Corporate Ownership, National Culture and
Firm Performance
Daniel Bogdanski
1Abstract
Building on the prevailing concept of separation of ownership and control, this study tests if certain cultural clusters, derived from grouping similar cultural characteristics, moderate the
relationship between managerial ownership and firm
performance. Most corporate governance literature presupposes that interests of principals and agents are naturally misaligned, ignoring that behavior and utility functions of individuals differ across the world and thus governance solutions may not be unidimensional. This study uses a sample of 15433 firms from 51 countries over a period of 8 years, yielding 75435 observations for the dependent variable, Tobin's Q. I tested several hypotheses
pertinent to the relation between ownership, performance and
national culture using panel data analysis while controlling for a multitude of firm- and country-level variables. Based on the
theory I expect that there are differences between country
clusters because of cultural differences. First, I did not find
support for the notion that managerially owned firms
significantly outperform outside owned firms in a world wide
sample. Second, I found strong support for the notion that the
function of the joint cultural values present in the Anglo-Saxon cluster (including the USA, UK, Australia etc.) positively moderates the relationship between managerial ownership and
firm performance. Third, I found strong support for the notion
that the function of the joint cultural values present in the
Confucian/Latin American cluster (China, South Korea, etc. plus South American) positively moderates the relationship between
managerial ownership and firm performance. Last, I found no
significant support for the notion that the function of the joint cultural values present in the Germanic cluster (Germany, Austria, Switzerland) negatively moderates the relationship between managerial ownership and firm performance. Consistent
with the models, I show that firm performance significantly
differs around the world, suggesting that national culture moderates the relationship between ownership and firm performance and that corporate governance should thus take national culture into account.
1Faculty of Economics and Business,
University of Groningen, the Netherlands
Correspondence: Daniel Bogdanski,
Noorderbinnensingel 59-8, 9712XE Groningen, the Netherlands; Student number: S2197375
Tel: +31 6 30 66 15 69
E-mail: d.h.k.h.r.w.bogdanski@student.rug.nl Final submission: June 13th, 2017
Keywords: Management, National Culture, Ownership, Firm Performance INTRODUCTION
effectively unaccountable to shareholders. For decades this question has triggered a tremendous amount of “managerialist” literature on accountability of managers (Baumol, 1959; Marris, 1964; Penrose, 1959; Williamson, 1964) and the separation of ownership and control in the field of corporate governance and corporate finance (Jensen and Meckling, 1976; Grossman and Hart, 1980). The image that Berle and Means painted has gradually loosened up; even within the original context of the United States its validity and precision regarding ownership distribution was questioned (Demsetz and Lehn, 1985; Shleifer and Vishny, 1986) and put into relation of how it affects firm performance (Morck et al., 1988). With aforementioned studies primarily focusing on the Anglo-Saxon background, a new level of complexity was introduced with the popularization of similar studies in other rich countries. Several of those revealed significant differences in ownership concentration and types in e.g. Germany (Edwards and Fischer, 1994; Gorton and Schmid, 1996), Italy (Barca, 1995), Japan (Prowse, 1994) or developing countries (La Porta et al., 1998). As La Porta et al. (1999) have clearly shown, there are large differences of ownership concentration and ownership types across countries.
With respect to these findings, the picture of rogue selfish managers with dictatorial control has clearly relaxed, however the core principle of managerial actions affecting firms has not. Perception and behavior of these agents affects firm performance and a similar pattern of country variant findings as for ownership distribution materialized in the academic literature. Empirical research finds the relationship between managerial ownership and firm performance in Anglo-Saxon countries to be s-curved (Morck et al., 1988; Short and Keasey, 1999), in Hong Kong the inverse of Anglo-Saxon findings (Ng, 2005), in Germany positive (Müller and Spitz, 2001) and in Singapore (Mak and Li, 2001) and Denmark (Rose, 2005) to be insignificant. Consequently, a picture emerges of conflicting results across countries, raising the question why no research has attempted to explain whether these differences in findings may depend on country specific effects moderating this relationship. Despite showing that national culture matters, its impact in finance has largely not been investigated (Greif, 1994; Stulz and Williamson, 2003) and while most empirical studies have been using US data, little corporate governance research has been done around the world (Shleifer and Vishny, 1997). Following this pattern, the Davis et al. (1997) argument that appropriate corporate governance mechanisms may not always be the same across different environments, shines in a new light. Neither separation of ownership and control nor CEO duality may be the single best solutions to maximize firm performance. A key question is thus whether national culture plays a role in the complex pursuit for firm performance, yielding the research question of this paper: Does national culture moderate the
relationship between firm performance and ownership?
This study uses a sample of 15433 firms from 51 countries over a period of 8 years, yielding 75435 observations for the dependent variable, Tobin's Q. I tested several hypotheses pertinent to the relation
between ownership, performance and national culture using panel data analysis while controlling for a multitude of firm- and country-level variables. Based on the theory I expect that there are differences between country clusters because of cultural differences. First, I did not find support for the notion that managerially owned firms significantly outperform outside owned firms in a world wide sample. Second, I found strong support for the notion that the function of the joint cultural values present in the Anglo-Saxon cluster (including the USA, UK, Australia etc.) positively moderates the relationship between managerial ownership and firm performance. Third, I found strong support for the notion that the function of the joint cultural values present in the Confucian/Latin American cluster (China, South Korea, etc. plus South American) positively moderates the relationship between managerial ownership and firm performance. Last, I found no significant support for the notion that the function of the joint cultural values present in the Germanic cluster (Germany, Austria, Switzerland) negatively moderates the relationship between managerial ownership and firm performance. Consistent with the models, I show that firm performance significantly differs around the world, suggesting that national culture moderates the relationship between ownership and firm performance and that corporate governance should thus take national culture into account.
To the best of my knowledge, there has been no other study attempting to find out how national culture moderates the effect of managerial ownership on firm performance in a world wide sample. While the bigger part of corporate governance research focuses on Anglo-Saxon countries, these findings may not be generalizable across the world. The findings of this study contribute to the literature in following ways; First, addressing the need for larger scale quantitative testing of national culture acting as a moderator (Licht, 2001). Second, theory building regarding the effect of national culture via joint cultural dimension characteristics as a moderator of ownership and firm performance, introducing elements from cultural research to better understand variation in firm performance around the world, addressing the large gap in the area of national culture and finance (Aggarwal and Goodell, 2014). Third, addressing the question more generally as to why firm valuations and ownership types vary around the world and what its implications are (Boyd and Solarino, 2016) by bringing different national cultures on a common denominator. Why are there, for example, patterns of higher valuations in one cluster and lower in another? What are the drivers of different managers’ behaviors in different environments? How are these related or is there a universal solution to corporate governance issues? To start theory development from a larger angle, cultural clusters provide a less arbitrary way of comparison for initial hypotheses testing. This study does not aim at giving a comprehensive solution to corporate ownership and governance issues around the world, but should provide intellectual nourishment about the importance of national culture in this framework and establish that national culture matters.
