Performance of Young Public Firms
Managerial vs Outside Shareholder Control
in an international context
/ Daniel Heinz Klaus Heinrich Richard Walter Bogdanski
/ 900912-T352
/ Double Degree MSc International Financial Management / Business and Economics
/ Faculty of Economics and Business
/ Rijksuniversiteit Groningen and Uppsala Universitet
/ Supervised by Dr. Wim Westerman
/ Co-Assessed by Dr. Halìt Gönenç
Abstract
This paper studies the relationship between firm performance, proxied by Tobin's Q, and two distinct ownership types, managerial owned firms and outside owned firms. The sample consists of 2005 young firms from Europe and the US that incorporated since the dotcombubble 2001. Very similar to the pre2001 period, young and highly funded firms are of popular concern. In particular their owners, founders and CEOs are topic of interest and serve as figurehead for their company, raising the question whether their firms perform better if they also own them or not and whether that differs with the institutional framework that the company is situated in. Thus the research question is the following: What is the effect of having management as majority shareholder(s) on the performance of the young firm in different environments? To find an answer, I used quantitative data from Orbis and analyzed it using timeseries panel data, recent information using simple OLS as well as multiple analyses of variance. I find evidence of higher valuations of firms owned by managers, especially in countries with common law and stronger shareholder rights. I also find evidence of relatively lower valuations of firms owned by their managers when these are situated in code law countries or countries with stronger creditor rights. A surprising addition to the findings were extreme values of Tobin's Q that may indicate another bubble in the making, coincidentally closing the circle of this study.
Key words: International, Financial, Management, Ownership, Firm Performance
JEL Classification: G32, L25, L26
1 Introduction
It is commonly accepted that managers are largely determining firm performance, but just as much that firms are largely determining this manager's job safety. These circumstances create uncertainty potentially leading the managers to alter their behavior. “Unfriend: What drove Zuck to fire Saverin” (CNET, 2012), “How Elon Musk Fired Tesla CEO and cofounder” (Business Insider, 2014) or the famous “Apple CEO John Sculley fires Steve Jobs” (Fortune, 1985) were some of the most famous headlines in past years and history. Those headlines and their background suggest that it is not exclusively performance that drives founders and managers out of their firms. Usually those stories start with “[Investor] Carl Icahn wants Yahoo CEO fired” (Fortune, 2008), followed by “Icahn Ramps Up Pressure, Vows To Get Jerry Yang Fired” (Business Insider, 2008) to end with “Jerry Yang Resigns From Yahoo, the Company He Founded” (WSJ, 2012), suggesting the power of controlling shareholders and implying forces other than performance. The question is so current and attracting so much attention, that US TV network HBO dedicated a whole award winning TV Series towards this topic, calling it “Silicon Valley” famously quoting “take the money or keep the company”. This raises the question: what is it actually that lets founders of young companies get into a position of facing to be ousted out of their own company? Do firms perform differently when separation of ownership and control does not exist?
Recent work in top journals took it upon them to investigate this subject: CEO turnover and their determinants. Research of Jenter and Kanaan (2015) suggests precisely the issue implied, turnover decisions irrespective of the pure nature of firm performance. Pointing out an issue that managers mostly sense one way or another, brings about frictions firms not necessarily desire. Looking at older research of Morck, Shleifer and Vishny (1989), they actually suggest that industry performance is filtered out of dismissal decisions, which additionally triggers the question whether this is a phenomenon that only applies to younger firms.
In order to grow their company, founders have to make crucial decisions in very early stages. Capital structures will haunt founders until the end of the days, because they imply control issues. Some founders therefore choose to base their financing decisions around control, whereas others choose to neglect this issue and first grow the company.
for the investor. Hart (1995) describes the feature of securities to be not only limited to cash flows, but also includes other rights that these securities give access to, amongst others the right to vote for directors. These rights become critical to acting managers, because their behavior is determined by the people who control their job. In the end, dividends are paid because shareholders control the directors and creditors are paid because otherwise they can demand their collateral (La Porta et al., 1998).
News coverage around young firms and startups picked up increasingly and developed into a trend. Many founders want to be part of the sharingeconomy, be the new Uber, Facebook, Google or AirBnB. The impact of young firms and the trend around it has been growing, fueled by venture capitalist and early series funding of high potential firms, even without revenues or tangible assets. While those crucial decisions around financing in young firms mostly happen in the period before public offering, the capital structure has to be carried with them during the process of going public. The longer the process takes, the more are shares of managing owners and founders watered down. With that in mind, looking at venture funding betting on exponential growth of cashflows, Modigliani and Miller’s findings do not hold anymore, triggering a new problem of being a young firm, disproportional cost of debt. “PreRevenue” companies with billion dollar valuations (socalled “unicorns”) are not the exception anymore. Even in 1976, Jensen and Meckling already argued that they believe ”the existence of agency costs provide stronger reasons for arguing that the probability distribution of future cash flows is not independent of the capital or ownership structure”, leading to agency costs.
The question I ask here is whether it is really a clever decision to give away shares and ownership early, to fuel fast growth or create more sustainable and potentially flat growth, risking to take too much time, but be part of the decision making process when the own firm is in the position that the founder always dreamed about. And how does that differ in between regions? If there are differences, what are those? La Porta et al. (1998) argue that the intrinsic characteristics inherent to the securities of investors is not equal across the world. According to their findings, capital structure differs around the world because of differences in enforcement. This raises the question: Do the findings of La Porta et al. still hold for determinants emanating from country regulatory differences and are they applicable to young firms?
just anecdotal evidence to what motivates founders and how they keep track of their ownership and exercising voting rights, I want to dive into a more generalized view by including the full sample of young companies that emerged after the dotcom bubble in the most developed regions in the world, the US and Europe. Is it really worth passing on early money, just to bootstrap the way up to glory?
“Lots of companies don’t succeed over time. What do they fundamentally do wrong? They usually miss the future. I try to focus on that: What is the future really going to be? And how do we create it? And how do we power our organization to really focus on that and really drive it at a high rate?” Larry Page, CEO Google (FT, 2014) on long term orientation, intrinsic motivation and the resulting inner power of founders who are free to chase their dreams within their own company.
