Master in Finance. The effect of ownership structure on firm performance: Are mutual funds actually monitoring?

Transcription

1 Master Thesis Finance The effect of ownership structure on firm performance: Are mutual funds actually monitoring? Abstract: In this thesis, the effect of mutual fund ownership on firm performance, as measured by Tobin s Q has been investigated. Mutual fund ownership is measured by the percentage of shares of a firm held by mutual funds. A firm fixed effects model and lagged ownership values have been used. The results show that firms with a higher percentage of shares held by mutual funds have significantly worse firm performance. This negative significant effect is even stronger for small and lowly levered firms and in times of recession. The only firms that mutual funds tend to monitor are highly levered firms. No moderating effect has been found by the amount of equity based compensation that firms give their directors. Most results are robust when using Return on Assets as a measure of firm performance. The results lead to conclude that mutual funds do not monitor the firms they hold shares of, leading to a reduction in firm performance. An explanation for this could be that mutual funds are short term investors that are not allowed to hold large holdings in firms. F. M. M. van Rooij ANR: Supervisor: Faculty: Study track: M. F. Penas Finance Master in Finance Hand in date: 13th of August,

2 Introduction Nowadays mutual funds have had more and more assets under their control. 10 years ago, in 2003, U.S. (open-end) mutual funds had 7.4 trillion dollars of net assets under their control. In 2009 this number has grown to 11.1 trillion dollars and in 2012 it reached a record high 13 trillion dollars. Also 44% of all U.S. households own mutual funds and those households hold 89% of all mutual fund assets in total (Investment Company Institute, 2013). So it shows that mutual funds have had a tremendous growth in the amount of assets they invest for their clients. Mutual funds can either hold actively managed portfolios, where an investment advisor combines different sorts of investments to meet an investment goal, or passively managed portfolios, where the investment advisor wants to track the performance of a given index or benchmark (Investment Company Institute, 2013). What makes mutual funds, also called open-ended funds, unique is that they commit themselves to buying back their shares at their net asset value (NAV) whenever their clients want to sell their shares. The net asset value is calculated by taking the total market value of the mutual fund s asset, subtracting the fund s liabilities and then dividing that number by the amount of mutual fund shares outstanding. (Investment Company Institute, 2013). This means that mutual funds require a lot of liquidity, because their clients can sell their shares back to them at any trading day and mutual funds must pay their investors within 7 days (Investment Company Institute, 2013). This is in contract to other institutional investors, like pension funds, who often have a certain amount of assets to invest for a longer period of time. Also, for mutual funds to not be taxed on their income or capital gains at the entity level, they have to qualify as a regulated investment company (RIC). A few of the requirements to be a regulated investment company under the Investment Company Act of 1940 include a mutual fund to have at least 50% of the fund s net asset value to consist of cash, cash items or securities, where they can t own more than 10% of a security s voting shares. A mutual fund can t invest more than 25% of their assets in one security either. At the end of each quarter mutual funds undergo a tax diversification test, where they have to hold securities from at least 12 different issuers to pass the test (Investment Company Institute, 2013). Both the liquidity requirements and the requirements of the Investment Company Act of 1940 make it hard for mutual funds to have large positions in firms for a longer period of time, in contrast to other institutional investors. This could indicate that mutual funds are poorer 2

3 monitors of the firms they own shares of. This in turn could lead to firms having many of their shares in the hands of mutual funds, who don t exercise any monitoring activities over the management of the firm. As a result, this could have negative implications for a firm s performance. In this thesis, I will investigate whether mutual funds monitor the firms they own shares of, by looking at the relationship between the ownership of a firm s shares by mutual funds and the firm s performance, as measured by Tobin s Q. I will focus on U.S. firms. My main research question is: Do mutual funds monitor the firms they hold shares of? First I ll provide a theoretical background and discuss the literature on this subject in section 1. Then I will create a model to test the relationship between mutual fund ownership and a firm s performance in section 2. I will include several factors to see if the (not existing) monitoring factor of mutual funds differs when looking at certain firm characteristics. The results will be explained in section 3. After that I will state my conclusions and recommendations for further research in section 4. Section 5 states the references and section 6 is the appendix. 3

4 Table of contents Introduction : Theoretic background and literature overview Agency Problems Institutional investors Mutual funds Firm characteristics : Data and methodology : Data : Methodology : Empirical results Relationship between Mutual Fund Ownership and Firm Performance Moderating effect of Firm Size on the relationship between Mutual Fund Ownership and Firm Performance Moderating effect of Capital Structure on the relationship between Mutual Fund Ownership and Firm Performance Moderating effect of Equity Compensation on the relationship between Mutual Fund Ownership and Firm Performance Moderating effect of Recession on the relationship between Mutual Fund Ownership and Firm Performance Robustness check : Conclusion and recommendations : Conclusion : Recommendations : References : Appendix

