Printer Friendly

Organizational form, ownership structure, and CEO turnover: evidence from the property-casualty insurance industry.

INTRODUCTION

A firm's corporate governance mechanisms play an important role in disciplining poorly performing chief executive officers (CEOs). Extant literature shows that stronger corporate governance mechanisms discipline management effectively by removing poorly performing CEOs and searching for the best candidate for the CEO position. Therefore, CEO turnover decisions provide an excellent setting to evaluate the quality of corporate governance mechanisms within a firm (Kang and Shivdasani, 1995).

Most previous studies examining the association between the quality of corporate governance mechanisms and CEO turnover decisions have focused only on publicly traded stock companies (e.g., Weisbach, 1988; Denis, Denis, and Sarin, 1997; Borokhovich, Parrino, and Trapani, 1996). Less is known about CEO turnover patterns and their association with performance within firms with organizational forms and ownership structure other than publicly traded stocks. Among 7 million corporate tax filers in the United States, only about 8,000 are publicly traded (Nagar, Petroni, and Wolfenzon, 2011). Because of the significance of nonpublicly traded firms in the U.S. economy, understanding the CEO turnover patterns within these firms would appear to be an important issue for investigation. This study intends to fill this research gap by examining the characteristics of CEO turnover in the U.S. property-casualty (P-C) insurance industry, in which various organizational forms and ownership structures coexist.

The present article focuses on how insurers' organizational forms and ownership structures affect the quality of corporate governance mechanisms, proxied by CEO turnover probabilities and turnover-performance sensitivity. In this regard, we distinguish between routine and nonroutine turnover and emphasize nonroutine turnover as especially indicative of governance differences among firms. The insurance industry provides a particularly rich environment for the analysis of organizational form because a variety of organizational forms coexist in the industry, including stocks, mutuals, and reciprocals (Mayers and Smith, 1988). Moreover, stock insurers have a full spectrum of ownership structures depending on the liquidity of ownership and the presence of controlling shareholders. Stock insurers owned by noninsurance holding companies provide an additional interesting ownership structure.

There has been one prior paper on the relationship between CEO turnover and performance in the insurance industry. He and Sommer (2011) examine the impact of ownership structure on the performance-turnover relationship for U.S. P-C insurers, focusing on the stock and mutual organizational forms. Our article extends He and Sommer (2011) by decomposing stock insurers into publicly traded and nonpublicly traded (closely held) entities and breaking down both types of stocks into family-owned and nonfamily-owned categories. We further subdivide family firms into those with family-member CEOs and those with nonfamily CEOs. We also consider stock insurers owned by noninsurance entities as a separate organizational type. This more detailed decomposition of stock insurers allows us to provide a richer analysis of the effects of organizational form on CEO turnover than has been provided in the prior literature. Thus, ours is only the second article to study organizational form and CEO turnover in the insurance industry and the first article to consider the effects of family ownership on CEO turnover for insurers. Our article also contributes more broadly as one of the limited number of papers to analyze CEO turnover for private firms and privately held family firms (see also Lausten, 2002; Coles et al., 2003; Bennedsen et al., 2007; Adams et al., 2013).

The analysis in this article is based on a large hand-collected data set, which covers 976 firms in the U.S. P-C insurance industry over the period 1993-2006. We conduct probit regression to study the impact of organizational form and ownership structure on CEO turnover decisions. Our findings suggest that the likelihood of CEO turnover, especially nonroutine turnover, is inversely related to firm performance in the insurance industry. Most important, we find that the magnitude of this association varies significantly by organizational form and ownership structure. Compared to publicly traded nonfamily-controlled stocks, mutuals have lower likelihoods of nonroutine CEO turnover. We also find that closely held nonfamily stocks have lower nonroutine turnover-performance sensitivity than publicly traded nonfamily stocks.

Finally, whether there is a controlling family matters in the CEO turnover decision. Controlling families exist widely in both publicly traded stocks and closely held stocks and have been shown to be an important corporate governance mechanism (e.g., Anderson and Reeb, 2003; Li and Srinivasan, 2011). On the one hand, controlling families in stock insurance companies are entrenched. The probability of CEO turnover is lowest for family-member CEOs in family-controlled firms, both closely held and publicly traded. On the other hand, controlling families can be effective monitors of management when they do not hold the CEO position. Nonfamily-member CEOs of publicly traded family firms have one of the highest likelihoods of turnover. This suggests that monitoring from both capital market and controlling shareholder (holding family) plays an important role in improving the quality of a firm's corporate governance.

The remainder of the paper is organized as follows. In the second section, we discuss the relevant literature. The third section develops the hypotheses based on prior literature. The fourth section describes the data set and methodology. The fifth section provides the results and discussion, and the sixth section concludes.

LITERATURE REVIEW

We first briefly review the literature about CEO turnover in publicly traded stock companies. We then consider the papers that also analyze nontraded firms.

Top Executive Turnover in Publicly Traded Companies

Over the past 3 decades, a large body of literature has developed on different aspects of CEO turnover in publicly traded industrial firms. The general consensus is that the likelihood of CEO turnover, especially nonroutine turnover, is negatively related to firm performance (Coughlan and Schmidt, 1985; Warner, Watts, and Wruck, 1988; Murphy and Zimmerman, 1993). (1) The magnitude of this relation, however, depends on the quality of corporate governance within the firm. For example, Weisbach (1988) finds that the negative relationship between turnover and performance is stronger when the majority of the board consists of outside directors. This is consistent with the argument of Fama and Jensen (1983a) that outside directors have incentives to enhance the value of their human capital by signaling to the managerial market that they are experts in decision control, while inside directors are less willing to challenge the CEO to whom their careers are tied. Ertugrul and Krishnan (2011) find that boards with higher equity-based compensation are more likely to be proactive in dismissing low-ability managers.

Other important determinants of performance-turnover sensitivity include managerial ownership, blockholder ownership, and the threat of takeover activity. Denis, Denis, and Sarin (1997) find that performance-turnover sensitivity is lower with higher management ownership but increases with the presence of an outside blockholder. They suggest that managers' ownership partially insulates them from the takeover market, which reduces the effectiveness of the internal monitoring mechanism in disciplining poorly performing CEOs. Cornelli, Kominek, and Ljungqvist (2013) argue that boards rely more on "soft" information than on "hard" information to learn about the CEO's ability and consequently to make CEO turnover decisions.

Top Executive Turnover in Nonpublic Companies

Although the extant literature mainly focuses on publicly traded stock companies, five primary papers consider nontraded firms. Examining data on closely held firms from the Forbes list of the top 500 private firms, Coles, Lemmon, and Naveen (2003) find no evidence that CEO turnover is more sensitive to changes in scaled profitability in closely held firms than in publicly traded firms. However, because they focus on only large closely held firms, their results do not necessarily generalize to other nontraded corporations. In contrast, Adams, Mansi, and Nishikawa (2013) study turnover of U.S. mutual fund managers and find that public sponsors are more sensitive to performance than private sponsors.

Lausten (2002) and Bennedsen et al. (2007) investigate private and public as well as family and nonfamily firms using data from Denmark. Lausten finds an inverse relationship between performance and CEO turnover and that family ties within management strengthen the CEO turnover-performance relationship. Bennedsen et al.'s main finding is that family successions are negatively correlated with firm performance around CEO successions.

In the paper most similar to ours, He and Sommer (2011) examine the sensitivity of CEO turnover to firm performance for stocks and mutuals in the U.S. P-C insurance industry over the period 1996-2004. They find that CEO turnover for stock insurers is inversely related to prior performance, but no such relationship is found for mutual insurers. They interpret these results as consistent with the managerial entrenchment hypothesis, such that mutual managers are able to insulate themselves from forces of corporate control, and also with the managerial discretion hypothesis that mutual managers are less likely to be held accountable for poor performance because they exercise less discretion than managers of stock firms. As mentioned above, we extend He and Sommer's analysis by also considering the distinction between publicly traded and nontraded stocks and between family-controlled and nonfamily-controlled stocks. Thus, our article is the first to consider the full range of ownership types and degree of family control for insurers as well as contribute to the general literature on CEO turnover and the turnover-performance relationship in both publicly traded and nonpublicly traded firms.

ORGANIZATIONAL FORMS AND OWNERSHIP STRUCTURES IN INSURANCE

Fama and Jensen (1983b) argue that the specific characteristics of the residual claims in each organizational form and ownership structure lead to efficient approaches to controlling agency costs between residual claimants and decision agents. According to Mayers and Smith (1988), there are three important functions within an insurer: the managerial function, the ownership/risk-bearing function, and the customer/policyholder function. The agency costs in an insurance organization arise mainly from conflicts of interest among these three functions. When the roles of owner and manager are separated, a potential incentive problem is created since managers do not bear the major wealth effects of their actions and generally have interests different from those of owners. The owner-policyholder agency problems are similar to those of shareholders and bondholders. The owners of the firm and/or their agents (managers) have incentives to pursue their own interests at the expense of policyholders after insurance policies are sold. Different organizational forms and ownership structures have comparative advantages in mitigating various incentive conflicts, as discussed in the following subsections.

Publicly Traded and Closely Held Stocks

Publicly traded stocks have advantages in relatively risky lines of business because of their complete separation of the decision management and risk-bearing functions (Fama and Jensen, 1983a). However, this separation creates agency problems between managers and shareholders. To mitigate these agency costs, publicly held stocks rely on monitoring from the capital market (e.g., financial analysts and institutional investors), the takeover market, and boards of directors (Fama and Jensen, 1983a). Most empirical evidence indicates that the probability of CEO turnover is inversely related to the firm's stock price performance in publicly traded firms (Coughlan and Schmidt, 1985; Denis, Denis, and Sarin, 1997).

Although publicly traded stocks have the most widely diffuse ownership and have been mostly widely studied in literature, the most popular firm structure in the United States is the closely held stock firm (Nagar, Petroni, and Wolfenzon, 2011). When a stock insurance company is closely held, the monitoring of managers by the owners is direct and simple. Fama and Jensen (1983a) argue that the most effective way to control the costs raised by separating owner and manager in the closely held stock company is restricting the ownership to managers. Ke, Petroni, and Safieddine (1999) posit that closely held insurers should have more direct monitoring of management by owners. Nagar, Petroni, and Wolfenzon (2011) indicate that a key feature of closely held firms is that shareholders are typically few in number, and they are also familiar with and involved in management. Mayers and Smith (1994) find that the costs of controlling the owner-manager conflicts are greater in widely held companies than in closely held companies.

Even in the case where ownership is not restricted to managers, those with special relations with managers may own the closely held corporation to control the agency problems efficiently, such as a controlling family. However, the lack of monitoring from the capital market and high information asymmetry between managers and (minority) shareholders might increase the entrenchment opportunity of incumbent managers. Overall, the effectiveness of disciplining CEOs of closely held stock firms stems mainly from monitoring by a relatively small number of owners who are familiar with and involved in the management, while the capital market enhances the effectiveness of corporate governance mechanisms of publicly traded stock firms. The sensitivity of CEO turnover to firm performance may be stronger in either closely held or publicly traded stock insurers since each type has its own specific advantage in terms of corporate governance. Accordingly, the relative turnover performance sensitivity of closely held and publicly traded stocks is ultimately an empirical issue. To avoid possible distortions from family control, we focus the hypothesis on the comparison between closely held nonfamily-owned stock insurers and publicly traded nonfamily-owned stock insurers.

Hypothesis 1: Closely held and publicly traded nonfamily-owned stock insurers are equally likely to remove the CEO when the firm is poorly performing.

Finally, there are a few stock insurers ultimately owned by noninsurance holding companies, including financial institutions and industrial firms. Control by a noninsurance parent might have some distorting effects on CEO turnover decisions. For example, there might be more turnovers due to the normal internal rotation of positions within subsidiaries of a holding company. Also, CEOs with good performance might be promoted to parent holding firms, leading to an relatively high CEO turnover rate. Therefore, the empirical analysis distinguishes this ownership type from firms whose ownership is exclusively within the insurance industry.