This study revolves around the key requirements set out by the master degree that it is intended to complete. It follows the papers of Jenter and Kanaan (2015), La Porta et al. (1998 and 2001) as well as large parts of the methodology and structure of Gönenç and Scholtens (2017). The papers employed serve individual purposes that are merged within this paper. While Jenter and Kanaan (2015) provide insights into the determinants of CEO turnover based on the performance of the firm the CEO manages, the papers of La Porta et al. (1998 and 2001) provide the international fundament that seeks to explain differences and/or similarities between the regions of the United States and Europe. Those papers introduce the variables of investor rights and legal origin.
The main research question I thus define as follows: What is the effect of having management as majority shareholder(s) on the performance of the young firm in different environments?
confirming H5. Surprisingly, I found results partly opposing H6, namely negative effects of creditor rights on firm performance, even though they are rather low and weak. On the other hand, stronger shareholder rights have a positive effect on firm performance, thus also just partly confirming H6 and providing support to the findings of La Porta et al. (2001). This study contributes to the literature by giving insights into effects of who owns companies and how this is in turn affected by the institutional framework, taking the general findings of La Porta et al. (1997, 1998, 1999a, 1999b, 2001 and 2008) to a further detailed level. I moreover confirm most of these findings and provide food for thought regarding the current situation in market valuation for a very specific type of firms, young firms.
To the best of my knowledge there has been no study trying to find out whether firms perform better if they are owned by their managers and whether that differs with the institutional framework that the company is situated in. This is the gap this study is supposed to fill with a fresh dataset of companies founded since 2001, while focusing on publicly listed firms from the US, Europe and together, to test effects in between. In order to investigate the research question, I make use of analyses of variance as defined by Mann and Whitney as well as multiple types of regression analyses and panel data to dig deeper into the effects of individual variables as well as investigate the effects of other variables affecting the findings.
2 Literature Review
Underlying this research is the agent, as in the manager(s), being the performance determinant of firms. Their impression of longterm ability to strategize and build their ideas depends on job safety, which I argue to create costs for nonowners, while owners of firms can focus on the job instead of building safety nets, leading to higher performance in firms owned by its managers. The overall approach takes strategies by other papers that I then apply to the topic of this paper. To avoid excessive repetition in literature review and hypothesis argumentation, this section focuses on the broad picture of available literature and outlines the factors that are relevant for this study per topic by weighing off findings and giving a wider impression. Specific findings including their hypothesized direction of impact applied to firm performance will mostly be laid out in section four “Hypotheses and Empirical Models”, which partly requires assembling multiple topics for the argument, hence requires to be separated from this literature review. Following literature builds the foundation of this paper:
2.1 Triggering Agency Costs CEO Turnover decisions
The decision to retain or fire a CEO is often outside the control of managers themselves and that is because of firm performance outside managerial control (Jenter and Kanaan, 2015), where managers get blamed for exogenous shocks on performance they can not control. This blame may come from outside pressure such as shareholders (Fisman, Khurana, and RhodesKropf, 2014), or the inside, whereas boards appear to be filtering out market shocks (Gibbons and Murphy, 1990; Kaplan and Minton, 2012). Other studies generally find that turnover decisions are filtered out by market shocks (Barro and Barro, 1990; Morck, Shleifer, and Vishny, 1989), yet not in those with older samples (Warner, Watts, and Wruck, 1988).
performance and turnover are potentially related, but one does not necessarily cause the other (Comte and Milhal, 1990). In the light of these circumstances, uncertainty created by conditions out of managerial power may inhibit managerial performance and lead to agency problems originating from the circumstances and selfprotection.
Interestingly, most literature on CEO turnover does not provide payforperformance as determinant for turnover, namely rather selecting the “optimal” person (Jovanovic, 1979), a type of match between the candidate and firm (Jenter and Kanaan, 2015). Most literature in the 90s does not focus or provide good explanations for turnover in good times (e.g Aggarwal and Samwick, 1999; Murphy, 1999). Under these circumstances CEO compensation surprisingly does not influence the argumentation of this paper, excluding the paper of Casamatta and Guembel (2010). They argue that legacy potential significantly determines CEO compensation, strategy and turnover Such legacies then increase replacement costs and thus strengthen certainty of managers, long term ideas and implicitly then strengthen their position, leading to similar effects argued to be an important performance determinant of managerial owners. Legacies are, however, not the norm. This rather provides reason to believe that familyfirms play an important role in describing ownership effects on firm performance.
2.2 Managerial ownership and firm performance
Relevant studies regarding managerial ownership and firm performance find a curvilinear relationship of firm performance and Tobin’s Q (Barnhart and Rosenstein, 1998; Morck, Shleifer, and Vishny, 1988; McConnell and Servaes, 1990). Stulz (1988) found that an increase in managerial voting rights leads to an increase in premium offered for takeovers, increasing shareholders wealth. Furthermore, Morck, Shleifer, and Vishny (1988) find Tobin’s Q to be increasing with larger ownership of board of directors, most importantly over the 25% threshold. Chen and Yu (2012) go further to base on Shleifer & Vishny’s findings (1994) for the relevance of ownership structure, that this is precisely what determines agency problems. Fama and Jensen (1983) argued that ownership of management in firms is counterproductive due to inability to supervise, which Morck, Shleifer, and Vishny (1988) refute. They also specifically point out the decrease in Tobin’s Q when the firm is run by a founding family member compared to outsider, yet only for older firms and second generation family firms. This is of vital importance for the following section and the line that this paper draws between familyfirms, managerial owners and the relevance of findings in familyfirm research.
