مالکیت در برابر اثرات مدیریت بر عملکرد در شرکت های خانوادگی و موسس: مصالحه بیزین

We employ agency theory to argue that the effects of family (and founder) ownership versus management will be quite different: the former is expected to contribute positively to performance, the latter is argued to erode performance. Previous studies, due to problems of multicollinearity have been unable to distinguish these effects. Using a Bayesian approach that avoids these problems, we find that whereas family and founder ownership are associated with superior performance, the results for family and even founder management are more ambiguous. Our study is the first to assess the distinctive performance effects of family and founder presence in both ownership and management using a Bayesian approach.

Previous research on family firm performance has been rife with disagreements, with some studies finding family firms to outperform (e.g., Anderson and Reeb, 2003, Maury, 2006, McConaughy et al., 1998, Villalonga and Amit, 2006 and Weber et al., 2003), other studies finding them to underperform (e.g., Claessens et al., 2002, Holderness and Sheehan, 1988 and Morck and Yeung, 2004), and others still finding that family firms do not differ in their performance from their non-family cohorts (Miller, Le Breton-Miller, Lester, & Cannela, 2007). Certainly, there has been some discussion that different types of family or founder businesses perform differently, depending on the generation of the family involved (e.g., Bennedsen et al., 2007 and Pérez-González, 2006) and the number of family members (Miller et al., 2007). Founder generation firms and those with a single founder present are said to outperform other “family” businesses. Results also vary according to the samples (Fortune 500/1000 vs. S&P 500) and periods studied, and the indicators of performance – market to book ratio (or Tobin's q) versus return on assets. We believe that for both theoretical and methodological reasons, one very important source of the conflicting findings on the performance of family firms has been ignored: namely that family ownership and family management have very different, and perhaps even opposite effects on performance. The theoretical rationale is as follows. Agency theory would suggest that concentrated ownership is associated with lower agency costs. Specifically, the monitoring power and incentives of major owners reduces information asymmetries between owners and managers, minimizing the chances that managers will misallocate firm resources to serve selfish purposes (Demsetz, 1988 and Jensen and Meckling, 1976). Because family firms have concentrated ownership, they will benefit from lower agency costs and outperform. However, those advantages may be absent under – or even negated by – family management. Family managers may decide to use the resources of the business for the particular benefit of their own families, and in the process oppress smaller shareholders ( Morck et al., 2005 and Shleifer and Vishny, 1997). They may also be less competent than outsiders as they are relatively immune to dismissal for incompetence and come from a smaller selection pool ( Burkart et al., 2003, Claessens et al., 2002, Morck and Yeung, 2003 and Volpin, 2002). In short, there are good conceptual reasons for believing that family ownership and family management may exert opposing influences on performance. The same might be said for founder ownership and management, which may be why founders are often released after their firms go public ( Boeker and Karichalil, 2002 and Shleifer and Vishny, 1997). Our argument is not only of conceptual significance to the development of theory for family and founder firms but it also has practical governance implications regarding ownership and management of family and founder firms ( Chrisman, Chua, & Litz, 2004). Our second concern is methodological. Although prior studies have taken into account both ownership and management, they have been required to do so in separate predictive models given the high correlations among family ownership and management (usually between 0.5 and 0.7). Typically, the results for ownership and management have converged, we suspect because of that high correlation (Anderson and Reeb, 2003, Miller et al., 2007 and Villalonga and Amit, 2006). Unfortunately, there may well be a confounding effect at work here: any perceived advantage to family (or founder) management might actually be due to family (or founder) ownership. As we have argued above, there may not be any advantage to family management. In short, whereas family and founder ownership reduce agency costs, family and founder management may negate those advantages. Thus these two dimensions may have very different effects on performance, a vital consideration in establishing appropriate governance mechanisms, and one that is often neglected in research on entrepreneurship ( Chandler and Lyon, 2001 and Connelly et al., 2010). As noted, because of the high correlation between family/founder ownership and family/founder management, previous research, using conventional methods of regression, has confounded these influences. The present study will avoid that problem by using Bayesian regression analyses which are able to encompass ownership and management in the same models, without incurring problems of multicollinearity. Bayesian methods also have an advantage in that their results are not point estimates but entire distribution functions of the effects of interest. Bayesian analysis allows statements of likely and unlikely effects. Thus, we are able to specify the probability of family and founder management and family and founder ownership having positive effects rather than merely reporting a regression coefficient which is either significant or not. That way, Bayesian analysis avoids the danger of non-significant, but potentially important, results not being reported ( Cohen, 1994, Schmidt, 1996 and Starbuck, 2006). This study of S&P 500 companies between 1994 and 2003 shows that over 90% of the time, a higher level of family or founder ownership increases the firm's financial performance. By contrast, family managed companies are outperformed by their non family managed companies in over 50% of the cases, although founder CEOs do somewhat better than that. In short, family influences are Janus-faced: ownership effects are positive, management effects ambiguous, or even negative. Subsequent studies would be well advised to take those differences into account. We must stress that classical random- and fixed-effects models obscured these differences and only the Bayesian analyses were sufficiently fine-grained to make these distinctions. We shall first elaborate agency arguments for the performance advantages of family and founder ownership, and then argue for the performance disadvantages of family and founder management. Then we shall present our sample and methods, describing the particular appropriateness of Bayesian regressions in eliminating the multicollinearity problem that has affected all prior studies in this domain. We conclude by presenting and discussing our results and their implications.