THEORY AND HYPOTHESES Firm Performance
First specified by James Tobin (1969), the Tobin's Q ratio serves as a predictor of investment profitability and is calculated by the ratio of (book value of total assets - book value of common equity + market value of common equity) to the book value of total assets. While the numerator represents the de facto market value of a firm, the denominator shows the book or replacement value leading to a ratio that indicates a direction which the market sees that investment moving towards. If the market value is above its replacement value, market pricing indicates an implicit or thus far unmeasured source of value that is not recorded in the book value of assets (Bharadwaj et al.,1999). This way of measuring firm performance opens the doors to further interpretation relevant to this study. Since the default assumption of Q is long-run parity, deviations indicate how the market perceives the value of assets and their profitability in the long-run; A Tobin's Q above 1 implies that the market is assuming increasing value and is therefore an indication for good firm performance, good managers, growth opportunities or monopoly power (Lindenberg and Ross, 1981), potentially also capturing market hype or speculation. A Tobin's Q below 1 indicates that the market believes that the company’s assets will not yield a sufficient return, punishing that firm with market valuation lower than replacement cost, indicating poor firm performance. Furthermore, it is important to note that value does not stem from profits or positive cash flows, but mobilization of resources in less efficient markets yielding competitive advantages (Hamberg, 2016), which is what drives shareholder actions and therefore Tobin's Q. A strong advantage of Tobin's Q is circumventing the difficulties of other firm performance measures based on e.g. marginal cost or rate of return that are hard to collect and may yield biased estimates in certain industries (Bharadwaj et al., 1999).
From a corporate governance perspective Tobin's Q combines the “what is” and “what may be” into one measure, recognizing the underlying idea of firm performance, namely creating shareholder wealth and their satisfaction. Tobin's Q has been the most utilized proxy for essentially all influential papers researching the effect of firm performance, especially in the managerial ownership and corporate governance arena. That includes the seminal work of La Porta et al. (2001) and Morck et al. (1988).
Ownership
As defined by La Porta et al. (1999), ownership in the context of this study is interpreted as voting rights rather than cash flow rights, which follows the fundamental idea of Berle and Means (1932), who wanted to know who controls modern corporations, managers or shareholders. This study too
wants to find out who performs better, based on factual ownership. The original argument of Berle and Means was that with separation of ownership and control, shareholders in principle can not control the agent anymore, who is supposed to act on behalf of the shareholders, because ownership of capital is dispersed among small shareholders, which coins the term “widely held”. While this finding was groundbreaking at the time, La Porta et al. (1999) found significant differences within the international sample of countries they studied, showing that ownership dispersion differs around the world.
The findings of La Porta et al. (1999) are of particular importance for theorizing the impact of managerial behavior. Before this contribution, ownership research revolved around and originated from Anglo-Saxon roots. In this framework, Fama and Jensen (1983) and Demsetz (1983) theorized that market discipline forces a manager towards value maximization via product market (Hart, 1983) labor market (Fama, 1980), and corporate control (Jensen and Ruback, 1983), even with small stakes in the company. If a manager, however, controls a substantial fraction of voting rights, he or she may abuse this power for actions that do not maximize value for the firm, building the entrenchment hypothesis. The same logic accounts for scenarios where ownership is so widely dispersed that shareholders are unable to supervise, effectively giving management full control. The convergence of interest hypothesis, on the other hand, suggests that market value should increase with increasing managerial ownership because of the stronger alignment with the interests of the company and the larger individual drawbacks when squandering with corporate wealth (Jensen and Meckling, 1976). Nevertheless, theoretical arguments can apparently not perfectly predict how ownership relates to firm performance (Morck et al., 1988); the behavior that is responsible for the actions of an agent is determined by individual preferences.
Prominent research, however, suggests an empirical relation between managerial ownership and firm performance (Barnhart and Rosenstein, 1998; Stulz, 1988). Morck et al. (1988) found an s-curve relationship where Tobin's Q rises from 0 to 5%, falls from 5 to 25% and continues to rise from a 25% threshold in US firms. In another Anglo-Saxon study, Short and Keasey (1999) found similar results for large firms from the UK. The opposite is found by a study conducted in Hong Kong (Ng, 2005). A German study (Müller and Spitz, 2001) found a different shape of the relationship, where performance only rises up to a threshold of 40% and then simply flattens, eliminating an entrenchment effect. Other studies from Singapore (Mak and Li, 2001) and Denmark (Rose, 2005) found no relationship of managerial ownership on firm performance. Fan and Wong (2002) and Ng (2005) thus argue that results from the Anglo-Saxon background are not applicable in the East-Asian background. Regarding the causality of findings, it must be noted that since the size of equity holdings are decisions made by boards and managers, correlations must not be interpreted as causal (Frydman and Jenter, 2010). Empirical research of the relationship between managerial ownership and firm performance is at a
point comparable to where La Porta et al. commenced in 1999 in the field of ownership dispersion, following from the one-dimensional approach addressing only large Anglo-Saxon companies (Morck et al., 1988; Short and Keasey, 1999). Sufficient international research regarding the impact of managerial ownership and behavior on firm performance is lacking, exposing the stunted bedrock upon which corporate governance solutions are developed outside of the US. Despite Eisenhardt’s (1989) early suggestion of exploring ownership problems outside the Anglo-Saxon perspective, Boyd and Solarino (2016) observe that the field has progressed beyond this perspective only to a limited extent.
To establish a benchmark for the following hypotheses and to put these into perspective, I measure the overall unmoderated relationship between managerial ownership and firm performance to find out whether there is a universal relationship across clusters between the two. An issue with comparing literature from different clusters and countries is that authors can only use theory to derive a potential conclusion about comparability and deviations in results, creating a significant gap due to lack of empirical findings. Yet as Morck et al. (1988) already indicated, theory itself can not adequately predict a relationship. This problematic is emphasized more strongly in quantitative studies where the data as a common denominator (here Tobin's Q) is available, but, as stated, not directly put into relation, because after all they are not part of the same sample and methodological strategies. To understand and comprehend the significance of the forces at work, I construct an a priori idealtypus (ideal-type; Weber, 1920), which is an abstract and hypothetical concept of a manager (or cluster later on). By highlighting the intrinsic values, norms and perceptions inherent to the managers, it intends to model an exaggerated paradigm and not perfectly represent an average individual. The idealtypus of a manager, following the theory of Jensen and Meckling (1976), acts in constant self-interest. Firm performance consequently is a result of the alignments of interest, leading to better firm performance the more a manager owns of the firm. This idealtypus is the yardstick of the following hypothesis. Expressed in dichotomous terms, I formally write:
Hypothesis 1: Managerially owned firms perform better than non-managerially-owned firms
National Culture
Precisely defining national culture has been an arduous task. The consensus is that key are shared understandings that manifest in artifact and actions (Franke et al. 1991; Kirkman et al., 2006). National culture includes behavioral patterns, values, norms and beliefs of a national group (Leung et al., 2005), does not converge and is stable over time (Beugelsdijk et al., 2006; Tabellini, 2010) stemming from the “culturalist perspective”, which states that values are rooted deeply within socioeconomic developments and history instead of vice versa (Beugelsdijk and Maseland, 2010;
Huntington, 1997). Values reflect “a broad tendency to prefer certain states of affairs over others” (Hofstede, 1980: 19), engaging judgement by the individual who carries those values, whose attitudes and beliefs deem modes of conduct desirable or not (Rokeach, 1973). Hofstede (2001: 9) defines culture as “the collective programming of the mind that distinguishes the members of one group or category of people from another.” This programming is not forced, but conditioned through shared history, religion, language and so forth. These differences among cultures are the most powerful force that divides the human race (Huntington, 1997), are key determinants relating to expansions of multinational enterprises and correlate with capital structure decisions as well as access to finance (Bartlett and Ghoshal, 2003; Chui et al., 2002; Li et al., 2011; Siegel et al., 2011). Hence, it is likely that financial and business decisions are influenced by national culture (Aggarwal et al., 2016).