2.3 Family firms and its comparability to managerial ownership
Research on family firms is widely available (Anderson and Reeb, 2003; Miller et al, 2007; Villalonga and Amit, 2006) and I argue that family firms are very similar to firms owned by the manager(s), much like Chen and Yu (2012). While there is no commonly accepted consensus as to what precisely constitutes a “familyfirm” (e.g. Kraiczy, 2013; Westhead and Cowling, 1998), two common uses emerged. Firstly firms with family CEO as successor in at least the second generation (Bennedsen et al., 2007) and secondly those with family ownership (Anderson and Reeb, 2003; Cronqvist and Nilsson, 2003; Miller, 2013; Villalonga and Amit 2006), whereas Anderson and Reeb (2003) and Villalonga and Amit (2006) go as far as defining a family firm as those with members of founders’ family or the founder himself in top management or as major shareholder, without specifying a threshold. This description builds the foundation for comparing managerial owners to family firms, as such a first generation family firm. Reversing this argument, managerial ownership would be defined as a family firm. Thus, I argue familyCEOs are similar or equal to founders and others managerial owners. In that case, characteristics of familyfirms are arguably features of managerial owned firms as well.
Benefits of family firms are lower agency cost, reducing those stemming from the separation of control and ownership (Jensen and Meckling, 1976), and the principle of making longterm decisions (GómezMejía et al., 2011). The authors found this to be especially significant in firms with concentrated ownership, which provides the opportunity to push the above mentioned ideas through that underlie familyfirms. Additionally, familyCEOs may behave altruistically with their family in mind (Schulze, Lubatkin and Dino, 2002) and controlling families are not found to expropriate wealth from minority shareholders (Croci, Gönenç and Ozkan, 2012), substantiating the similarities between family firms and managerial owners who are thus interested in firm performance. 2.4 Implicit forces on managers, founders, family and widelyheldfirms
they were unable to enforce maximization of value, because control is concentrated in the hands of managers. Jensen and Meckling (1976) then further developed this theory into the principalagent theory that constitutes the basis of the now known problem. According to them, it is in the interest of the both principal and agent to maximize their own interest instead of the others and thus an agent in the form of a manager does not always work towards the interest of the principals. The work a manager does is then influenced by his or her own perception of what is the best to save the job. Thus, according to Adams and Ferreira (2009), shareholders prefer risky project to increase wealth, whereas managers are trying to reduce taking risks that could put them in a bad position.
Stewardship theory differs in the structure of control. While agency theory underlies a separation of board and CEO, who is put on track by incentive schemes set out by the board, this separation does not exist in stewardship theory, putting the CEO in full charge, acting as “steward” of the firm. Donaldson and Davis (1991) find significant improvements in ROE of firms that have combined chair and CEO positions, compared to companies with separated ones. This separation exists in the same sense when firms are owned by their managers instead of widely held.
On the other side this effect of stewardship may be weaker depending on institutional context, moving to the downsides of family ownership, much like managerial owners could. Even though La Porta et al. (1999 and 2000) did not explicitly mention the idea, I would like to use these papers to show in what direction I am pointing here, taking findings of legal origin and the investor rights effects to demonstrate performance effects.
The above mentioned agency problems may, however, also be prevented by controlling shareholders other than managers as controlling shareholders themselves. Research of Shleifer and Vishny (1997) as well as La Porta et al. (1998) suggest that such controlling shareholders are able to implement better controlling mechanisms to balance out governance issues to avoid bad decisions of the firm, much like the governance mechanisms described in Lin and Hu (2007), emanating from families as controlling shareholders. Different to agency in families who do not expropriate from the firm (Croci, Gönenç and Ozkan, 2012) in this case the authors write that it may well be that controlling shareholders act against other shareholders of a firm. These findings may emphasize again the differences in issues related to control and (longterm) motivation of the controller.
Considering the overall picture that I use to proxy family firms for managerial ownership performance, the relevant points made are those originating from the overall performance determinants of ownership by the managers. Firstly, control issues leading to uncertainty avoidance. Secondly age, namely founders and not secondgeneration or later stage family firm managers. Those aspects firm up the basis of this study relating to familyfirms.
2.5 Entrenchment as performance inhibitor for notowners
who have not been founder, to engage in entrenchment, especially their intermediate years. Additionally, their study shows weakening effects over time, which in turn expressly underlines the importance of a safe position and manager concerns of uncertainty.
2.6 Regional differences and investor protection
Different ownership structures are commonly found in between the regions of the US and Europe, for example family control is comparatively dominant in continental Europe compared to the United States (Faccio and Lang, 2002). Multiple papers of La Porta et al. (1998, 1999a, 1999b, 2000, 2001 and 2008) go into more detail of how individual companies finance themselves, what capital structures they are using and what determinants are responsible for these outcomes.
La Porta et al. (1998) found strong differences between the legal background of countries and their legal layout of investor protection. The separation is found to be especially strong between common and civil law countries, with weakest investor protection in countries with French legal origin and strongest in those countries whose legal system originates from traditions of the United Kingdom. More specifically, they found differences in ownership structures depending on the ability of an investor to enforce their rights. Two types of protection emerged, the first being creditor rights, the second shareholder rights.
Shareholder rights are measured by the authors as the socalled “antidirectorrights”, which provides an indicator of the ability of minority shareholders to take part in the decision making process and how the law protects these rights to vote. The index is made up of six dummy variables providing a range of values from 0 to 6, the latter being the strongest, representing individual rights that the investor may use. Oneshare onevote practices are not relevant in this study due to the countries in the sample.
Next to shareholder rights, the papers of La Porta et al. explain creditor rights, which is again a number of (five) dummy variables representing a total value, in this case ranging the outcomes from 0 to 4. The difference between the two lies, as mentioned in the introduction, in the security of the investor. For creditors the security is collateral in the firm's assets and for shareholders their voting rights, whereas there are more different types of creditors with largely varying aims. The essence is, however, that creditor security is based around the collateral claim. Making reclaiming of collateral in default comparably difficult, would disincentivize an investor to invest (La Porta et al., 1998) .
these investor rights. Looking at differences in enforcement for creditor and shareholder rights then explains differences of capital structures inbetween countries. This leads to La Porta et al. (1999a), who partly oppose the picture that Berle and Means (1932) painted of separation of control and widely held ownership, finding relatively few firms which are widely held outside the United Kingdom and the United States. To give another example, Edwards and Fischer (1994) find Germany to have strong banks, but a weak stock market, which is supported by the findings of La Porta et al. (1998) rating Germany with relatively strong creditor rights and relatively weak shareholder rights, leading to relatively more debt financing versus equity financing for German firms.