Family-Owned Stocks

An important type of stock company in the insurance industry is the family firm, which can be either closely held or publicly traded. Ours is the first article to investigate family ownership for insurance companies, and we consider both closely held and publicly traded family stock firms. Because firm performance has substantial wealth effects on the controlling family, the controlling family is more likely to actively monitor managers compared to other types of large shareholders, such as institutional shareholders (Anderson and Reeb, 2003; Li and Srinivasan, 2011). Ownership by a controlling family can effectively mitigate owner-manager incentive conflicts, either by monitoring a nonfamily-member CEO or appointing a family member as CEO. We posit that the dominant factor determining turnover-performance sensitivity in family stock firms is whether a family member holds the CEO position.

CEOs in family-owned stock firms who are not family members are likely to be closely monitored by the controlling family besides the normal disciplining of the corporate monitoring mechanism. The controlling family generally serves as a more effective monitor than other blockholders due to the linkage of firm performance and family wealth. Therefore, we expect the sensitivity of CEO turnover to firm performance for nonfamily CEOs in family firms to be higher than for CEOs of nonfamily-owned stock firms. (2) This leads to the following hypotheses:

Hypothesis 2-1: Compared to peers in nonfamily-controlled publicly traded stock firms, a nonfamily CEO in a family-controlled publicly traded stock firm is more likely to be removed and has higher turnover-performance sensitivity.

Hypothesis 2-2: Compared to peers in nonfamily-controlled closely held stock firms, a nonfamily CEO in a family-controlled closely held stock firm is more likely to be removed and has higher turnover-performance sensitivity.

Appointing a family member to the CEO position can also effectively mitigate agency problems between owners and managers by combining ownership and management. However, a new agency problem between majority and minority shareholders is created. This problem may be even more severe in family firms since many effective corporate control mechanisms might not function well for family firms. Poorly performing family-member CEOs have relatively high entrenchment opportunities and are more difficult to remove due to the family's ownership stake (Morck, Shleifer, and Vishny, 1988). Higher managerial ownership causes less effective external and internal managerial monitoring and thus weaker links between performance and managerial turnover (Denis, Denis, and Sarin, 1997). Thus, we expect family-member CEOs to have lower turnover probabilities and performance-turnover sensitivity than nonfamily-member CEOs of family stock firms and CEOs of nonfamily-owned stock firms. If there is turnover, family firms with a family-member CEO might more likely take the form of routine rather than nonroutine turnover. This leads to the following hypothesis:

Hypothesis 2-3: Among stock insurers, family-member CEOs of family firms have the lowest likelihood of nonroutine turnover and lowest turnover-performance sensitivity.

Mutuals and Reciprocals

Mutuals are an important organizational form in insurance. The major benefit of the mutual organizational form is the control of the policyholder-owner agency conflict by merging the policyholder and ownership functions. However, some effective corporate control mechanisms for publicly traded stock insurers are not available in mutuals because the ownership rights of mutuals are restricted in comparison with stock corporations (Mayers and Smith, 1988). Because they do not have traded shares, it is not possible to form a capital market for mutuals, and their managers are not subject to monitoring by stock analysts, institutional investors, and blockholders. The inability to transfer ownership rights in mutuals also makes takeover activity for mutuals costly and ineffective. Because of the inalienability of ownership rights in mutuals, it is not possible to align incentives of board members with those of owners by giving them ownership interests in the firm. (3) Inalienability also prevents individual policyholders from concentrating ownership, while shareholders of stocks can do so by purchasing shares. The wide diffusion of ownership also gives policyholders little ability to monitor managers in mutuals (Hansmann, (1985)). Finally, withdrawing ownership in a mutual does not constitute liquidation of a share of the insurer's equity capital. (4) Therefore, the management-control mechanisms are much weaker in mutuals than in publicly traded stocks.

We specifically compare CEO turnover between mutual and publicly traded nonfamily-owned stock insurers to control for the possible confounding effects of close ownership or family control. In other words, nonfamily-owned publicly traded firms may provide a better benchmark than all stocks to evaluate the effectiveness of the corporate governance mechanisms of mutuals. As mentioned above, the main corporate governance mechanisms in publicly traded stock companies are internal monitoring from boards of directors, external monitoring from the capital market, and the threat of takeover for poorly performing firms. In mutuals, the main corporate governance mechanism is monitoring from the board of directors. However, we argue that board monitoring alone is less effective than capital market monitoring. Hence, we posit:

Hypothesis 3: Mutuals have lower probability of nonroutine CEO turnover and lower sensitivity of nonroutine turnover to firm performance than publicly traded nonfamily-owned stocks.

In reciprocals, another important organizational form in insurance, the subscribers exchange insurance contracts through an attorney-in-fact to share or spread their risk. (5) The daily affairs of the reciprocals are under the management of the attorney-infact. Similar to a mutual, the reciprocal merges the policyholder and ownership functions into the subscribers, who are the members of the reciprocal. Therefore, mutuals and reciprocals have similar owner-manager agency problems. Even though reciprocals in their original form differed from mutuals in some important respects, such as having withdrawable deposits, modern reciprocals are virtually indistinguishable from mutuals. For this reason and because our tests reveal that mutuals and reciprocals behave similarly with respect to CEO turnover, we combine the two organizational forms in the remainder of the study and refer to the joint category as "mutuals." (6)

Organizational Form and Ownership Structure: Conclusions

According to the aforementioned discussion about organizational forms and ownership structures, insurers in the United States generally can be classified into two principal categories of organizational forms: stocks, on the one hand, and mutuals and reciprocals ("mutuals"), on the other hand. (7) We further categorize stock insurers into four types of ownership structures: publicly traded family-owned firms, publicly traded nonfamily-owned firms, closely held family-owned firms, and closely held nonfamily-owned firms. Finally, family-owned firms can be categorized depending upon whether the CEO is a family member or not. Figure 1 provides a branch diagram summarizing the types of insurers covered in this study. The control of owner-manager agency problems depends on the characteristics of the different ownership structures. Generally, each organizational form and ownership structure has specific features to discipline management, and no consensus has been reached with regard to their comparative effectiveness.

SAMPLE AND METHODOLOGY

Sample Selection

The sample for the study consists of U.S. licensed P-C insurance companies. Our main data sources are A.M. Best's Insurance Reports: Property/Casualty Edition (Best's Insurance Reports), (8) the National Association of Insurance Commissioners (NAIC) annual statement database, and proxy statements of the publicly traded insurers. The sample consists of U.S. P-C insurers who have CEO information in the "Management" section of Best's Insurance Reports from 1993 to 2006. (9) CEO information for publicly traded companies is cross-checked with executive biographies in the proxy statements. Because Kang and Shivdasani (1995) indicate that foreign companies have corporate governance and accounting practices different from U.S. firms, 70 foreign-owned insurance companies are dropped from the sample. (10) We also exclude 110 companies that disappear from Best's Insurance Reports before 2006 due to the regulatory actions for reasons such as insolvency or liquidation that could distort the CEO turnover results. (11) Following Denis and Denis (1995), we eliminate 289 insurance companies that were merged or acquired during the sample period. (12)

Our primary sample includes all unaffiliated single insurers and groups of insurers under common ownership. Treating each insurance group as an independent decision-making unit can minimize sample bias because the subsidiaries within a group share the same ownership and almost always have the same management (Beatty, Ke, and Petroni, 2002). (13) If we treat these subsidiaries as individual decision-making units, a single CEO turnover event for the group might be counted several times among its subsidiaries, biasing the analysis. In the final sample, we have 976 firms and 8,755 firm-years. The sample firms represent 65 percent of total industry premiums in 2000 and comparable percentages for other years.

Variable Construction

To identify the factors affecting CEO turnover, we construct four sets of variables: CEO turnover events, which serve as dependent variables; organizational form and ownership structure dummy variables; firm performance measures; and control variables.

CEO Turnover Events. We define the top executive as CEO when she holds the title of CEO in the firm. (14) If an insurer has no individual listed as CEO, the executive who has the title of president is selected. If no individual is listed as either CEO or president, we define the chair of the board of directors as the top executive. (15) We identify CEO turnover events first according to the information reported in Best's Insurance Reports and company proxy statements. If the information about a CEO turnover event is not revealed explicitly, we identify the event by tracking the names of relevant officers in Best's Insurance Reports and the proxy statements through 1992-2006. If there is any change of CEO names between 2 consecutive years (t and t + 1), we define a CEO turnover event in the tth year. We delete turnover of interim CEOs who hold CEO positions less than 1 year. Following Kang and Shivdasani (1995) and He and Sommer (2011), we treat turnover events where the CEO does not remain in the company as a director or in another capacity for more than 2 years as nonroutine turnover, and all others as routine turnover. (16)

Organizational Form and Ownership Structure. We define the firm's organizational form and ownership structure using the information provided by A.M. Bests' Key Rating Guide, Best's Insurance Reports, the NAIC database, and proxy statements for the publicly traded companies. This enables us to identify stocks, mutuals, and reciprocals. (17) If these sources do not reveal the ultimate owner of a stock insurance company, we further check the company's website and news sources on the Internet. A closely held stock firm is defined as a family firm if the information from the "Management" section of Best's Insurance Reports gives explicit detailed information about family ownership. A publicly traded insurance company is classified as a family firm if more than 5 percent of the firm's shares are owned by the family (Anderson, Reeb, and Zhou, 2012; Villalonga and Amit, 2006). If the ultimate owner of an insurer is not from the insurance industry, we classify it as a "stock insurer owned by a noninsurance holding company" to control its effect on the CEO turnover decision. (18)

Performance. We use book return on assets (ROA) as the primary proxy for firm performance and conduct robustness checks using book return on equity (ROE). (19) Industry performance is also used to control for industry-wide shocks to performance and for the effects of the insurance underwriting cycle. Industry performance is proxied by median ROA. We use 1-year lagged values of the performance measures, as boards are believed to react relatively fast to poor performance in the CEO dismissal decision (Weisbach, 1988; Murphy and Zimmerman, 1993). (20)

Extant literature suggests that industry factors, which are out of the control of individual firm CEOs, are usually filtered from dismissal decisions by the board of directors; that is, CEO compensation is driven by relative performance evaluation (Morck, Shleifer, and Vishny, 1989). This argument suggests a positive relationship between turnover and industry-wide performance. That is, CEOs are more likely to be dismissed following bad years for their firm but less likely to be dismissed following bad years for the industry.

However, Jenter and Kanaan (2014) argue that the corporate boards cannot effectively filter exogenous shocks from CEO dismissal decisions and find an inverse relationship between turnover and industry performance. Arguably, performance might be more informative about CEO skill in soft markets associated with prolonged years of declining profitability and intense price competition in the insurance industry. Boards thus may act more quickly to remove underperforming CEOs during soft markets than in hard markets. Thus, the sign of the industry performance variable remains an empirical issue.

Control Variables. Consistent with prior empirical studies, we control for board and other firm characteristics that may affect turnover decisions. Following Weisbach (1988) and Denis, Denis, and Sarin (1997), we create a board independence dummy variable. Firms in which outsiders make up no more than 40 percent of the directors are classified as insider-dominated firms, and firms in which at least 60 percent of the directors are outsiders are classified as outsider-dominated firms. For each firm-year in the sample, we identify the composition of the board of directors from Best's Insurance Reports for nonpublicly traded insurers and from proxy statements for publicly traded firms. (21) Following the independence requirement of the New York Stock Exchange, we define outside directors in nonpublicly traded insurers as those who are not listed as executives in the company or in the same insurance group, are not retired CEOs, and do not have the same last name as any executive listed in the Management section of Best's Insurance Reports. If the director is from the controlling family, she is also classified as an inside director. We expect a positive relationship between board independence and CEO nonroutine turnover.

We also include another board characteristic variable--logarithm of board size. We expect a negative relationship between board size and the likelihood of CEO turnover (Jensen, 1993; Yermack, 1996). We include firm size (Denis, Denis, and Sarin, 1997), measured by the log of total net premiums written. We control for firm risk and potential regulatory costs using leverage, measured as the ratio of liabilities to total admitted assets.