To measure such national culture, cultural dimensions emerged as the most utilized form of quantification (Aggarwal and Goodell, 2014; Aggarwal et al., 2016). One of the first to consistently and parsimoniously quantify national culture in dimensions was Geert Hofstede (1980, 2001), whose taxonomy of national culture is also the most extensively used (Beugelsdijk et al., 2017; Kirkman et al., 2006). Consequently, to increase comparability to prior research, his dimensions will be used in this study. He originally constructed the framework from attitudinal surveys in IBM subsidiaries across 72 countries in the 1970s. The five dimensions that are used in this study are; Firstly, the level of Power Distance (PDI), which denotes the level of acceptance of unequally distributed power; Secondly, the level of Uncertainty Avoidance (UAI), which denotes the level to which people are comfortable in situations that are unstructured and unpredictable; Thirdly, the level of Individualism (IDV), which denotes the level to which societies see individuals as collective members of closely knit communities or where individuals are rather looking after themselves; Fourthly, Masculinity (MAS), which denotes the level to which people care rather for success and achievement versus emphasizing quality of life, solidarity and caring for others; Lastly, Long-Term Orientation represents a future orientation, betoken the acceptance of a figurative delayed gift reception.
A pivotal issue in the literature is treating cultural dimensions as a unidimensional construct, which can not be isolated without consequences. As Hofstede (2001) suggested, a country needs to be viewed as a whole. One cannot simply pick one trait out without taking the others into account; cultural dimensions are not entirely independent and individual dimensions are elements that reflect only part of the taxonomy of a country as a whole. One may look at it as baking a cake, where the individual ingredients do not reflect the end product, yet the mixture of these ingredients to differing amounts are what is creating the end result. Simply isolating the individual amounts of flour that go into the cake and comparing them with each other does not allow me to draw conclusions about how the cake tastes, because also e.g. the amounts of sugar or the size may differ, creating a completely different end product. Nevertheless, most literature uses isolated national cultural dimensions
individually, raising severe construct validity concerns. This is an important aspect to be considered when reviewing findings about national culture. Using the dimension of individualism, Chui et al. (2010) find a positive relation to profit volume and volatility from momentum strategies. Using the dimension of uncertainty avoidance, Aggarwal and Goodell (2009) find a negative relation to transaction costs and Frijns et al. (2013) find an association with lower M&A across borders and higher takeover premiums. Using the dimension of masculinity, uncertainty avoidance and power distance, Zheng et al. (2012) find a positive association with usage of short-term debt.
Drawing on the notion of collective programming of minds (Hofstede, 2009), reflecting the function of the joint characteristics of underlying cultural values that are present within a certain framework, this study uses country clustering. Clustering emanates from the “families of nations” concept in sociology, law and political science literature (Castles and Flood, 1991; Glendon, 1987) and can guide cross-cultural sampling to investigate potential boundary conditions for theory testing (Gupta et al., 2002). It is the ordering of entities together that are relatively similar, resulting in a relatively homogeneous set of relatable entities, which helps avoiding to measure cultural distance and applying sampling biases (Ronen and Shenkar, 1985, 2013). The idea of clusters is to mitigate opportunism and uncertainty and not to perfectly represent their member countries on average, but embedding key elements of in-cluster commonalities surrounding institutional pillars (Scott, 2001). To arrive at cultural commonality clusters, Ronen and Shenkar (2013) employed ecocultural variables (religion, geography and language) as well as ten major studies based on raw data and the dimensions utilized in those, most notably included is the seminal work of Geert Hofstede (2001) and the GLOBE study (House et al., 2004).
In the following I develop hypotheses on the relation between ownership and performance (H1), and how national culture moderates this relationship. Based on theoretical arguments I distinguish between several national culture clusters which I expect to affect the relation between ownership and performance differently. Specifically, I argue that firm performance has a stronger positive relationship in managerially owned Anglo-Saxon and Confucian/Latin American firms and a negative relationship in managerially owned Germanic firms as each compared to all other firms. I describe the clusters that are used in this study and develop hypotheses based on their national cultural character as a whole, this character interacting with ownership is argued to be better or worse in certain clusters. The descriptions include the countries that are within the cluster and how they fit together. Following this groundwork is an explanation of the traits that are most prevalent within the cluster, intending to provide a deeper understanding of what makes this cluster special. The argument as to why one country or cluster would behave and thus perform differently to all others is a function of the joint characteristics of underlying cultural values that are present within a certain framework.
Anglo-Saxon Cluster
The Anglo-Saxon cluster includes the United States, United Kingdom, New Zealand, Australia, Ireland and Canada. While some studies also include white South Africans, this line is hard to draw with firm data and is thus excluded from this study. Despite the extreme geographical distribution of the countries, this cluster is a very cohesive one, showing a value of 0.07 (Ronen and Shenkar, 2013). The cohesiveness measure indicates how close-knit the countries cultures are from 0 (strong) to 1 (weak). Therefore, except for excluding white South Africans, there is theoretically no need to break this cluster up, because of their shared history, religion, language and still stable behavioral characteristics. Distinguishing institutional characteristics in all Anglo-Saxon countries are market-based financial systems as well as common law, defining rights for the three most involved actors: management, shareholders and directors, owing to the strong equity finance market. Overall, strong equity-based financial systems are argued to be home in more risk tolerant countries (Hofstede, 2003). These features partly stem from the embedded cultural orientations of individuality, doing, assertiveness and indulgence. On the other side, the Anglo-Saxon cluster is characterized with low power distance and uncertainty avoidance compared to others. The Anglo-Saxon cluster can again be illustrated by an a priori idealtypus (ideal-type; Weber, 1920) that highlights the intrinsic values, norms and perceptions inherent to the joint value characteristics present in the cluster. This idealtypus is one that acts as an individual striving for individual success in the sense that if everyone cares for themselves then everyone is cared for. Market forces and “moral sentiment” induce commonly accepted “fairness” where “one reaps what one sows” (Smith, 1761) and individual achievement depicts what makes other people respect the individual, creating the socially legitimized way of behaving. These goals are generally perceived as more important than collective bonds or family and in turn are the glue that keep people together. Thisidealtypus is self-focused, superficial, risk-tolerant, hard working, independent, present-oriented and materialistic.