The key takeaway of this section is the divergence of incentives in each country emanating from the law and enforcement within its borders, which spills over to the market valuation of firms governed by these countries. The legal origin that persists in the countries is a strong determinant of these outcomes. The papers of La Porta et al. emphasize that countries differ and that this difference has an impact on firms and corporate finance. I would also like to emphasize that incentivization is the red line that leads through this study and provides the push and pull mechanism that finally leads to performance. These mechanisms are involved throughout all parties, from binary legal origin providing the direction; to individual countries using the direction and creating a unique framework; to firms working with the framework and adapting their capital structure to accommodate their needs; and then the firm’s managers who are bound by the framework set out by countries and firms, trying to extract as much as possible for personal needs, loosely adapted from the principle maximum utility function: “if everyone thinks of themselves, everyone is sought after”.
3 Data and Methods
Following Croci, Doukas and Gönenç (2011) and Gönenç, Hermes and van Sinderen (2013), I collect the data from the ORBIS database on all information I need. Especially the ability to distinguish between ownership thresholds is convenient for dummy variable creation. I apply a three step approach in order to increase validity and decrease potential measurement errors. After the first step, data collection in ORBIS, I secondly, edit the data in EXCEL, which leads to thirdly, the final data analysis in STATA. Interestingly, I find two datasets of almost equal size. A distribution of countries included in the dataset can be found in table A3. A distribution observations per countries can be found in figure A1
Following the suggestions from Chen and Yu (2012), managerial ownership is created as a dummy variable where managers either have full (or ultimate) control or not. To do so, I apply filters in ORBIS for either >50% ownership of managers or <50% ownership of managers. These datasets are then merged per country and later for the whole set.
3.1 Data collection process
For the international perspective of the research, to obtain comparable data, the same settings are downloaded into four different sets (each region and each ownership type). The main reason for doing this is to have four sets that can be analyzed individually and put together into one to analyze the fifth as a whole. Additionally, Orbis has better internal measures to distinguish between ownership types in their search strategy options than would be possible to evaluate by hand or judgement of other variables provided by Orbis. Hence, following search strategy is pursued:
1. Firms are excluded that are in the financial service, gambling and betting, insurance, reinsurance, (pension) funding, those that are similar and those auxiliary to these services because of their potentially distorted Tobin’s q values. Their complete code according to SIC categorization (SIC, 2016) is excluded.
2. The year of incorporation is set to between 2001 up to and including 2015, excluding companies for which the year of incorporation is unknown, to be precise. Data for 2016 is excluded, because most of the data is not available yet.
This concludes the general search strategy that applies to all of the four datasets. Following selections are created once for each combination:
1. The two possible regions are first, the geographical region of the United States, which is at the same time one political region. This is important to mention, because for the counter variable, the political region of the all areas that are considered as belonging to the European Union in a wider spectrum, namely western and eastern Europe as well as Scandinavia, the Baltic, Nordic and Balkan states are chosen. Currency issues should not be relevant within the framework of this research as the performance measure, Tobin’s Q, is merely a ratio. Hence, this measure is independent of currencies.
2. The possible ownership scenarios are twofold here; first scenario are managers owning less than 50% of the business and the second scenario is managers owning more than 50% of the business. Orbis provides this selection criteria in their selections strategy options and for this I include companies for which the manager is also the ultimate owner only for companies >50% managerial ownership.
3. Only absolute year values for annual data are used and all values are denominated in US dollar.
Final size of datasets is the following: Firstly, US firms with managers owning more than 50%; before cleaning raw data size is 409 firms. Secondly, US firms with managers owning less than 50%; before cleaning raw data size is 1117. Thirdly, European firms with managers owning more than 50%; before cleaning raw data size is 279. Lastly, European firms with managers owning less than 50%; before cleaning 997. The combined set with added dummy variables for the two different specifications (ownership and region) includes a raw amount of 2802 firms. The absolute final sample size is 2005 firms. The US dataset includes 1131 and the European dataset 878 firms. The difference of 4 firms stems from double counted firms that are split in two, but run under the same firm name.
Finally, I insert the original datasets of La Porta et al. (1998 and 2008), to obtain the crosscountry variables of legal origin, antidirector rights and creditor rights.
3.2 Treatment of missing data
not known for what reason no last value is listed. This provides a more realistic number of firms that are actually in the dataset. This is done by creating a function that gives an output of either 0 (for available data in 2015) or 1 (for no data available in 2015) and then filtering by 1, which rows are deleted. Secondly, those that arise because of “late IPO”, namely stemming from late public offering that results in available data from a later point as this research includes only young firms and not all are going public at the same time. In these cases regularly data is available from that point on, but not before. This information is only necessary for panel data. Thirdly, values missing for specific years. It is rare but does happen that a single year is missing. In these cases the average of the year before and after is taken. This information is also only needed for panel data and not for the main test, MannWhitney U test. Additionally, two extreme outliers were dropped.
3.3 Treatment of outliers
Only applied to the nonpanel data tests, to finalize the data treatment process and avoid outliers affecting results, trimming to the 5th and 95th percentile is being conducted as the extremity of some values does not reflect the overall picture of the dataset and it may well be that the data collected is at times without foundation. I chose trimming in place for winsorizing, because a test with winsorizing led the medians that replaced the extreme values still be outliers due to the extreme outliers that I want to correct for. It is also my intention to test for realistic results that others would obtain as well with different datasets or in different regions. Hence, it appears obvious that this dataset needs to be trimmed. For the joint dataset the lower threshold is .09, the higher 224.95, leaving 2005 companies. For the European dataset the lower threshold is .083, the higher 5.62, leaving 878 companies. For the US dataset, the lower threshold is .1, the higher 651.48, leaving 1131 companies. I left the dataset I use for panel data as is and only trimmed values for robustness and sensitivity tests.