Business complexity also might affect CEO turnover decisions (Parrino, 1997). We first proxy business complexity using the proportions of net premiums written in personal lines and long-tail commercial lines, with short-tail commercial lines omitted to avoid singularity. Long-tail commercial lines are the most complex type of P-C insurance, and personal lines are the least complex. An insurer's business complexity is also positively correlated with business diversification. Insurers can diversify risk by writing across many different product lines and/or across different geographic areas. Therefore, Herfindahl indices for product mix and geographical diversification are included in the model as alternative proxies for complexity. (22) The lower the Herfindahl, the higher is the expected degree of business complexity.

RESULTS AND DISCUSSION

Descriptive Statistics

Table 1 reports the annual turnover numbers and rates for the full sample and the various subsamples. There are 976 firms and 8,755 firm-years in our sample. The largest subset of firms consists of mutuals (4,463 observations), and the smallest consists of publicly traded family stock firms where the CEO is not a family member (183 observations). Roughly 75 percent of closely held stock insurers are owned by controlling families. The first data column of Table 1 also reports (in parentheses) the number of unique firms in each organizational form category. The largest number of unique firms is for mutuals (439 firms) and the smallest is for stock firms owned by noninsurance holding companies (28 firms).

Our results indicate 572 turnovers, 243 of which are identified as nonroutine turnover. The overall annual turnover rate for the full sample is 6.53 percent, and the nonroutine turnover rate is 2.78 percent. (23) Because we are mainly interested in nonroutine turnover, we focus the discussion on the results in the last column of Table 1. By far the lowest nonroutine turnover rates are for closely held and publicly traded family stock firms where the CEO is a family member (0.58 percent and 1.06 percent, respectively). These turnover rates are significantly lower than for all nonfamily member CEO stocks (4.37 percent). These results suggest that family-member managers are difficult to remove, due to their power and entrenchment or greater willingness to invest their long-term human capital in the company, consistent with Denis, Denis, and Sarin (1997). Mutuals have a significantly lower average nonroutine turnover rate (2.60 percent) than either publicly traded or closely held nonfamily-owned stock firms (4.67 percent and 4.53 percent, respectively).

Panel A of Table 2 presents the medians of board characteristics and financial variables for the overall sample and subsamples. Panel B of Table 2 presents Kruskal-Wallace tests of differences between the medians of various pairs of sub-samples.

Mutuals have larger boards at the medians (nine board members) than closely held family stock firms (six board members where there is a family-member CEO and five where there is a nonfamily CEO) and stock insurers owned by noninsurance companies (seven board members). Mutual boards are not significantly different at the median than the boards of closely held nonfamily stocks and publicly traded family and nonfamily stocks. Mutuals have about the same proportion of outside board members (73.3 percent) as closely held and publicly traded nonfamily stocks (77.8 percent and 75.0 percent, respectively), and these three types have higher board independence than all other organizational forms. Thus, the conclusion by Mayers, Shivdasani, and Smith (1997) that mutuals have higher proportions of outside directors than stocks is mainly driven by the relatively low board independence of family-owned stocks.

Table 2 also presents the medians of the financial variables for the firms in the sample. All categories of publicly traded insurers have significantly larger average premiums than the other firm types, with the largest being for publicly traded, nonfamily-owned firms ($544.2 million), and publicly traded family stock firms with nonfamily CEOs ($260.8 million). Closely held family firms are relatively small in terms of net premiums written compared to all other types of firms ($7.3 million with a family-member CEO and $10.7 million with a nonfamily CEO), indicating the importance of controlling for firm size in our regression analysis. Mutuals are larger at the median ($20.4 million) than closely held family-owned stock firms but smaller than all other categories of stock firms. Mutuals also have lower leverage (0.576) than all categories of publicly traded stock firms (ranging from 0.653 for publicly traded family firms with a family-member CEO to 0.707 for publicly traded nonfamily stocks). The lower leverage of mutuals is an operating strategy that reflects their more limited access to external capital.

Turnover Likelihoods: Regression Results and Economic Significance

Turnover Regression Results. Column (1) of Table 3 presents the probit regression results with turnover as the dependent variable, and columns (2) and (3) present the multinomial probit regression results for the outcomes of routine and nonroutine turnover with no-turnover used as the base outcome (Parrino, 1997). (24) We omit the dummy variable for publicly traded nonfamily-owned stock firms, and this category thus serves as the benchmark firm type. We include the interaction terms of ROA with organizational firm dummies and report a linear probability model in columns (4)-(6) "to ease interpretation" when predicting CEO turnover involving interaction terms (Cornelli, Kominek, and Ljungqvist, 2013). (25)

We focus the discussion on the nonroutine turnover results presented in columns (3) and (6) of Table 3 and mention the main differences between the routine and nonroutine turnover results where applicable. The coefficient of lagged ROA is negative and significant at the 1 percent level in column (3) and the 10 percent level in column (6). This implies a significant negative relationship between firm performance and the likelihood of nonroutine turnover, consistent with most previous CEO turnover studies (e.g., Weisbach, 1988; Evans, Nagarajan, and Schloetzer, 2010; He and Sommer, 2011). (26) In contrast, ROA is not significant at the conventional level for routine turnover in column (2) and positive and significant in column (5). The latter result could suggest that CEOs of well-performing firms have higher probabilities of obtaining employment elsewhere, or it could be a statistical anomaly. (27)

The industry performance variable (median of industry ROA) is positive and statistically significant in all models shown in Table 3 except for the interacted, nonroutine turnover model (column (6)) where it is positive with a p-value of 11.7 percent. Thus, there is some support for the argument that the board of directors can determine whether firm performance is industry-wide or firm-specific, that is, the relative performance evaluation hypothesis. Our results are not consistent with Jenter and Kanaan's (2014) finding that there are more turnovers when industry operational results are worse.

The dummy for nonfamily-owned closely held stock firms is insignificant in all equations, consistent with Coles, Lemmon, and Naveen (2003). However, the interaction of ROA and the dummy for nonfamily-owned closely held stocks is positive and significant in column (6), rejecting Hypothesis 1 that closely held and publicly traded nonfamily-owned stocks have the same nonroutine turnover-performance sensitivity and providing evidence that stock market monitoring leads to higher turnover-performance sensitivity than monitoring by a small group of owners.

The dummy variable for publicly traded family stock firms where the CEO is not a family member is positive and significant in column (6), providing evidence consistent with Hypothesis 2-1's prediction that nonfamily CEOs in publicly traded family stock firms are more likely to be removed than CEOs of publicly traded nonfamily stock firms. The coefficient on the interaction of the publicly traded, nonfamily-CEO dummy variable with ROA is significant and negative, supporting the part of Hypothesis 2-2 predicting that the turnover-performance sensitivity is higher for publicly traded family firms with nonfamily CEOs than for publicly traded nonfamily firms. Indeed, the coefficient of the interaction term of ROA and the dummy variable for publicly traded stocks with a nonfamily member CEO is the smallest of all ROA interaction terms, implying that this organizational form has the highest nonroutine turnover-performance sensitivity. Support for Hypothesis 2-1 implies that controlling families provide additional monitoring on top of capital market monitoring in public firms with nonfamily CEOs.

The dummy for closely held family stock firms with nonfamily CEOs is negative and significant, whereas the dummy variable for closely held nonfamily stock firms is insignificant. Thus, the results do not support the part of Hypothesis 2-2 predicting that nonfamily CEOs in closely held family stock firms are more likely to be removed than CEOs of nonfamily-owned closely held stock firms.

The interaction of ROA and the dummy for closely held family stock firms with nonfamily-member CEOs is not significant. Combining this result with the significant positive interaction of ROA and the dummy for closely held nonfamily-owned stocks, we find evidence supporting the part of Hypothesis 2-2 predicting that turnover-performance sensitivity is higher for family-owned closely held firms with nonfamily CEOs than for nonfamily-owned closely held stock firms.

The dummy variables for publicly traded and closely held family stock firms with a family-member CEO are significant and negative in the nonroutine turnover equations ((3) and (6)). Further, the coefficients of these two variables are significantly smaller than dummy variable coefficients for any other organizational forms and ownership structures, except for the coefficient of publicly traded family stock firms with family-CEOs in equation (6), which is not significantly different from the coefficients for mutuals and closely held family stock firms with nonfamily CEOs. Taken as a whole, these results provide evidence that family-member CEOs are the least likely to be removed. The interaction between ROA and the dummy for closely held family stock firms with a family-member CEO is positive and significant in column (6), suggesting the lowest nonroutine turnover-performance sensitivity except for the closely held nonfamily-owned stocks. The interaction term of ROA and the dummy for publicly traded family stock firms with a family-member CEO also carries the expected positive sign in equation (6), although it is not significant. Overall, the evidence supports Hypothesis 2-3 and is consistent with managerial ownership of the firm resulting in entrenchment (Denis, Denis, and Sarin, 1997; Chen, Cheng, and Dai, 2013). These results imply the presence of a family-member CEO is the most important factor in CEO turnover decisions.

The mutual dummy has a significant negative coefficient in the nonroutine turnover equations. 28 Thus, the results provide support for the part of Hypothesis 3 predicting that mutuals have lower probability of nonroutine turnover than publicly traded nonfamily stocks. Mutuals also have significantly lower nonroutine turnover probabilities than all other types of stock firms except closely held family stock firms with family or nonfamily CEOs and publicly traded family stock firms with family-member CEOs. The fact that mutuals have lower nonroutine turnover than some but not all types of stock insurers again emphasizes the importance of using a more detailed categorization of stock firms.

The interaction term of ROA and the mutual dummy is positive but not significant in column (6) (p-value of 13.0 percent). Thus, the results do not provide significant support for the part of Hypothesis 3 predicting that mutuals have lower turnover-performance sensitivity than publicly traded nonfamily stocks. Mutuals do have significantly lower turnover-performance sensitivity than publicly traded family stock firms with nonfamily CEOs. (29)

He and Sommer (2011) also find a significant negative coefficient on the mutual dummy variable in their nonroutine turnover regressions, where mutuals are compared to all stocks. They do find a significant positive coefficient on the interaction of the mutual indicator and ROA, whereas our mutual indicator-ROA variable is insignificant. The difference in findings may reflect our splitting the stock dummy variable into different categories, because turnover-performance sensitivity does vary significantly across the types of stock firms.

The dummy for stock insurers owned by noninsurance holding companies is positive and significant in the nonroutine turnover equations ((3) and (6)). Considering the magnitude and significance of the other dummy variables, this result suggests that noninsurance-owned stocks experience more nonroutine turnover than any other type of firm. However, the interaction term of ROA and the dummy for stock insurers owned by noninsurance holding companies is not significant. This may indicate that noninsurance parents are more likely to rotate executives among businesses to gain experience with different types of operations, whereas such moves make less sense within an insurance organization. This variable is not tested in He and Sommer (2011) because they group all stock firms into a single category. (30)

The line-of-business Herfindahl index has a significant positive relationship with nonroutine turnover, consistent with He and Sommer (2011). This may indicate that boards have difficulties in precisely evaluating the performance of CEOs in more diversified firms (with lower Herfindahls) than in focused firms, and is also consistent with the argument the costs of replacing a CEO are higher in more complex firms (Parrino, 1997; Berry et al., 2006). The significant negative coefficient on the commercial long-tail variable provides further evidence of lower turnover probabilities in complex insurers. Therefore, business complexity does seem to affect nonroutine turnover decisions for insurers. Like He and Sommer, we find that the geographical Herfindahl is insignificant, providing evidence that line-of-business complexity is more important than geographic diversification in determining nonroutine CEO turnover.