By accepting the precondition as a society that individuals striving for success are the norm, appropriately aligning goals is the burden that needs to be overcome between contractual parties, to mitigate a conflict of interest and risk preferences, when the manager is no (longer) entrepreneur (Boyd and Solarino, 2016). Most literature addresses this interest alignment dilemma with agency theory. Such a relationship is defined as a contract where the principal (as in a company) engages an agent (as in a manager) by authorizing him or her to perform a service (in terms of decision making) on behalf of that principal, or as Shleifer and Vishny (1997) call corporate governance mechanisms: “how investors get the managers to give them back their money”. If both parties intend to maximize their own utility, it is likely that the agent is not always acting in a principal's best interest, creating agency cost. These cost are the sum of bonding costs by the agent, monitoring costs by the principal
and the residual loss resulting from the divergence of optimal welfare improving decisions and actual decisions made by the agent (Jensen and Meckling, 1976). In other words, principal and agent are separate parties with diverging interests that are aligned via mechanisms, which pose a cost in themselves plus the costs of suboptimal actions by the actor. Thus, agency theory tends to be based on the assumption of individualistic and opportunistic types of homines economici, where the parties involved are merely self-serving and goals naturally misaligned (Davis et al., 1997). Owing to the threat of being replaced, even outside of their control (Jenter and Kanaan, 2015), Shleifer and Vishny (1989) argue that managers may engage in entrenching actions that makes them more valuable to the firm and more costly to replace by diverting shareholder funds into acquisitions or strategies that are more valuable with that manager in place. These actions are only necessary when the manager faces the threat of dismissal and is especially critical for managers with low stakes (Shleifer and Vishny, 1989). If, to ensure their own power managers entrench themselves, the organization they manage is likely losing sight of its competitive environment, which threatens performance in the long term (Walsh and Seward, 1990). In this scenario, managers build their own utility at the expense of the principal's utility (Jensen and Meckling, 1976). Opportunities to engage in such actions are diverse. While some managers cinch their position by foregoing risky projects (Fama, 1980; Lambert, 1986) to prevent failure and save face, others may engage in empire building (Jensen, 1986) for private benefits. These and similar actions actively destroy shareholder value and hinder performance and growth (Shleifer and Vishny, 1989). In terms of the agency cost equation, the threat of dismissal and potentially resulting entrenchment of the manager negatively impacts firm performance by posing a cost through control mechanisms and residual loss (Walsh and Seward, 1990).
In this Anglo-Saxon environment high individualism and low long-term orientation are the drivers of success, compared to all other clusters, subject to the condition that a mechanism is in place that aligns the individual goals of principal and agent. I overall argue that the function of these joint cultural value characteristics moderate firm performance of managerially owned Anglo-Saxon firms compared to all others in following directions; First, by dealing with “other people’s money”, equity financing induces a risk-function and may lead to myopic actions of managers as many shareholders are unable to directly control behavior of managers. In such individualistic countries, these individuals are more aggressive in their drive for performance as the outcome is more directly attributed to an individual (Beugelsdijk and Frijns, 2010). This should lead to a positive moderation of firm performance when natural alignment (managerially owned firms) and cultural traits interact. Second, the particular nexus of cultural traits within the Anglo-Saxon cluster of high individualism, low uncertainty avoidance and low power distance leads to individual pursuit of wealth and success. This cultural nexus positively moderates success of the firm if goals are aligned (in managerially owned firms). Yet this is potentially only the case in relation to each other, because otherwise the
combination of values rewards successes, which should let the Anglo-Saxon cluster outperform other clusters in general. Third, the aforementioned prevailing corporate governance theory, agency theory, stems from the Anglo-Saxon background, suggesting that agency problems may be pertinent to the Anglo-Saxon cultural background. Formally, I write:
Hypothesis 2: The relationship between ownership and performance as hypothesized under H1 is
moderated by national culture, such that firm performance has a stronger positive relationship in managerially owned Anglo-Saxon firms as compared to all other firms.
Chinese Confucian/Latin American Cluster
In the following I group two clusters together that theoretically match in their perception of what is important and how they structure their utility function, these are Confucian Chinese and Latin American countries. The Latin American countries include Colombia, Ecuador, Mexico, Venezuela, Argentina, Bolivia, El Salvador, Chile, Uruguay, Peru, Costa Rica, Guatemala and Brazil. This cluster is despite a wide range of countries with varying political systems and spanning the two continents of Middle- and South-America, relatively cohesive, showing a value of 0.40 (Ronen and Shenkar, 2013). Albeit the size, within this cluster only two relatively similar languages are prevalent, Spanish and Portuguese. Latin American countries set themselves apart with high scores for collectivism and power distance as well as low scores for uncertainty avoidance and doing orientation. The Confucian cluster, as defined by Ronen and Shenkar (2013) includes China, Singapore, Taiwan, Hong Kong, South Korea and Japan. This cluster emphasizes collectivism and power distance strongly. Yet this type of national culture has a specific way of dealing within its collective, by emphasizing power distance through relationships and hierarchies and their resulting perception of family. The combination of Confucian and Latin American culture can again be illustrated by an a priori
idealtypus that highlights the intrinsic values, norms and perceptions inherent to the joint value
characteristics present in the clusters. The Confucian/Latin American idealtypus is one that acts in groups, in the sense that people strive for a collective, including a clear sense of power distance. This
idealtypus is a family-oriented, recognition- and group-seeking subordinate accepting authority until
having worked the way up by seniority. Work is done as one is told and not necessarily how one perceives it to be correct, thinking is done and rules are set by the leader. This is strictly obeyed. Success is defined over group cohesiveness, relationships and particularly family, which is of supreme importance.
The idealtypus and its implied utility function presupposes a different motivation than agency theory where the agent is not the problem, but the solution, called steward, who values collective and thus organizational success higher than individual ones. One may compare a steward with a father,
who assumes to be accepted as the person in charge and expects to be listened to and followed. By accepting these preconditions, the father is willing to do his best to achieve a maximum collective result by raising his child and catering for it, naturally aligning goals. Based on this notion, research in sociology and psychology suggests that there are limits about the assumptions regarding individual utility in agency theory (Perrow, 1986; Hirsch et al., 1987). Opposing the idea of agency theory, stewardship theory argues against the inherent theoretical claim that the interests of principals and agents are naturally misaligned (Jensen and Meckling, 1976) and outlines principal-agent relationships based on behavioral premises, in which the utility function is valuing collectivistic and pro-organizational outcomes higher than self-serving and individualistic ones (Donaldson and Davis, 1989, 1991). On the one hand, the authors argue that there are non-financial motivators such as recognition or need to achieve. On the other hand, they argue that if managers feel that their financial interests are bound to the company, interests are automatically aligned. A steward believes his or her goals are met when the organizational goals are met. Under this assumption of motivated managers, firm performance depends on whether the executives are able to implement their ideas, empowering and facilitating action by the executives, because their actions are company centered and collectivistic either way. Hence, not motivation is the issue adversely affecting firm performance, but structural conditions. As a result, Donaldson and Davis (1991) found that firms with CEO duality (executives simultaneously being chairman of the board) are performing better than separation of executive and board, using ROE to measure firm performance. This raises the question why there is not always a stewardship relationship, which Davis et al. (1997) answer with the risk-profile of principals and their perception of whether the agent is self-serving or views alignment with the organization as the best tool to reach personal goals as a collective. This suggestion presupposes an inherent relation of an agent's personal characteristics with the appropriate governance solution to maximize firm performance.