3.4 Empirical methodology
In my empirical analysis, I use two measures of firm performance: Tobin’s Q and the natural logarithm of Tobin’s Q. Adding the natural logarithm of Tobin’s Q gives me the opportunity to normalize the distribution of the dataset and obtain elasticity results between the variables when conducting multiple regression analysis.
3.5 Dependent Variable Tobin’s Q and LN Tobin’s Q
The dependent variable is a proxy for firm performance. For this research, I use the natural logarithm of Tobin's Q (LNQ), which is market value/total assets and used commonly as a proxy in similar studies (e.g. Barnhart and Rosenstein, 1998; Gönenç and Scholtens, 2017; La Porta et al., 2001; Morck, Shleifer and Vishny, 1988). Additionally to following similar studies on ownership, I argue that using Tobin’s Q as a measure of intrinsic value of a firm reflects future potential better than other types of accounting based measures because of the young age of firms. Especially in the current times it is relatively common to have large prerevenue companies that live off a potential without actual revenues such as Facebook or Google for a long time, to name prominent examples. Therefore, using ROA or ROE as firm performance indicators would be wrong, which will be shown and inferred upon in robustness tests. Note that the assumption holds that Tobin's Q is a measure that implies time (as in expectancy) within its value, therefore the sets for last years values and panel data should be similar enough to obtain the same results, which is also found in the robustness tests. To be comprehensive, the main tests will still be using panel data analysis.
3.6 Relevant Independent Variables
The main independent variables are managerial ownership (OWNER) and region (REGION). For both variables I created dummy variables where the dummy takes one for nonmanagerial owned firms and zero otherwise as well as the region dummy, which takes the value for zero if the firm is located in the US and one for Europe. This separation is based on findings of different concentrations of ownership and family owners and provides a benchmark for the following variables. To test the international perspective, this paper follows the papers of La Porta et al. (1998, 1999 and 2001), who developed creditor rights scores and categorized the legal origins of the countries involved in this research. Within the framework of this research, investor rights (CR and ADR) are dealt with as a dummy for the individual creditor rights levels. The same procedure applies to legal origin (CIVIL and COMMON). The reason behind this is that creditor rights is not a continuous variable in itself. While there is strength indicated by having a smaller number, the individual dummies that make up the creditor rights value may not be of equal weight (La Porta et al., 1998).
3.7 Control Variables
Previous research shows that other variables may also influence the outcome of this type of research. In the regression analyses at a later stage, I therefore control for multiple factors that may influence the dependent variable firm performance and the independent variables, managerial ownership and region. These control variables are categorized into size and countrylevel variables.
With respect to the sizelevel control variables, I control for firm size (FIRMSIZE), measured as the natural logarithm of market capitalization. Another firm size proxy is controlled for, employee “size” (EMPLOYEES), measured as the natural logarithm of number of employees. Further, to include the managerial aspect into firm size and the focus on employees, I control for number of managers (MANAGERS) and again use the natural logarithm for this measure.
Additionally, I will use controls for industry and test with switching firm performance proxies to see whether the Tobin’s Q is the driving factor instead of other firm performance proxies for the results. To do so, I use the Standard Industrial Code categorization that can be found in ORBIS. Due to size and amount of variables within the table as well as their results, those tests are moved to robustness.
3.8 Variance analysis models
In order to test whether there are actually significant differences between the means of different dependent and independent variables, I conduct MannWhitney U tests. A similar approach has been conducted in Gönenç and Scholtens (2017). For hypothesis 1, I check for significant differences between Tobin's Qs under managerial and under outside ownership. Further, I want to test whether there are significant differences in means that predict higher values for managerial ownership, compared to outside ownership, which is hypothesis 2. Additionally I test for variances between regions for managerial ownership, followed by testing whether one region has a stronger stronger effect of managerial ownership on Tobin's Q. The selection of this test originates from the significant tests for homogeneity of variance (Levene’s test) and normality (Skewness/Kurtosis and ShapiroWilk) for all datasets and for both each Tobin's Q as well as the natural logarithm of Tobin's Q.
3.9 LogLog regression models
similar papers again (e.g. Barnhart and Rosenstein, 1998; Gönenç and Scholtens, 2017; La Porta et al., 2001; Morck, Shleifer and Vishny, 1988). The data utilized within these equations are panel data, which includes static information that is timeinvariant (legal origin, etc.). Throughout the models, I use the natural logarithm of Tobin's Q (LNQ) as dependent variable. Due to the still extreme values of Tobin's Q from the United States, further trimming may not be ideal and not be equally possible, hence the logarithmic transformation to normalize distribution. The methods are regressed using random effects.
3.10 Random Effects model
To avoid STATA dropping timeinvariant variables out of the dataset due to collinearity with the ID, I use random effects models instead of fixed effects. Most of the dependent variables used are timeinvariant and fixed effects do not allow timeinvariant variables. Additionally, it is possible that differences across the firms in the sample have some influence on firm performance. To ensure that the results are valid, random effects are used across firms.
Table 1
Definition of relevant variables, their respective measurement and part within this study. Variable Unit of Measurement Usage Description
LN Tobins Q Ratio Dependent Variable The natural logarithm of Tobins Q as described below Tobins Q Natural Logarithm Alternative Dep.