Board independence is measured in our regressions by a dummy variable equal to one if the fraction of outside directors is greater than 0.6. This variable is positive and significant in the nonroutine turnover regression, implying that a higher fraction of outside board members increases the probability of nonroutine CEO turnover. (31) This is not consistent with He and Sommer (2011), who find a negative coefficient for board independence, and also not consistent with some of the regression results presented in Weisbach (1988). However, our finding is consistent with recent papers theorizing a positive relationship between board independence and CEO turnover (e.g., Laux, 2008). In our regressions, the interaction term of ROA with the board independence dummy is not statistically significant. Hence, we do not find evidence of a higher degree of turnover-performance sensitivity with independent boards. (32)

Firm size is not statistically significant in our regressions. (33) Board size also is insignificant in our regressions. (34) Board size is not tested in He and Sommer (2011). The leverage variable also is insignificant, suggesting that firm leverage does not play an important role in CEO turnover decisions. (35) He and Sommer do not include a leverage variable.

In addition to the variables discussed above, He and Sommer (2011) also test a group indicator variable for group membership. Because we analyze the data at the group and unaffiliated firm levels to avoid double-counting turnover events at subsidiaries, the group membership variable is not relevant in our study. We did test an unaffiliated firm dummy in some regressions, but it was not significant and was excluded from our final regressions.

We also followed He and Sommer (2011) in testing earnings volatility, based on rolling 5-year standard deviations of net income for the 5 years prior to year t, and reinsurance usage, defined as reinsurance premiums ceded divided by direct premiums written plus reinsurance assumed. Like He and Sommer, we found the volatility variable to be statistically insignificant, and it was excluded from our final regressions. He and Sommer find reinsurance usage to be positively related to nonroutine turnover. This variable was not significant in our regressions and had no qualitative effect on the other variables included in the models. Therefore, this variable too was excluded from the final version of our regressions. Where there are differences between our results and He and Sommer (2011), we believe they are mostly attributable to our breaking down stock firms into a wider variety of categories and to our use of groups and unaffiliated firms rather than companies as the decision making units.

Economic Significance by Organizational Form and Ownership Structure. We would like to compare the economic significance of the change in ROA on likelihoods of CEO turnover, routine turnover, and nonroutine turnover for the full sample and subsamples for each type of organizational form and ownership structure. We mainly focus the discussion on nonroutine turnover. First, we rerun the multinomial probit regression for the full sample without firm organizational form and ownership structure dummies. (36) We then estimate the multinomial probit regression for each type of organizational form and ownership structure separately. Based on these regression results, we calculate the likelihood of CEO turnover when ROA is in the top 10 percent ([greater than or equal to] 90th percentile) and bottom 10 percent ([less than or equal to] 10th percentile) using medians of the subsamples for the other regressors. The results are reported in Table 4.

The results in the first row of Table 4 for the full sample show that the likelihood of CEO nonroutine turnover increases from 1.862 percent to 3.404 percent when an insurer's ROA falls from the top decile to the bottom decile, an 83 percent increase in the nonroutine turnover likelihood. This difference is statistically significant at the 1 percent level. (37)

We compare the CEO turnover patterns between closely held and publicly traded nonfamily-owned stock firms (rows (3) and (8)). When ROA is in the bottom decile, the probability of nonroutine turnover is 5.325 percent for closely held nonfamily-owned firms (row (3)) and 5.923 percent for publicly traded nonfamily-owned firms (row (8)). This difference is statistically significant at the 1 percent level. This result rejects Hypothesis 1 that nonfamily-owned publicly traded stock insurers have similar nonroutine turnover rates and likelihoods of removing poorly performing CEOs in comparison to nonfamily-owned closely held stock insurers. Public market monitoring does seem to lead to higher turnover rates when the firm is poorly performing.

We further compare the nonroutine turnover probability between nonfamily-member CEOs in family-owned stock firms and in nonfamily-owned stock firms to see if controlling families play a role as active monitors. When ROA is in the bottom decile, the likelihood that CEOs are forced to resign is about the same and not statistically different in publicly traded nonfamily-owned stock firms (5.923 percent in row (8)) than in publicly traded family-owned stocks with nonfamily CEOs (5.964 percent in row (7)). However, for poorly performing publicly traded family-owned stock companies with nonfamily CEOs, the likelihood of nonroutine turnover significantly increases from 0.392 percent to 5.964 percent (row (7)) when firm performance deteriorates from the top decile to the bottom decile of ROA. This is much higher than the increase in the rate for nonfamily-owned publicly traded stock firms (from 4.848 percent to 5.923 percent, row (8)). This suggests that the performance-turnover sensitivity is higher for publicly traded family-owned stock companies with nonfamily CEOs than for nonfamily-owned publicly traded stock firms. These results support the part of Hypothesis 2-1 regarding turnover-performance sensitivity.

For closely held stock firms in the bottom ROA decile, the nonroutine turnover rates are 5.325 percent for nonfamily owned firms (row three) and 3.683 percent for family-owned firms where the CEO is not a family member (row five). The difference is statistically significant at the 1 percent level. This result does not support the turnover probability part of Hypothesis 2-2. However, the increase in turnover as performance deteriorates from the top decile to the bottom decile is from 3.791 percent to 5.325 percent for nonfamily owned firms (a 40 percent increase) and from 2.024 percent to 3.683 percent for family-owned firms where the CEO is not a family member (an 82 percent increase), supporting the turnover-performance sensitivity part of Hypothesis 2-2.

The results indicate that both the capital market and controlling families play important roles in disciplining poorly performing CEOs in publicly traded companies. If there is only the presence of a controlling family without monitoring from the capital market (row (5)), bottom-decile nonfamily-member CEOs are significantly less likely to be replaced than for bottom-decile publicly traded family-owned stock firms with nonfamily CEOs (row (7)) (3.683 percent vs. 5.964 percent, respectively). In addition, the percentage increase in turnover rates as performance deteriorates from the top to the bottom ROA decile is much larger for publicly traded than for closely held family-owned stocks with nonfamily CEOs. Hence, adding stock market monitoring increases the performance sensitivity of nonroutine turnover.

In the case of nonroutine turnover, when the ROA is in the bottom decile, the turnover likelihood is the lowest in family stock firms with a family-member CEO (0.065 percent for closely held and 0.149 percent for publicly traded firms) among all organizational forms and ownership structures. The differences between these bottom-decile turnover rates and the nonroutine turnover rates for all other types of firms are statistically significant at the 1 percent level. A comparison of row (11) (all family stock firms with family-member CEOs) and row (12) (all family stock firm with nonfamily CEOs) confirms this conclusion. These results support Hypothesis 2-3 that among common stock insurance companies, family-owned firms with family-member CEOs have the lowest likelihoods of nonroutine turnover.

We next analyze the likelihoods of nonroutine CEO turnover in the bottom deciles of ROA by organizational form and ownership structure. Mutuals have significantly lower nonroutine turnover probabilities than publicly traded nonfamily-owned stock firms (row (2), 3.155 percent, vs. row (8), 5.923 percent) when firm performance is at the bottom decile of ROA. Mutuals in the bottom decile also have significantly lower nonroutine turnover than nonfamily-owned closely held stock firms (row (3), 5.325 percent) and closely held family stock firms where the CEO is not a family member (row (5), 3.683 percent). However, mutuals have significantly higher bottom-decile nonroutine turnover than both closely held and publicly traded family stock firms with family-member CEOs (row (4), 0.065 percent and row 6, 0.149 percent, respectively). Thus, the turnover probability part of Hypothesis 3 is supported for insurance-owned stock firms with nonfamily CEOs. In comparison to mutuals, insurance-owned stocks are more likely to use nonroutine turnover to discipline management, unless the CEO is a family member. This provides evidence that mutuals have lower nonroutine turnover rates than stocks with nonfamily CEOs, even though mutuals have a higher fraction of outside members on the board than most types of stocks, and suggests that the stock organizational form provides more effective discipline than boards.

Overall, the results in Table 4 suggest that the sensitivity of nonroutine turnover to performance is generally consistent with the predicted pattern that poor performance increases the likelihood of turnover for all firm subsamples except where the CEO is a family member, where the turnover rates are trivial. The magnitude of the turnover sensitivity varies across organizational forms and ownership types in patterns generally consistent with the hypotheses.

CONCLUSIONS

A large body of research literature studies CEO turnover in publicly traded stock companies. Our article contributes to this literature by exploring the pattern of CEO turnover in both publicly traded and closely held stock insurance companies as well as mutual insurers. We subdivide both major types of stock firms into family-owned and nonfamily-owned firms, and we subdivide family-owned firms into those that do and do not have a family-member CEO. This study thus conducts a much more detailed analysis of organizational and ownership types than provided in the existing literature. This article provides evidence on how organizational forms and ownership structures affect corporate governance mechanisms in insurance companies. We find that the likelihood of CEO turnover is inversely related to firm performance. We also find that the CEO turnover rate and its sensitivity to performance vary by organizational form and ownership structure in patterns consistent with most of our hypotheses.

The main findings of this article are the following. First, we provide evidence that closely held nonfamily stocks have lower nonroutine turnover rates and lower turnover-performance sensitivity than publicly traded stocks, suggesting that stock market monitoring provides more effective managerial discipline than monitoring by a small group of owners. Second, on balance the evidence shows that poorly performing family-member CEOs in both closely held and publicly traded family stock firms are the most difficult to remove, suggesting that the controlling shareholders in such firms are entrenched. Third, nonfamily-member CEOs in publicly traded family stock firms have the highest nonroutine turnover-performance sensitivity. This suggests the effectiveness of monitoring of managers by both the capital market and controlling shareholders. A fourth finding is that poorly performing mutuals have lower likelihoods of CEO turnover than any other category of firm except closely held family stock firms with family or nonfamily CEOs and publicly traded family stock firms with family-member CEOs. This suggests that a higher proportion of outside members on the board in mutuals does not fully compensate for the lack of capital market monitoring and takeover threat.

The quality of an insurer's corporate governance mechanisms potentially affects the policyholders' rights because both owner-manager and owner-policyholder conflicts of interest exist in insurance companies. Our findings carry implications for insurance regulators regarding how to protect policyholders' rights by allocating more resources to monitoring firms with relatively weak corporate governance.

Unlike publicly traded stock firms, which are required to reveal certain information to the public, nonpublicly traded firms have more limited disclosure requirements. Similar studies on other industries would help to establish broader information about the patterns of CEO turnover in companies with various organizational forms and ownership structures. Additional research on the causal relationship between organizational forms and ownership structures and corporate governance mechanisms, including CEO turnover, would also be valuable. In addition, our research could be extended to consider the impact of CEO turnover on firm performance, adding to the existing insurance literature on the topic (He, Sommer, and Xie, 2011). Future research should also consider the relationship between CEO succession and organization/ownership type.

DOI: 10.1111/jori.12083

REFERENCES

Adams, J. C., S. A. Mansi, and T. Nishikawa, 2013, Public Versus Private Ownership and Fund Manager Turnover, Financial Management, 42(1): 127-154.

Ai, C., and E. C. Norton, 2003, Interaction Terms in Logit and Probit Model, Economic Letters, 80: 123-129.

Anderson, R., and D. Reeb, 2004, Board Composition: Balancing Family Influence in S&P 500 Firms, Administrative Science Quarterly, 49: 209-237.

Anderson, R. D., W. Reeb, and Zhou, 2012, Family-Controlled Firms and Informed Trading: Evidence From Short Sales, Journal of Finance, 67: 351-385.

Beatty, A. L., B. Ke, and K. R. Petroni, 2002, Earnings Management to Avoid Earnings Declines Across Publicly and Privately Held Banks, Accounting Review, 77: 547-570.

Bennedsen, M., K. M. Nielsen, F. Perez-Gonzalez, and D. Wolfenzon, 2007, Inside the Family Firm: The Role of Families in Succession Decisions and Performance, Quarterly Journal of Economics, 122: 647-691.

Berry, T. K., J. M. Bizjak, M. L. Lemmon, and L. Naveen, 2006, Organizational Complexity and CEO Labor Markets: Evidence From Diversified Firms, Journal of Corporate Finance, 12: 797-817.

Blackwell, D. W., J. A. Brickley, and M. S. Weisbach, 1994, Accounting Information and Internal Performance Evaluation, Journal of Accounting and Economics, 17: 331-358.

