Efficiency and return in Norwegian family firms

Transcription

1 Efficiency and return in Norwegian family firms - Are family firms more efficient than non-family firms? - Caroline Brudvi Brøsholen GRA19003 Master Thesis Supervisor: Øyvind Bøhren Hand-in date: Campus: BI Oslo Programme: Master of Science in Business Major in Finance This thesis is a part of the MSc program at BI Norwegian Business School. The school takes no responsibility for the methods used, results found and conclusions drawn.

2 Abstract This paper examines if there is a relationship between family ownership, efficiency and profitability. Total factor productivity will be used to measure efficiency and Return on Asset will represent the profitability of the firms. This study finds family firms to be more efficient and more profitable than non-family firms. The results regarding efficiency also vary with different thresholds of ownership stake and an increasing family ownership stake gives higher efficiency. Firms that are fully owned by family members are most efficient. In addition the results imply that family firms are more capital constrained and smaller in size than other firms. It appears to be a positive relationship between efficiency and Return on Asset, however this effect most prominent in non-family firms. This study is based on both large and small privately held Norwegian companies. Acknowledgement I would like to thank my supervisor Øyvind Bøhren for his guidance and constructive feedbacks and the Centre of Corporate Governance Research for help with extracting data. 1

3 1. INTRODUCTION FAMILY FIRMS EFFICIENCY DIFFERENCES PERFORMANCE DIFFERENCES RESEARCH ON NON-LISTED FIRMS MAIN HYPOTHESES/RESEARCH QUESTION MOTIVATION EARLIER RESEARCH AGENCY PROBLEMS AND ADVANTAGES DIFFERENCES BETWEEN FAMILY FIRMS AND NON-FAMILY FIRMS EFFICIENCY TOTAL FACTOR PRODUCTIVITY MEASURE TOTAL FACTOR PRODUCTIVITY AND FAMILY FIRMS IN SPAIN RETURN ON ASSET METHODOLOGY EFFICIENCY PERFORMANCE EXPECTATIONS EFFICIENCY PERFORMANCE DATA PANEL DATA DATA SOURCE FILTERS: RESULTS DESCRIPTIVES Differences between family and non-family firms Financial constraint and capital structure EFFICIENCY PERFORMANCE OTHER CONSIDERATIONS AND WEAKNESSES CONCLUSION APPENDIX MODELS Model: Capital constraint model ANALYSIS

4 8.2.1 Descriptives Exhibit 1.1 Mean and Median Efficiency results Exhibit 2.2 Efficiency results with all independent variables lagged Exhibit 2.3 Efficiency results without control variables Exhibit 2.4 Redundant fixed effects test Exhibit 2.5 Fixed effects test Ownership effects Exhibit 3.1 Efficiency result with various ownership structure Correlation matrices Exhibit 4.1 Correlation matrix without lagged governance variables Exhibit 4.2 Correlation matrix with lagged governance variables Model testing Exhibit 5.1 Augmented Dickey Fuller test Exhibit 5.2 Goodness of fit statistics for efficiency regression REFERENCE LIST:

5 1. Introduction This section will present the most important factors in my analysis in addition to my research question and motivation for writing this thesis. 1.1 Family firms Earlier research has shown that Norwegian family firms have higher return on assets (ROA) than other firms (Bøhren, 2011). This study is based on numbers for all Norwegian AS and ASA from When controlling for age, industry and size these difference in return on asset becomes very small and is caused by small companies with one owner. Other family firms are not more efficient than non-family firms. Higher return on asset have been found in other countries, such as the US (Anderson and Reeb, 2003) and Western Europe (Maury, 2006). The remaining question is why family firms have a higher ROA. High return on assets is beneficial for the firm, and exploring how these firms obtain this advantage is very interesting and useful. Knowing what factors that affect return on assets can help managers to improve the profitability in their firm. It is therefore of interest to look at which factors that makes family firms special. Throughout this paper my definition of family ownership is an ownership stake of more than 50% by blood or marriage relatives, with exception of the times I relate to previous studies where a variety of thresholds have been used. This threshold is lower than in several of the previous studies conducted on family firms. The reason for this high ownership stake is that I am using privately held firms and the ownership stake needs to be higher to maintain control of the firm than in public firms. Lastly it is worth mentioning that family firms is the most common organization form in the world (Lee. 2004) 1.2 Efficiency differences To investigate the difference in profitability further I have chosen to look at efficiency. Higher efficiency can result in a higher net operating profit, and hence give a higher return on assets. In addition I believe that a capital constrain and lower agency problems in family firms should increase the investment 4

6 efficiency in family firms by investing in the most profitable projects. This will be discussed further under the section 2.2. A study on efficiency and profitability has been conducted on U.S data on large listed companies (Lee, Jim. 2004) and Spanish listed companies (Gorriz and Fumas. 2011), but I do not know of any earlier studies of family firms and efficiency using Norwegian data. In addition this study will be based on private held companies and the sample consists of both large and small companies. Another important part of the efficiency story is that previous research has shown that family firms might be financially constrained due to their wish to maintain control over the company. Their need for majority ownership affects their access to both debt and equity. With simple majority ownership the family have the power to control the company though general ensembles. This means that they have power related to selection of board, which ultimately chooses the manager of the firm (Bøhren, 2011). Without majority ownership they will loose some of the benefits they have regarding control of the firm. I will further elaborate on this under section Performance differences As mentioned earlier, there has been found a difference in profitability between family and non-family firms and Bøhren (2011) provides four potential explanations for why family firms have higher return on asset. Three of these are related to agency problems and the last explanation is related to a long-term perspective. The agency problems that Bøhren (2011) suggests that are less prominent in family firms are the conflict between owner and management (A1), majority- and minority shareholders (A2) and owners and other stakeholders (A4). I will elaborate on this under the paragraph agency problems and these agency problems will be an important part of this paper. It is important to notice that there is one disadvantage regarding the governance advantages family firms have with regard to agency problems. In order to maintain control of the company family firms might be more capital constrained. 5

7 1.4 Research on non-listed firms Berzins, Bøhren and Rydland (2008) raise the issue that very little research exists on private firms within the area of corporate finance and corporate governance. They suggest two reasons for this. Firstly, financial economists might prefer publicly listed firms because they can observe the market value as oppose to the book value. Secondly, information on unlisted firms is harder to obtain because of differences in accounting requirement and openness around their accounting. They also point out two reasons why the lack of this kind of research is problematic. In a macro economy perspective, the market consists of a much larger fraction of the non-listed firms than listed firms. Out of all limited liability firms in Norway from , non-listed firms on average employed 78% of the people working in limited liability firms. Non-listed firms also accounted for 78% of total sales, 69% of total assets and represented 99,8% of all limited liability firms in this period. Secondly generalization of existing literature on listed companies to unlisted companies is problematic because of essential differences including transparency, financing of the firm and access to future funds, high transaction costs associated with trading of shares, and protection of minority shareholders. Bøhren (2011) also points out that ownership and control in listed firms are clearly divided. There are many owners, low proportion of the owners represented on the board and the managers usually have a low stake in the company. These factors may, according to corporate finance theory, affect behavior regarding real investments, capital structure, dividend policy and risk management. Growth and return on capital invested may also be affected (Berzins, Bøhren and Rydland, 2008). An important difference between public and private firms is that shareholders of public firms are diversified and usually there are a larger number of shareholders than in private firms. With a large number of shareholders, it is reasonable to believe that some of these owners do not take an active interest in the company and the decision making. This might imply that family members need a higher ownership stake to maintain control over the family firm if the firm if private. Most previous studies on differences between family and non-family firms define family firms with a lower threshold regarding ownership than I will 6