Variable
ORBIS' ratio of market capitalization/assets; Confirmed by the data specialist of ORBIS Bureau van Dijk
Ownership Dummy Independent Variable 1 if the firm is owned by outside shareholders (managers <50% ownership) and 0 if managers own company (>50% combined)
Region Dummy Independent Variable 1 if the firm is from the Europe and 0 for companies from the United States
Creditor Rights Categorical Independent Variable Strength of creditor rights as defined by La Porta et al. (1998), ranging from 0 to 4; 0 being low, 4 being high investor protection
Shareholder Rights Categorical Independent Variable Strength of shareholder rights as defined by La Porta et al. (1998), ranging from 0 to 6; 0 being low, 6 being high investor protection
Legal Origin Categorical Independent Variable Legal origin of firm country as described in La Porta et al. (1998) LN Market Cap Natural Logarithm Control Variable Market capitalization of firm
LN Employees Natural Logarithm Control Variable Number of employees in firm (estimation) LN Managers Natural Logarithm Control Variable Number of managers in firm
4 Hypotheses and Empirical Models
In this study I focus on publicly listed firms that incorporated after the dotcombubble 2001, which I call “young firms”. I use data of firms from the United States and Europe. The structure applied is a funnel approach. In the wake of having a large amount of comparable and potentially pivotal proxies for the international perspective, empirical testing will be done stepwise, introducing variables successively and adopt different combinations to better comprehend their influence. Using this strategy aims to improve the validity from a reader’s view. In the following I use the findings outlined in the literature review to point into a direction, which then creates hypotheses. Additionally, for ease of read and to prevent repetition of theory, I include the regression equations and analyses methods following from the hypothesis, whose methodology is explained in section three and is applied in section five. This way it should become clear immediately how this section (four) is put into practise in the following section (five). A definition of variables can be found in Table A1 and a description of the variables utilized in the model can be found in the respective table descriptions. Equations using analyses of variance testing are denoted (1) thru (5) and panel data regressions are denoted (6) thru (13).
4.1 Null Hypothesis
4.2 Ownership Difference and Direction Model and Hypotheses
As discussed before, there are multiple forces affecting behavior of the agents, which leads to the hypothesis of difference between the two groups. The forces of ownership types mostly lean on tacit and soft facts. PrincipalAgent theory (Jensen and Meckling, 1976) provided evidence that managers as agents not always work toward the same goals as their principles, especially if governance does not align these. Stewardship theory (Donaldson and Davis, 1991) argues too, that there are differences in firm performance when goals are automatically aligned by having the manager in full charge. Both agency and stewardship theory are leaning on Berle and Means (1932), that the selfinterest of managers combined with their ability to set directions, may affect firms. Additionally, Morck, Shleifer, and Vishny, (1988), Chen and Yu (2012) and Fama and Jensen (1983) provide plenty evidence that firms with different types of owners are not performing equally. Due to the effects of agency cost, stewardship theory and past research on family firms, especially insights into the comparable first generation family firms, I argue that there are also differences when managers own a firm, much like there are differences when families own firms, compared to nonfamily firms. Consequently, I hypothesize the following:
H1: There are differences in firm performance between firms owned by managers and those who are not owned by managers
Hence, for hypothesis 1, I check for significant differences between Tobin's Q under managerial and under outside ownership:
μ Q1 * ManagerialOwnership = Q
/
μ2 * OutsideOwnership (2)
that goes the “extra mile”, outperforming others, consequently creating a positive effect. These findings are heavily inspired by findings on familyfirms, who have lower agency costs (Jensen and Meckling, 1976), think more longterm (GómezMeija et al., 2011), resulting in a potentially more altruistic course of conduct (Schulze, Lubatkin and Dino, 2002) and still are not found to be expropriating wealth from minority shareholders (Croci, Gönenç and Ozkan, 2012). Rounding up the argumentation, agents may entrench themselves into the company, with exclusively negative repercussions, creating another type of friction through agency costs (see section 2.6). By implication, I argue that on one hand negative effects of agency costs and on the other hand positive effects of stewardship benefit one side more than the other, not only explaining that the groups are different, but also that there are significant firm performance differences in favor of those that are owned by the managers. Thus, I hypothesize that:
H2: Managerial owned firms perform better than outside owned firms
Hence, I want to test whether there are significant differences in means that predict higher values for managerial ownership, compared to outside ownership, which is H2:
μ Q1 * ManagerialOwnership> μ Q2 * OutsideOwnership (3)
Further, I argue that there is a significantly positive and economically relevant relationship for ownership on firm performance in a simple model that serves as benchmark for following tests, leading to following panel regression equation, estimated using random effects:
LN Q i,t = α 0 + γ1OW N ER ′i + γ2REGION ′i + (c i+ u ) i,t (6)
4.3 International Perspective
Europe (4.3.1), the legal origin hypotheses follow (4.3.2), finalized by investor protection, which is both creditor and shareholder rights (4.3.3).
Even in the hypothetical case of not finding differences in firm performance by purely separating ownership types in the whole dataset, these findings may still be swayed by the regions used. If there are differences in ownership effects on firm performance, but with different directions, they would not be visible in the tests before, balancing each other out. Hence, the international perspective testing adds another significance level and robustness.