Borokhovich, K. A., R. Parrino, and T. Trapani, 1996, Outside Directors and CEO Selection, Journal of Financial and Quantitative Analysis, 31: 337-355.

Cannella, A. A., D. R. Fraser, and D. S. Lee, 1995, Firm Failure and Managerial Labor Markets: Evidence From Texas Banking, Journal of Financial Economics, 38:185-210.

Chen, X., Q. Cheng, and Z. Dai, 2013, Family Ownership and CEO Turnovers, Contemporary Accounting Research, 30: 1166-1190.

Coles, J. L., M. L. Lemmon, and L. Naveen, 2003, A Comparison of Profitability and CEO Turnover Sensitivity in Large Private and Public Firms, Working paper, Temple University, Philadelphia, PA.

Cornelli, F., Z. Kominek, and A. Ljungqvist, 2013, Monitoring Managers: Does It Matter? Journal of Finance, 68(2): 431-481.

Coughlan, A., and R. Schmidt, 1985, Executive Compensation, Management Turnover, and Firm Performance: An Empirical Investigation, Journal of Accounting and Economics, 7: 43-66.

Denis, D. J., and D. K. Denis, 1995, Performance Changes following Top Management Dismissals, Journal of Finance, 50: 1029-1057.

Denis, D. J., D. K. Denis, and A. Sarin, 1997, Ownership Structure and Top Executive Turnover, Journal of Financial Economics, 45: 193-221.

Ertugrul, M., and K. Krishnan, 2011, Can CEO Dismissals Be Proactive, Journal of Corporate Finance, 17: 134-151.

Evans, J. H., III, N. J. Nagarajan, and J. D. Schloetzer, 2010, CEO Turnover and Retention Light: Retaining Former CEOs on the Board, Journal of Accounting Research, 48: 1015-1047.

Fama, E. F., and M. C. Jensen, 1983a, Separation of Ownership and Control, Journal of Law and Economics, 26: 301-325.

Fama, E. F., and M. C. Jensen, 1983b, Agency Problems and Residual Claims, Journal of Law and Economics, 26: 327-349.

Fee, C. F., and C. J. Hadlock, 2004, Management Turnover Across the Corporate Hierarchy, Journal of Accounting and Economics, 37: 3-38.

Hansmann, H., 1985, The Organization of Insurance Companies: Mutual Versus Stock, Journal of Law, Economics, and Organization, 1: 125-153.

He, E., and D. W. Sommer, 2011, CEO Turnover and Ownership Structure: Evidence From the U.S. Property-Liability Insurance Industry, Journal of Risk and Insurance, 78: 673-701.

He, E., D. W. Sommer, and X. Xie, 2011, The Impact of CEO Turnover on Property-Liability Insurer Performance, Journal of Risk and Insurance, 78: 583-608.

Jensen, M. C., 1993, The Modern Industrial Revolution, Exit, and the Failure of Internal Control Systems, Journal of Finance, 68: 831-880.

Jenter, D., and F. Kanaan, 2014, CEO Turnover and Relative Performance Evaluation, Stanford University and National Bureau of Economic Research Working Paper, Stanford, CA.

Kang, J. K., and A. Shivdasani, 1995, Firm Performance, Corporate Governance, and Top Executive Turnover in Japan, Journal of Financial Economics, 38: 29-58.

Ke, B., K. Petroni, and A. Safieddine, 1999, Ownership Concentration and Sensitivity of Executive Pay to Accounting Performance Measures: Evidence From Publicly and Privately-Held Insurance Companies, Journal of Accounting and Economics, 28: 185-209.

Lausten, M., 2002, CEO Turnover, Firm Performance, and Corporate Governance: Empirical Evidence on Danish Firms, International Journal of Industrial Organization, 20: 391-414.

Laux, V., 2008, Board Independence and CEO Turnover, Journal of Accounting Research, 46: 137-171.

Li, F., and S. Srinivasan, 2011, Corporate Governance When Founders Are Directors, Journal of Financial Economics, 102: 454-469.

Mayers, D., A. Shivdasani, and C. W. Smith, 1997, Board Composition and Corporate Control: Evidence From the Insurance Industry, Journal of Business, 70: 33-62.

Mayers, D., and C. W. Smith, 1988, Ownership Structure Across Lines of PropertyCasualty Insurance, Journal of Law and Economics, 31: 351-378.

Mayers, D., and C. W. Smith, 1990, On the Corporate Demand for Insurance: Evidence From the Reinsurance Market, Journal of Business, 63: 19-40.

Mayers, D., and C. W. Smith, 1994, Managerial Discretion, Regulation, and Stock Insurer Ownership Structure, Journal of Risk and Insurance, 61: 638-655.

Morck, R., A. Shleifer, and R. W. Vishny, 1988, Management Ownership and Market Valuation: An Empirical Analysis, Journal of Financial Economics, 20: 293-315.

Murphy, K., and J. L. Zimmerman, 1993, Financial Performance Surrounding CEO Turnover, Journal of Accounting and Economics, 16: 273-316.

Nagar, V., K. Petroni, and D. Wolfenzon, 2011, Governance Problems in Closely Held Corporations, Journal of Financial and Quantitative Analysis, 46: 943-966.

Parrino, R., 1997, CEO Turnover and Outside Succession: A Cross-Sectional Analysis, Journal of Financial Economics, 46: 165-197.

Villalonga, B., and R. Amit, 2006, How Do Family Ownership, Control, and Management Affect Firm Value, Journal of Financial Economics, 80: 385-417.

Warner, J. B., R. L. Watts, and K. H. Wruck, 1988, Stock Prices and Top Management Changes, Journal of Financial Economics, 20: 461-492.

Weisbach, M. S., 1988, Outside Directors and CEO Turnover, Journal of Financial Economics, 20: 431-460.

Yermack, D., 1996, Higher Market Valuation of Companies With a Small Board of Directors, Journal of Financial Economics, 40: 185-211.

Jiang Cheng is Associate Professor at Shanghai University of Finance and Economics. J. David Cummins is the Joseph A. Boettner Professor of Risk Management and Insurance at Temple University. Tzuting Lin is Assistant Professor at National Taiwan University. The authors can be contacted via e-mail: cheng.jiang@shufe.edu.cn, cummins@temple.edu, and tzuting@ntu.edu.tw. All authors are Research Fellows at the Risk and Insurance Research Center, College of Commerce, National Chengchi University. We would like to thank Patricia Born, Georges Dionne, Elyas Elyasiani, Vidhan Goyal, Martin Grace, and participants at the 2013 AEA meeting, 2012 ARIA meeting, 2012 APRIA meeting, 2012 NTU International Conference on Finance, and seminars at National Taiwan University, Southwestern University of Finance and Economics, Tsinghua University, and Yuanze University for helpful comments and suggestions. We are also grateful to two anonymous reviewers for constructive comments, which helped to improve the article. This paper won the Harold Skipper Best Paper Award at 2012 APRIA meeting and the Best Paper Award at 2012 NTUICF meeting.

(1) While many top executive turnover studies focus on publicly traded industrial firms and generally exclude the highly regulated financial institutions, such as banks and insurance companies, others do not exclude financial institutions. For example, Warner, Watts, and Wruck (1988), Weisbach (1988), Denis, Denis, and Sarin (1997), and Parrino (1997) do not explicitly mention whether they exclude financial institutions from their sample. Fee and Hadlock (2004) and Kang and Shivdasani (1995) exclude banks, insurers, and financial holding companies. There is limited literature about top executive turnover that focuses only on financial institutions. One exception is Cannella, Fraser, and Lee (1995), who analyze a sample of failed and surviving Texas banks.

(2) Li and Srinivasan (2011) find that nonfamily-member CEO turnover sensitivity to performance is higher than in other firms when the founder of a family firm serves on the board of the company.

(3) Although managers of mutuals can be awarded bonuses for good firm performance, they cannot enjoy the wealth effects associated with increases in the value of equity ownership as can managers of publicly traded stocks.

(4) Fama and Jensen (1983b) suggest that the policyholders can surrender an insurance policy, an exercise of redemption rights that deprives managers of control of assets. However, the mutual P-C policyholder who surrenders a policy may receive a partial return of the premium for the current coverage period but receives no other compensation. Therefore, the threat of surrenders has little effect on managers of insurers.

(5) The attorney-in-fact in a reciprocal is a company or organization that is given power of attorney to act for the members of the reciprocal in conducting the business of insurance.

(6) We use the term "mutual" because mutuals are more numerous than reciprocals (4,073 mutual observations vs. 390 reciprocal observations in our sample).

(7) Other organizational forms and ownership structures exist in the insurance industry, such as Lloyds and risk retention groups (RRGs). We do not include Lloyds due to its relatively small role in the U.S. P-C insurance industry (Lloyds accounts for less than 1 percent of total P-C premiums). We do not study RRGs because they are small (less than 1 percent of total premiums) and are an atypical organizational form operating under a special federal law.

(8) A.M. Best Company, various years, Best's Insurance Reports: Property/Casualty Edition (Oldwick, NJ).

(9) The sample ends in 2006 because it was the most recent year available at the time the authors began the hand-collection of data from Best's Insurance Reports and proxy statements. The proxy statements are from www.sec.gov.

(10) The turnover rate of these foreign firms is 17.2 percent and most turnovers are nonroutine turnover (12.1 percent). However, the high nonroutine turnover rate might be due to foreign firm characteristics rather than performance. For example, a foreign firm might have a corporate policy to replace the CEO of its U.S. subsidiary company every 3 years. Further, the performance of the subsidiary of a multinational company partly depends on its parent company's global strategy, and profit transfers internally might also distort subsidiary profit reported on financial statements.

(11) Generally speaking, the CEO naturally steps down in a liquidation event because the firm no longer exists. In some cases, CEOs remain in place after the firm is conserved or rehabilitated but the firm is actually under regulatory control. In most cases, we cannot find information after the firm is announced to be under regulatory action. Thus, firms involved in regulatory actions are not included in the analysis.

(12) There are 350 M&A events and 226 cases are associated with CEO turnover. CEOs of acquired firms have a very high turnover rate of 64.6 percent, as expected. Interestingly, 38.5 percent of replaced CEOs remain on the board of the merged firm, suggesting that the new entity still relies on these former CEOs' experience to some extent.

(13) Blackwell, Brickley, and Weisbach (1994) argue that the holding-company board rather the boards of subsidiary banks plays a dominant role in deciding the retention or removal of subsidiary executives. They find evidence that it is quite common for one person to simultaneously hold several executive positions in different subsidiary banks. If we treat these subsidiaries as individual decision-making units, a single CEO turnover event for the group might be counted several times among its subsidiaries, biasing the analysis. We do not treat stock-owned mutuals and mutual-owned stocks as independent observation units since they are subsidiaries of an insurance group and share the same management with the lead company in the group. Further, it is arguably the ultimate organizational form and ownership structure of the controlling parent that determines the corporate governance mechanism.

(14) In a limited number of association related closely held stock firms and mutual insurance firms, the officer with CEO title is only responsible for the daily administration, and the president of the company, usually the president of the association, is the decision maker. In this case, however, we still code the CEO as top executive to be consistent. Results do not change if we code the president as top executive in these firms.

(15) Usually, Best's Insurance Reports and proxy statements have the names of company officers with the following titles: chairman of the board, president, CEO, senior vice president, secretary, CFO, vice president, and treasurer. All companies in our sample have at least one of the three titles: CEO, president, or chairman of the board.

(16) When we collected the management and board information from Best's Insurance Reports, we found that the management information and board information on an insurance company may not be updated at the same time. The update of board information may lag 1 year behind the management information. Thus, we define the turnover event that previous CEOs stay on the board for at least 2 years as routine turnover to avoid possible bias. We also search for the reason for CEO departure from the company's website and news articles on LexisNexis database and the Internet. This helps us to further identify whether the CEO departs the company because of normal retirement, death, health issues, or comparable appointment elsewhere. Such events are classified as routine turnover. As in prior research, our approach does not perfectly classify routine and nonroutine turnover.