8 use in this study. These studies are based on public firms and this supports my argument above. I also believe that there might be more focus on long-term profits in privately held firms. A large fraction of the shareholders in public firms are only interested in the return on their investment and not the company itself. A consequence of a long-term perspective is that in the short-term the company might be less profitable than companies with a short-term perspective of generating returns Pressure from shareholders might lead to focus on creating high returns, which might contribute to a short-term focus of generating profits. Shareholders of privately held firms are often less diversified and their stock is less liquid, which should incentivize a long-term perspective because they usually have a long investment horizon and are interested in the long term survival of the company. Overall this suggests that managers in privately held companies are less concerned about the short-term perspective than public firms Most of the studies conducted on family firms, regarding returns and efficiency, are done on publicly traded companies. Due to the data that is available through the Centre of Corporate Governance database it is now possible to investigate the unexplored area of private held family firms. 1.5 Main hypotheses/research question A. Family firms are more efficient than non-family firms B. Family firms have higher return on assets, and efficiency contributes to this higher profitability 1.6 Motivation The goal of this paper is to investigate if family firms are more efficient than other firms, and if the efficiency difference is an important reason for higher profitability. Since there has been conflicting evidence in studies conducted on data from different countries, it will be interesting to look at the effects in Norway, especially because there has been found a slightly higher return on asset for Norwegian family firms compared to Norwegian non-family firms. In addition I will use data on privately held firms, which is not very common in pervious 7

9 papers. For previous mentioned reasons under section 1.4 it might not be correct to generalize the result found for public firm and apply them for privately held firms. Private firms often have fewer shareholders and less access to both debt and equity because of illiquidity and lower transparency. In addition I believe that the effect of access to capital will be less prominent using privately held companies. The illiquidity associated with privately held firms, makes it more difficult for non-family firms to raise equity as well. When I comes to debt it might be more difficult to get access to debt because privately held companies does usually not disclose as much information as public firms. This paper will therefor be a contribution both to the new line of research on private family firms and also to the distinction between efficiency and profitability for Norwegian family firms. 2. Earlier research This section will present earlier research done on important topics included in this paper, including explanation of agency problems, which is a key factor in this paper. This section also explains the total factor productivity, which is an important part of the methodology. 2.1 Agency problems and advantages Agency problems and costs are important topics related to family ownership. Agency cost is the value loss caused by an agent being more informed and having different preferences than the principal. This cost will be zero when the principal and the agent are the same. This problem is not accounted for in the standard microeconomic model because this model assumes no conflict between the parties (Bøhren, 2011). Bøhren (2011) refers to four types of agency problems in his book. The first problem (A1) is between the owner and the management. The second problem (A2) is between majority and minority shareholders. The third problem (A3) is between the owners and the creditors and the last agency problem (A4) is between the owner and other stakeholders than management and creditors. Agency problems A1, A2 and A4 might be lower in family firms according to Bøhren (2011). One way to reduce agency costs of A1 problems is to 8

10 make the decision makers the residual claimants. This means that the group of people or person making the decision will also be the ones benefiting from enhanced return in the company (Gorriz and Fumas, 1996). When the decision maker and the residual claimant is the same person or group of people there is no need for discipline and monitoring of the decision maker. It is common for family firms to elect a family member as their CEO. Because a family firm is defined as a firms with family ownership above 50% or more in this paper, the family is often able to elect a family CEO, because of their majority power at general ensembles. This means that the management and owner often is one and the same or at least have the same objectives and interest for the company s future. This implies that A1 should be low. The problem between majority and minority shareholders (A2) is often the possibility for majority shareholders to extract special benefits from the company because they as majority shareholder have the power to make that kind of decision. The majority shareholder will per definition be the family in this thesis. The reason why this problem is expected to be low in family firms is that the extraction of non-pecuniary benefits also comes at a cost for the company (Bøhren, 2011). In family firms the family often have a higher ownership concentration than in other firms and this implies that the family members can extract benefits, but also that they carry a large part of the cost of these benefits. This implies that family members are expected to extract less private benefits from the company because of the cost and that the A2 problem should be lower in family firms. Bøhren (2011) also points out that there might be a problem regarding value loss because of risk reduction, which can cause a conflict between majority- and minority shareholders. Since family firms are not diversified in the same way as other owners they would want to reduce their risk through other channels, like investing less to avoid more debt financing. This risk reduction might not benefit minority shareholders and might reduce the overall profitability of the firm. This gives rise to the conflict between majority and minority owners (A2). A4 is connected to corporate social responsibility (CSR). I expect that family firms will be more concerned with preserving their reputation than other firms because their family name is connected to the company. This means that the agency conflict between owner and other stakeholders should be lower for family 9

11 firms and I would expect family firms to have higher focus on CSR than other firms. In addition connection with other stakeholders might create economic value for the firm and this relationship should be maintained (Bøhren, 2011). Magnussen and Sundelius (2011) points out that the third agency problem is expected to be high in family firms because family firms have less incentive for transparency, which will cause the banks to be more reluctant to grant them loans. Family ownership facilitates several advantages over other organizational forms. As mentioned earlier, monitoring is an important part of the difference between family and non-family firms. This might lead to improved investment efficiency, which will contribute to maximizing the performance of the firm. James (1999) found that family firms often have longer investment periods than firms owned by other shareholders. When looking at performance, the objective of the decision maker is important, because the performance will not be maximized if the decision maker has other objectives. With family ownership, succeeding generations is an important factor when defining the goals for the company. For non-family firms the goal is often to maximize the profit for shareholders, but for family firms goals are also relate to survival rate, reputation and succession. This can result in either forgoing positive NPV projects because they are careful or that they make efficient investment decision because the managers use their limited funds wisely and do not overinvest. Andersson and Reeb (2003) investigate agency problems and their paper suggests that family ownership reduces overall agency problems without severely reducing efficiency in decision making in their sample of public firms. Some of the arguments described above also apply to other majority owners that are not family, but there are also some differences between family ownership and other majority ownership that makes the statements above more prominent for family firms. Bøhren (2011) points out that family firms more often have one of the owners involved in the managing if the company. On average 96% of the ownerships is represented on the board of directors and management in family firms, compared to 54% for other companies. In family firms the largest owner (the family) is more often the Chairman of the board (89% of the cases) and CEO (83% of the cases) than in other firms (38%, 39% of the cases). The largest shareholder is also more often a part of the board of director in family firms (94%) than in other firms (66%). The largest owner has a very large 10

12 fraction of the board position in family firms (86% of the board), where the largest owner in other firms only have 36% of the positions. These statistics show that the A1 agency problems should be close to nonexistent in family firms because the owner often is the decision maker as well. The A2 problem should also be smaller in family firm because the ownership stakes is often a larger in family firms. If an owner only have simple majority the A2 problem becomes more important because even though the owner only owns 50,1% of the firm, it is still possible for that group or person to control the company. The owner can then take advantage of non-pecuniary benefits, like overspending employee benefits, even though they only paid 50,1% of the cost. With a high concentrated ownership the benefits of exploiting minority shareholders becomes small, because the majority owner covers a large part of the cost of these non-pecuniary benefits. 2.2 Differences between family firms and non-family firms We know that there are several variables that make family firms differ from non-family firms. Earlier research gives an indication that dividend (Setia- Atmaja, 2010) and growth (Magnussen and Sundelius, 2011) significantly differ between family and non-family firms. Magnussen and Sundelius (2011) found a positive difference in growth rates between non-family and family firms and also that family firms have a higher variation in their growth. A lower growth rate for family firms is consistent with Gorriz and Fumas (2011) article, which I will elaborate on under section 2.5. Magnussen and Sundelius (2011) also found that family firms have a higher debt/asset ratio which can mean that family firms do not forgo growth opportunities, as Gorriz and Fumas suggests in their article, but use debt financing for new projects instead of equity. As mentioned earlier the CEO in family firms is often a family member. The disadvantage with hiring a family member as a CEO is that you might miss out on a more qualified manager, which will probably result in lower performance than the company s full potential. The advantage of having a family CEO is that the CEO usually know the company and the business very well, which gives an entrepreneur effect. Having a family CEO provides specials knowledge that cannot be taught in a short-time perspective. 11

