The effect of ownership structure on short sale disclosure
returns: a research on the European market
F.H.P. Uithoven 5974364
07/06/15 Utrecht
Abstract
This paper examines the effect of the ownership structure at a firm when a large short sale is announced. There is a significant cumulative abnormal return on short sales around the announcement date on the European market. The average cumulative abnormal return is positive at 0.6%. There is no evidence found that the ownership types have a difference in influence on the cumulative abnormal of short sale announcement returns. The degree of ownership concentration at a firm is negatively correlated with the short sale announcement returns and is significant a 5% level.
Key words: Short sale returns; type of ownership, ownership concentration
Dr. S. Arping
Economics and business: Finance University of Amsterdam
2
Statement of Originality
This document is written by Student Ferdy Uithoven who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 1. Introduction
A short sell is a bet on the decline of a stock and investors seek to profit from it (D’Avolio, 2002), (Brent, et al. 1990). It enhances the law of one price for securities because investors have the opportunity to take offset positions (Brent, et al. 1990). The announcement of a short position is a bearish signal and when the market follows, it results in a negative return. (Desai, Ramesh and Thiagarajan, 2002).
As of November the first, 2012 it is by European law mandatory to disclose a substantial short position on a publicly listed company when it exceeds 0.5%. For example, on the Dutch market this has to be done at the Authority of the financial markets. The reasons for this new regulation is reducing systemic risk, preventing market fragmentation, reduce the risk of financial instability and reducing the scope for regulatory arbitrage and compliance costs. This new regulation is a good opportunity to see how the market reacts to a short sell position on the market.
The importance of the ownership structure of a firm with regard to the firm performance is derived from a corporate governance perspective. (Berle and Means, 1932) (Shleifer and Vishny, 1996), (La Porta et al., 1999), (Thomsen and Pedersen, 2000). Corporate Governance deals with the allocation of rights and responsibilities among stakeholders. Berle and Means (1932) and Shleifer and Vishny (1996) state that corporate governance can be seen as an agency problem with the separation of ownership and control. Corporate governance issues occur when there is an agency problem (Hart, 1995). Given by Fama and Jensen (1983), agency cost include the cost of structuring, monitoring and bonding a set of contract among agents with conflicting interests. When a company is largely held by small owners two problems arise. First, the voting rights of small individuals are not large enough to gain control. Second, small investors have no incentive to monitor the managers of a company. Monitoring is costly and a public good. If one investor benefits from doings so, all will. This way each shareholders will do nothing with the hope for a free ride and thus somebody else will pay the costs (Hart, 1995). One way of reducing agency costs is by having a large block holder. Large block holders are inclined to monitor managers and thus limit possibility that managers act in their own interest (Shleifer and Vishny, 1986). Another agency problem is present if the incentives of majority and a minority shareholder do not correspond. The majority shareholder can be an institution and can use his power to improve his own position (Hart, 1995). These institutions hire managers to act on their behalf resulting in a new agency problem (Hart 1995). Different type owners of a firm have different objectives because there is a unique mix between the owner’s wealth, risk aversion and the emphasis they put on shareholder value (Shleifer and Vishny, 1997). This study examines the effects of ownership structures on a short sale announcement return.
4
Although previous studies have been done regarding short selling, an examination of the European market is relevant and contributes to prior literature because of the following reasons. As of my knowledge, there is no previous extensive research done regarding short selling on the European market. Also, the effect different types of owners and the degree of shareholder concentration at a firm has on the short sale announcement has never been tested.
The following research question is formulated to further examine the short sale announcement returns: Does the ownership concentration and ownership type have influence on the return of a European listed company when a large short position is announced?
To give a substantiated answer on the research question there are three hypotheses formulated. The first hypothesis examines if a short position announcement has effect on the stock return. This is done by performing an event study. The second and third hypotheses test whether the different type of owners and the degree of ownership concentration at a firm have influence on the short position announcement returns.
In this study a data set is constructed of 397 unique short sale announcements. The short sale announcements are obtained from the site of the European financial authorities from the period 01/11/2012 till 05/01/2013. They give information on the position holder, the issuer, the net short position and the date of the announcement. Performing an event study there is a positive abnormal return found of 0.6% for a three day event study significant at a 5% level. The effects of the types of shareholder and the degree of shareholder concentration is examined by and ordinary least square method and an analysis of variance with the obtained cumulative abnormal returns of the short sale announcement returns as the dependent variable. The results are insignificant for the different types of owners but are significant for the degree of shareholder concentration at a firm at a 5% level. The degree of shareholder concentration is negatively related to the short position announcement
The outline of this paper is as follows. Section 2 is a literature review based on previous studies done on short sales, ownership types and ownership concentration. Section 3 gives the hypothesis and method used. Section 4 describes the data and gives the descriptive statistics. In section 5 the results are presented. The conclusion is found in section 6 were also recommendation for further research is proposed. Appendix A defines the variables used in this paper.