(17) Some firms change organizational forms and ownership structures during the sample period, through mutualization, demutualization, or going public. There are 38 IPOs and 13 demutualizations during our sample period. Our definition of organizational forms and ownership types varies to reflect these changes. For example, if the firm is demutualized and conducts an IPO to become a public stock company in 2000, it is coded as a mutual before 2000 and as a public stock insurer after 2000. The results remain unchanged if we drop these firms from our sample.

(18) In this category, besides stock insurers owned by industrial parents, we have 15 firms owned by noninsurance financial firms, for example, American Express. These firms were not separately distinguished in our analysis because our tests showed that insurers owned by industrial firms behave similarly to those owned by noninsurance financial firms with respect to CEO turnover decisions.

(19) Book value performance measures are utilized because our sample includes many nontraded insurers. Our performance measures are after tax, paralleling the prior literature. We conducted robustness checks where pretax ROA and ROE are used instead, giving similar results.

(20) We also tested 2-year lagged values of performance measures as predictors for turnover with similar results.

(21) For some publicly traded firm-year observations where proxy statements are unavailable, we obtained board information from the proxy statement in the nearest available year. Weisbach (1988) indicates that this approximation is appropriate because board composition remains stable over time.

(22) Specifically, the variables are Herfindahl indices based on net premiums written for the NAIC lines of business and geographically for total net premiums written by state.

(23) Coles, Lemmon, and Naveen (2003) find that turnover rates for large private and public firms are 8.1 percent and 11.5 percent, respectively. He and Sommer (2011) find that turnover rates for mutuals and stocks in the U.S. P-C insurance industry are 9 percent and 19 percent, respectively. The extant literature focusing on publicly traded stock companies finds the following turnover rates: 7.8 percent in Weisbach (1988), 12.9 percent in Kang and Shivdasani (1995), 9.3 percent in Denis and Denis (1995), and 9.6 percent in Fee and Hadlock (2004). Most prior papers report total turnover rates rather than distinguishing between routine and nonroutine turnover, so most of the rates reported in this footnote are overall turnover rates.

(24) The results remain virtually unchanged if logit and multinomial logit are used instead of probit models. We also apply Cox semiparametric proportional hazard regressions to reestimate our main model (Table 3). The magnitude and statistical significance of the coefficients in the hazard model and multinomial probit model are roughly similar.

(25) Ai and Norton (2003) show that the coefficients of interaction terms in probit models are difficult to interpret. As a robustness test, we also conducted the regressions in columns (4) through (6) using probit models. The probit models support the same conclusions as the linear probability models.

(26) To avoid any problem with endogeneity of turnover and performance, we utilize lagged values of insurer performance in our turnover regressions. The lagged values of the performance variables are expected to be exogenous with respect to current-period turnover. Nevertheless, we also employ standard methods to test for endogeneity of the performance variables. Specifically, we conduct Hausman's specification test and an inverse Mill's ratio test. Both tests indicate that endogeneity between turnover and performance is not a concern in our study. We also conducted several robustness tests with alternative measures of firm performance, such as return on equity. The conclusions of the analysis are unchanged based upon the alternative measures. The robustness tests are discussed in an earlier working paper version of this paper available from the authors.

(27) He and Sommer (2011) also find positive (but insignificant) coefficients on ROA in their routine turnover regressions.

(28) As explained above, we combine mutuals and reciprocals in our tests. When we separate mutuals and reciprocals in our regressions, F-tests confirm that there is no significant difference between mutuals and reciprocals, and all other results remain virtually unchanged. The results also are virtually unchanged when we eliminate reciprocals.

(29) We recognize that firms might endogenously select their organizational form, presumably for reasons of efficiency. For example, the board of a mutual is more likely to be able to directly observe managerial performance without having to infer this performance from aggregate measures of firm performance because mutual insurers tend to be more geographically concentrated and operate in lines of business that require less managerial discretion. Alternatively, mutuals might consider policyholder service but not financial performance as their firm strategy objective. As a result, a mutual might exhibit a lower sensitivity of turnover to measures of firm performance even if the quality of internal governance is quite high. However, ownership structure is arguably a predetermined variable with respect to CEO turnover from an econometric perspective in this study due to the infrequency and difficulty of changing organizational form (Mayers and Smith, 1990).

(30) We also conducted robustness tests where we included a dummy variable for duality, that is, a dummy set equal to one when the CEO is also the chairman of the board and to zero otherwise. Because we did not have information on duality for all of the firms in the sample, the sample size was smaller in these regressions. The duality variable was never statistically significant and the results with the other variables were robust to the inclusion of this variable and the smaller sample size. Regressions with the duality variable are not reported in Table 3 to conserve sample size.

(31) We also conducted some additional robustness tests with regard to the board independence variable. These tests are reported in an earlier working paper version of this paper available from the authors.

(32) He and Sommer (2011) do not include an interaction of ROA and board independence in their regressions.

(33) He and Sommer (2011) find a significant positive relationship between firm size and nonroutine turnover. It is possible that size is partially proxied in our regressions by our breaking down the stock organizational form into categories, in view of the Table 1 finding that firm size differs significantly across the types of stock organizations.

(34) When we replace the log of board size with board size, all results remain virtually unchanged. We also tried adding the square of board size. The square of board size is never significant across columns in the Table 3 models, and all other results remain virtually unchanged.

(35) To further explore the possible role of regulation in CEO turnover, a dummy variable for the firm's risk-based capital (RBC) ratio falling below 250 percent is tested rather than leverage. The RBC variable also was not significant.

(36) The results on board characteristics, performance, and control variables are quantitatively similar to Table 3. The results are available from the authors.

(37) He and Sommer (2011) find the CEO turnover probability increases from 4.9 percent in the top decile to 8.6 percent in the bottom decile. Because they use companies rather than groups and unaffiliated single firms in their analysis, it is possible that some of the difference in turnover rates between the two studies results from double counting specific turnover events in He and Sommer's company-based sample.

Caption: FIGURE 1 Insurer's Organizational Form and Ownership Structure in This Study

TABLE 1
Descriptive Statistics of Turnover Type by Organizational
Form and Ownership Structure

                                      Number of Obs.
                                      (Unique Firms)   Turnover

Panel A: Sample Description

(1) Full sample                        8,755 (976)         572
(2) Mutual                             4,463 (439)         299
(3) Closely held nonfamily-owned         750 (89)           58
  stock firms
(4) Closely held family stock          1,723 (180)          49
  firms, CEO is a family member
(5) Closely held family stock            475 (63)           35
  firms, CEO is not a family member
(6) Publicly traded family stock         378 (55)           14
  firms, CEO is a family member
(7) Publicly traded family stock         183 (36)           26
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned      514 (86)           49
  stock firms
(9) Stock insurers owned by              269 (28)           42
  noninsurance holding companies
(10) All family-member CEO stocks      2,101 (235)          63
  (4 + 6)
(11) All nonfamily-member CEO          1,922 (274)         168
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Tests of Equality Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)
Row (3) versus row (5)
Row (3) versus row (8)
Row (7) versus row (8)
Row (10) versus row (11)

                                      Routine    Nonroutine
                                      Turnover    Turnover

Panel A: Sample Description

(1) Full sample                           329          243
(2) Mutual                                183          116
(3) Closely held nonfamily-owned           24           34
  stock firms
(4) Closely held family stock              39           10
  firms, CEO is a family member
(5) Closely held family stock              20           15
  firms, CEO is not a family member
(6) Publicly traded family stock           10            4
  firms, CEO is a family member
(7) Publicly traded family stock           15           11
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned        25           24
  stock firms
(9) Stock insurers owned by                13           29
  noninsurance holding companies
(10) All family-member CEO stocks          49           14
  (4 + 6)
(11) All nonfamily-member CEO              84           84
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Tests of Equality Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)
Row (3) versus row (5)
Row (3) versus row (8)
Row (7) versus row (8)
Row (10) versus row (11)

                                                         Routine
                                        Turnover        Turnover
                                          Rate            Rate

Panel A: Sample Description

(1) Full sample                              6.53%           3.76%
(2) Mutual                                   6.70%           4.10%
(3) Closely held nonfamily-owned             7.73%           3.20%
  stock firms
(4) Closely held family stock                2.84%           2.26%
  firms, CEO is a family member
(5) Closely held family stock                7.37%           4.21%
  firms, CEO is not a family member
(6) Publicly traded family stock             3.70%           2.65%
  firms, CEO is a family member
(7) Publicly traded family stock            14.21%           8.20%
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned          9.53%           4.86%
  stock firms
(9) Stock insurers owned by                 15.61%           4.83%
  noninsurance holding companies
(10) All family-member CEO stocks            3.00%           2.33%
  (4 + 6)
(11) All nonfamily-member CEO                8.74%           4.37%
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Tests of Equality Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)                 (0.017) **        (0.416)
Row (3) versus row (5)                   (0.815)         (0.355)
Row (3) versus row (8)                   (0.259)         (0.132)
Row (7) versus row (8)                 (0.080) *        (0.096) *
Row (10) versus row (11)              (<0.0001) ***   (0.0003) ***

                                       Nonroutine
                                        Turnover
                                          Rate

Panel A: Sample Description

(1) Full sample                              2.78%
(2) Mutual                                   2.60%
(3) Closely held nonfamily-owned             4.53%
  stock firms
(4) Closely held family stock                0.58%
  firms, CEO is a family member
(5) Closely held family stock                3.16%
  firms, CEO is not a family member
(6) Publicly traded family stock             1.06%
  firms, CEO is a family member
(7) Publicly traded family stock             6.01%
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned          4.67%
  stock firms
(9) Stock insurers owned by                 10.78%
  noninsurance holding companies
(10) All family-member CEO stocks            0.67%
  (4 + 6)
(11) All nonfamily-member CEO                4.37%
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Tests of Equality Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)                 (0.007) ***
Row (3) versus row (5)                   (0.232)
Row (3) versus row (8)                   (0.909)
Row (7) versus row (8)                   (0.476)
Row (10) versus row (11)              (<0.0001) ***

Notes: This table reports summary statistics for our sample during
1993-2006. The table presents the observation number of turnover
event, routine turnover event, and nonroutine turnover event, by
organizational form and ownership structures, respectively. We treat
turnover events where the CEO does not remain in the company as a
director or in another capacity for more than 2 years as nonroutine
turnover and all others as routine turnover. Turnover (routine
turnover, nonroutine turnover) rates are defined as the ratio of
turnover (routine turnover, nonroutine turnover) events to the number
of observations. Panel B compares the turnover rates between different
organizational forms and ownership structures. The p-values for
Kruskal-Wallis tests are reported in parentheses in Panel B. *, **,
and *** indicate statistical significance at the 10%, 5%, and 1%
levels, respectively.