The high correlations between family/founder ownership and family/founder management have made it impossible to assess the separate effects of each on performance using traditional regression techniques. Thus previous studies which have relied exclusively on these techniques have not been able to evaluate the predictions of agency theory (e.g., Anderson and Reeb, 2003, Miller et al., 2007 and Villalonga and Amit, 2006). This study has attempted to overcome that problem by employing a more robust Bayesian methodology. Our results not only support agency predictions related to the separation of ownership and control in both family and founder firms, but also inform the governance practices for these firms. Reasoning from agency theory, we hypothesized that family ownership and management would show diverging effects on corporate performance. Our findings indicate that family influences are indeed Janus-faced: family ownership shows a positive effect whereas family management has a neutral effect. Our results suggest that the mixed findings of earlier performance studies might be explained by the particular definitions of family businesses and the way performance effects were measured. Some studies defined family influence according to family ownership (Barth et al., 2005, Claessens et al., 2002 and Faccio and Lang, 2002), others have used family management presence (Bennedsen et al., 2007 and Fahlenbrach, 2009), while others still have employed both criteria (Anderson and Reeb, 2003 and Villalonga and Amit, 2006). Since family ownership and management have diverging firm performance effects, the way these variables are reflected in the definition of a family business can influence performance results. To date, two different but potentially biased approaches have been used to assess the performance effects of family ownership and management (e.g., Miller et al., 2007 and Villalonga and Amit, 2006). If one examines only family ownership or family management, or conducts separate regressions for either one, an omitted variables bias may occur. By contrast, if one combines both dimensions using traditional regression analyses, the effects attributed to family ownership and management may be biased as both indicators tend to be highly correlated. This problem of confounding ownership and management effects may account for the many conflicting findings among studies of family firm performance (compare Anderson and Reeb, 2003, Claessens et al., 2002, Holderness and Sheehan, 1988, Miller et al., 2007, Villalonga and Amit, 2006 and Weber et al., 2003). As we have argued, classical regression techniques that rely on null hypothesis testing cannot resolve these conflicts. Our Bayesian approach, however, represents an alternative that is robust to multicollinearity ( Hahn and Doh, 2006 and Leamer, 1973). Thus we were able to estimate the performance effects of family management and family ownership variables in a single regression with unbiased predictors. This approach was shown to lead to far more discriminating and informative results than that of classical regression. According to our results, the performance enhancing effect of family ownership is in line with positive agency expectations for family blockholders – parties argued to have information advantages, higher incentives for management control, and lower monitoring costs (Fama and Jensen, 1983 and Jensen and Meckling, 1976). Family owners have also been shown to pursue a long-term orientation and to focus on sustainable growth and profits (Miller & Le Breton-Miller, 2005). In our analyses, founder and family ownership contribute positively to firm performance. These findings significantly qualify earlier research by authors such as Anderson and Reeb (2003), Miller et al. (2007) and Villalonga and Amit (2006) who found that family-owned firms did not outperform other companies after the founder had left. We believe that is because this and indeed all previous research in the area confounds family ownership with the less salutary impact of family management. As to family management, our findings suggest at best a neutral but neither clearly inferior nor superior performance effect, again challenging much of the previous research. Some studies have argued that family management adversely affects firm performance due to nepotism, cronyism, the pursuit of socio-emotional wealth, and entrenchment problems. Family managers are said to lack skills and experience and to pursue private benefits and hyper-conservative strategies (Berrone et al., 2010, Block, in press, Bloom and Van Reenen, 