5 2. Literature review
2.1 Short selling
A short sale is a transaction at which an investor sells a borrowed security with the intention to buy it back in the future (D’Avolio, 2002), (Brent, et al. 1990). This yields in a profit for the investors if the price declines. With a short sale an investor thus expects a security to decline and hopes to gain from it. Short sales play an important role in the theory of asset prices models. The option of taking offset positions enforces the law of one price (Bris, A et al., 2007). Short selling bears extra cost and therefore short sellers are more likely to be informed traders (Diamond and Verrecchia, 1987). These traders will not participate unless they will be paid for the increased costs associated with short selling. Short traders tend to be more informed. They reflect negative information to the market and this puts pressure on the stock price which ultimately lowers it permanently (Desai, Ramesh and Thiagarajan, 2002). It is important that not all shares can be lend out. Because ultimately someone has to buy back the shares. We therefore expect equilibrium to keep the supply and demand steady. Hence, the investors are heterogeneous (Charles and Lamont, 2002). Saffi and Sigmurdson (2011) argue that stocks with stringer short sell constraints face a lower price efficiency. This means that being able to short a stock helps to set a subsequent stock at its fundamental value.
Asquith, Pathak and Ritter (2005) argue that there are two main motives for short selling. These are value or for arbitrage motives. Thus if investors believe that a stock is fundamentally overpriced the main motive to short is because they believe that the price of the stock is going to decline in the future. The other motive is where investors see arbitrage opportunities, for instance convertible bond arbitrage.
2.2 Arbitrage
When linked with other assets, establishing a short position may yield in arbitrage opportunities (Brent, et al. 1990). Going long and short on a subsequent asset can never lead to a profit and therefore it is required that this is combined with a security that is related with the stock. One can think of an option on the stock, a convertible security, a stock index future or any other instrument that is correlated with the value of a stock (Brent, et al. 1990). Also the arbitrage opportunities in convertible bonds are exploited by investors and are becoming increasingly popular. A convertible bond is a security like debt but the bond can be converted to a share. Convertible bond arbitrageurs buy the undervalued convertible bonds and hedge their position by shorting the subsequent stock. They wait until the price of the convertible bond rises till there is no more undervaluation before they unwound their position. In a takeover arbitrage an investor short the acquiring
6
firm and goes long on the firm that is been taken over. Convertible bond and takeover arbitrate are the most practiced reasons to short a stock with regard to exploiting arbitrage opportunities (Asquith, Pathak and Ritter, 2005).
2.3 Value short
An announcement of a short position signals bad news to the market and results in a negative return (Desai, Ramesh and Thiagarajan, 2002). This paper confirms the expectation that bearish information is reflected in the stock price and the adjustment of the stock price is immediate upon announcement date (Desai Ramesh Thiagarajan, 2002). This immediate reaction is due to the efficient market hypothesis. This hypothesis states that all information known to the market is directly incorporated in the share price (Bodie, Kane and Marcus, 2011). Since it is mostly informed investors that do short selling it is common belief that the price reaction is permanent.
2.4 Short Squeeze
An aspect of short sale is that the borrower must give back the shares on demand if, for instance, the lender wants to sell the stock. The short seller must find a new lender to maintain their position. If the short seller is unable to do so he must repurchase the shares in the open market to meet the requirements of the lender, this is a liquidity cost for the short seller (Dechow, Hutton, Meulbroek and Sloan, 2001). This can drive the price of the shares upwards and is known as a squeeze risk for the short seller. A short squeeze is less likely for highly liquid stocks because an alternative lender can be found more easily. High market capitalization and institutional ownership are indicators of liquidity (Dechow, Hutton, Meulbroek and Sloan, 2001). An alternative view is that a short interest represents a bullish signal when the short seller’s reputation is not sufficient. Since people whom are short will buy the stock in the future, the short interest reflects an increased demand later on (Desai, Ramesh and Thiagarajan, 2002). But their findings are not consistent with most existing literature.