Drawing on the notion of seeing stewardship as analogous to a father, firms in these countries should strive under a steward leadership that fits their cultural pattern. In this environment, high power distance and collectivism are the drivers of success, subject to the condition that a stewardship environment is in place, where a manager can act as a steward and is not constrained by limitations that question his or her authority and the ability to cater for his or her employees. Gorodnichenko and Roland (2010) argue that such a collectivistic environment facilitates collective action, which is what drives the performance. Note that the argument is only managerial performance stands out, not the cluster overall. As opposed to the Anglo-Saxon cluster, the inherent utility function in this cluster is not trimmed to succeed overall, by discouraging to stand out as an individual and emphasizing collective success. This argument also constitutes the striking difference between the agency and stewardship argument, which only work in combination with the appropriate cultural values. While
the argument for the Anglo-Saxon cluster was that individuals always strive to improve, irrespective of ownership form, but firms profit more when managers are more aligned, stewardship works the other way around from an alignment perspective. Managers do not necessarily need to be tamed, but encouraged and respected, thus not particularly financially aligned, but culturally. Merely when the optimal governance mode is in place, namely a steward manager, firm performance is moderated positively compared to all other countries and clusters, because then the managerial ownership form utilizes the cultural norms and values working in symbiosis, creating a particularly efficient combination of traits and governance, the positive interaction effect. Formally, I write:
Hypothesis 3: The relationship between ownership and performance as hypothesized under H1
is moderated by national culture, such that firm performance has a stronger positive relationship in managerially owned Confucian/Latin American firms as compared to all other firms.
Germanic Cluster
This cluster includes the countries of Germany, Switzerland and Austria. It is based on institutional characteristics of bank-based financial systems (Institute of International Bankers, 1997), high ownership concentration (La Porta et al., 1999) and codified law (La Porta et al., 2001), features which are dominant in countries that are comparably risk averse (Hofstede, 2003). Germanic culture can again be illustrated by an a priori idealtypus that highlights the intrinsic values, norms and perceptions inherent to the joint value characteristics present in the Germanic cluster. The Germanic
idealtypus is one that acts rational, in the sense that uncertainty needs to be avoided and clear rules are
to be created and strictly followed. This idealtypus is an orderly, punctual, stiff, frumpy, fairness seeking and rule abiding typus, which lacks a sense of humor, outright confronts issues at hand and avoids innovation and change as they create uncertainty. This description is in line with the institutional regulatory forces that emerged within the cluster, where law is codified and its bank-based system functions as risk management vehicle.
The joint value characteristics within this cluster lead to and can be illustrated by some of the highest ownership concentrations worldwide (Franks et al., 2006). Ownership concentration is made responsible for a large part of the divergent findings in the relationship of managerial ownership and firm performance across countries (Morck et al., 1988; Fan and Wong, 2002; Ng, 2005). From a regulatory institutional perspective, the law and finance thesis (La Porta et al., 1998, 2004) conjectures that such concentration of ownership differences are brought about by differences in investor and shareholder protection. However, the theory in this case only holds when these regulatory institutional forces can also explain the historical development, which it can not (Franks et al., 2006). When studying the history of legal rights and their influence, Germany and the UK (as Germanic and
Anglo-Saxon examples, respectively) started similarly in the beginning of the 20th century when private enforcement and strong anti-director provisions were absent. Yet despite today's differences in said regulations, suggesting differences in past investing behavior, both markets were equally active. In fact, the problematic of separation of ownership and control as described by Berle and Means (1932) around this time was similar in Germany (Franks et al., 2006). Nevertheless the banking sector grew strongly, which Franks et al. (2006) argue to be based on trust, because at the time banks were utilized not to finance, but to cast proxy votes for investors. Banks were a tool to mitigate the uncertainty through control. Consequently, I argue that the following different paths of rationalization were a product of the joint value characteristics within clusters, leading to different results between clusters. Uncertainty avoidance, stemming from a perception of a hostile environment, necessitates social and formal norms to reduce uncertainty (Beugelsdijk et al., 2015; Hofstede, 2001), inducing and maintaining stability, where ambiguity about being subject to hostility and new ideas is not appreciated (Hofstede, 1980; Hofstede et al., 2010). Therefore, in this environment formal and informal norms originating from uncertainty avoidance are the primary driver of decisions and utility function structures, thus performance.
Owing to the strong and stiff regulatory framework and behavior of individuals, I argue that the relationship between managerial ownership and firm performance must consequently be moderated negatively. In this environment the key to the negatively moderated performance is shared values of uncertainty avoidance working through the individuals and institutions. Part of the uncertainty avoidance trait leads to the argument of why both agency and stewardship approaches may not perfectly work. Agency theory is a shareholder focused model that does not explicitly create certainty for all stakeholders involved and thus from a Germanic perspective may be deemed “unfair”. Stewardship implies trusting a “leader”, which under the argued circumstances of a perceived hostile environment creates uncertainty and needs to be avoided. Therefore, the Germanic stakeholder approach goes a different and controlled path through defined terms and rules. Following arguments lead me to hypothesize a negative moderation of firm performance in these countries; First, managers are theoretically driven to (cinch their position by) foregoing risky projects (Fama, 1980; Lambert, 1986). These and similar actions actively destroy shareholder value and hinder performance and growth (Shleifer and Vishny, 1989). Similarly, people are driven into risk averse organisations instead of engaging in entrepreneurship and innovation, which is inherently risky (Knight, 1922; Thomas and Müller, 2000). Second, giving employees a stronger voice about managerial actions (e.g. codetermination or union rights) dilutes the pool of expertise to pursue a searingly straightforward path to pure firm performance and triggers the threat of firing, leading to suboptimal decision making (e.g. entrenching actions). Such welfare capitalism and two-tier boards are not meant for volatile stock performance and engagement of employees creates tougher controls on executives; e.g. in terms of
pay (Morck et al., 1988), limiting the pool of (especially outside) talent. Third, banks provide long-term stability through the mechanism of debt financing, pooled proxy voting and their behavior as shareholders, curtailing myopic managerial actions and limiting short term risk-taking. Formally, I write:
Hypothesis 4: The relationship between ownership and performance as hypothesized under H1 is
moderated by national culture, such that firm performance has a stronger negative relationship in managerially owned Germanic firms as compared to all other firms
METHODOLOGY Sample and Data Collection
The aim of this study is to identify whether national culture moderates how ownership affects firm performance. Therefore, I collected data from the financial database Orbis for firm-level financial data, the World Bank (2017) for country data, the article of Ronen and Shenkar (2013) for a first theoretical benchmark of cluster level country separation and country level cultural data from Hofstede (2001). I apply a three step approach to merge the relevant data on their respective levels. This strategy follows the logic of this section. Firstly, I explain how I collected the data in a step-by-step approach and the respective variables that this collection led to. Secondly, I explain what steps I have taken to work with the data within Microsoft Excel. Lastly, I explain what methods of analysis I have used in Stata 14. While the individual sections for the variables explain the respective variables, a more mechanical description and how they are being employed in the empirical analysis can be found directly in the description of the empirical models.