13 A very important difference between family firms and non-family firms is the incentive to maintain control. The company often has sentimental value to family members and their need to maintain control and keeping the firm in the family might inhibit their access to funds. Raising equity through other shareholders is difficult in family firms because this would mean diluting their own stake in the company. Bøhren (2011) points out that family firms often only have two ways of raising equity. Either they have to invest more of the family s wealth in the company or retain profit from operations. The firm will avoid issuing new shares to outside investors if this will jeopardize the family s control over the company. The concern regarding losing control over the company provides a financial constraint that non-family firms do not have. This might inhibit growth and family firms might not reach its full potential because of insufficient funds. Gorriz and Fumas (2011) imply that there is a binding financial constraint for family firms if they choose a lower ratio of capital to labour than non-family firms. Earlier research has also shown that family firms often are smaller than non-family firms (Bøhren, 2001). I find that family firms have less assets and fewer employees than non-family firms, thus they are smaller in size. Large firms often have lower risk than small firms. This is because large firms often are less dependent on individual customers and individual industries. Their customer basis is often larger and involved different segments, which makes them more diversified. There are some risk factors related to family ownership. Since family members often are both owners and work as CEO or are a member of the board of directors they have invested their fortune and get their income from the same company. There stake is highly undiversified and it is therefore natural that they want the firm to perform well in a long-term perspective. Overall I would say that family firms have more incentive to maximize the performance of the company and not just maximizing dividend or payouts to the shareholders. With a majority stake I would also say that they have the opportunity to achieve these objectives. In addition the family s social reputation is linked to survival of the company, which should indicate a long-term perspective. One way to reduce this risk is focus on financial health and reduction of operational risk. For example low fixed 12

14 costs, stable income, diversified production and a low debt ratio can be important factors. Investments is also an important factor, which might indicate that family firms should invest more efficiently as suggested earlier. Hence I believe that family firms have a longer perspective and that they in the long run probably will have higher profitability and higher efficiency than other organization forms. Overall, the most prominent differences between family and non-family firms is the access to equity, which is constrained because of their need to maintain majority stake. But also a long term perspective in family firms and that they are smaller in size. This is implied through lower economies of scale, more focus on risk reduction and the capital constraint do to the limited access to equity. 2.3 Efficiency Gorriz and Fumas (1996) have conducted a study on how performance relate to ownership structure by comparing public family firms with public nonfamily firms. They divided performance into productive efficiency and profitability and found that family firms have the same profitability as non-family firms, but on average higher productive efficiency. The reason for this is their argument about a size-growth constraint in family firms. Where they argue that family firms are more financial constrained than non-family firms because they do not want to jeopardize their controlling stake in the company. By issuing new shares to raise equity, they risk diluting their own stake in the company. By buying the newly issued shares themselves they will invest even more of their private wealth in the company and not be able to diversify their stake. When it comes to debt, family firms and non-family firms should be equally restricted by restrictions set by the bank. However Gorriz and Fumas (2011) claim that nonfamily firms are larger in size than family firms and therefore should also have easier access to debt. With a high stake invested in the company, the investors should also be more concerned about the risk associated with debt, implying that family firms should undertake less debt than non-family firms. Fumas and Gorriz have also conducted a more resent study with improved methodology on this subject. I will elaborate on this under section 2.5, since the 13

15 main methodology in this paper will be based on Gorriz and Fumas paper from A study conducted by Lee (2004) suggests that family members are more loyal and hence improve the firm s productivity. A paper based on U.S. data from the S&P 500 has found that family firms are more efficient in their use of labour and capital resources (Martikainen, Nikkinen and Vähämaa 2009). Another important factor is the agency problems that arise in firms. Bøhren (2011) refers to four important agency problems; which is elaborated under agency problems and advantages. These agency costs are related to indirect cost as defined earlier, and hence, they are also related to the efficiency of the firm. I believe that reduced conflict between different stakeholders in a company will contribute to increased efficiency. Lower conflict related to A1 problem can for example make investment policies more efficient. I expect that the family will choose to invest in the most profitable projects without taking excessive risk. If the CEO is a family member, he or she will most likely have the same preference as the owners and avoid empire building and avoid unnecessary risk. Palia and Lichtenberg (1999) found that managerial ownership changes and changes in productivity are positively related. 2.4 Total factor productivity measure To measure efficiency I will use the total factor productivity (TFP) measure. This measure has been used in several other studies of efficiency. Earlier research has shown a relationship between production efficiency and efficient use of labour and capital resources (Palia and Lichtenberg; 1999, Martikainen, Nikkinen and Vähämaa; 2009 and Gorriz and Fumas; 1996 and 2011). These four studies use the total factor productivity measure and this approach is often said to refer to technological efficiency or utilization of technology. There are several definitions of total factor productivity measure. The TFP measure is based on Solow residual, which defines an increase in productivity as an increase in output when capital and labour are held constant. TFP accounts for other determinants of economic growth than the defined input factors, which in this case is capital and labour (Holm and Trengereid, 2011). According to Solow (1957) the relationship between output and capital, labour and time form the 14

16 aggregated production function, where the time component allows for technical changes that might cause a shift in the production function. This can for example be slowdowns, speed-ups or more educated employees. Holm and Trengereid (2011) points out that technology is usually nonrivalry; hence higher TFP is probably caused by the employees or company s ability to use the available technology, rather than the technology itself. This ability can be related to knowledge regarding producing efficiently, reducing costs and ability to apply the technology across several aspects of the company. However, in this case the access to capital might not be the same for family and non-family firms. Because of this I use different production functions for family and non-family firms. I will also test if the use of different production functions is appropriate. The TFP measure will be used as an independent variable to check for efficient use of capital and labour. Since the family might not have the same access to capital, they might try to use the capital they have available more efficiently. A capital constraint implies that family firms do not have sufficient funds to invest in all positive NPV projects, and should therefor only invest in the projects with the highest NPV they can afford with their constraint. Gorriz and Fumas (2011) argue that family firms might forego positive NPV investments because they do not have excess to sufficient funds. I believe this often is the case, but also that this might make the firm more efficient. By only investing in high NPV projects, family firms will obtain a higher average NPV to capital invested ratio than firms that invest in all positive NPV projects. This illustrates one advantage of having low A1 problems. Owners want a high return on their investment and if manager s objectives are aligned with the owner s objectives, I believe that there will be more efficient use of capital. I will not directly test what part of the TFP that is higher in family firms, but provide some explanations. Efficient use of capital will be related to agency problems as argued above. The efficient use of labour can be related to employee s devotion to the shareholders. An earlier study conducted by Lee (2004) has shown that family firms have more loyal employees and employee productivity should therefore be higher. Another argument is that family firms might be knowledge companies and that technology might not be as important. I will look further into this by looking at the elasticity s of labour and capital for 15