2.5 Tax motives
Other reasons for going short, rather than the belief that an asset is overpriced or arbitrage motives, is tax related (Brent, et al. 1990). The ability of going short and long on the same asset gives the investor an opportunity to defer capital gain taxes and thus reduces its cost. By going long and short on the same asset the investor cannot yield a profit but delay the recognition of a capital gain (Brent, et al. 1990). This may be profitable when taxes are expected to be lower in the future. A seasonal pattern in short interest is a weak indication that reducing taxes is a motivation for short selling (Brent, et al. 1990).
7 2.5 Short sell constraints
Asquith, Pathak and Ritter (2005) state that there are three main empirical literatures on short selling sell constraints. The first implies that a high short interest forecasts low future returns. In this case the market believes that the stock is above its fundamental value. The second looks at the actual cost by looking at the interest rate or rebate rate. If the supply of the share that is available for borrowing exceeds the demand the rate is approximately equal to the Fed rate. If the demand is high and the supply is low the rebate rate can become negative. The last one argues that the short sales depend on the stock ownership by institutional investors. The degree of institutional ownership serves as a proxy for the supply in lending for short sales. In this paper I will further examine the first and the last literatures to give a substantiated answer on my research question.
2.6 Direct and indirect costs
There are two types of short sell constraints that limit the possibilities of investors that are willing to borrow and could lead to an overvaluation of a stock. The two types are the direct and indirect short sell constraints.
The indirect cost of going short the culture of institutional constraints. For example, 30% of mutual funds are allowed to short their stock and roughly 2 a 3% do actually engage in short selling (Charles and Lamont, 2002), (Asquith, Pathak and Ritter (2005). Also the institutional investors are more likely to be professional in their investment decisions and therefore it is more likely that individual, small, investors are not rational in their behavior. Hence, individual investors are less likely to lend stock. Another reason is the fact that institutional investors have better access to the capital market because they can exert economies of scale. It therefore makes it easier to engage in stock lending (Asquith, Pathak and Ritter, 2005).
The immediate cost of going short is a direct short sell constraint (Nagel 2005). In order to sell short there must be a lender that is willing to lend the subsequent stock. The short seller must leave collateral and pays an interest rate on this collateral, hence in a perfect capital market this would be zero (Nagel, 2005). An extra cost is incurred when the stock, in contrary to the beliefs of the short selling side, goes up. In this case an investor receives a margin call that may be costly and can ultimately result in the unwound of the deal.
2.6 Institutional ownership
Nagel (2005) finds that the most active lenders are passive mutual funds and for that reason a low institutional ownership increases the direct short sale constraints. Also the paper of Chen, Hong and Stein, (2002) use the institutional ownership as a short sell
8
constraint. They argue that short sell constraints are linked to the amount of share that is available for borrowing. This is consistent with the findings of D’Avolio (2002). This paper finds that 55% of the cross sectional lending variation is explained by the institutional investor ownership. Thus when the degree of institutional ownership is high the possibilities to sell short also increase. If, for example, a negative free cash flow announcement becomes publicly available the subsequent firm that is owned by an institutional investor reacts more negatively to this news. But if there is a low degree of institutional ownership, hence less possibilities to short, the market under reacts (Asquith, Pathak and Ritter, 2005).
2.6 Ownership structure
Mintzberg (1983) proposed two dimensions of ownership. These are involvement and detachment. It differentiates between owners who play an active role in the decision making and owners who are passive and do nothing. The other part is the degree of concentration whereas a high concentration yields in a stock that is closely held and a low concentration yields in a firm that is widely held. Combining these two dimensions gives four types of ownership. Namely, dispersed-detached, dispersed-involved, concentrated-detached, concentrated-involved (Mintzberg, 1983).
2.7 Different goals
Different investors have different objectives (Levin and Levin, 1982) and is due to the fact that there is a unique mix between the owner’s wealth, risk aversion and the emphasis they put on shareholder value (Shleifer and Vishny, 1997). The difference in company strategies is contributed to the fact that there are differences in goals, capital structure, growth rates, dividend policies and profits that arises from the objectives of their owners (Thomsen and Pedersen, 2000).