For the ownership perspective of this study, the same search strategies are downloaded twice, once for firms with managerial ownership and once without. This step is necessary to create dummy variables, as there is no built in feature to distinguish dichotomously. The following search strategy is applied; First, firms are excluded that are not listed publicly as their information is mostly unavailable. Second, I excluded based on SIC codes, namely the following: 60 to 65, 67, 93 and 99, because valuation ratios for such firms are not comparable to non-financial firms (La Porta, 2001). Third, the time frame for data collection is set to 8 years, between 2007 and 2015, which is the maximum in Orbis. As mentioned, this strategy has been applied twice, once with each of the following settings included in the boolean search function: Companies in which shareholder is also manager; owning together between 0 and 49.99%. This creates the dummy variable “0” for managerial ownership and companies in which shareholder is also manager; owning together between 50.01 and 100% or ultimate shareholder. This creates the dummy variable “1” for managerial ownership.
I trimmed the data to improve quality and limit the impact of complete outliers that do not represent the dataset. Following procedures have been undertaken: First, I deleted all companies that do not at least have 2 values for Tobin’s Q for each larger than 0.1 to prevent results dragged down by “dead” companies or those that are not active, as they do not represent “firm performance” in that scenario as well as year over year changes relevant for panel data and certain control variables (from 21845 firms to 17130). Second, I sorted out all “n.s.” and “n.a.” by using a triple nested IF-function in Excel to clean up data and prevent manual mistakes in the editing process. Third, I dropped all companies with a combined market capitalisation of less than 1 million, adding all available years
together (from 17130 to 16846); while the value line is drawn arbitrarily and could be any other combined value, I believe one million dollar over multiple years in combined value represents a threshold that is reasonable to assume for companies that are not “dead” or used for alternative purposes than pursuing firm performance. Fourth, I dropped firms without “Last Market Cap” as this is an Orbis proxy for whether Market Capitalization data is available, which is needed to calculate Tobin's Q (from 16846 to 16833). Fifth, I dropped countries from tax havens and outside of cultural clusters, namely countries such as Bermuda or the Cayman Islands. These firms do not represent the respective national culture and raise flags as to why they are listed in these countries (from 16833 to 15501). Last, I dropped all duplicates to convert the data in Stata 14 from wide to long (from 15501 to 15329).
Variables
Dependent variable. The dependent variable throughout is Tobin's Q as proxy for firm performance,
defined as the ratio of (book value of total assets - book value of common equity + market value of common equity) and book value of total assets (Q). As most utilized measure of firm performance in most important studies in this field (e.g. La Porta, 2001; Morck et al., 1988), this accounting based measure incorporates the complex reality of past performance in combination with its future outlook. Essentially, one can separate Tobin's Q into two parts, past and future. While the past is the departure point, based on accounting information, subject to disclosure and auditing for the listed firms that are used in this study, the future is subject to how investors perceive their investments to yield returns that outperform the market.
Independent variables. Underlying the logic of La Porta et al. (1999) and Berle and Means (1932),
I create dummy variables for what is factual ownership of a company to leave any doubts about the interpretation of effective ownership, defined as either representing ownership by managers (showing in the model as “1”) or ownership not by managers (showing in the model as “0”), to capture the effect that managers have on firm performance. While La Porta et al. (1999) used a 20% threshold to create a distinction between who controls companies, their findings also indicated the logic that followed this threshold, namely that companies tend to be widely held (Berle and Means, 1932). This, however, varies across countries, more so in some than others. Therefore, in the context of the diversity of countries that are used in this study and in order to reduce uncertainties where the widely held assumption applies and where it does not, I use a 50% threshold to reduce the uncertainty of actual ownership. Since the managerial uncertainty created by not-ownership plays a pivotal role in this study, one can be certain that a 50% ownership of managers on the other hand eliminates this uncertainty fully. Regarding the idea that prior studies found a curvilinear relationship between managerial ownership and firm performance (Morck et al., 1988), I argue that it is still appropriate to
use dummies to separate ownership types as did La Porta et al. (1999), because of the aforementioned logic. Following agency theory (Jensen and Meckling, 1976), alignment of interests between principals and agents is the core line of reasoning that explains differences in performance.
Following the work of Ronen and Shenkar (2013), I created a dataset that interprets the work in terms of codes, where codes are applied to separate the empirically determined regions from another. To arrive at their cultural commonality clusters, Ronen and Shenkar (2013) employed ecocultural variables (religion, geography and language) as well as ten major studies based on raw data and the dimensions utilized in those. Additionally, for individual theoretical changes of clusters that follow the logic of this study, I added codes for every single country in the dataset to simply apply functions in STATA. The Anglo-Saxon cluster follows the classification of Ronen and Shenkar (2013) and includes the United States, the United Kingdom, New Zealand, Australia, Ireland and Canada. The Chinese Confucian/Latin American cluster includes the countries of the respective clusters as classified by Ronen and Shenkar (2013); namely Colombia, Ecuador, Mexico, Venezuela, Argentina, Bolivia, El Salvador, Chile, Uruguay, Peru, Costa Rica, Guatemala and Brazil from Latin America and China, Singapore, Taiwan, Hong Kong and South Korea from the Confucian cluster. The Germanic cluster includes Germany, Switzerland and Austria (Ronen and Shenkar, 2013).
Control Variables. Previous research has shown that there is the possibility of various factors
jointly affecting the dependent and independent variables, which may induce spurious correlation and thus needs to be dealt with. The control variables used in this study follow the logic of previous studies (La Porta et al., 2001; Morck et al., 1988). Those too recognized the separation of Tobin's Q in numerator and denominator or observable and unobservable measures of intangible assets, as previously explained. Note that relevant ratios follow the matching principle of ratio analysis (Hamberg, 2016), which here also extends to the compatibility of control variables with Tobin's Q itself.
I control for firm level effects, employing following control variables: First, the natural logarithm of sales growth, measured as a ratio of change within a year, calculated per year for all firms and all available observations of revenue. La Porta et al. (2001) and Black et al. (2003) use this control as a proxy for the growth opportunities value of Tobin's Q. In this study, it substitutes the ratio of R&D expenditures divided by assets as well as advertising expenditures divided by assets (Morck et al., 1988) as the underlying idea is the same and data availability is extremely limited for the R&D and advertising, reducing the sample size by over 90% if employed. The time variant data of revenue was obtained from the financial database Orbis. Second, the natural logarithm of debt divided by assets, calculated as the ratio of market value of long-term debt divided by total assets and is additionally responsible for capturing tax shield benefits. Following Morck et al. (1988) and Black et al. (2003), this control also captures the negative correlation of debt with profitability of the firm. The time
variant data of the two components was obtained from the financial database Orbis. Third, the natural logarithm of total assets, the replacement cost of assets. This control is a proxy for size and follows the logic of capturing the potential effect of unobservable assets, which may correlate with size (Morck et al., 1988). Including the natural logarithm of total assets is common practice for firm performance studies (Black et al., 2003; Durnev and Kim, 2002; Shin and Stulz, 2000). Additionally it controls for the (lower) probability of owning a large share in bigger firms. The time variant data was obtained from the financial database Orbis. Fourth, the natural logarithm of number of managers and the natural logarithm of number of employees to control for number of employees and managers as a measure of firm size and to add another novel control in this study for ownership. The time invariant data was obtained from the financial database Orbis.