5 1: Theoretic background and literature overview 1.1 Agency Problems Ever since the existence of separation in firms between ownership and control, there have been discussions about ownership structure. Already in 1932, Berle and Mean (1932) devoted an entire book to the problems of this separation between ownership and control. They argue that ownership in large companies is so dispersed that no single shareholder has the incentive or ability to monitor and exercise control over the company. The dispersion leads from shareholders wanting to diversify their idiosyncratic firm risk and therefore they don t want to take a large stake in any company. The lack of incentives and abilities to monitor and control the firm leads to the company being inefficiently run and managers reaping too many private benefits at the cost of shareholders. Jensen and Meckling (1976) developed a model of how the interests of managers and shareholders diverge when ownership and control are separated. They show that the utility function of a manager consists only of on the job consumption and firm value, where on the job consumption destroys firm value. Outside shareholders on the other hand only have firm value in their utility function. This is where agency costs arise, when controlling managers consume on the job at the cost of the firm value for outside shareholders. This leads people to believe that shareholders with a larger stake in the firm could indeed better monitor managers and reduce on the job consumption. Kang and Sorensen (1999) also argue that a firm is an optimal competitor in the market when it minimizes the amount of contracts it has and optimizes the relationship between the manager and shareholders of the firm. This means fewer shareholders in total, but each with higher stakes in the firm. However, as argued above, shareholders should be reluctant to take a large stake in a firm, because of idiosyncratic risk. 1.2 Institutional investors The efficient monitoring hypothesis claims that shareholders who own large blocks of shares of one firm are more incentivized and have lower costs to monitor management (Hu & Izumida, 2008). Institutional investors, for example pension funds, insurance companies or hedge funds are considered as good candidates for this role, because they have large amounts of money to invest for a longer period of time. However, since investing a large part of their money in a small amount of firms involves taking extra idiosyncratic risk, something must drive large institutional shareholders to not optimally diversify their portfolio. According to 5

6 Holderness (2003), shareholders are only willing to hold a large stake in a firm (and become a blockholder) for two reasons. One reason is that they receive the private benefits of control as well as the shared benefits of control in the firm. A second reason is that these benefits must exceed the costs that the extra risk that comes with holding the large stake brings. Holderness then argues that firms with at least one large shareholder must have a better performance than other firms without large shareholders, to cover the costs for the large shareholder of holding a large stake in that firm. Grossman and Hart (1986) also argue that large shareholders are more willing to monitor the firm, because they reap some of the benefits of their monitoring efforts. So the efficient monitoring hypothesis would imply a positive relationship between institutional ownership and firm performance. Evidence in accordance with the efficient-monitoring hypothesis is found by McConnel and Servaes (1990). They find a strong positive relationship between the percentage of shares held by institutional investors and firm performance, as measured by Tobin s Q. Agrawal and Mandelker (1990) also find evidence for the efficient-monitoring hypothesis. They find a positive relationship between the amount of shares owned by institutional investors and the change in shareholders wealth around the announcements of antitakeover charter amendments. More evidence for the monitoring role is found by Bushee (1998). He found that more institutional ownership reduces the probability that managers reduce their investments in research and development to use that money for achieving short-term earnings goals. They find an opposite effect when institutional investors have a higher portfolio turnover. Similar evidence is found by Gaspar et al. (2005), who found that firms with many long-term shareholders as opposed to short-term shareholders experience higher takeover premiums as targets and lower takeover premiums as bidders. This also indicates that long-term shareholders are monitoring the decisions of managers effectively. Ellili (2011) also finds a positive effect between firm performance and the percentage of shares owned by institutional investors. There is also other evidence that conflicts with the efficient-monitoring hypothesis. A research, done by Mikkelson and Regassa (1991), looks at premiums for the trade of large blocks of shares. They document an average premium of 9.2 percent. Chang and Mayers (2012) report premiums that average 13.6 percent. They suggest that large blockholders are willing to pay a premium that large, because they anticipate a takeover, not because they expect to increase firm value by monitoring the firm. Another reason that Mikkelson and Regassa (1991) give for the large premiums is that blockholders expect to reap private 6