2.8 Types of ownership
Thomsen and Pedersen (2000) distinguish between family, financial firms, institutional investor, government and other companies as main identifiers of ownership structures. For example, an institutional investor’s main goal is shareholders wealth (Thomsen and Perdersen, 2000) and may care less about social outcomes. Family owner types are typically wealthy and may have a long term commitment to the firm (Thomsen and Pedersen, 2000). When a government holds a large share in a firm it may put emphasis on social outcomes and goals rather than shareholder wealth. In Germany some companies are owned by banks, which are financial firms, and have internalized their banking relationships (Thomsen and Perdersen, 2000). Financial firms have an incentive to monitor firms because they also provide credit. This means that the financial owners have less
9
informational asymmetry with regard to the subsequent firm. Industrial owned firms are owned by other, non-financial, companies and also cross holdings are included. This is the case when the company holds shares of their own stock. Industrial owned firms are part of conglomerate and are mostly undervalued and are traded at a discount (Lins and Servaes, 1999).
2.9 Ownership concentration
The degree of shareholder protection is different among the countries in Europe. For example the German, French and the Dutch market have a lower degree of investors’ protection because civil law is imposed. A paper by Porta, Lopez-de-Silanes, Shleifer and Vishny (2000) states that strong investor protection is associated with effective corporate governance. This implicates that there is more concentrated ownership compared to the UK, where common law is practiced. Shleifer and Vishny (1997) state that large ownership concentration may be a substitute for reducing the agency costs, when there is less investor protection. It means that an owner that holds a significant portion of the shares has an incentive to monitor and forces managers to take actions favorable to the effective owner. The Paper of Clark and Wojcik (2005) shows that there is a negative correlation between ownership concentration and the risk adjusted stock return. This analysis on the German market shows that investors discount the value of the firms with a high ownership concentration, associated with bad corporate governance, because the cost of capital is higher (Clark and Wojcik, 2005). Also the paper of Slovin and Sushka (1993) show a negative relationship between ownership concentration and firm performance. They examine the deaths of inside block holder owners and see that the shareholder value increases when such an ‘event’ occurs. They find no evidence for the incentive of large shareholders to monitor and in effect share the same interest as minority shareholders (Slovin and Sushka, 1993)
3. Hypothesis development and Methodology
This section gives the hypothesis development and the subsequent methods to answer the research question: Does the ownership concentration and ownership type have influence on the return of a European listed company when a large short position is announced? 3.1 Hypothesis 1 and methodology
Previous papers show that there is a negative abnormal return associated with the announcement of a short position (Diamond and Verrecchia, 1987), (Desai, Ramesh and Thiagarajan, 2002), (Asquith, Pathak and Ritter, 2005). Following these studies I expect a negative abnormal return for short sale announcements
10 The first hypothesis looks as follows,
Ho = there is no abnormal return on the announcement date of a short position (1) Against
Ha ≠ there is a negative abnormal return on the announcement date of a short position (1)
The number between the brackets (.) refer to the hypothesis
To test the first hypothesis, There is a negative abnormal return associated with the announcement of large short positions, I will test whether the Ho (1) can be rejected An event study will be performed to test the first hypothesis and examines if a negative abnormal return is associated with the announcement of large short sale positions of institutional investors. There is a time lag of maximum one day between the actual short sale and the announcement of the short sale. The announcement of the short sale is made public after the market closes. An event study is a method to test the influence of an event on the value of a firm. The event study will be done as follows. First a benchmark period is chosen of at least 30 days prior to the announcement return with a minimum estimation 170 days to establish the expected return of a subsequent stock. This return will be established by using the CAPM model. The expected return is then subtracted from the announcement return of the shorted stock. The announcement date of the short position is chosen because all information known to the market will be incorporated immediately upon the announcement date (Bodie, Kane and Marcus, 2011).
The formula (1) to estimate the abnormal return looks as follows,
𝐴𝑅
𝑖= 𝑅
𝑖− 𝐸[𝑅
𝑖|𝑅𝑚] (1)
Where 𝐴𝑅𝑖 stand for the abnormal return and i is the subsequent firm, 𝑅𝑖 stands for the actual return on the event date and 𝐸[𝑅𝑖|𝑅𝑚] stands for the expected return based on the estimated benchmark where Rm is estimated alpha and beta of the subsequent firm. The formula (2) to test cumulative abnormal returns looks as follows,
CAR
𝑖𝑡= ∑
𝑡=𝑛𝑡=−𝑛𝐴𝑅𝑖
(2)
.Where CAR𝑖𝑡 stands for the cumulative abnormal return of firm 1, t=n stands for the days after the event and t=-n stands for the days prior to the event. If the investors prior to the announcement date already know about the short sell event the CAR𝑖𝑡 should be equal to the
expected return on the market and therefore should not bias the results (Kanay and
11
Kronlund, 2009). In total 3 event windows will be tested to see if there is any difference. Also the CAR𝑖𝑡 is divided into subgroups of ownership types and ownership concentrations to examine the effect further.