4.3.1 Ownership Country Difference and Direction Model and Hypotheses
Additionally to ownership, I argue that there are differences between the region in which the firm is located. I derive this argument from the same studies that focused on family firms and managerial ownership, who also limited or compared their data sets to regions. However the main focus for this part, using the plain regional separation, serves as methodological and theoretical benchmark for the following sections and the separations according to La Porta et al. (1997, 2001 and 2008). Here, I hypothesize that:
H3: There are differences between managerial owned firm performances in US and European firms
Hence, I want to test whether there are significant differences in means of managerial ownership in the United States, compared to managerial ownership in the Europe, which is H3:
μ Q3 * ManagerialOwnershipU S = Q
/
μ4 * ManagerialOwnershipEurope (4)
provide a ground to argue for international differences. The basic panel regression is estimated using random effects:
LN Q i,t = α 0 + γ1OW N ER ′i + γ2REGION ′i + γ3OW N ER′i* REGION ′i + (c i+ u ) i,t (7)
Leading further, I argue that the effects of managerial ownership are stronger for firms owned by managers in the United States, again serving as benchmark. Thus, I hypothesize that:
H4: Managerial ownership effects are stronger for US firms than European firms
Hence, I want to test whether there are significant differences in means that predict higher values for managerial ownership in the US, compared to managerial ownership in Europe:
μ Q3 * ManagerialOwnershipU S> μ Q4 * ManagerialOwnershipEurope (5)
4.3.2 Legal Origin Models and Hypothesis
Additionally to ownership direction and the regional hypothesis, I want to go into more detail and argue that the legal origin of a company is a performance determinant of a company. In the legal origins theory of La Porta et al. (1997 and 2008), the authors argue that two different types of legal tradition, common and code law, affect economic outcomes based on their legal specific determinants; better protection of investors, less government regulation, less government ownership and better enforcement through judicial systems that are more independent and less formalized. In the outlined scenario, common law countries outperform code law countries. The stronger protection of investors that La Porta et al. found in common law countries determines a better breeding ground for investors with better developed capital markets (La Porta et al., 1997), because enforcement of their rights incentivizes these investors to finance firms. On the other hand, weaker rights in for example french law somehow pose a bottleneck, which is supported by the findings of higher concentration of ownership in french civil law countries. Following these findings, the fifth hypothesis is formulated as follows:
H5: The effect of managerial ownership firm performance is positive for firms located in countries with
Hence, I am creating dummy variables for each legal origin to obtain individual results, which can then be assessed upon economic relevance and significance. To prevent overloading this model with too similar variables, I replace REGION with the two variables, COMMON and CIVIL to be region indicators:
LN Q i,t = α 0 + γ1OW N ER′i+ γ2COM M ON ′i + γ3CIV IL ′i + (c i+ u ) i,t (8)
Further, to assess the specific effect to managerial ownership, I again create an interaction variable of ownership with the other relevant independent variable, common law and civil law, respectively. To further substantiate potential findings, I include the control variables FIRMSIZE, MANAGERS and EMPLOYEES to determine the impact of legal origin on firm performance:
LN Q i,t = α 0 + γ1OW N ER′i+ γ2COM M ON ′i + β3F IRM SIZE′i,t+ β4M AN AGERS ′i,t
β5EM P LOY EES γ6OW N ER OM M ON (c ) + ′i,t + ′i* C ′i+ i+ ui,t
(9)
LN Q i,t = α 0 + γ1OW N ER′i+ γ2CIV IL ′i + β3F IRM SIZE′i,t+ β4M AN AGERS ′i,t
β5EM P LOY EES γ6OW N ER IV IL (c ) + ′i,t + ′i* C ′i+ i+ ui,t
(10)
4.3.3 Creditor and Shareholder Rights Models and Hypothesis
Additionally to investor protection and legal origins, the missing puzzle piece following from the famous papers of La Porta et al. (1997, 1998, 1999a, 1999b, 2001 and 2008) are investor rights. While the regional separation model differs from code vs common law mainly in switching the United Kingdom from one side to the other (compared to dummy REGION), investor rights are specific to their respective countries. Creditor rights are ranging from 0 to 4 and shareholder rights are ranging from 0 to 6, each indicating stronger rights of investors with increasing values (La Porta et al., 1998). Further, the authors explain weaker investor rights to penalize firms with lower valuations, which is crucial to this study, using Tobin's Q as firm performance proxy. Consequently, I hypothesize:
In the following models, CR are used as continuous variable as the methodology of La Porta et al. (1998) may be laid out. A high value for creditor rights being strong, a low value of creditor rights being weak. Hence, the higher the CR coefficient, the stronger the effect of creditor rights, while the sign indicates the direction, e.g. negative signage indicating weak creditor rights to be bad for firm performance and vice versa. Further, to assess the specific effect to managerial ownership, I again create an interaction variable of ownership with the other relevant independent variable, creditor rights (CR) and shareholder rights (ADR), respectively. To further substantiate potential findings, I include the control variables FIRMSIZE, MANAGERS and EMPLOYEES to determine the impact of investor rights on firm performance:
LN Q i,t = α 0 + γ1OW N ER′i+ γ2CR ′i + β3F IRM SIZE′i,t+ β4M AN AGERS ′i,t
β5EM P LOY EES γ6OW N ER R (c ) + ′i,t + ′i* C ′i+ i+ ui,t
(11)
To test the effects of shareholder rights and obtain directly comparable results, I switch creditor rights with shareholder rights
LN Q i,t = α 0 + γ1OW N ER′i+ γ2ADR ′i + β3F IRM SIZE′i,t+ β4M AN AGERS ′i,t
β5EM P LOY EES γ6OW N ER DR (c ) + ′i,t + ′i* A ′i+ i+ ui,t
(12)
Lastly, all variables employed in this study combined into one regression equation to substantiate potential economic significance of the variables:
LN Q i,t = α 0 + γ1OW N ER′i+ γ2OW N ER′i* REGION ′i + γ3COM M ON ′i + γ4CIV IL ′i
5F IRM SIZE 6EM P LOY EES 7M AN AGERS 8CR 9REGION (c ) + β ′i,t + β ′i,t + β ′i,t+ γ ′i+ γ ′i+ i+ ui,t
(13)
5 Empirical results
This part of the study presents the results of the aforementioned analyses carried out. Descriptive statistics begin with an impression of the data at hand by showing the summary statistics as well as correlation matrix of the selected relevant variables of the panel data in table 2. A full description can be found in table A1. In the following, I start with testing the distributions and variances combined, because they deliver the same results for the US and Europe (5.1). Afterwards, individual samples for the US firms are tested (5.2), then the European firms are tackled (5.3), to get a first individual impression of their results, following the research of La Porta et al. (1998) suggesting that different countries are heterogeneous. These two sections paint the dry picture of the results received, which are interpreted subsequently. After doing so, I dig deeper and explain further, based on the whole sample. The same strategy applies to the explanation of empirical results that was laid out in section four; I am building up the argumentation stepwise to empirically explain the independent variables used and understand their impact, finally leading to a conclusion. The intention is to let the reader observe how the variables and setups affect ownership impacts on firm performance, which is the main aim of this study. The dummy “ownership” is thus the only variable that stays constantly in the regressions. The empirical results are concluded by various robustness tests in different setups.