17 family and non-family firms. If family firms are more labour intensive the total factor productivity measure might measure the ability to motivate and recruit employees. 2.5 Total factor productivity and family firms in Spain Gorriz and Fumas (2011) points out that if family ownership gives access to governance advantages that non-family ownership does not provide, and that firms have the same technological knowledge no matter what organization form, then family firms will transform inputs into outputs more efficiently. In addition to this they point out that this efficiency advantage does not necessarily result in higher profitability if family firms and non-family firms have different competitive advantages in the product market they operate in, or if family firms have an additional constraint related to maintaining control. This reasoning is illustrated in section 8.1 as model Gorriz and Fumas (2011) raise the question about a financial constraint related to family firms. They point out that as a result of family members wish to remain in control of the family company they face a capital constraint, which causes a size growth constraint. The company might need funding beyond family wealth and because family firms want to remain in control they are not able to raise equity in the same manner as non-family firms. This is mainly because of the fear of diluting their own stake in the company and maybe also forgoing potential private benefits related to family ownership. Gorriz and Fumas also argues that if the governance advantages of having family ownership have a higher or equally positive effect on performance then the negative effect of the capital constraint, then family companies will perform better or equally well as companies without this constraint. Such advantages might be less or no agency cost since family ownership provides either control through family members managing the company or by supervision. Gorriz and Fumas (2011) results show that family owned firms are smaller than non- family owned firms, but that they have the same profitability as non-family firms. Larger firms are related to higher profitability and it is therefore interesting to investigate why family and non-family firms have the same profitability. Gorriz and Fumas (2011) also found higher total factor productivity 16

18 in family firms and argues that this can be the reason why we see the same profitability in family and non-family firms. Another interesting factor is that studies have found that Norwegian family firms are more profitable than non-family firms, however this study is based on private firms (Berzins, Bøhren, Rydland, 2008), and hence Norwegian privately held family firms might not face the size constraint explained by Gorriz and Fumas (1996, 2011) However, this profitability difference almost disappears when the results are controlled for effects of size, industry, and age (Bøhren, 2011). Only family owned firms with only one owner are significantly more profitable than other firms. This on the other hand might imply that Fumas and Gorriz (2011) assumption of a capital constraint is correct. Similar to Gorriz and Fumas (2011), Jensen and Mekling (1979) also propose that different production possibility sets depends on ownership structure. They claim that the production function consists of technology, productive recourses and ownership structure. ROA should be highly affected by the firm s investment efficiency. Positive NPV investments generate a higher inflow than outflow of money and hence it will improve a firm s economical performance. It is preferable to invest in project with the highest net present value. If it is correct that family firms have limited access to capital, they might not be able to invest in all positive NPV projects that are available. By assuming that the managers are rational and that the manager s objective is to create value for the firm, I assume that the manager will invest in the projects with the highest NPV first. Consequently, family firms should have a higher average NPV divided with their capital investment, which is what I define as higher investment efficiency. If firms invest in negative NPV projects, they will destroy value and get a lower return on asset. This is the case for both family firms and non-family firms. I believe that the probability of investing in negative NPV projects is higher for non-family firms because they might have more funds available and the management s objectives might not be aligned with the owners objectives. On the other hand this does not mean that family firms create more value than non-family firms. 17

19 2.6 Return on asset A study on Norwegian data has shown that Norwegian family firms have a higher median ROA than Norwegian non-family firms (Bøhren, 2011). The average difference across the years 2000 to 2008 is 2% (Family firms: 6,7%, nonfamily firms 4,7%). However when the results are controlled for ownership mechanisms, size, age and industry the difference is only 0,8% in favor of family firms and this difference is caused by small family firms with one owner. There is no significant profitability difference for other family firms compared to nonfamily firms. The data used in this study involves almost exclusively non-listed firms. Lee (2004) raises an important issue regarding research done on higher return in family firms. He points out that it is a common mistake to confuse higher returns with higher productivity and efficiency. He refers to profitability as the financial performance of the firm, while efficiency refers to how well the company exploits their resources. Several studies have been done on family firms and higher returns but few studies have been done on productivity and efficiency. Andersson and Reeb (2003) investigates whether or not family firms preform better than non-family firms based on ROA and Tobin s q. They find that family firms have higher ROA and a higher Tobin s Q, and from this result they suggest that family firms are more efficient. This is an example of the problem Lee (2004) raises in his study. It is therefore interesting to investigate whether or not we can rightfully assume that family firms are more efficient if they are more profitable. Gorriz and Fumas (2011) also examine profitability and family ownership. They points out that family firms can operate with a lower return than their opportunity cost of capital even in competitive markets if the loss is compensated by amenities potential and non-pecuniary benefits. Since family firms often are majority shareholders they have the possibility to exploit these kinds of benefits. They also argue that this is probably not applicable to their study because they use public firms, but this might be true for private family firms. This means that privately held family firms may have lower performance than non-family firms and still survive in the market. If the stake in the company is very high, these benefits might not be exploited because the stakeholders also carry the cost of these benefits. 18

20 This means that the non-pecuniary benefits might differ in family firms and non-family firms. With a higher ownership stake the majority owner will also carry a larger part of the cost for the private benefits. If the ownership stake in family firms is higher than majority ownership stakes in other firms, family members are more likely to carry a larger part of the cost of private benefits and it would be reasonable to believe that the use of private benefits will be lower in family firms than other firms with a large (majority) owner. This again implies that A2 agency problem should be lower in family firms. Bøhren (2011) found that when it comes to ownership stake there is a difference between family firms and other firms 1. In family firms the individual ownership share is 78% compared to 53% for other companies. When it comes to largest family ownership, family firms have substantially higher ownership with 93% compared to 37% for other firms. Family firms might have higher incentives to invest long-term and establish long-term profitability, due to their high stake in the company, succession and that the family s reputations is closely related to the company. A long-term perspective might give lower short time performance, but higher longterm performance. A majority stake in the company makes the shareholders more concerned with survival rate than short horizon dividend payments and other payouts. Maury (2006) found that family firms with active family control have higher return on asset and Tobin s Q and in addition they have higher valuations if shareholder protection is high. However passive family ownership does not provide higher profitability. This result is related to agency problems A1 (owners and management) and A4 (owners and other stakeholders). If the family both owns and also has a top position in the company they have less A1 problems but there is a problem related to extracting private benefits. With high protection of minority shareholders the benefits of A1 is larger than the potential cost of A4. Lauterbach and Vaninsky (1999) conducted a study on the effect of separation of ownership and management on 280 listed Israeli companies. This study shows that family owner-managed firms are the least efficient 1 Other firms are defined as firms where the largest owner has a higher ownership stake than the family. This means family firms where the largest owner is not a family. 19