7 benefits from their control rights, which could be at the cost of firm value. This indicates that large shareholders are not always just there to monitor the firm and increase firm value, conflicting with the efficient-monitoring hypothesis. The big influence of large shareholders in a firm could also cause conflicts of interest between them and minority shareholders. This happens when the controlling shareholders expropriate the smaller shareholders. This is called the expropriation-of-minority-shareholders hypothesis (Hu & Izumida, 2008). Large shareholders will then do everything in their power to take care of their own interests, which don t necessarily coincide with other smaller shareholders. They could use a pyramidal control structure or super-voting rights to allow them to get more control rights than are in accordance to their cash flow rights (La Porta et al. 1999). When small investors fear this will happen, they will pay less for the firm s shares, which will make the firm s value drop. The expropriation-of-minority-shareholders hypothesis thus implies a negative relationship between institutional ownership and firm value. Holderness and Sheehan (1988) find evidence against the expropriation-of-minority-shareholders hypothesis. They look at the survival rates of companies with majority stockholders. They find that firms with majority shareholders are still surviving, which would not happen if they systematically consume the firm s value. They also find that the accounting rates and Tobin s Q for the majority-owned firms are equal to that of firms that have diffuse ownership. Two other hypotheses that suggest a negative relationship between institutional ownership and firm value are the conflict-of-interest hypothesis and the strategic-alignment hypothesis. The conflict-of-interest hypothesis claims that because of business relationships with the firm they are holding shares of, the institutional investors are forced to vote their shares in accordance with management. Voting against management might negatively affect their business relationships with the firm (Pound, 1988). The strategic-alignment hypothesis also states that institutional investors vote with management and do not monitor them effectively. However this hypothesis argues it s advantageous for the institutional investors and the managers of the firm they hold shares of to work together. This cooperation leads to worse firm performance than when the large institutional investors would monitor management. More mixed evidence about the relationship between institutional investor ownership and firm performance has been found. Al-Fayoumi and Abuzayed (2009) find no evidence of active monitoring by institutional investors. Demsetz and Lehn (1985) found no significant 7

8 relationship between accounting profit rates and the ownership concentration in a firm. Mehran (1995) didn t find any relationship between the percentage of shares held by institutional investors and firm performance. Also Woidtke (2002), found a negative relationship between activist public pension fund ownership and firm performance, as measured by Tobin s Q, but a positive effect between firm performance and private pension funds. 1.3 Mutual funds The hypotheses discussed above are theories for institutional investors in general. Mutual funds, however, differ from the typical institutional investor. Therefore it is interesting to look at them separately. As discussed in the introduction as well, mutual funds are different from example pension funds in that they buy back their shares at their net asset value whenever their clients want to sell their shares. In the case of investors selling their shares back to the mutual funds, the mutual funds have to pay their investors within 7 days (Investment Company Institute, 2013). They also need to comply with several requirements to not be taxed on their income or capital gains at the entity level. Maybe one of the most important requirements is that mutual funds are not allowed to own more than 10% of a security s voting shares. This means that mutual funds can become blockholders and have influence if they monitor. Also, because mutual funds often have no other business relationships with the firms they hold shares of, the conflict-of-interest hypothesis and the strategic-alignment hypothesis would not be applicable to them. This is why Brickley et al. (1988) argue that mutual funds are so called pressure-insensitive institutions, which are less exposed to conflicts of interest than other institutional investors that have business relationships with the firm, like banks and insurance companies. According to Davis (2008), mutual funds do not monitor effectively at all. He found that Fidelity and American Funds, two of some of the biggest mutual fund complexes, liquidate their large shareholdings quickly. For Fidelity, about one quarter of the amount of firms in which it held 5% or more of its shares in 1999, was liquidated 1 year later. After five years, two-thirds of those holdings were dropped. For American funds, also two-thirds of its 5% holdings from 1999 were dropped in (Davis 2008). This means that even some of the largest mutual fund complexes do not hold their positions for long. Also Fidelity tended to vote with management on almost all cases. Davis (2008) proposes 3 reasons for this: legal restrictions. Funds that have a 10% ownership in a firm are sometimes treated as insiders under the law, so trying to exercise control puts the fund at risk of being accused of 8