2007, Claessens et al., 2002, Gomez-Mejia et al., 2007, Miller et al., 2010, Miller et al., 2011 and Morck et al., 2005). They may hire incompetent relatives or cronies for key positions (Bloom & Van Reenen, 2007), avoid strategic and product-market renewal (Block, 2010, Block, in press, Le Breton-Miller et al., 2011 and Miller et al., 2011), and diversify family wealth via inappropriate acquisitions (Miller et al., 2010). Cucculelli and Micucci (2008) reported that family successors underperform founders in previously high performing founder firms. Pérez-González (2006) too found that family-successor led firms significantly underperformed other firms – especially if family CEOs lacked a college education. Finally, Bloom and Van Reenen (2007) showed significant underperformance of family firms but only where the eldest son became CEO “by rule”. Other literature has suggested the opposite – that family managed firms may actually outperform other firms (Anderson and Reeb, 2003, Arregle et al., 2007, Miller and Le Breton-Miller, 2005, Ward, 2004 and Weber et al., 2003). Collectively, these studies found that family managed firms either underperformed or outperformed other companies – rather dogmatic conclusions perhaps motivated by a use of conventional statistical approaches. By contrast, our analysis shows a far more moderate outcome: the performance effects for family management do not reach the 90%/10% probability levels in Bayesian analysis – levels which would reflect an unambiguous effect on performance. The probability that family management will positively affect firm performance is only 43% or 34% (Table 4 and Table 5). In other words, family management is ambiguous rather than negative (or positive) in its impact. Our findings also inform the literature on founder-managed firms which again has produced mixed performance findings. Using conventional regression analysis, He (2008) reported that only founder-CEO status enhances firm performance whereas CEO ownership is detrimental. In contrast, Jayaraman, Khorana, Nelling, and Covin (2000) found no performance effect of founder-CEO status. Again, these mixed performance results may stem from multicollinearity that disguises the true effects of founder ownership and management. Our Bayesian results indicate that only founder ownership is generally associated with higher firm performance (94% or 78% probability of a positive effect). The effect of founder presence in management is less clear (74% or 58% probability of a positive effect). These more fine grained findings suggest that continued founder management does not unambiguously enhance firm performance – a finding that neither supports nor clearly contradicts studies arguing that a firm's growth overwhelms founder expertise ( Boeker and Karichalil, 2002, Wasserman, 2003 and Willard et al., 1992), and is also at odds with research advocating the replacement of founders with outside, “professional” managers ( Daily and Dalton, 1992 and Flamholtz, 1990). In a nutshell, our findings suggest that founders can pass on firm ownership to their families but should consider being succeeded in leadership by professional rather than family management. Bayesian statistics have allowed us to separate the effects of ownership and management. We believe that it is more important to place probabilities on outcomes of interest than to feel obliged to assert flatly that there is a positive, negative or “null” effect. The latter assertions often give rise to interminable pro-con debates, the former simply inform decision makers about the probabilities of focal outcomes. Our results consolidate previously mixed findings on the link between family (founder) involvement and firm performance and provide a foundation for further theory development and practice in the field of family business research (Chrisman et al., 2004 and Dyer, 2003). The Bayesian approach also holds promise for entrepreneurship research. Indeed, despite considerable progress in research methods in entrepreneurship (Chandler and Lyon, 2001 and Dean et al., 2007), almost all related research studies rely exclusively on classical regressions using null hypothesis significance testing (Connelly et al., 2010). This contrasts with developments in other disciplines (Hahn and Doh, 2006 and Rossi and Allenby, 2003), and can lead to misleading conclusions (Schwab et al., 2011). Indeed, many variables in entrepreneurship and family business research show significant inter-correlation, which is why it is often difficult to distinguish their effects. The Bayesian approach can overcome such problems.