Additionally, I control for industry and country level effects, employing the following control variables: First, three-digit SIC code dummies as the standard industry classification used by Morck et al. (1988) and Black et al. (2003), which control for spurious correlation between ownership and Tobin's Q emanating from industry effects. This time invariant data was obtained from the financial database Orbis. Second, the natural logarithm of GDP per capita is used to control for country-level development states and is a standard control variable in international studies. In a dataset largely dominated by US firms, this control variable also serves to capture a potential bias from more developed countries as well as account for the positive correlation of GDP per capita with disclosure transparency, shareholder protection and corporate policy as well as the negative correlation with Tobin's Q and Return on Assets (Griffin et al., 2014; La Porta et al., 1999, 2001; Sethi et al., 2003; Feinberg and Gupta 2009). The time variant data was obtained from the World Bank (2017). The aforementioned variables are summarized in table 1, an extended version is available in table B2.
Table 1
Definition of relevant variables applied in the following analytic methods
Variable Code Unit of Measurement Usage Description
LN Tobin’s Q LNQ Natural Logarithm Dependent Variable Ratio of (book value of total assets - book value of common equity + market value of common equity) and book value of total assets
Tobin's Q Q Ratio Alt. Dep. Variable The natural logarithm of Tobin's Q as described above
Owner Manager OWNERMAN Dummy Independent Variable 1 if the firm is owned by managers (>50% ownership combined) and 0 if managers do not own company (<50% combined)
Leverage Ratio LNLEV Continuous Control Variable The natural logarithm of the ratio of total debt to total equity Total Assets LNTAS Continuous Control Variable The natural logarithm of the total book value of assets Managers LNMAN Categorical Control Variable The natural logarithm of the number of managers in firm Employees LNEMP Discrete Numerical Control Variable The natural logarithm of the number of employees in firm (estimation) Sales Growth LNSAG Discrete Numerical Control Variable The natural logarithm of year over year sales growth (Revenue) per company GDP per Capita LNGDP Ratio Control Variable The natural logarithm of the ratio of GDP per Capita; World Bank (2017) Industry Dummy SIC Categorical Control Variable Firm category of industry according to SIC classification, three digits Anglo-Saxon Cluster AN Dummy Independent Variable 1 if the firm is within specific cluster and 0 for outside; Employed for
Anglo-Saxon Cluster
Latin American/Confucian Cluster LACO Dummy Independent Variable 1 if the firm is within specific cluster and 0 for outside; Employed for Latin American and Confucian Cluster
Germanic Cluster GE Dummy Independent Variable 1 if the firm is within specific cluster and 0 for outside; Employed for Germanic Cluster
Analytic Methods
The gathered data, which is both cross-sectional and time-series, resulting in a panel data structure, allows to include effects that may drive the results. It is desirable for this research to be able to assume variation across entities to be random and uncorrelated with dependent or independent variables as the data includes time-invariant variables (data for the control variables of industry, number of employees as well as managers was not available over time) and all regressions were estimated using the Breusch-Pagan (1980) lagrange multiplier, which strongly rejects the null hypothesis (p<0.001) that errors are independent within entities, therefore random effects are employed for the panel data regressions. Such random effects allow generalization of inferences beyond the sample utilized and does not treat entities as independent observations. In this section I examine the effect that the variables of analysis have on the dependent variable, Tobin's Q, by estimating the following empirical models:
NQ α γ OW NERMAN (CONT ROLS)
L i,t = 0 + 1 i+ ∑8
j=2βj + ui,t+ εi,t
(1)
NQ α γ OW NERMAN γ CLUST ER γ CLUST ER W NERMAN
L i,t = 0 + 1 i+ 2 i+ 3 * O i (CONT ROLS) + ∑10 j=4 βj + ui,t+ εi,t (2)
Equation (1) estimates the aforementioned benchmark for comparison of ownership effects. Equation (2) reflects the applied regression equation for the managerial ownership hypotheses in the respective clusters. Where NQ L i,tis the dependent variable (the natural logarithm of Tobin's Q) in which i represents entity and t represents time; W NERMAN O i is a dummy variable that indicates 1 if the company is owned by their manager(s) and 0 otherwise; LUST ER C i is a variable that indicates 1 for all countries within a country cluster as defined under the respective hypotheses and 0 otherwise; is the interaction variable of the respective cluster and the
LUST ER W NERMAN
C * O i
variable; is the unknown intercept; represents the coefficient of the
W NERMAN
O i α 0 γ
time-invariant independent variables; βrepresents the coefficient of the time-variant control variables; represents the between-entity error since this is a random effects panel data regression equation
ui,t
and ε i,t represents the within-entity error.
For a more straight forward assessment and more appropriate dealing with the non-normal distribution and sharp skewness of the data (table C1 and C2), the continuous variables have been logarithmically transformed. The exception to this rule are the dummy variables. Therefore, in order to correctly interpret the impact of the moderator and independent variables on the unlogged
dependent variable (Tobin's Q), the guiding rule is that a one unit change in the variable leads to a change of the dependent variable, which is simplified by the fact that the dummy
100 )%
( * eγ− 1
variables can only take 0 or 1 (Benoit, 2011). This is relevant as I want to find out to what extent the presence of (a combination of) the independent variable(s) changes the unlogged and not the logged Tobin's Q. An application follows in the interpretation of the regression output. Variance inflation factors (VIFs) were estimated to assess multicollinearity, all of which were lower than 10, indicating no multicollinearity issues (table C3). For improved readability I suppress the industry codes, which were estimated according to 3-digit SIC codes. In the following, I report summary statistics and correlation matrix (table 2) and an overview of the relevant independent and dependent variables (table 3); for a more detailed map of the data, see appendix-A.