3.3 Hypothesis 2: Ownership types
Investment decisions vary by the different governance systems because of the optimization problem. Namely the profit and utility maximization (Fama and Jensen, 1983). Meaning that there is a tradeoff between profit and utility maximization. This utility may include, for example, the salary of the workers and the contribution towards sustainability. Levin and Levin (1982) finds that different shareholders have different objectives. The reason why there is a difference in goals is because there is a unique mix between the owner’s wealth, risk aversion and the emphasis they put on shareholder value (Shleifer and Vishny, 1997).
Ownership types are the different types of owners distinguished by Thomsen and Pedersen (2000). Namely, family, financial companies, institutional investor, government and industrial. Family and government owner types are typically wealthy and may have a long term commitment to the firm (Thomsen and Pedersen, 2000). Financial companies, institutional investors and industrial companies lay emphasis on profit maximization Thomsen and Pedersen, (2000). I expect family and government owned firms to react less negative on a short sale announcement because of the long term commitment to the firm and the utility maximization. Government owned firms care about social outcomes more than shareholder value (Thomsen and Pedersen, 2000).
The second hypothesis looks as follows,
Ho = There is no significant difference in return when there is different type of owner (2) Against
Ha ≠ There is a significant difference in return when the owner is family or government owned (2)
The number between the brackets (.) refer to the hypothesis 3.2 Hypothesis 3: Ownership concentration
Ownership concentration is measured as a percentage of shares held by the majority shareholder. The difference in ownership concentration stems from the difference in corporate governance mechanisms. The agency theory, which is explained by separation of ownership and control, explains this difference (Hart, 1995). Strong investor protection is associated with effective corporate governance and a low concentration of ownership. If there is a strong investor protection system, the need to monitor the managers is reduced. The rights of the shareholders are protected by a legal system that provides contracts and
12
law enforcement (Porta, Lopez-de-Silanes, et al. (2000). Previous research has been done regarding the link between ownership concentration and firm performance. Porta, Lopez-de-Silanes, et al. (2000), Shleifer and Vishny, (1997), Clark and Wojcik, (2005) and Slovin and Sushka, (1993) all find a negative relationship between the ownership concentration and firm performance. I therefore expect the degree of ownership concentration negatively related to the short shale announcement return. If a short position is announced and a subsequent firm has better corporate governance mechanisms I expect the market to react less negative on this announcement. The chance that a stock will rise if the firm is badly governed is less likely (Dechow et al., (2001)
The second hypotheses looks as follows,
Ho = The degree of ownership concentration has no impact on the abnormal return (3)
Against
Ha ≠ The degree of ownership concentration has a negative impact on the abnormal return (3)
The number between the brackets (.) refer to the hypothesis 3.4 Methodology for hypothesis 2 and 3
A multiple regression test with the obtained CAR𝑖𝑡 as the dependent variable is regressed on a constant and several variables as shown in equation (3) and (4) to test how each of the owner type and the owner concentration has effect on the short sale announcement.
The regressions look as follows,
CAR
𝑖𝑡= β
0+ β
1× 𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝑡𝑦𝑝𝑒
𝑑𝑢𝑚𝑚𝑦+ β
3× Size + β
4× Volume + β
5× Leverage + β
6×
Book
market
+ β
7× AUM + β
8× Institutional ownership
+ β
9× Net Short + β
813
CAR
𝑖𝑡= β
0+ β
2× 𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛
𝑑𝑢𝑚𝑚𝑦+ β
3× Size + β
4× Volume + β
5× Leverage + β
6×
Book
market
+ β
7× AUM + β
8× Institutional ownership
+ β
9× Net Short + β
8× Convertible bonds outstanding
𝑑𝑢𝑚𝑚𝑦(4)
To examine the second and third hypotheses I will test whether the Ho(2) and Ho(3) can be rejected by using an analysis of variance model analysis (ANOVA) and an analysis of covariance model (ANCOVA) with the obtained CAR𝑖𝑡 as the dependent variable. These methods are best to test the differences in means between the various groups when looking at categorical variables. The ANCOVA method adds covariate variables to the analysis. When the covariates are added it reduces the variance of the dependent variable and it accounts for the differences between groups thus giving more power to comparison of means of the categorical variables. The following assumptions must be met for conducting the ANOVA. The variance across the whole sample must be, approximately, the same for each variable. This implicates that the variance of the CAR𝑖𝑡 of short sale announcements must be the same for each of the ownership types and across the different levels of ownership concentration. Another assumption is that the observations must be independent. I expect this to be the case in this dataset because all observations are unique events. The ANCOVA adds variables to the analysis and also some additional assumptions must be satisfied to give meaning to the inference. The covariates must not correlate with one another. The covariates must have the same effects across all the groups. Meaning that, for example, the size of the firm has the same effect on the CAR of short sale announcements across all types of owners and concentration levels.