21 organizational form with regards of generating profits. They also find that nonowner-managed firms preform better than owner-managed firms. In addition higher debt is related to higher risk and family members should demand a higher return on capital if the debt level is raised. One important reason for this is that they are not diversified. This implies that if family firms should have higher profitability than non-family firms if they have the same debt level. This implies that family firms are restricted in their ability to obtain debt. There are also disadvantages with family ownership that might affect the profitability. Poor management might lead to lower performance and by electing a family member as the CEO or family members having a large fraction of the board seats, family firms might not recruit the most qualified decision makers. Access to equity is often also a problem as mentioned earlier because the family is reluctant to issue new shares and the only possible access to more equity is investing more of the family wealth or retaining profits. For the private firms used in the valuation difference caused by less transparency will also reduce the likelihood of issuing new shares. The capital constrained explained earlier should also reduce the size of the firms and there access to capital, which leads to less growth opportunities and gives lower total value creation. In addition to finding a connection between managerial ownership and productivity, Palia and Lichtenberg (1999) also found that firm value in the stock market increased with increased productivity. 3. Methodology This section will explain the methodology used in this paper, including the steps used to attain the main regression and proxies used in this paper. 3.1 Efficiency To examine whether family firms are more efficient or not, I am using a multiple that shows the added value relative to the resources being used as dependent variable. The main recourses I will use are labour and capital and I will use total factor productivity as a measure of efficiency. I will adopt the methodology used by Gorriz and Fumas (2011) and also use age and industry as control variables. 20

22 I will use total factor productivity as a proxy for efficiency throughout this paper. To measure efficiency I examine how sensitive variations in resources like capital and labour are and in addition how well the different types of firms exploit these recourses. I assume different production functions for family and non-family firms because of the difference in access to capital implied by Gorriz and Fumas (2011). I believe that the access to labour will be approximately the same for family and non-family firms and hence the main difference between the two types of firms should be access to capital and total factor productivity. To test my hypothesis the following definition of total factor productivity will be used: (1) (1.1) Numerator shows output while the denominator show input factors as a cobb- Douglas function, which does not include the total factor productivity measure. Where refers to the total factor productivity, Q refers to output and is a Cobb-Douglas production function with capital (K) and labour (L) as input factors. The TFP measure is used to capture effects on productivity that are not accounted for by these two input factors. By rearranging expression (1) I obtain the following equation: (2) Where is the TFP measure and α and β represent elasticity s to output for capital and labour respectively. Using a Cobb Douglas production function allows family firms and non-family firms to have different production functions, because of the elasticity s measured by the exponent. This is used to see if family and nonfamily firms have different production technologies. And the elasticity s are used to look at how sensitive the two groups of companies are to capital and labour. If output for family firms is sensitive to labour then this indicates that family firms are more labour intensive. If on the other hand the output for family firms is sensitive to capital this means that family firms should be highly affected by a potential capital constrained related to family ownership and also that they are more labour intensive. 21

23 The TFP measure will be used as an independent variable in the analysis and ownership will be added as a part of the input function. I will also include the effects ownership structure has on total factor productivity. There can be difference related to ownership because of the previous mentioned arguments regarding agency costs and investment efficiency. In addition the management of family and non-family owned company might have different objectives and there might also be a difference in their ability to motivate and implement effective coordination systems (Gorriz and Fumas, 1996). To obtain an appropriate regression to use in this analysis the ownership effects are included and the expression is divided by labour (L) to obtain the capital to labour ratio. This ratio can give an implication whether family firms have a binding capital constraint. I assume a linear relationship between family ownership and the total factor productivity measure. Finally I take the natural logarithm of both sides of the equation to obtain the elasticity for capital and labour. The new specified regression will look like this: (3) The key to this regression is the part marked in bold, which is the family firm part of the equation. This part includes the same variables as for non-family firms, but shows the increase or decrease of the different factors for family firms compared to non-family firms. Solow (1957) wants to measure growth in production and refers to two sources of labour productivity; growth in TFP and growth in capital per worker. Hence this is included in our model and the effects are measured by lna, γ, α and η in equation 3. This model is designed to show the different effects on value added (Q) to labour for family firms and non-family firms related to TFP ( ), capital use and size ( ). The coefficients in the parts of the equation that does not include FF measures the effects related to non-family firms. The coefficients related to the parts including FF measures the additional effects family ownership has on these three variables. This means that to obtain the effects the different variables have on the added value to labour for family firms we need to sum the coefficient for non-family firms and the additional effects for family firms. We include separate 22

24 variables for family firms to account for differences in production functions between family and non-family firms and to obtain the difference in efficiency. show the natural logarithm of total factor productivity measure for non-family firms and shows the higher/lower total factor productivity measure for family firms; hence the total factor productivity measure for family owned firms is. This means that if is significant and positive, family firms have higher total factor productivity and if it is significantly negative family firms will have lower total factor productivity than non-family firms. The same reasoning goes for the other variables. measure output to capital and is the elasticity of output to capital. The additional output to capital for family firms is measure by, hence the total elasticity for family firms regarding capital is. If we get a negative this could imply that an increase in capital effect the productivity less in family firms then in non-family firms. One potential explanation for this is that family firms use less capital-intensive technology in their production. Labour is a proxy for size and measure return to scale for non-family firms expressed by, where α and β are elasticity for capital and labour respectively. If this measure is positive, negative or equal to zero we have increasing, decreasing or constant return to scale respectively. The return to scale for family firms is therefore expressed by, where is the difference in return to scale for family firms compared to non-family firms. This measure also lets us find the measure of elasticity to labour β, since is expressed as. CV represents the variables that controls for industry and age effects. It is also worth noticing that the efficiency and elasticity s might represent other characteristics regarding family firms than the governance advantages. The control variables for age and industry should capture some of these effects. The value added is the value that is added by the company. In other words the price of the product less intermediate output bought outside. Empirical proxies are shown in the following table. 23

25 Table 3.1 Theoretical Variable Empirical Proxy Output/value added (Q) Revenue acquisition cost of goods sold Capital (K) Total assets- accounts payable Labour Number of employees FA 1 if Family ownership > 50% 0 if Family ownership < 50% I will also test for differences in efficiency using different threshold for family ownership to see if the different levels of ownership stakes effects the difference in efficiency. It is important firm size because large firms might possess more market power and greater economies of scale and this might even out the efficiency difference between family firms and non-family firms because family firms often are smaller than non-family firms. 3.2 Performance To look at performance I will use the profitability measure return on asset (ROA). Bodie, Kane and Marcus (2012) define ROA as: (4) (5) Since the sample consists of unlisted family firms, I do not have access to market values, hence the ROA measure is based on book values and is already calculated when extracting the data from the CCGR database. I will compare family and non-family firms to see if there is higher return on asset in family firms compared to non-family firms. It is also interesting to compare the results regarding return on asset to the efficiency measure to see if there is a relationship and if higher efficiency affects the return on asset. 24

26 To connect ownership characteristics, efficiency and performance I will look at the differences between family and non-family firms and look at the relationship between return on asset and the total factor productivity measure. This will be done by splitting the sample between most profitable companies and the least profitable companies and compare the efficiency means between these. I will compare the top 1%, 5%, 10% and 15% firms regarding return on asset and compare these to the rest of the sample with lower profitability. In addition I split the sample into family and non-family firms and conduct the same procedure as described above to see if there is a difference between family and non-family firm. 4. Expectations This section will show and explain what results I expect to find regarding both efficiency and performance. The expectations of the different coefficients will be displayed in a table followed by en explanation regarding the expectations. 4.1 Efficiency Here I will provide some economical intuition towards which direction we can expect the different factors to effect value added. Table 4.1 presents the sign prediction of the different coefficients regarding efficiency. 25