Table 2
Panel A: Summary Statistics of joint sample
Variable Observations Mean Std. Dev. Min Max
LN Tobins Q 75,435 0.27 0.92 -6.65 13.17 Tobin's Q 75,435 49.07 3,203.10 0.00 524,686.20 Owner Manager 124,353 0.11 0.31 0.00 1.00 Leverage Ratio 90,783 -0.91 0.98 -10.63 11.65 Total Assets 92,600 11.08 2.40 -4.61 19.83 Managers 124,029 2.23 0.81 0.00 7.63 Employees 76,770 5.73 2.39 0.00 13.32 Salesgrowth 42,706 -1.81 1.54 -12.25 11.15 GDP per Capita 124,353 9.68 1.33 6.86 11.54 Anglo-Saxon Cluster 124,353 0.32 0.47 0.00 1.00
Latin American/Confucian Cluster 124,353 0.35 0.48 0.00 1.00
Germanic Cluster 124,353 0.02 0.12 0.00 1.00
Panel B: Correlation Matrix LN Tobin's Q Tobin's Q Owner Manager Leverage Ratio Total Assets
Managers Employees Sales Growth GDP Anglo-Sax on Latin Confucian Ger LN Tobin's Q 1 Tobin's Q 0.1789 1 Owner Manager 0.0356 0.003 1 Leverage Ratio 0.2644 0.129 0.0406 1 Total Assets -0.3129 -0.0854 -0.1029 -0.0632 1 Managers -0.1575 -0.0229 -0.071 0.077 0.3959 1 Employees -0.0528 -0.0146 -0.1481 0.1584 0.7843 0.3829 1 Sales Growth 0.1696 0.0245 0.0036 -0.0694 -0.2159 -0.1115 -0.1859 1 GDP per Capita 0.163 0.0139 0.0082 -0.0089 0.0847 -0.0395 -0.1635 -0.0463 1 Anglo-Saxon Cluster 0.2864 0.0268 0.0066 -0.0012 -0.2041 -0.1625 -0.2496 0.0899 0.5809 1 Latin/Confucian Cluster -0.097 -0.0117 -0.1226 -0.0614 0.2817 -0.1348 0.2505 -0.0605 -0.0378 -0.5084 1 Germanic Cluster 0.0007 -0.0022 0.0532 0.0099 0.0472 0.0944 0.0364 -0.0345 0.1087 -0.087 -0.0928 1
Table 3
Description of cultural clusters by owner (independent variable) and firm performance (dependent variable)
Manager Observations Mean Std. Dev. Min Max
Anglo-Saxon Cluster 0 35,766 190.98 6,862.01 0.00 524,686.17 1 4,374 405.66 3,064.05 0.04 58,230.38 Confucian Cluster 0 40,302 1.44 1.98 0.00 127.35 1 2,088 1.53 1.61 0.08 34.02 Germanic Cluster 0 1,485 1.65 1.50 0.14 21.33 1 459 1.44 1.12 0.01 10.15
Latin American Cluster 0 1,026 1.34 1.01 0.04 8.72
1 306 3.04 13.48 0.07 121.87
EMPIRICAL RESULTS Default hypothesis effects
The default hypothesis aims to create a benchmark upon which inferences can be made and show the standard level of ownership impacts, thus indicate whether there is a universal relationship between the two. The method of analysis is panel data analysis with random effects. The ownership impact is insignificant and has a positive coefficient of 0.042, a finding that applies throughout for managerial ownership in general. The control variables of GDP (0.050; p<0.001), leverage (0.061; p<0.001) and number managers in a firm (-0.051; p<0.001) are significant. This finding refutes major previous findings of a theoretically positive relationship (Jensen and Meckling, 1976), empirically significant linear (Demsetz and Lehn, 1985) or otherwise positive relationship (Morck et al., 1988; Short and Keasey, 1999) that built the backbone of ownership performance studies and does not seem to hold, when taken outside the Anglo-Saxon context, suggesting that the findings may in fact be bound to the institutional and cultural context of their sample (Table H2a). In sum, taking all listed firms together, there is no significant relationship between managerial ownership and firm performance, meaning that deviations in the following tests indicate cluster specific effects, compared to all other firms.
The moderating effect of national culture in the Anglo-Saxon cluster
Model H2 presents the effect of ownership on firm performance moderated by the cluster of Anglo-Saxon countries, measured by panel data analysis with random effects. Since the large body of literature measuring ownership effects stems from countries that are within this cluster too, I expected a similarly positive relationship to take place (Morck et al., 1988; Short and Keasey, 1999). I test hypothesis 2 by using an interaction variable of the respective ownership type and the cluster, comparing the effect of managerial owned firms of the cluster with all other firms.
In line with aforementioned studies, effects for managerially owned firms are strongly positive (p<0.001) with a coefficient of 0.291, indicating that when the managerial ownership and cluster interaction triggers, Tobin's Qs are 33.8% higher. This effect is even stronger for the whole Anglo-Saxon cluster with a significant (p<0.001) positive effect of 43.6% on Tobin's Q, compared to all others. This finding clearly shows that overall, the combination of cultural values in the Anglo-Saxon cluster positively affects the relationship between managerial ownership and firm performance. Additionally, the latter finding indicates that Tobin's Q are generally higher in the Anglo-Saxon cluster. This confirms hypothesis 2 that firm performance has a stronger positive relationship in managerially owned Anglo-Saxon firms as compared to all other firms. Having established that there is no universal relationship between managerial ownership and firm performance (H1), this finding shows that the theory of Jensen and Meckling (1976) as well as
empirical finding of Morck et al., (1988) and Short and Keasey (1999) does apply in their context (see also Table G1a and H2a), but may then also be limited to it. In other words, under the hypothesized circumstances, findings do apply and this relationship is restored when interacting the cultural framework of the original study with the relationship and it does not exist when all firms are treated as equals. Regarding the effect on the control variables, interestingly not much changed, especially considering the strong effect sizes of the interaction and cultural variables. The expectation of the dominating Anglo-Saxon firms is also fulfilled, as the Anglo-Saxon regressions are the only ones where the constant is significant (p<0.05) and the only one where GDP is indifferent and turned insignificant.
Table 4
Regression results. Dependent variable: Tobin's Q
Model H1 H2 H3 H4
Description Default Hypothesis Anglo-Saxon Cluster Confucian/Latin
American Cluster Germanic Cluster Owner Manager 0.044 0.034 -0.043 0.057* [-0.03] [-0.03] [-0.03] [-0.03] LN GDP 0.050*** 0.002 0.055*** 0.053*** [-0.01] [-0.01] [-0.01] [-0.01] LN Leverage 0.061*** 0.063*** 0.055*** 0.062*** [-0.01] [-0.01] [-0.01] [-0.01] LN Assets -0.026*** -0.015* -0.017** -0.026*** [-0.01] [-0.01] [-0.01] [-0.01] LN Number of Managers -0.051*** -0.094*** -0.123*** -0.049*** [-0.01] [-0.01] [-0.01] [-0.01] LN Number of Employees 0.019** 0.011 0.022*** 0.019** [-0.01] [-0.01] [-0.01] [-0.01] LN Sales Growth 0.034*** 0.032*** 0.034*** 0.034*** [0.00] [0.00] [0.00] [0.00]
Interaction Anglo-Saxon Cluster * Manager Owner 0.291***
[-0.08]
Anglo-Saxon Cluster 0.362***
[-0.02]
Interaction Confucian/Latin American Cluster * Manager Owner 0.228***
[-0.06]
Confucian/Latin American Cluster -0.206***
[-0.02]
Interaction Germanic * Manager Owner -0.085
[-0.1] Germanic Cluster -0.100* [-0.05] constant 0.145 0.576** 0.222 0.130 [-0.22] [-0.22] [-0.22] [-0.22] Dependent Variable LNQ LNQ LNQ LNQ
Random Effects Yes Yes Yes Yes
Industry Effects Yes Yes Yes Yes
Observations 21912 21912 21912 21912
Individual Entities 6136 6136 6136 6136
Sigma_E 0.40 0.40 0.40 0.40
rho 0.70 0.70 0.70 0.70
Chi Squared 1371.60 1766.50 1485.80 1380.60
Notes: Robust standard errors in parentheses. *** p<0.001, ** p<0.05, * p<0.1