3.5 Variable definitions
Ownership type denotes the different types of owners. Namely, financial companies, family owned, governmental, industrial companies and institutional investors.
Ownership concentration denotes the ownership concentration dummies at a subsequent firm. The 3 levels of ownership concentration are firms with none of the shareholder has more than 25%, firms with shareholders between 25% and 50% and firms with a recorded shareholder that exceeds 50%. The degree of information publicly available is important because it has a negative impact on the announcement return (Kalay and Kronlund, 2009). The determinants of publicly available information are the Size of the firm and the Volume by which a share is traded. A higher size and volume are indicators of less informational asymmetry and would thus have a negative coefficient. Also a high size serves as a proxy for high liquidity, reducing the risk of a short squeeze (Dechow, Hutton, Meulbroek and Sloan, 2001). Leverage denotes the debt to equity ratio. The cost
14
bankruptcy is bigger when this ratio is high (Altman, 1968). I thus expect a negative coefficient for this variable. The book-to-market ratio distinguishes between the growth, low book-to-market, and the value, high book-to-market, firms. I expect that a short position announcement on a growth firm has more negative impact on the magnitude of the abnormal return because riskiness of the assets differ (Barber and Lyon, 1997). Fama and French (1992) also show that the book-to-market ratio is positively related to future earnings. I therefore expect the coefficient book-to-market to be positive. The net short position is the short position in percentages and reflects strength of the signal to the public and a negative coefficient is expected (Desai Ramesh Thiagarajan, 2002). AUM denotes the asset under management of the position holder and measures the reputation Asquith, (Pathak and Ritter, 2005). Institutional investors that have a long history tend to have a higher asset under management. Let me clarify this with an example. If for instance a small one time player is going short on a stock the market reacts but does not take it for granted because they don’t know their history. But if a big player, take Goldman Sachs for instance, goes short on a stock the market tends to follow them because of their long established reputation. I expect that he variable AUM to have a negative effect on the announcement return. Institutional ownership serves as a proxy for the supply in lending for short sales (Asquith, Pathak and Ritter, 2005). I expect this variable to be negatively related with the announcement return. Because when there is enough supply of lending the market can fully react. Convertible bond outstanding is a dummy for arbitrage hedging. When a company has convertible bonds outstanding this is an indicator that the short sale announcement has to be associated with arbitrage opportunities rather than the value purpose (Asquith, Pathak and Ritter, 2005). Note that this is a mere approximation because both value and arbitrage shorting are not mutually exclusive. I expect this dummy to be positively correlated with the announcement return. To exclude takeover arbitrage shorting motives I exclude all firms that are engaged in a takeover and are also the acquirer. The variables AUM, Size, Volume, Leverage, Book-to-Market are put in log form. A 1% increase in the beta of an explanatory variable will increase the dependent variable, cumulative abnormal return, also with 1%.
4 Data and descriptive statistics
4.1 Data
The main data sources are: Sites of the financial authorities whom are members of the European Union, DataStream, Thomson One, EDGAR, Zephyr and Amadeus.