27 Table 4.1 Coefficients Measures Sign prediction TFP for non-family firms - Factors that effect productivity other than labour and capital Elasticity of output to capital for non-family firms Elasticity of output to labour for non-family firms - Effect of return to scale Excess TFP for family firms - Factor that affect productivity other than labour and capital that are special for family firms Expresses the difference in elasticity of output to capital between family and non-family firms Expresses the difference in elasticity of output to labour for family and non-family firms - Effect of return to scale I expect both and to be positive. This means that the total factor productivity measure should be positive for both family and non-family firms. A positive also means that family firms have higher growth in TFP than nonfamily firms. The TFP is the measure the captures other effects than labour and capital; hence for example affects of governance variables will appear under this variable. Family firms are expected to have some advantages related to their organizational form because of reduced agency problems among other issues. measures the elasticity of output to capital for non-family firms and is expected to be positively related to output. An increase in access to capital is expected to increase the output. This is because access to capital can be used to make employees and processes more efficient by for example investing in more efficient machines or educate employees further. Magnussen and Sundelius (2011) found that family firms grow slower and this might indicate that we find these types of companies in industries with less capital intensive production than non-family firms. Because of this we expect to be negative. This does not mean 26

28 that an increase in capital has a negative effect of output, but rather that capital has a lower elasticity to output for family firms than non-family firms. Hence, the need for capital per produced unit should be lower, which is consistent with Gorriz and Fumas (2011) expectations regarding the capital constraint. Gorriz and Fumas (2011) expected firms to operate at the optimal size in the long run perspective and that families are capital constrained. Assuming optimal size would indicate that that and should be negatively related to output. In other words, when firms are operating at their optimal the return to scale should be decreasing. Increasing the number of employees would therefor negatively affect the output per unit of labour ratio. This might be because by hiring another employee the output does not increase value added sufficiently because we are already operating at the optimal size and market conditions makes if difficult to increase value added to the same extent. This might be due to organizational issues or other constraints that inhibit the efficiency of production. If the firm is capital constrained they are not operating at their optimal size and return to scale should be increasing for these types of companies, hence should be positive when we expect family firms to be financially constrained. 4.2 Performance I expect family firms to have a higher return on asset than non-family firms when not controlling for age, industry and size. This has been shown in previous research done on data for Norwegian firms (Bøhren, 2011). As mentioned earlier there is a difference related to agency problems between family and non-family firms. It has been argued that family firms should experience less agency costs, and less agency problems can result in higher efficiency like for example investment efficiency. Hence reduced agency problems should lead to higher profitability for the firm. Since performance is measured as ROA, we does not only look at net income by itself, but also see how well the firm performs with the funds they have available. Unlike Gorriz and Fumas (2011) I expect to find a higher return on asset for family firms and expect the difference to be caused by their findings regarding the financial constraint. In other word I expect their argument for family firms not being more profitable to be the reason why they are more profitable. I expect the financial constrain to be less binding for privately 27

29 held firms because both family and non-family firms are constrained with regarding their access to equity, because of illiquid private market. Regarding the relationship between efficiency and profitability I expect there to be a positive relationship and that the firms with the highest profitability are also the firms with the highest efficiency. This means that the mean difference in efficiency when comparing top performing firms and the rest of the sample should be highest when comparing the top 1% performing firms to the 99% rest of the firms. The mean difference should be decreasing with an increased percentage of top performing firms in the comparing group. This means that the difference should be lower when comparing the top 15% with the rest than when I compare the top 10%, 5% and 1% top firms with the rest of the sample. I also expect the difference in means to be equally large for family firms and non-family firms. Even though I expect family firms to be more efficient that this will contribute to higher return on asset than in non-family firms, I expect this to be the case for the entire sample. This means that the difference in efficiency between the top performing firms and the rest of the sample should not be different for family and non-family firms. 5. Data This section includes information about data extraction and the filters used on the data before applying it in my analysis. 5.1 Panel data Since the data set I have consists of firms over several years, in other words longitudinal data, I find it appropriate to use panel data in this study. This method will control for any unobservable firm heterogeneity (Palia, Lichtenberg, 1999) and also mitigate omitted variable bias (Brooks, 2008) I will also use a fixed effects model, since the data I use effectively constitute the entire population (Brooks, 2008). For the fixed effect model to be appropriate the error term and the constant of each company should not be correlated with each other. I used the redundant fixed effects likelihood ratio test to determine if the fixed effects are necessary. The results are shown in exhibit 2.4 and show that a fixed effect model is appropriate and that I should not estimate 28

30 my regression on a pooled sample. Also applied the Hausman test to se if it is appropriate to use a fixed effect model rather than a random effect model. The Hausman test is displayed in exhibit 2.5 and the p-value show that the fixed effects model is a more appropriate specification of the model than a random effects model. Since I do not have data for the same number of years for all firms the regression will be based on an unbalanced panel. In addition to using panel data I lag the governance variables in the regression to avoid endogeneity. This will be further explained in section Data source I want to look at companies with a family ownership above 50%, which will be accounted for by using dummy variables. Family ownership includes ownership by any blood or marriage relation. By using Centre for Corporate Governance Research (CCGR) we can obtain accounting data from 1994 to 2011 and governance data from 2000 to 2011, hence both accounting variables and corporate governance variables are available. Since governance data is only available from 2000, I will only use data for 2000 to The database includes all Norwegian limited liability firms (Berzins, Bøhren, Rydland, 2008) With the data obtained from the database there is a high percentage of small firms and large difference in to what extent the individual firms contribute to activity such as employment, assets or sales. Berzins, Bøhren and Rydland (2008) specify in their report that 92% of all firms in the data have less than 20 employees and that more than half of these employ less then 5 people. They also point out that firms with more than 99 employees are 1% of all firms in the sample, but this 1% accounts for roughly 60% of activity as measured by employment, assets or sales. 5.3 Filters: The data found using Centre for Corporate Governance Research is very large and hence, filters are needed to obtain consistency in the study. First of all, the data includes different types of enterprise forms like partnership, mutual 29

31 foundations and limited liability firms; hence, we use the first filter to include the relevant sample Berzins, Bøhren and Rydland (2008). 1. Excludes firms without limited liability We also need to eliminate non-active firms and we do that by applying filter Positive sales/required sales 3. Positive assets 4. Delete firms with no employees or missing data with regards to employees over the sample period 5. Current assets have to be greater than cash equivalents 6. Assets have to be greater than working capital - Working capital = current assets current liabilities 7. Exclude financial sector 8. Exclude subsidiaries 9. Exclude listed firms a. Even though the listed firms only constitute a small part of the data, we choose to exclude them to maintain consistency in our study 10. Exclude firms with less than four years of data The first filter is used to exclude partnership, mutual foundations so that our sample only consists of limited liability firms. One advantage of only using AS and ASA is that family firms like farms are not included in the sample. Filter 2-4 ensures that the firms in the sample are active (Almli and Søndergaard, 2011) and Berzins, Bøhren and Rydland (2008) applies filter 5 and 6 to put consistency restrictions on the relationship between a sum and its components. Filter 7 is applied because valuation ratios and accounting profit rates are not comparable for financial firms to those of non-financial firms (Maury, 2006). Financial sector also tend to provide outliers, which will bias the data set. I exclude subsidiaries because I use consolidated accounting number where this is available. Another important reason is that one of my arguments involves a potential financial constraint. Subsidiaries often have a strong parent company and I believe that both the financial constraint and the risk related to operation will be different for the subsidiaries. This means that if subsidiaries were to be included it necessary to look at the subsidiary as a whole with the 30