5.1 Overlap US and Europe for Homogeneity of Variance and Normality
For both data sets I begin by testing for normality. Both the regular as well as natural logarithm of Tobin's Q show significant results for their normality tests, therefore I reject the null hypothesis of normal distribution. Additionally, I conduct a Levene’s test for equality of variances, for which I also obtain significant results and thus again fail to reject the null hypothesis, affirming homoscedasticity (see tables AppendixC). Based on the findings, I conduct nonparametric tests instead, to compare the two independent groups of ownership types, namely the MannWhitney U, which does not assume normality. To test robustness, I also include ttests. To get a clear picture of the values in this study, both panel and single year values of both the logarithmized and raw values of Tobin's Q are used to visualize their comparability. The assumption holds that Tobin's Q implies time within its value. To be precise and comprehensive, the main regressions are using exclusively panel data.
Table 2
Descriptive statistics in panel A report and compare number of observations (N), mean, standard deviation, minimum values and maximum values for the whole sample, with manager as an owner and for outside owners. This sample comprises 2005 observations for both US and European firms with a sample period between 2001 and 2015. A full description of variables is presented in table A1. Panel B of this table reports the correlations of the relevant variables, defined in table 1. Panel A: Summary Statistics of the variables used in the joint sample after reshape for panel analysis
All Sample Manager as Owner Outside Owner
Variables N Mean Std. Dev. Min Max N Mean Dev.Std. Min Max N Mean Dev.Std. Min Max LNTobinsQ 26,065 0.534 1.544 2.386 5.412 4,511 0.927 1.953 2.333 5.412 21,554 0.452 1.430 2.386 5.403 LNMarketCap 26,065 16.806 3.110 1.833 26.387 4,511 15.444 3.196 1.833 26.387 21,554 17.091 3.015 2.259 25.204 LNEmployees 21,333 4.121 2.608 0 13.323 3,289 3.116 2.516 0 11.043 18,044 4.304 2.582 0 13.323 LNManagers 25,766 2.139 0.802 0 4.812 4,407 1.671 0.829 0 4.812 21,359 2.236 0.761 0 4.804 Ownership 26,065 0.173 0.378 0 1 4,511 1 0 1 1 21,554 0 0 0 0 Region 26,065 0.459 0.498 0 1 4,511 0.450 0.498 0 1 21,554 0.461 0.499 0 1 Owner x Region 26,065 0.078 0.268 0 1 4,511 0.450 0.498 0 1 21,554 0.000 0.000 0 0 Shareholder Rights 17,108 4.521 1.049 0 5 3,016 4.375 1.236 0 5 14,092 4.553 1.002 0 5 Creditor Rights 25,428 1.708 1.282 0 4 4,316 1.283 0.860 0 4 21,112 1.794 1.335 0 4 Common Law 25,558 0.762 0.426 0 1 4,342 0.608 0.488 0 1 21,216 0.794 0.405 0 1 Civil Law 26,078 0.203 0.403 0 1 4,511 0.360 0.480 0 1 21,554 0.171 0.376 0 1 Panel B: Correlations matrix 1 2 3 4 5 6 7 8 9 10 11 1 LNTobinsQ 1.0000 2 LNMarketCap 0.1606 1.0000 3 LNEmployees 0.3513 0.6771 1.0000 4 LNManagers 0.3643 0.6414 0.6068 1.0000 5 Ownership 0.1165 0.2004 0.1645 0.2653 1.0000 6 Region 0.3355 0.1922 0.1783 0.4396 0.0090 1.0000 7 Owner x Region 0.1338 0.0554 0.0189 0.0037 0.6349 0.3151 1.0000 8 Shareholder Rights 0.2201 0.1600 0.1145 0.3290 0.0645 0.9109 0.4407 1.0000 9 Creditor Rights 0.2156 0.2371 0.2298 0.3643 0.1497 0.5956 0.0212 0.4064 1.0000 10 Common Law 0.2164 0.0007 0.0145 0.1657 0.1638 0.5997 0.4170 0.9173 0.2052 1.0000 11 Civil Law 0.2145 0.0206 0.0307 0.1318 0.1781 0.5484 0.4313 0.8342 0.1905 0.9155 1.000 5.2 US Firms
results indicate that the medians are different with a significance level of 1%, I thus reject the null hypothesis that the means of managerial ownership and outside ownership on Tobin's Q are equal, indicating differences between the two types of ownership, showing z = 8.404 (7.649), p = 0.000 (0.000). Additionally, the output shows a porder value of 0.596 (0.668), indicating that in 59.6% (66.8%) of random draw cases, Tobin's Q performance would be higher if there is managerial ownership present, providing support for hypothesis 2 in the US sample. The same results are received for the logarithmized as well as raw forms of Tobin's Q and also show the comparability of Tobin's Q values at one point in time to timeseries panel data.
Panel B of table 3 adds some robustness to the rank sum tests performed and further substantiates the findings. The results of the twotailed ttest on the natural logarithm of Tobin's Q show that in the United States, firms with owners who are also managers have statistically significant higher Tobin's Q (1.9815 +/ 2.26682) than those owned by outsiders (1.1562 +/ 2.0651), t(4889) = 9.569, p=0.000. Those results are also supported by testing the original value of Tobin's Q on last year’s values, obtaining statistically significant higher Tobin's Q (70.1332 +/ 131.1039) than those owned by outsiders (20.3622 +/ 67.0274), t(876) = 7.889, p=0.000.
5.3 European firms
I am testing a sample of 4766 (878) observations, of which 601 (145) are management owned and 4165 (733) outside owned. The results indicate that the medians are different with a significance level of 1%, thus I reject the null hypothesis that the means of managerial ownership and outside ownership on Tobin's Q are equal, indicating differences between the two types of ownership, showing z = 2.884 (2.985), p = 0.004 (0.003). Additionally, the output shows a porder value of 0.464 (0.422), indicating that in only 46.4% (42.2%) of random draw cases, Tobin's Q performance would be higher if there is managerial ownership present, thus not providing support for hypothesis 2 in the European sample and indicating differences for the meaning of managerial ownership in between those regions.