There are 501 unique disclosures with regard to the announcement date of a short position and are obtained from the sites of the European financial authorities from the period 01/11/2012 till 05/01/2013. Of the 27 members of the European Union there are 15
15
countries found where short position announcements are disclosed. The data on the sites of the financial authorities contain information about the position holder, the issuer, the net short position and the announcement date of the short position in percentages. The data of the type of owners of the company and the ownership concentration are retrieved from Amadeus. The five dummy variables used to measure this are Financial companies, Industrial Companies, family owned firms, government owned firms and institutional investors. If, for example, the firm is owned by a family then the dummy variable family yields 1 and zero otherwise. The owners that are named ‘publicly listed’ are left out of the sample because Amadeus states that they are unable to exert control over a company. Also the percentage of institutional ownership is retrieved from Amadeus. To measure the ownership concentration the BvD independence indication is used, A, B, C, D and U. It measures degree of independence of a firm with respect to its shareholders. Indicator A states that none of the shareholder has more than 25% of shares and is deemed independent companies by Amadeus. Indicator B has one or more shareholders with more than 25% but not more than 50%. Indicator C has a shareholder with a recorded total ownership above 50%. Indicator D has a shareholder with a recorded direct ownership above 50%. In this paper I will not distinguish between the difference of a total and direct ownership and thus pool these two variables together. Indicator U has an unknown degree of ownership and I remove these samples from the dataset. DataStream is used to get the stock price history of the firms, gather the information of the subsequent benchmark markets to estimate the expected return, find variables to control for size, leverage trading volume and book-to-market ratio. The size of the firm is the market capitalization which is the share price multiplied with the shares outstanding. The market capitalization is taken 10 days prior to the event. Leverage is the debt-to-equity ratio. The trading volume is the amount by which a share is trading per day. The monthly average trading volume is taken prior to the event. The book to market ratio is the book value of the firm divided by the market value of the firm where the market value is the market capitalization. The data of the convertible bonds outstanding is obtained in Thomson one. The total asset under management is obtained from the 13f fillings on EDGAR. If EDGAR did not yield any results the asset under management was obtained by hand. Zephyr is used to check whether an issuer is engaged in a takeover.
16 4.2 Descriptive statistics
This section describes the data of the ownership types and the ownership concentration and the control variables
Table 1 reports the descriptive statistics for the ownership type and the ownership concentration. As can be seen in panel A of table 1 the firms associated with the short sale announcement are largely owned by Financial, Industrial and Institutional companies. They make up for 79.65% of the sample. Contrary to the paper of (Thomsen and Pedersen, 2000) most companies have no recorded shareholders above 25% and firms with shareholders that have a total or direct control above 50% make up for 19,85%, indicating that the firms are widely held. Panel B shows the cross table of the number of observation between the types of ownership and the ownership concentration. Industrial owners have the most shares of 50% and over, namely 34 out of 97. Financial and institutional owners seem to invest in widely held firms; they make up for 164 out of 239. The industrial owner type is fairly even matched across the sample and the family owner type seems to have a bias towards widely held firms.
Ownership type and ownership concentration
Panel A:
#obs
%
#obs
%
Family
47
11,81
<25
239
60,3
Financial
124
31,16
Goverment
34
8,54
25>50
79
19,85
Industrial
95
23,87
Institutional
97
24,62
>50
79
19,85
Total
397
100
397
100
Panel B:
<25
25>50
>50
Type
Family
24
10
13
Financial
86
19
19
Goverment
16
13
5
Industrial
35
26
34
Institutional
78
11
8
239
79
79
:398
Pa nel A reports the number of obs erva tions of owners hi p type a nd the owners hi p concentra tion. Al s o the percenta ges a re gi ven.
Pa nel B reports the cros s ta bl e of the number of obs erva tions owners hi p type a nd owners hi p concentra tion
Type
concentration
concentration
Table 1
17
Table 2 reports the summary statistics of the control variables. Panel A of table 2 reports the mean, median, minimum and maximum value. The average size of the firms being shorted on is 14094 and the minimum value is 1.4 and the maximum value is 1933590, suggesting a high variation in the firm sizes. The leverage ratio and book-to-market seems to be biased upward.