TABLE 2
Descriptive Statistics (Medians) on Financial and Board
Characteristics

                                      Observations         ROA

Panel A: Sample Medians

(1) Full sample                          8,755            0.028
(2) Mutual                               4,463            0.025
(3) Closely held nonfamily-owned stock    750             0.028
(4) Closely held family stock            1,723            0.028
  firms, CEO is a family member
(5) Closely held family stock             475             0.035
  firms, CEO is not a family member
(6) Publicly traded family stock          378             0.037
  firms, CEO is a family member
(7) Publicly traded family stock          183             0.030
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned       514             0.029
  stock firms
(9) Stock insurers owned by               269             0.041
  noninsurance holding companies
(10) All family-member CEO stocks        2,101            0.029
  (4 + 6)
(11) All nonfamily-member CEO            1,922            0.030
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Test of Equality of Medians Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)                     --          (0.019) **
Row (3) versus row (5)                     --          (0.003) ***
Row (3) versus row (8)                     --            (0.996)
Row (7) versus row (8)                     --            (0.946)
Row (10) versus row (11)                   --            (0.425)

                                            Net
                                         Premiums         Leverage =
                                          Written        Liabilities/
                                        (millions)          Assets

Panel A: Sample Medians

(1) Full sample                            20.5              0.596
(2) Mutual                                 20.4              0.576
(3) Closely held nonfamily-owned stock     27.2              0.613
(4) Closely held family stock               7.3              0.568
  firms, CEO is a family member
(5) Closely held family stock              10.7              0.584
  firms, CEO is not a family member
(6) Publicly traded family stock           104.4             0.653
  firms, CEO is a family member
(7) Publicly traded family stock           260.8             0.657
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned        544.2             0.707
  stock firms
(9) Stock insurers owned by                26.3              0.586
  noninsurance holding companies
(10) All family-member CEO stocks          10.1              0.589
  (4 + 6)
(11) All nonfamily-member CEO              53.4              0.642
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Test of Equality of Medians Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)                 (<0.0001) ***     (<0.0001) ***
Row (3) versus row (5)                 (<0.0001) ***      (0.002) ***
Row (3) versus row (8)                 (<0.0001) ***     (<0.0001) ***
Row (7) versus row (8)                  (0.011) **       (<0.0001) ***
Row (10) versus row (11)               (<0.0001) ***     (<0.0001) ***

                                                          Herfindahl
                                        Herfindahl         Index of
                                      Index of Lines       States of
                                        of Business        Business

Panel A: Sample Medians

(1) Full sample                            0.473             0.934
(2) Mutual                                 0.372             1.000
(3) Closely held nonfamily-owned stock     0.623             0.841
(4) Closely held family stock              0.619             0.997
  firms, CEO is a family member
(5) Closely held family stock              0.608             0.951
  firms, CEO is not a family member
(6) Publicly traded family stock           0.378             0.178
  firms, CEO is a family member
(7) Publicly traded family stock           0.329             0.121
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned        0.265             0.111
  stock firms
(9) Stock insurers owned by                0.517             0.437
  noninsurance holding companies
(10) All family-member CEO stocks          0.580             0.892
  (4 + 6)
(11) All nonfamily-member CEO              0.502             0.454
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Test of Equality of Medians Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)                 (<0.0001) ***     (<0.0001) ***
Row (3) versus row (5)                    (0.345)        (<0.0001) ***
Row (3) versus row (8)                 (<0.0001) ***     (<0.0001) ***
Row (7) versus row (8)                    (0.111)           (0.260)
Row (10) versus row (11)               (<0.0001) ***     (<0.0001) ***

                                         % of NPW          % of NPW
                                           From         From Long-Tail
                                         Personal         Commercial
                                           Lines             Lines

Panel A: Sample Medians

(1) Full sample                            33.9               4.4
(2) Mutual                                 48.0               3.8
(3) Closely held nonfamily-owned stock      0.0              10.3
(4) Closely held family stock              13.4               6.3
  firms, CEO is a family member
(5) Closely held family stock              52.8               1.4
  firms, CEO is not a family member
(6) Publicly traded family stock           17.6              14.5
  firms, CEO is a family member
(7) Publicly traded family stock           31.9              27.2
  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned        20.6              14.9
  stock firms
(9) Stock insurers owned by                 0.0               4.9
  noninsurance holding companies
(10) All family-member CEO stocks          15.5               2.5
  (4 + 6)
(11) All nonfamily-member CEO              12.0              11.1
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Test of Equality of Medians Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)                 (<0.0001) ***     (<0.0001) ***
Row (3) versus row (5)                 (<0.0001) ***     (0.0009) ***
Row (3) versus row (8)                 (<0.0001) ***        (0.406)
Row (7) versus row (8)                  (0.006) ***       (0.001) ***
Row (10) versus row (11)               (0.0009) ***      (<0.0001) ***

                                           Board             Board
                                           Size          Independence

Panel A: Sample Medians

(1) Full sample                             9.0              0.667
(2) Mutual                                  9.0              0.733
(3) Closely held nonfamily-owned stock      9.0              0.778
(4) Closely held family stock               6.0              0.375
  firms, CEO is a family member
(5) Closely held family stock               5.0              0.250
  firms, CEO is not a family member
(6) Publicly traded family stock            9.0              0.594
  firms, CEO is a family member
(7) Publicly traded family stock            9.0              0.571

  firms, CEO is not a family member
(8) Publicly traded nonfamily-owned        10.0              0.750
  stock firms
(9) Stock insurers owned by                 7.0              0.571
  noninsurance holding companies
(10) All family-member CEO stocks           7.0              0.429
  (4 + 6)
(11) All nonfamily-member CEO               8.0              0.646
  stocks (3) + (5) + (7) + (8)

Panel B: Kruskal-Wallis Test of Equality of Medians Between
Organizational Forms and Ownership Structures

Row (2) versus row (8)                   (0.085) *        (0.018) **
Row (3) versus row (5)                 (<0.0001) ***     (<0.0001) ***
Row (3) versus row (8)                  (0.047) **          (0.221)
Row (7) versus row (8)                  (0.014) **       (<0.0001) ***
Row (10) versus row (11)               (<0.0001) ***     (<0.0001) ***

Notes: This table reports summary statistics (medians) for our sample
during 1993-2006. Board size is the number of directors on the board
of directors. Board independence is the proportion of independent
directors on the board of directors. ROA is the ratio of net income
after taxes and extraordinary items to total admitted assets and is
Winsorized at 1 and 99 percentile to remove excess effects of
outliers. Leverage is the ratio of liabilities to total admitted
assets. Herfindahl indexes of lines and states of business are
Herfindahl indexes of net premiums written by product line and by
state, respectively. Fraction of NPW from commercial long-tail lines
(workers' compensation, other liability, and commercial automobile
liability) is the proportion of net premiums written (NPW) in
long-tail lines to total NPW. Fraction of NPW from personal lines is
the proportion of NPW in personal lines (farmowners multiple peril,
homeowners multiple peril, automobile physical damage, and personal
automobile liability) to total NPW. Panel B compares the difference in
variable medians between different organizational forms and ownership
structures. The p-values for Kruskal-Wallis tests are reported in
parentheses in Panel B. and *** indicate statistical significance at
the 10%, 5%, and 1% levels, respectively.

TABLE 3
Multinomial Regression Results of Routine and Nonroutine Turnover

                                                          Routine
                                   [Turnover.sub.t]   [Turnover.sub.t]
                                         (1)                (2)

Intercept                             -1.820 ***         -2.871 ***
                                       (0.000)            (0.000)
[ROA.sub.t-1]                         -1.033 **            -0.596
                                       (0.027)            (0.440)
Median of industry [ROA.sub.t-1]       8.916 **          11.243 **
                                       (0.024)            (0.075)
Firm [size.sub.t-1] (Log of             0.017              0.029
  net premiums written)                (0.218)            (0.200)
[Leverage.sub.t-1]                      0.035              0.068
  (Liability/Assets)                   (0.790)            (0.758)
Herfindahl index of lines of            0.029              -0.205
  [business.sub.t-1]                   (0.710)            (0.131)
Herfindahl index of states of           -0.048             -0.163
  [business.sub.t-1]                   (0.471)            (0.139)
Fraction of NPW from personal           -0.018             -0.022
  [lines.sub.t-1]                      (0.789)            (0.842)
Fraction of NPW from                    -0.088             -0.068
  commercial long-tail                 (0.197)            (0.544)
  [lines.sub.t-1]
Log of board [size.sub.t]               0.067              0.131
                                       (0.283)            (0.237)
[(Dummy equal to one if               -0.229 **          -0.631 ***
  fraction of outside directors        (0.024)            (0.001)
  > 0.6).sub.t]
ROA*[(Dummy equal to one if
  fraction of outside directors
  > 0.6).sub.t]
Dummy for [mutual.sub.t-1]]             -0.135             0.047
                                       (0.106)            (0.761)
ROA*Dummy for [mutual.sub.t-1]

Dummy for closely held                  -0.070             -0.116
  nonfamily-owned stock                (0.495)            (0.567)
  [firms.sub.t-1]
ROA*Dummy for closely held
  nonfamily-owned stock
  [firms.sub.t-1]
Dummy for closely held family         -0.584 ***         -0.479 **
  stock firms, CEO is a family         (0.000)            (0.012)
  [member.sub.t-1]
ROA*Dummy for closely held
  family stock firms, CEO is a
  family [member.sub.t-1]
Dummy for closely held family           -0.121             -0.029
  stock firms, CEO is not a            (0.347)            (0.897)
  family [member.sub.t-1]
ROA*Dummy for closely held
  family stock firms, CEO is not
  a family [member.sub.t-1]
Dummy for publicly traded             -0.495 ***         -0.441 **
  family stock firms, CEO is a         (0.000)            (0.052)
  family [member.sub.t-1]
ROA*Dummy for publicly traded
  family stock firms, CEO is a
  family [member.sub.t-1]
Dummy for publicly traded               0.187              0.269
  family stock firms, CEO is not       (0.181)            (0.239)
  a family [member.sub.t-1]
ROA*Dummy for publicly traded
  family stock firms, CEO is not
  a family [member.sub.t-1]
Dummy for stock insurers owned        0.353 ***            0.156
  by noninsurance holding              (0.007)            (0.538)
  [companies.sub.t-1]
ROA*Dummy for stock insurers
  owned by noninsurance holding
  [companies.sub.t-1]
Sample size                             8,372              8,372
Turnover event counts                    545                315
Log likelihood/[R.sup.2]              -1,948.473         -2,297.259

                                      Nonroutine
                                   [Turnover.sub.t]   [Turnover.sub.t]
                                         (3)                (4)

Intercept                             -3.065 ***           0.024
                                       (0.000)            (0.546)
[ROA.sub.t-1]                         -2.264 ***           0.097
                                       (0.009)            (0.721)
Median of industry [ROA.sub.t-1]      12.437 **           1.147 **
                                       (0.092)            (0.016)
Firm [size.sub.t-1] (Log of             0.010              0.002
  net premiums written)                (0.708)            (0.284)
[Leverage.sub.t-1]                      -0.008             0.003
  (Liability/Assets)                   (0.973)            (0.884)
Herfindahl index of lines of           0.367 **            0.003
  [business.sub.t-1]                   (0.019)            (0.770)
Herfindahl index of states of           0.079              -0.005
  [business.sub.t-1]                   (0.557)            (0.563)
Fraction of NPW from personal           0.012              -0.002
  [lines.sub.t-1]                      (0.923)            (0.852)
Fraction of NPW from                    -0.186             -0.012
  commercial long-tail                 (0.164)            (0.183)
  [lines.sub.t-1]
Log of board [size.sub.t]               0.049              0.002
                                       (0.671)            (0.825)
[(Dummy equal to one if                 0.151              0.001
  fraction of outside directors        (0.458)            (0.886)
  > 0.6).sub.t]
ROA*[(Dummy equal to one if                                -0.117
  fraction of outside directors                           (0.364)
  > 0.6).sub.t]
Dummy for [mutual.sub.t-1]]           -0.476 ***           -0.015
                                       (0.002)            (0.328)
ROA*Dummy for [mutual.sub.t-1]                             -0.272
                                                          (0.304)
Dummy for closely held                  -0.128             -0.014
  nonfamily-owned stock                (0.452)            (0.417)
  [firms.sub.t-1]
ROA*Dummy for closely held                                 0.107
  nonfamily-owned stock                                   (0.736)
  [firms.sub.t-1]
Dummy for closely held family         -1.299 ***         -0.058 ***
  stock firms, CEO is a family         (0.000)            (0.001)
  [member.sub.t-1]
ROA*Dummy for closely held                                 -0.039
  family stock firms, CEO is a                            (0.890)
  family [member.sub.t-1]
Dummy for closely held family           -0.319             -0.002
  stock firms, CEO is not a            (0.154)            (0.939)
  family [member.sub.t-1]
ROA*Dummy for closely held                                 -0.359
  family stock firms, CEO is not                          (0.277)
  a family [member.sub.t-1]
Dummy for publicly traded             -1.037 ***         -0.053 **
  family stock firms, CEO is a         (0.001)            (0.013)
  family [member.sub.t-1]
ROA*Dummy for publicly traded                              -0.152
  family stock firms, CEO is a                            (0.671)
  family [member.sub.t-1]
Dummy for publicly traded               0.251             0.062 **
  family stock firms, CEO is not       (0.283)            (0.014)
  a family [member.sub.t-1]
ROA*Dummy for publicly traded                            -0.714 **
  family stock firms, CEO is not                          (0.092)
  a family [member.sub.t-1]
Dummy for stock insurers owned        0.666 ***          0.072 ***
  by noninsurance holding              (0.001)            (0.002)
  [companies.sub.t-1]
ROA*Dummy for stock insurers                               -0.053
  owned by noninsurance holding                           (0.877)
  [companies.sub.t-1]
Sample size                             8,372              8,372
Turnover event counts                    230                545
Log likelihood/[R.sup.2]              -2,297.259           0.015