32 parent company. Hence I find it most appropriate to exclude subsidiaries and instead use consolidated accounting numbers. I apply filter 9 to achieve consistency in the sample and filter 10 to create a lower bound for available accounting information. I will use all the data from the first to the last year a firm passes the filters. After applying the filters above my sample consists of companies, with data for between 4 and 12 years for the time period 2000 to In total the sample consists of observations for the companies over these years. The sample consists of observations for family firms and observation for non-family firms. 6. Results This section includes the results from my quantitative analysis of efficiency and performance differences between family and non-family firms. The first section displays descriptive statistics and also related numbers to the question whether or not family firms are capital constrained. Section 6.2 and 6.3 show the results regarding efficiency and performance and the final section includes weaknesses regarding the analysis. 6.1 Descriptives The most important descriptive statistics can be found in exhibit 1.1. To see if the variables are significantly different for family and non-family firms, I have used in independent sample t-test, and the main results can be found in table 1. Overall the results show that there is a significant difference between most of the variables with the exception of some variables regarding financing of the firm and NOPAT. 31

33 Table 1 Mean difference between family and non-family firms This table shows the mean difference and if there is a significant difference in the variables between family and non-family firms in my sample. The first column displays the mean difference between the firms (family firms non-family firms). The second column display the p-value and the third column shows whether there can be assumed equal variance between the two groups. Mean Equal difference P-value variance Efficiency 0,035 0,000 No ROA 2,241 0,000 Yes Revenue ,000 No Age 1,265 0,000 No NOPAT ,120 No Number of employees -1,044 0,000 No Value added ,000 No Value added per employee ,009 No Capital per emoloyee ,000 No Capital ,000 No Assets ,000 No Equity ,071 No Liabilities ,000 No Debt to Assets ratio 0,030 0,187 No Equity to Assets ratio -0,030 0,187 No Debt to Equity ratio -0,491 0,026 No Differences between family and non-family firms On average family firms have significantly higher return on asset. The descriptive statistics regarding ROA supports Bøhren s (2011) findings and display a higher return on asset for family than non-family firms. It is important to notice that these statistics are not controlled for other variables and other firm characteristics. On average family firms have a return on asset that is 2,24% higher for family firms. We see that the mean and median differ a lot, hence the median might reflect the real difference better. This gives us a difference between family and non-family firms of 0,19%. This is low but since the median ROA is 3,36% it should be regarded as a high difference. However no test of significance is provided regarding the median numbers. There is also a significant difference between efficiency between family and non-family firms. This difference represents a 0,035 higher total factor productivity for family firms. We see that the median efficiency is higher than the mean and also that the differences in efficiency is higher between the medians (0,051). 32

34 Revenues are higher in non-family firms, hence the negative difference in means. The difference is also very large compared to the average revenues for the firms. Which might indicate that family firms do not have as high value creation than non-family firms. Net operating profit after tax (NOPAT) is included in the descriptive to examine if there is a higher value creation in family firms. Both the mean and median show a higher NOPAT for non-family firms, indication that they have a higher overall value creation. However I do not find any significant difference between family and non-family firms when applying the t-test for differences in means. There is also a very large difference between the mean and the median, indicating that it might be more appropriate to test for differences in medians. The medians show a higher NOPAT for non-family firms, but no significance test is provided. The measure for value added could also give us an indication whether or not non-family firms create more overall value. Value added is significantly lower in family firm, which gives us an indication that non-family firms create overall higher value than non-family firms. Combined with the results regarding ROA these results show that family firms create less value, but also that they create higher value with the recourses they have available. This is also an implication that there should be a differences in efficiency and that they use the capital they have available more efficiently. This also implies that family firms have less capital available, which will be discussed later in this section. I also find that family firms are older and that they have fewer employees than non-family firms. On average non-family firms are 1,265 years younger than and have 1,044 more employees than family firms. Both these differences are quite large compared to their mean value, especially the difference in number of employees. This will affect the results regarding efficiency because labour is an important part of the regression I use. To avoid biases in the regression I have deleted extreme outliers, and the numbers shown in exhibit 1.1 represent the number of employees used in my regression analysis. The large number of firms with one employee causes the large difference in the mean and medians. Having 1 employee less is a very high difference when the average number of employees is 3,685, and indicates that family firms might be smaller than non-family firms. We can also see that the effect of difference in employee s affect the ratios for value added per employees and capital per employee. Both of these numbers are significantly higher in family firms even though both value added and capital are 33

35 significantly lower in family firms. This means that there is more capital available for per employee and the value each employee creates is higher than in non-family firms. The mean and median show conflicting evidence regarding differences in value added per employee, but the t-test shows a highly significant difference between the two groups and nonfamily firms have a higher value added to employees than non-family firms. A significant higher capital to employee ratio indicates that there is not at binding financial constraint related to family firms (Gorriz and Fumas, 2011) Financial constraint and capital structure The descriptives also show the differences in capital structure between family firms and other firms and gives implications regarding a potential financial constraint. Capital represents assets less account payables and capital, assets, liabilities and equity are all significantly lower in family firms. The differences are quite high for capital, assets and liabilities indicating that family firms have less access to capital and that they are smaller in size. This might also imply that they operate in less capital-intensive industries or that they are not operating at their optimal size. The difference is equity is smaller, but still significant at the10% level. The difference in mean and median is also quite large for all of these financial measures, and the difference is especially large for equity. However looking at financial ratios provides a different insight regarding the difference in how firms are financed. Both the debt to asset and equity to asset ratio are not significantly different between the two groups. Indicating that there is no difference in the proportion of equity and debt financing between the two firms. This indicates that if there is a capital constraint regarding equity, family firms do not increase their debt to finance new investments as suggested by Gorriz and Fumas (2011) and Magnussen and Sundelius (2011). In addition there is a significant difference between the debt to equity ratios, showing a lower ratio for family firms. Indicating that family firms use less debt compared to their equity. This is consistent with the expectation that family firms should have less debt, because of risk reduction since their stake in the company is high and that they are not diversified. Overall the results regarding how firms are financed show that family might be constrained because they have 34

36 lower assets, but also that family firms are constrained regarding both debt and equity, since there is no significant difference in capital structure. This may be cause by their fear of diluting their own stake in the company, and also by their need for risk reduction because of an undiversified stake in the firm, indicating a self imposed financial constraint. Combined with the regarding higher return on asset in family firms this might imply that family firms invest more efficiently and avoid biases like for example overconfidence and empire building. These results also imply that my expectations regarding return to scale should be correct. Lower assets for family firms imply that family firms are smaller in size and might not be operating at their optimal size. Hence they should benefit from growing the company if the capital is available and obtain an increasing return to scale. Another interesting aspect of the results is that size is often positively related to profitability. I find that family firms are smaller than non-family firms regarding both number of employees, but still they have a higher ROA than nonfamily firms. This again implies that there is some governance advantages with family ownership compared to non-family ownership. 6.2 Efficiency The redundant fixed effects test is displayed in exhibit 2.4 and show that a fixed effect model is appropriate and that I should not estimate my regression on a pooled sample. The Hausman test is displayed in exhibit 2.5 and the p-value show that the fixed effects model is the appropriate specification and that the random effects model is not appropriate. Table 2 shows my results regarding efficiency. The first column presents the regression result when it is assumed that family and non-family firms have the same production function. The second column shows the regression result when I allow family and non-family firms to have different production functions. As we can se from the two regressions, family and non-family firms do not have the same production function and by using the same production function, family firms would be less efficient than non-family firms. I choose to interpret the model using different production functions for family and non-family firms. This model gives a slightly higher R-squared and 35