9 expropriating minority investors. A second reason is conflicts of interest: Some of the big mutual fund complexes also manage pension funds for corporations. This causes conflicts of interest when funds have to vote against management, because it could hurt their other business relationships with the firm. The third reason Davis (2008) gives is that a mutual fund that can at most own 10% of a firm s shares still has to pay for all the costs of monitoring the firm, while only reaping 10% of the added firm value. In 2000, the 5 largest mutual fund complexes held 32% of the total net assets of all mutual funds. The 10 largest mutual fund complexes held 44% of all mutual funds total net assets under their control. In 2010 this grew to 40% for the 5 largest mutual fund complexes and 53% for the 10 largest mutual fund complexes (Investment Company Institute, 2013). This shows that most mutual funds assets are concentrated with a small amount of large mutual fund complexes. However, not one of the mutual fund complexes has been able to really dominate the market (Investment Company Institute, 2013). Some empirical work has been done on the effect of the percentage of shares held by mutual funds as an institutional investor on firm performance. Yuan et al. (2008) found a positive relationship between the percentage of shares held by mutual funds and a firm s performance for firms in China. Bhattacharya and Graham (2009) divide institutional investors in Finland in 2 categories: pressure-sensitive investors (insurance companies, banks and non-bank trusts) and pressure-resistant investors (public pension funds, mutual funds, endowments and foundations). They argue that pressure-resistant investors have no business relationships with a firm and therefore should be able to monitor better. They find a negative relationship between ownership concentration and firm performance for both groups, but the effect is larger for pressure resistant investors. Evidence of mutual funds monitoring is given by Kroll et al. (2006), who found that both short and long term investors moderate the form of the relationship between research and development spending and firm performance. They put mutual funds in the category of quick-entry-and-exit institutional investors. Brickley et al. (1988) also divided institutional investors in the pressure-sensitive and pressure-resistant investors group. They found that the group of pressure-resistant investors (mutual funds, public pension funds, foundations and endowments) is more likely to oppose management than the group of pressure-sensitive investors (banks and insurance companies). It is clear that there is no real consensus in the empirical literature about whether mutual funds monitor effectively or not at all. While Yuan et al. (2008) find a positive relationship between 9

10 the percentage of shares of a firm held by mutual funds in China, Bhattacharya and Graham found a negative effect in Finland. The effect may be different in the US. Brickley et al. (1988) found a positive relationship between the ownership by pressure-resistant investors, under which they categorized mutual funds, and firm performance. Davis (2008) shows that many of the largest mutual fund complexes, who make up for most of the assets held by mutual funds, liquidate their holdings quickly and tend not to interfere with management. This leads to believe that mutual funds do not monitor the firms they hold shares of effectively and more shares of a firm owned by mutual funds could have a negative effect on firm performance. Therefore the first hypothesis that I will investigate in this thesis is: Hypothesis 1: Mutual fund ownership of a firm has a negative effect on firm performance. 1.4 Firm characteristics Demsetz (1983) came up with the initial argument about the ownership structure of a firm being endogenous. This means that every firm in theory has an optimal ownership structure that maximizes firm value. This optimal ownership structure in theory varies with certain observable and unobservable firm characteristics. Himmelberg et al. (1999) also claim that some firms have characteristics which make them are more difficult to monitor. Therefore it is important to include observable firm characteristics that make firms need more or less monitoring than other firms. A few of those characteristics that I will now discuss are size, capital structure, stock and option compensation. I will also discuss general uncertainty for firms. As explained by Demsetz and Lehn (1985) size has an inverted relationship to ownership structure. The larger a firm gets the higher it s market value generally gets. Obtaining or retaining the same percentage of shares in that firm then becomes more expensive for shareholders. This indicates that larger firms should in general have more diffused ownerships. This is called the risk-neutral effect of size on ownership. Himmelberg et al. (1999) add to that that in large firms, agency and monitoring costs can be greater due to their complex structures. On the other hand they claim that large firms might benefit from economies of scale in monitoring by the highest layers of management. Also large firms are gathering more attention from rating agencies, which also have a monitoring effect. This means that firm size could have a moderating effect on the relationship between mutual fund ownership and firm performance. I have formulated the following hypothesis to test this effect: 10

11 Hypothesis 2: Firm size has a positive moderating effect on the relationship between mutual fund ownership and firm performance. Jensen and Meckling (1976) argue that debt holders of a firm are motivated to monitor managers to prevent managers of levered firms to take on projects that have high outcomes but a very low probability of success. In the event the project turns out well, the manager will gather the most value and when the project fails, the debt holders will incur most of the project s costs. This incentivizes debt holders to monitor managers with for example covenants that limit managerial behavior. Also, highly levered firms face more risk of bankruptcy and management can become unemployed when the firm goes bankrupt or gets taken over by another firm (Jensen and Meckling, 1976). There is also empirical evidence on the effect of leverage in a firm. Gilson (1989) investigated management turnover in financially distressed firms. He found that often a change in management was related to the firm s high debt ratio. In over 10% of the cases bank lenders initiated a change in management. Also in 52% of the cases management turnover happens when firms have to restructure their debt in private to avoid bankruptcy or go bankrupt. James (1987) researched that shareholders achieve a positive abnormal stock return of 2% when a firm has announced new bank loan agreements. He claims this increased value is because of the monitoring activities by banks. Following the theory and empirical literature, there is reason to believe that firms that are highly levered need less monitoring than other firms, since these firms are also monitored by debt holders. I will investigate this effect with the following hypothesis: Hypothesis 3: Firm debt ratio will have a positive moderating effect on the relationship between mutual fund ownership and firm performance. Equity based compensation like option plans and stock awards are often used by firms to align the interests of managers with shareholders (Smith and Watts, 1982). These options and stock awards increase in value when firm value increases. Some options awarded by option plans can only be exercised after the manager has remained with the company for a certain amount of time and are forfeited when the manager leaves the firm early. This goes for restricted stock awards as well, which can t be sold before the conditions are met (Smith and Watts, 1982). This equity based compensation should incentivize managers to increase firm value for shareholders. Empirical evidence provides support for the monitoring role of equity based compensation. 11

12 Aboody and Kasznik (2008) found that the structure of equity based compensation for managers can help align the payout choices of managers with the preferences of shareholders. Cornett et al. (2007) found a positive moderating effect of the amount of a CEO s compensation that consists of options on the relationship between the percentage of shares held by institutional investors and the return on assets. They claim that the equity based compensation is a monitoring device for a firm. Theory and empirical literature implicate that firms that give managers more equity based compensation could have less agency problems. As a result these firms need less monitoring since the equity based compensation aligns managers interests with shareholders interests. This means equity based compensation could have a positive moderating effect on the relationship between mutual fund ownership and firm performance, as measured by Tobin s Q. I will test this effect by using the following hypothesis: Hypothesis 4: Equity based compensation has a positive moderating effect on the relationship between mutual fund ownership and firm performance. Demsetz and Lehn (1985) explain that for firms that operate in markets with stable prices, stable market shares and stability in general, management can be monitored at lower cost, because their performance can be measured more accurately. In unstable environments, it becomes harder to monitor management. More uncertainty in a firm s operating environment leads to increased benefits for monitoring the firm effectively (Demsetz and Lehn, 1985). This indicates that there could be a link between the uncertainty of the firm s environment and time it operates in and the amount of monitoring a firm requires. The United States had a negative real global domestic product growth in 2008 and a big part of 2009 (United States Department of Commerce, 2013). Also the National Bureau of Economic Research (2010) announced in 2010 that a through in June 2009 marked the end of the recession that started in December Since firms experience extra uncertainty during recessions they could benefit from extra monitoring in these times. This leads to the hypothesis that years of recession, in this case 2008 and 2009, have a negative moderating effect on the relationship between mutual fund ownership and firm performance. Hypothesis 5: Recession years have a negative moderating effect on the relationship between mutual fund ownership and firm performance. 12

13 2: Data and methodology 2.1: Data To gather the data needed for this research I used the databases of Wharton Research Data Services (WRDS). WRDS is a data research service from the Wharton School at the University of Pennsylvania. In 1993 the service has been developed and has since then been used by over 290 institutions around the world. From WRDS, the Thomson Reuters and Compustat databases will be used. This thesis focuses on North American firms from 2002 to The Thomson Reuters database has been used to gather the data to determine the ownership of firms by mutual funds and institutional investors in general. Thomson Reuters provides data analyses and information tools to their clients to help them make better financial decisions and achieve better financial results. Their institutional holdings database consists of data of the institutional stock holdings and transactions, as they are reported with the US Securities and Exchange Commission (SEC) on form 13F. The database was previously known as CDA/Spectrum 3 4 database and provides ownership information for institutional managers with over 100 million dollars in assets under their management. This database will be used to gather the ownership information of institutional investors. To separately look at mutual funds ownership information as well, I will use the mutual fund holdings database of Thomson Reuters. This database provides information about security holdings for all mutual funds that that report their holdings to the SEC and also for 3000 global mutual funds. This database was previously known as CDA/ Spectrum 1 2 database. The Compustat North America database consists of more than 100 quarterly and 300 annual statement of cash flow, balance sheet, income statement and supplemental data items. For most companies, the quarterly data goes back to 1962 and the annual data goes back to From Compustat I have gathered general financial data and equity based compensation data from their Execucomp database to use in this thesis. To determine the amount of shares held by mutual funds for each firms, I added all the data of the mutual fund holdings database of Thomson Reuters together for each firm. This gave a total amount of shares held per firm by all the mutual funds in the database. I then divided this amount by the total amount of shares outstanding of the firm to determine the percentage of 13

14 shares held by mutual funds. After that I merged the dataset, using each firm s unique CUSIP code as the identifier, with the institutional holdings database of the Thomson Reuters database. This database provides the percentage of shares held by all institutional investors per firm. I then subtracted the earlier calculated percentage of shares held by mutual funds from the mutual funds holdings database from the percentage of shares held by all institutional investors to get the percentage of shares held by all institutional investors that are not mutual funds. This way the percentage of shares held per firm by mutual funds and other institutional investors can be looked at separately. The next step was to merge the data with the financial data from the Compustat database. Again the CUSIP code was used as the identifier. The Compustat database uses a 9-number CUSIP code while the databases from Thomson Reuters use an 8-number one, so I removed the last number of the CUSIP code from Compustat to be able to merge the datasets. Companies that were missing either financial data needed to calculate Tobin s Q as a proxy for firm performance, or had missing ownership data, were removed from the analysis. A number of control variables from Compustat, that will be described later, still had missing variables. Instead of removing those observations from the analysis, leading to a smaller dataset, I used a method used by Himmelberg et al. (1999). They also had trouble with some control variables from Compustat being missing for certain observations. Especially for Research and Development (R&D) expenses and Advertising expenses, they argue that the most common reason that firms to not comply with disclosing the expenses is that the amount is negligible. As they do as well, I include dummy variables that are 1 if the data for R&D and Advertising expenses are missing and are 0 otherwise. This is to control for the possibility that firms that do report those expenses are different from the ones that don t (Himmelberg et al. 1999). Also, I have replaced the missing values with 0. The same method has been applied to other missing control variables. For most variables however, this was only needed for a small amount of observations. Observations that had measurement errors like a mutual fund or other institutional investor ownership percentage of above 100% or negative dividends have been removed from the dataset. Finally, I removed firms that had less than 3 observations in the entire dataset. This is because firms with only 1 or 2 observations are definitely not useful for testing relationships, especially when using lagged values and firm fixed effects. I didn t want to remove more firms that have a small amount of observations (for example 3 or 4), because that would induce a survivorship bias in my dataset. In the end I 14

15 am left with observations of 3352 firms. The observations range from a minimum of 1861 in 2002 to a maximum of 2317 in As the dependent variable I will use Tobin s Q, as a proxy for firm performance. Following McConnell and Servaes (1990) and Himmelberg et al. (1999) I have calculated Tobin s Q as a ratio of the firm s value divided by the replacement value of the firm s assets. As firm value I use the sum of the market value of common equity plus the book value of total liabilities plus the market value of preferred stock. As the replacement value of the firm s assets I use the book value of total assets. Tobin s Q has been winsorized at the 1% level to account for outliers. To reduce the probability that any results I will get are due to other factors that are not accounted for, I use several control variables. Following Bhattacharya and Graham (2009), McConnell and Servaes (1990) and Mehran (1995) I have included several control variables that are described in table 1b For my second hypothesis, investigating the moderating effect of firm size on the relationship between mutual fund ownership and firm performance, I use the natural logarithm of the replacement value of a firm s assets as a proxy for firm size. I use the natural logarithm to control for outliers. For my third hypothesis, a firm s debt ratio will be calculated as the amount of debt divided by the replacement value or book value of a firm s assets. Firms with a higher debt ratio than 1 have been removed from the dataset, For the fourth hypothesis, a dummy called recession has been created that is 1 if the observation year is 2008 or 2009 and 0 otherwise, to incorporate the effect of the recession years. For the fifth hypothesis, I used the compensation data from the Execucomp database from the Compustat database. I calculated the equity based compensation as a sum of option awards and stock awards, based on fair value, given out to a director in a given year. I then divided this sum by the total compensation given to a director in a given year. This provided me with a variable called Equity Compensation, which gives the percentage of total compensation for a director that consists of equity based compensation. The describing variables can be found below. 15

16 Table 1a Description Variables Company name SIC-code CUSIP Year Tobin s Q ROA Mutual Fund Shares Other Institutional Shares Shares Outstanding Mutual Fund Ownership Other Institutional Ownership Recession Ln(Assets) Debt Ratio Equity Compensation Market Value Sales Assets EBIT Depreciation and Amortization Expenses Preferred Stock Name of the company Industry code to which the company belongs. Unique identifier code of each company. Year of the measured values. This variable is a ratio that is defined by dividing the firm s value by the replacement value of the firm s assets. As firm value, the sum of the market value of common equity plus the market value of preferred stock plus the book value of total liabilities has been used. As the replacement value of assets I have used the book value of total assets. Tobin s Q is used to measure firm performance. Variable obtained by dividing EBIT plus the depreciation and amortization expenses by the book value of the assets of a company. ROA, Return On Assets, is being used to measure company performance for a robustness check. The amount of shares of a firm that are owned by mutual funds The amount of shares owned by institutional investors other than mutual funds The amount of common shares outstanding for a firm. The amount of shares of a firm that are owned by mutual funds, divided by the total amount of common shares outstanding for that firm. This variable can have values between 0 and 1. A variable that is calculated by dividing the amount of shares of a firm that are owned by institutional investors other than mutual funds by the amount of common shares outstanding for that firm. This variable can have values between 0 and 1. A dummy variable to indicate the years of recession in the US. It is 1 in recession years, 2008 and 2009, and 0 otherwise. The natural logarithm of a firm s book value of assets. This is used as a proxy for a firm s size. This ratio measures a firm s capital structure. It is calculated by dividing a firm s book value of liabilities by the firm s book value of assets. The percentage of equity based compensation a director of a firm was given in a given year. It is calculated by adding together the fair value of the option awards and stock awards given to the director for that year. This number is then divided by the total compensation the director got that year. The resulting number can have values between 0 and 1. Market value of a company (measured in millions) Revenues (measured in millions) Total assets of a company (measured in millions) Earnings Before Interest and Taxes (measured in millions) The depreciation and amortization expenses for a firm in a given year, because of firms spreading the costs of tangible (depreciation) or intangible (amortization) assets over several years (measured in millions) The value of the preferred stock of a firm (measured in millions) 16

17 Table 1b Control Variables Other Institutional Ownership Equity Compensation Ln(Assets) Debt Ratio CAPX/Assets R&D/Assets Dividend/Assets Advertising/Assets PPE/Assets Cash/Assets Sale/Assets DummyR&D DummyCAPX DummyTotal Debt DummyDividend DummyCash DummyPPE DummyAdvertising DummyDepreciation and Amortization DummyPreferred Stock Time dummies A variable that is calculated by dividing the amount of shares of a firm that are owned by institutional investors other than mutual funds by the amount of common shares outstanding for that firm. This variable can have values between 0 and 1 The percentage of equity based compensation a director of a firm was given in a given year. It is calculated by adding together the fair value of the option awards and stock awards given to the director for that year. This number is then divided by the total compensation the director got that year. The resulting number can have values between 0 and 1. The natural logarithm of the book value of the assets of a firm. The ratio that divides the total book value debt by the book value of assets. The ratio of Capital Expenses divided by a firm s book value of assets. The ratio of Research & Development Expenses divided by a firm s book value of assets The ratio of Dividend paid out to shareholders in a given year divided by a firm s book value of assets. The ratio of Advertising Costs divided by a firm s book value of assets. The ratio of the gross value of the firm s Property, Plant and Equipment divided by a firm s book value of assets The ratio of Cash divided by Assets The ratio of a firm s Sales divided by a firm s book value of assets. Has a value of 1 if the value of Research and Development is missing and has a value of 0 if the value of Research and Development is not missing. Has a value of 1 if the value of Capital Expenses is missing and has a value of 0 if the value of Capital Expenses is not missing. Has a value of 1 if the book value of debt of a firm is missing and has a value of 0 if the book value of debt of a firm is not missing. Has a value of 1 if the value of Dividend is missing and has a value of 0 if the value of Dividend is not missing Has a value of 1 if the value of Cash is missing and has a value of 0 if the value of Cash is not missing Has a value of 1 if the gross value Property, Plant and Equipment of a firm is missing and has a value of 0 if the gross value of Property, Plant and Equipment is not missing Has a value of 1 if the value of Advertising Costs is missing and has a value of 0 if the value of Advertising Costs is not missing. Has a value of 1 if the value of Depreciation and Amortization is missing and has a value of 0 otherwise Has a value of 1 if the value of Preferred Stock is missing and 0 otherwise. Dummies that have a value of 1 if the observation belongs to the year the dummy belongs to and has a value of 0 if the observation belongs to another year 17

18 Table 2 Describing Variables In this table I present the means, standard deviations, minimums and maximums of the most important variables I have used. As can be seen below, the mean Mutual Fund Ownership is 0.158, which means that about 15.8% on average of the shares of firms are owned by mutual funds. With a standard deviation of 11.8% and a maximum of 84.4%, the percentage of shares owned by mutual funds varies quite a lot. Also the percentage of shares held by other institutional investors varies by quite a lot, with a standard deviation of 20% and a maximum of 96.7%. The amount of observations for Equity Compensation is a lot lower than for other variables, because the Execucomp database did not have that much data that matched with my other data from the Thomson Reuters and Compustat databases. The large difference between the minimum and maximum of the R&D/Assets variable can be explained by the fact that innovating/growth firms spend relatively a lot of money on Research and Development to invent new technologies and products but don t have many tangible assets and little or no sales. Variable Observations Mean Standard Deviation Minimum Maximum Tobin s Q ROA Mutual Fund Ownership Other Institutional Ownership Debt Ratio Ln (Assets) Equity Compensation Dividend/Assets CAPX/Assets PPE/Assets Cash/Assets R&D/Assets Advertising/Assets

19 Graph 1: Mutual Fund Ownership per year Below you can see a graph, containing the average percentage of shares of a firm held by mutual funds per year. As can be seen there is an upwards trend from 2004 and on. In 2004 the average mutual fund ownership was 12.31% of a firm s common shares. In 2008 this has grown to 16.68% and in 2011 to 20.13%. A probable explanation for the increase in mutual fund ownership is the increase in assets that mutual funds have under their control now compared to several years ago, as also discussed in the introduction of this thesis. In 2004, for example, mutual funds in the US had 8 trillion dollars of net assets under their control. In 2011, this number has grown to 11.6 trillion dollars (Investment Company Institute, 2013) Year 19

20 Table 3 Mutual Fund Ownership in different industries The distribution between industries of the firms being researched in this thesis is given below. Descriptions for the different one digit sic codes are given by the SIC manual of the United States Department of Labor (1987). The largest part of my sample consists of firms in the Manufacturing industry. Less than one percent of the observations come from the Agricultural, Forestry and Fishing industry and from the Public Administration industry. These industries are therefore not considered when comparing the industries in this table. Looking at the other industries, it is clear that the Mining and Construction industry has on average the firms with the highest percentage of their shares held by mutual funds. On the other hand, the Financing, Insurance and Real Estate industry has on average the lowest percentage of their shares held by mutual funds. The other industries have quite average percentages of shares held by mutual funds. One digit SIC Industry Amount of observations N Mutual Fund Ownership & & 8 9 Total Agricultural, Forestry and Fishing Mining and Construction Manufacturing Transportation, Communication & Public Utilities Firms Wholesale and Retail Trade Financing, Insurance and Real Estate Services Public Administration : Methodology In this section I will describe the different models that I will use in to get to my Empirical Results in section 3. With regard to explaining firm performance with ownership structure as an independent variable, endogeneity may exist. Demsetz (1983) first claimed this argument of the endogeneity of a firm s ownership structure. He argues that a firm s ownership structure is an endogeneous outcome that is most likely influenced by each firm s characteristics and environment under which it operates. Himmelberg et al. (1999) argue that this endogeneity might be caused by the unobserved heterogeneity in the environments of firms. This unobserved heterogeneity means that some firms might have different ownership structures than other firms, because for those particular firms, it might work better. The different ownership structures and their effect on firm performance could then be determined by observable and unobservable firm characteristics. As did Himmelberg et al., to account for possible unobservable firm characteristics and to reduce possible endogeneity, I will use a firm fixed effects model for my regressions. 20

21 Researchers also argued that there might me more than just a one-way causality between a firm s ownership structure and firm performance. Holderness (2003) claimed there could be a reverse causality issue. He explains firm performance may cause a certain ownership structure for a firm, and not the other way around. An explanation Holderness gives is that investors seek firms with high firm value, because these firms might have higher private benefits of control. This is also probable for mutual funds, since they are constantly watched and chosen (or not) by clients because of their ability to reach certain returns for their portfolios. Therefore they could tend to try and chase firms with high firm values. To account for the possible effect of reverse causality I will use lagged values for my Mutual Fund Ownership and Other Institutional Ownership variables. This way I can observe what effect of a move first in ownership structure has on a subsequently observed firm performance a year later. For my first hypothesis I will use the Ordinarly Least Squares (OLS) regression method. I will use the following model: Where is Tobin s Q for each firm i at time t. MutualFundOwnership is determined for each firm i and is lagged one year. The same goes for OtherInstitutionalOwnership. The other control variables X are also determined for each firm i at time t. is the unobserved timeinvariant individual firm effect and is the error term for each firm i at time t. For my second hypothesis I use the following model: ( ) For this hypothesis the interaction variable ( ) has been added to observe a possible moderating effect of a firm s size on the relationship between mutual fund ownership and firm performance. This effect has also been lagged one year to reduce the possible effect of reverse causality. The third hypothesis uses the following model: 21

22 ( ) The moderating effect is now given by the variable, ( ) which is also lagged. With this model I will investigate any moderating effect of a firm s capital structure on the relationship between mutual fund ownership and firm performance. The following model is used for the fourth hypothesis: ( ) To test the moderating effect of the amount of equity based compensation directors get, the variable ( this model. ) has been added for Finally, the fifth model for testing the fifth hypothesis is: ( ) In this model, the moderating effect of the recession years 2008 and 2009 are being tested with the variable( ). It will show whether the relationship between mutual fund ownership and firm performance was stronger, weaker or the same in the recession years as in other years of the sample. After doing the regressions for model two through five, I will further test the moderating effects of the moderating variables used in those regressions. For the continuous moderating variables ( ), ( ) and ( ) I will follow Cohen et al. (2003) in the way I will further examine the effect of the moderating variables. I will look at the effect of mutual fund ownership on firm performance at three 22

23 levels of the continuous independent variables: Ln(Assets), Debt Ratio and Equity Compensation. The first level I will test the effect will be at a moderate level of the corresponding independent variable, the mean. The second level is a high value of the independent variable, mean + one standard deviation. The third level will be a low value of the independent variable, mean one standard deviation. On the next page you can find the correlation table of the most important variables I use in my research. The next section will present my Empirical Results. 23