                                       Routine           Nonroutine
                                   [Turnover.sub.t]   [Turnover.sub.t]
                                         (5)                (6)

Intercept                               -0.002             0.025
                                       (0.954)            (0.326)
[ROA.sub.t-1]                          0.437 **          -0.340 **
                                       (0.038)            (0.059)
Median of industry [ROA.sub.t-1]       0.653 **            0.493
                                       (0.076)            (0.117)
Firm [size.sub.t-1] (Log of             0.002              0.000
  net premiums written)                (0.253)            (0.779)
[Leverage.sub.t-1]                      0.001              0.002
  (Liability/Assets)                   (0.938)            (0.897)
Herfindahl index of lines of          -0.015 **           0.018 **
  [business.sub.t-1]                   (0.080)            (0.013)
Herfindahl index of states of           -0.010             0.005
  [business.sub.t-1]                   (0.160)            (0.442)
Fraction of NPW from personal           -0.000             -0.001
  [lines.sub.t-1]                      (0.958)            (0.826)
Fraction of NPW from                    -0.001           -0.011 **
  commercial long-tail                 (0.900)            (0.062)
  [lines.sub.t-1]
Log of board [size.sub.t]               0.002              -0.000
                                       (0.750)            (0.969)
[(Dummy equal to one if                 -0.009            0.010 **
  fraction of outside directors        (0.123)            (0.043)
  > 0.6).sub.t]
ROA*[(Dummy equal to one if             -0.095             -0.023
  fraction of outside directors        (0.344)            (0.791)
  > 0.6).sub.t]
Dummy for [mutual.sub.t-1]]            0.020 **          -0.034 **'
                                       (0.092)            (0.001)
ROA*Dummy for [mutual.sub.t-1]        -0.537 ***           0.265
                                       (0.009)            (0.130)
Dummy for closely held                  0.003              -0.018
  nonfamily-owned stock                (0.819)            (0.135)
  [firms.sub.t-1]
ROA*Dummy for closely held              -0.281            0.388 **
  nonfamily-owned stock                (0.251)            (0.063)
  [firms.sub.t-1]
Dummy for closely held family           -0.005           -0.053 ***
  stock firms, CEO is a family         (0.701)            (0.000)
  [member.sub.t-1]
ROA*Dummy for closely held              -0.358            0.319 **
  family stock firms, CEO is a         (0.106)            (0.092)
  family [member.sub.t-1]
Dummy for closely held family           0.024            -0.026 **
  stock firms, CEO is not a            (0.130)            (0.059)
  family [member.sub.t-1]
ROA*Dummy for closely held            -0.622 **            0.263
  family stock firms, CEO is not       (0.015)            (0.229)
  a family [member.sub.t-1]
Dummy for publicly traded               -0.010           -0.043 ***
  family stock firms, CEO is a         (0.552)            (0.002)
  family [member.sub.t-1]
ROA*Dummy for publicly traded           -0.369             0.216
  family stock firms, CEO is a         (0.184)            (0.362)
  family [member.sub.t-1]
Dummy for publicly traded               0.027             0.035 **
  family stock firms, CEO is not       (0.162)            (0.039)
  a family [member.sub.t-1]
ROA*Dummy for publicly traded           -0.183           -0.531 **
  family stock firms, CEO is not       (0.577)            (0.058)
  a family [member.sub.t-1]
Dummy for stock insurers owned          0.010            0.063 ***
  by noninsurance holding              (0.589)            (0.000)
  [companies.sub.t-1]
ROA*Dummy for stock insurers            -0.242             0.188
  owned by noninsurance holding        (0.366)            (0.409)
  [companies.sub.t-1]
Sample size                             8,372              8,372
Turnover event counts                    315                230
Log likelihood/[R.sup.2]                0.007              0.020

Notes: This table reports regression results for the sample period
during 1993-2006. Column (1) reports bivariate probit regression
results for turnover with no turnover as the base outcome. Columns
(2)-(3) provide multinomial probit regression results for the outcomes
routine turnover and nonroutine turnover, respectively, with no
turnover as the base outcome. The dependent variables are listed on
the top of the columns. Column (4)-(6) report the linear probability
models due to the presence of interaction effects. The dependent
variables are listed at the heads of the columns. In columns (4)-(6),
we interact return on assets (ROA) with dummy variables for ownership
forms and organizational structures. We omit the dummy for the
nonfamily-owned publicly traded firms, which serves as the base case.
The other independent variables are as defined in Tables 1 and 2. The
p-values are reported in parentheses below each coefficient estimate
using robust standard errors controlling for firm-level clustering. *,
**, and *** indicate statistical significance at the 10%, 5%, and 1%
levels, respectively.

TABLE 4
Implied Annual Turnover Probabilities Based on Probit
Regressions by Organizational Form and Ownership Structure

                                             Turnover
                                       Probability (Percent)

                                (1)            (2)            (3)

                                Top           Bottom
                               Decile         Decile
                               of ROA         of ROA      (2)-(1)/(1)

Panel A: Organizational Form and Ownership Structure

(1) Full sample               5.254          7.415         41.13 ***
(2) Mutual                    5.321          8.191         53.94 ***
(3) Nonfamily-owned           7.679          6.709        -12.63
  closely held stock
  firms
(4) Closely held family       1.878          2.036          8.41
  stock firms, CEO is a
  family member
(5) Closely held family       6.802          7.957         16.98
  stock firms, CEO is not
  a family member
(6) Publicly traded           1.946          4.027        106.94
  family stock firms, CEO
  is a family member
(7) Publicly traded           3.122         15.913        409.71 *
  family stock firms, CEO
  is not a family member
(8) Nonfamily-owned          10.278         12.031         17.06
  publicly traded stock
  firms
(9) Stock insurers           25.835         19.142        -25.91
  owned by noninsurance
  holding companies
(10) All stocks (3 + 4        4.304          5.842         35.73 *
  + 5 + 6 + 7 + 8)
(11) All family-member        1.879          2.218         18.04
  CEO stocks (4 + 6)
(12) All                      7.380         10.002         35.53 *
  nonfamily-member CEO
  stocks (3+5 + 7 + 8)

Panel B: f-Test of Equality Between Organizational Forms
and Ownership Structures

Row (2) versus row (5)      (28.53) ***     (4.94) ***         --
Row (2) versus row (8)      (83.25) ***    (72.86) ***         --
Row (3) versus row (5)       (6.19) ***    (10.81) ***         --
Row (3) versus row (8)      (16.34) ***    (41.35) ***         --
Row (5) versus row (7)      (16.32) ***    (25.30) ***         --
Row (7) versus row (8)      (26.70) ***    (11.58) ***         --

                                        Routine Turnover
                                      Probability (Percent)

                                (4)            (5)            (6)

                                Top           Bottom
                               Decile         Decile
                               of ROA         of ROA      (5)-(4)/(4)

Panel A: Organizational Form and Ownership Structure

(1) Full sample               3.373           3.846        14.02
(2) Mutual                    3.300           4.845        46.82 **
(3) Nonfamily-owned           3.877           1.167       -69.90
  closely held stock
  firms
(4) Closely held family       2.029           1.940        -4.39
  stock firms, CEO is a
  family member
(5) Closely held family       4.345           4.027        -7.32
  stock firms, CEO is not
  a family member
(6) Publicly traded           3.979           1.833       -53.93
  family stock firms, CEO
  is a family member
(7) Publicly traded           5.065           6.748        33.23
  family stock firms, CEO
  is not a family member
(8) Nonfamily-owned           3.881           3.391       -12.63
  publicly traded stock
  firms
(9) Stock insurers           20.755           0.795       -96.17 *
  owned by noninsurance
  holding companies
(10) All stocks (3 + 4        3.292           2.730       -17.07
  + 5 + 6 + 7 + 8)
(11) All family-member        2.133           1.720       -19.36
  CEO stocks (4 + 6)
(12) All                      4.515           3.581       -20.69
  nonfamily-member CEO
  stocks (3+5 + 7 + 8)

Panel B: f-Test of Equality Between Organizational Forms
and Ownership Structures

Row (2) versus row (5)      (24.15) ***    (22.02) ***         --
Row (2) versus row (8)      (13.18) ***    (37.96) ***         --
Row (3) versus row (5)       (3.43) ***    (39.04) ***         --
Row (3) versus row (8)       (0.03)        (29.04) ***         --
Row (5) versus row (7)       (2.96) ***    (10.96) ***         --
Row (7) versus row (8)       (4.80) ***    (13.28) ***         --

                                       Nonroutine Turnover
                                      Probability (Percent)

                                (7)            (8)            (9)

                                Top           Bottom
                               Decile         Decile
                               of ROA         of ROA      (8)-(7)/(7)

Panel A: Organizational Form and Ownership Structure

(1) Full sample              1.862           3.404          82.81 ***
(2) Mutual                   1.901           3.155          65.97 **
(3) Nonfamily-owned          3.791           5.325          40.46
  closely held stock
  firms
(4) Closely held family      0.045           0.065          44.44
  stock firms, CEO is a
  family member
(5) Closely held family      2.024           3.683          81.97
  stock firms, CEO is not
  a family member
(6) Publicly traded          6.22e-07        0.149           2.40e+05
  family stock firms, CEO
  is a family member
(7) Publicly traded          0.392           5.964        1421.42 **
  family stock firms, CEO
  is not a family member
(8) Nonfamily-owned          4.848           5.923          22.17
  publicly traded stock
  firms
(9) Stock insurers          11.958          19.245          60.94
  owned by noninsurance
  holding companies
(10) All stocks (3 + 4       1.189           2.772         133.14 ***
  + 5 + 6 + 7 + 8)
(11) All family-member       0.059           0.216         266.10
  CEO stocks (4 + 6)
(12) All                     3.022           6.100         101.85 ***
  nonfamily-member CEO
  stocks (3+5 + 7 + 8)

Panel B: f-Test of Equality Between Organizational Forms
and Ownership Structures

Row (2) versus row (5)       (4.09) ***    (15.19) ***         --
Row (2) versus row (8)      (69.29) ***    (74.95) ***         --
Row (3) versus row (5)      (19.36) ***    (15.42) ***         --
Row (3) versus row (8)       (8.96) ***     (5.39) ***         --
Row (5) versus row (7)      (14.85) ***    (12.98) ***         --
Row (7) versus row (8)      (23.96) ***     (0.22)             --

Notes: This table reports the economic significance for the sample
period during 1993-2006. Medians of other independent variables are
used in calculating the predicted turnover probability. Tests of the
equality of turnover possibility when ROA at 10 and 90 percentile are
conducted and results are given in column (3), (6), and (9). Panel B
compares the turnover possibility at 10 and 90 percentile of
performance of subsamples between different organizational form and
ownership structure based on hypotheses developed in section 4. The
t-statistics are reported in parentheses in Panel B. *, **, and ***
indicate statistical significance at the 10%, 5%, and 1% levels,
respectively.
COPYRIGHT 2017 American Risk and Insurance Association, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Cheng, Jiang; Cummins, J. David; Lin, Tzuting
Publication:Journal of Risk and Insurance
Geographic Code:1USA
Date:Mar 1, 2017
Words:16766
Previous Article:Portfolio optimization under solvency II: implicit constraints imposed by the market risk standard formula.
Next Article:Self-insurance with genetic testing tools.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters