of 6.70% and partial R² values of 0.40% for HHI and 0.78% for ΔIND_VOL. With both significant at the 1% level, this suggests that both variables are sufficiently exogenous to satisfy the exclusion restrictions criteria. Table 6, Panel B presents the second-stage regression results. The coefficient of FF is significantly negative (β = −0.274, p < 0.01) in Model (1), while the coefficient of FF×TRANS is significantly positive (β = 0.050, p < 0.05) in Model (2). In addition, the coefficients of the inverse Mills ratio (IMR) are statistically significant in both models, indicating that our results hold after addressing self-selection bias. Overall, the results from applying Heckman’s (1979) model corroborate our main findings.

Our study examines the moderating role of corporate transparency in the relationship between family firms and CEO turnover. Recognizing that corporate transparency is exogenous to family firms yet potentially influenced by industry dynamics, such as volatility and proprietary costs of disclosure, we adopt a simultaneous equation model to address potential endogeneity issues. Model (1) examines the joint effects of family control and corporate transparency on CEO turnover and focuses on explaining corporate transparency, while Model (2) focuses on explaining corporate transparency. We employ the three-stage least squares (3SLS) technique to estimate these simultaneous equations, ensuring robust estimation by addressing the potential correlations between the error terms of the equations.

We considered using an exogenous shock to analyst coverage as a basis for constructing a shock-based instrumental variable, following He and Tian (2013). They highlight an event in August 2012 when Barclays expanded its Taiwan-based equity research team by recruiting a senior managing director and a group of analysts from Credit Suisse Group AG. This expansion represents a credible exogenous shock, as it is unlikely to be correlated with unobserved factors influencing CEO turnover in family firms. We created a variable, ANA_SHOCK, as an indicator variable that takes the value of 1 for the years after 2012 and 0 for the years before 2011. Additionally, following Chen et al. (2014), we incorporate industry competition (HHI) as an instrumental variable to account for industry dynamics that influence corporate transparency. Table 7 presents the simultaneous equation regression results, indicating that the coefficient of FF×TRANS is significantly positive (β = 0.010, p < 0.01) in Model (1), suggesting that our results are robust after using an exogenous shock to analyst coverage as a determinant of corporate transparency.

While our study accounts for various firm-specific variables and includes industry and year fixed effects in all regression models, the potential for time-invariant omitted variable bias remains. To address this issue, we employ firm fixed effects regressions which remove the effects of omitted time-invariant firm characteristics that could otherwise lead to spurious correlations between CEO turnover and family firm status. The results of these firm fixed effect regressions are presented in Table 8. In Model (1), the coefficient of the family firm variable (FF) is significantly negative (β = −0.281, p < 0.01), indicating that lower CEO turnover is experienced in family firms. In Model (2), the interaction term between family firm status and corporate transparency (FF×TRANS) is significantly positive (β = 0.012, p < 0.10), suggesting that increased corporate transparency in family firms is associated with a higher likelihood of CEO turnover.

In our main analysis, we construct a corporate transparency index using principal component analysis (PCA) based on four proxies: analyst forecast error (AFERROR), analyst forecast dispersion (AFDISP); analyst coverage (ANALYST), and bid–ask spread (SPREAD). To align these variables with the notion that higher values indicate greater transparency, we multiply AFERROR, AFDISP and SPREAD by minus one (−1). Further analysis involves examining each proxy individually, denoted as RAFERROR, RAFDISP and RSPREAD. More specifically, higher values for RAFERROR, RAFDISP and RSPREAD, as well as for ANALYST, are interpreted as indicating a higher level of corporate transparency (Kerr, 2019). The regression results for these individual proxies are presented in Table 9, Models (1)–(4). The coefficients of the interaction term FF×TRANS are significantly positive in all models except Model (3). These results demonstrate that our main findings remain robust when using individual proxies for corporate transparency, thus supporting our H1 and confirming the consistency of our findings.