37 since the coefficients for return to scale and elasticity of output to capital are highly significant, I find it appropriate to include them in the model. From table 2 it can be seen that my expectations is consistent with the result I have obtained. The results show a positive, significant at 5% level coefficient for the family dummy variable, which implies that family firms are more efficient than non-family firms. I find a total factor productivity measure of 2,025 2 and 2,105 3 for non-family and family firms respectively. Table 2 Main efficiency results Family firms Same production Individual production function functions Efficiency lna 7.690*** 7.573*** ( ) ( ) Capital/Labour α *** 0.460*** ( ) ( ) Labour δ *** *** ( ) ( ) Efficiency ϒ *** ** ( ) (2.2640) Capital/Labour η *** ( ) Labour ϕ *** *** Significant at 1% level, **significant at 5% level, *significant at 10% level This table shows the results regarding efficiency (equation 3), where only the family ownership variables are lagged. The dependent variable is value added to number of employees. The first column display the results when it is assumed that family and non-family firms have the same production function and equal elasticity s. The second column displays the results when it is assumed that family firms have a different production function than non-family firm. The results are also controlled for age and industry effects, but the results are not reported. T-stats are shown in brackets. (9.3641) R-squared I expected an increasing return to scale for family firms than for nonfamily firms; hence should be positive. The coefficient is positive, and significant at the 1% level. Implying that return to scale is significantly different for family firms compared to non-family firms. Also the other variables are consistent with my expectations. The variables estimating elasticity of output to capital are consistent with my expectations and I find an elasticity of output to capital of 46,0% and 44,8% for non-family and family firms respectively. The 2 Ln(7,573) 3 ln(7,573)

38 elasticity of output to labour is 32,9% 4 and 37,9% 5 for non-family firms and family firms respectively. These findings indicate the non-family firms are slightly more capital intensive than family firms and also that family firms are more labour intensive than non-family firms. Since family are less sensitive to changes in capital they are less affected by a possible capital constrain than nonfamily firms. However this difference in elasticity of output to capital is very small and since the results shown under descriptives imply that family firms are capital constraint, this similar elasticity implies that family firms operate in less capital-intensive industries and that the capital constraint might not be binding. This is also consistent with the implication made by Magnussen and Sundelius (2011), stating that family firms grow slower and that we therefor should find these types of companies in less capital-intensive industries. Since family firms appears to operate in more labour intensive industries the differences in efficiency might be caused by the managers ability to motivate employees or higher devotion of employees in family firms. Family firms obtain a positive coefficient, but overall decreasing return to scale. Since return to scale for family firms are less decreasing than for nonfirms, this implies that family firms are more capital constraint than non-family firms. If firms are not operating at their optimal size they would benefit from growing and hence get an increasing return to scale. However the return to scale for family firms is also decreasing, which implies that they are operating at their optimal size. This means that by growing the company and increasing the number of employees the value added per employee will decrease the value added to number of employees. In other words this means that one extra employee would create less value than the current employees. This does not mean than the extra hired employee would create negative value added, only that the ratio will be reduced. In addition this is an implication that family firms are not capital constrained, since a decreasing return to scale implied that the firm is operating at the optimal size. Regression results from running the regression without control variables is found in exhibit 2.3. From these results we can see that the efficiency measure for family firms is higher when we exclude the control variables. This implies that 4 1-0,2113-0, ,172-0,448 37

39 some of the effects regarding efficiency differences can be explained by industry and the maturity of the company (age). As mentioned earlier Bøhren (2011) found that profitability differences between family and non-family firms are highly reduced when controlling for age, industry and size. Since size in included in the regression as the measure for return to scale I do not include this as a control variable. To further investigate the effects of ownership I have also looked into efficiency differences with equal or greater than 25%, 75% and 100% ownership stake. These results can be found in table 3. Table 3 Comparing effects of ownership stake *** Significant at 1% level, **significant at 5% level, *significant at 10% level This table shows the relationship between ownership stake and efficiency, capital elasticity, labour elasticity and return to scale. These numbers are based on regression results from exhibit 3.1 and labour efficiency is calculated using the following formula: Table 2 and appendix 3.1 show that family owned firms with 100% ownership stake are more efficient (TFP of 2.14) than firms with an ownership stake above 50% (2.105). The measure for family efficiency also becomes significant at 1% level compared to significance at 5% level for above 50% ownership coefficient. This is also the case for 75% ownership, which gives a total factor productivity measure of 2,12 and with a higher p-value than above 50% ownership, but rejected at the 5% level. One potential reason for this can be lower agency problems between majority and minority shareholders. With a higher ownership stake the benefits of extracting private benefits becomes lower because the family carries a large part of the cost of these benefits. In firms fully 38

40 owned by the family this problem will be non-existent. In the case where I define family firms with an ownership stake of 25% or more, I obtain an insignificant coefficient for the family dummy variables, which means that efficiency is not significantly higher for family firms with 25% or more of the company owned by the family. This implicate that the governance advantages that effect efficiency in family firms is not present when the family does not have a controlling stake in the company. The elasticity s to capital and labour and return to scale measure for family firms also becomes insignificantly different from zero, implying that family and non-family firms operate in equally labour intensive industries and are equally capital constrained. There is also a difference between the other ownership stakes when looking at elasticity of output to capital and the return to scale measure. Both the measure of capital elasticity and return to scale decreases with increasing family ownership stake. This means that family firms with a high family ownership stake have a value added measure that is less sensitive to changes in capital. This might imply that family firms with a high ownership stake might not be as capital intensive and may be less financially constrained. One potential reason for this is that they own a higher stake in the company and are able to raise capital without losing their controlling stake in the company. It also implies that they are operating at their optimal size. The elasticity s of output to capital is higher for 75% or higher ownership stake compared to 100% and over 50% ownership stake. The difference is small but implies that firms with an ownership stake of 75% or more are more labour intensive than firms with 100% or above 50% ownership stake. I find it surprising that 100% ownership does not provide higher labour intensity when they have lower capital elasticity. One potential explanation is that 100% owned firms may be small companies with few employees and maybe only family members employed. Increasing the number of employees or growing the company might therefor reduce the value added per employee. It is also worth noticing that by excluding the control variables (exhibit 2.3) the efficiency measure for family firms increases and becomes significant at the 1% level. This might imply that age and industry effects is related to family ownership and that some of the higher efficiency in family firms might be caused by their choice of industry and how old the firms are. 39

41 My regression results show a very high R-squared, which implies that the model has high explanatory power. One reason for this can be because the regression is specified as a production function including labour, capital and efficiency and the model should have high explanatory power. In addition I include governance effects in the model and control for age and industry effects. 6.3 Performance The comparison of means between firms with high return on asset and the rest of the sample can be found in table 4 and 5. Table 4: Relationship between efficiency and profitability This table shows the differences in efficiency when comparing the top 1%, 5%, 10% and 15% most profitable firms with the rest of the sample. The first column shows the difference between the top firms and the rest of the sample, the second column shows the p-value regarding the significance of the difference and the third column shows show if there is equal variance between the two samples. The measure of profitably used to split the samples is Return on Asset. Table 4 shows results for three different samples, family firms, non-family firms and the entire sample. Top 1% displays the comparison of the 1% most profitable companies and the rest of the sample. The same reasoning goes for the columns top 5%, top 10% and top 15% which compare efficiency for the top 1%, 5%, 10% and 15% and the rest of the sample. The results indication that there is a link between efficiency and return on asset since the difference in efficiency between the top firms is significantly higher than the efficiency for the rest of the sample. The results also imply that firms with the highest return on asset are not necessarily the firms with the highest efficiency. This implication is drawn from the increase in difference in means when including a higher percentage of top firms. Comparing the firms with the 5% highest return on asset and the rest of the sample I achieve a higher mean difference than comparing top 1% firms with the rest of the sample. 40