Table 2
Descriptive statistics control variables
Panel A: summary statistics for the control variables
Mean Median Min Max Std.Dev
Size 14094 1513 1,4 1933590 114294 Volume 2.450 677 10 57749 5241 Leverage 3 1,3 -146,9 284 18,2 B/M 3,5 1,7 -21,1 266,2 14,6 AUM 44299 8500 8 744000 109560 InstOwner 19,50% 14,97% 0% 100% 19,52% Net Short 0,85% 0,60% 0,48% 5,33% 0,63%
Panel B: summary statistics for the control variables on ownership types
Family (n=47) Financial (n=124) Gov (n=34) Industrial (n=95) Institutional (n=97)
Mean Median Mean Median Mean Median Mean Median Mean Median
Size 9.985 1424 13089 1364 11635 2492 27245 1538 4432 1289 Volume 1904 337 2728 855 2531 1097 2367 568 2411 710 Leverage 1,73 1,29 4,3 1,3 -2 1,4 4,88 1,44 2,1 1,08 B/M 3,88 1,74 5,5 1,68 1,9 1,04 1,99 1,32 3 2 AUM 66082 7000 36328 8890 85441 9650 28444 6500 45044 8300 InstOwner 12,45% 8,32% 19,10% 15,50% 11% 9,80% 10,11% 6,32% 35,50% 32,50% Net Short 0,85% 0,62% 0,83% 0,62% 1,17% 0,75% 0,77% 0,59% 0,86% 0,60%
Panel C: summary statistics for the control variables on ownership concentration
<25 (n=236) 25>50 (n=79) >50 (n=79)
Mean Median Mean Median Mean Median
Size 12541 1207 9685 1874 23202 1896 Volume 2163 728 2700 507 3068 615 Leverage 4 1,25 0,72 1,2 2,46 1,7 B/M 4,12 1,75 2,1 1,47 3,24 1,52 AUM 44478 8300 37162 8500 50898 9500 InstOwner 23,80% 20,45% 16,35% 8,32% 9,40% 2,32% Net Short 0,85% 0,61% 0,88% 0,61% 0,6 0,82
Panel A reports the summary statistics for the control variables
Panel B reports the summary statistics for the control variables on ownership types
Panel C reports the summary statistics for the control variables on ownership concentration
18
Where, for example, the mean of the leverage ratio is 3.04 and the median is 1.3. When looking at the standard deviation in panel A we see that there is a lot of variation in the leverage and book-to-market ratio. Panel B and C of table 2 describes the subsamples of the variables ownership types and the ownership concentration. Panel B of table 2 shows that the average size of firms is the highest when the largest owner is an industrial company (27245) and is the lowest for institutional investors (4432). For financial and industrial owned firms the average leverage ratio is the highest. Family and government owned firms have low average leverage ratios and the average leverage ratio of the government is -2. This means that the return on borrowed money yields less than the cost of borrowing. This can be a minor indication that the governmental owners do not have the sole goal of shareholder wealth maximization. The book to market ratio is, regarding panel B, fairly even with a median value varying between 1 and 2. Panel C of table 2 reports the control variables divided between the different scopes of ownership concentrations. Widely dispersed firms have the highest leverage ratio on average. The high ratio of debt and equity may be an indication that these firms have better access to debt capital because of better corporate governance mechanisms. Also the degree of institutional ownership is higher for widely dispersed firms. The variable convertible bonds outstanding is, not tabulated, is distributed as follows. 350 firms have no convertible bonds outstanding and 47 have.
Table 3 shows the Pearson correlation between the independent variables. The correlation coefficients can vary between -1 and 1 where the both maximum values implicate a perfect linear relationship and a correlation of 0 means that the two variables have no relationship. The bold figures indicate that they are significant at a 5% level. The ownership types have little influence on the ownership concentration. Firms owned by Industrial companies are slightly negatively correlated with widely dispersed firms and slightly positively correlated with concentrated ownership firms above 50%. The opposite seems to be the case with institutional owned firms. These firms are positively correlated with widely dispersed firms and negatively correlated with the firms that have a high concentration of ownership. The ownership types are, in overall, not correlated with the control variables. Only the degree of institutional ownership, net short and convertible bonds outstanding have significant influence. The firms owned by institutional owners and the degree of institutional ownership have a correlation of 0.5. This positive correlation is logical because when owned by an institutional investor it is apparent that the degree of institutional ownership is also altered. There is no danger of multicollinearity because the coefficient does not exceed 0.7 (Keller, 2008). Government owned firms are positively correlated with the net short position and the variable financial owned firms are negatively correlated with the dummy convertible bonds outstanding. The correlations of ownership concentration on the control
19
variables seem to have little to no influence. Only the degree of institutional ownership is significant for widely dispersed firms and large concentration. Widely dispersed firms seem to have a positive correlation of 0.28 and large concentrated ownership has a negative correlation of -0.03. This implicates that institutional investors tend to invest more in firms that have an ownership concentration below 25%. The control variables have little influence on each other. The degree of institutional ownership is negatively correlated with the average trading volume. The average trading volume is positively correlated with the convertible bonds outstanding. The degree of institutional ownership is negatively correlated with the dummy convertible bonds outstanding. From the table we see that the correlations between the control variables and the ownership type factors and the ownership concentration factors are insignificant for the variables size, volume, leverage, book-to-market and AUM. The degree of institutional ownership, net short position and convertible bonds outstanding yield some significant outcomes. The variables that are insignificant will be included in the ANCOVA.
20
Table 3
Correlation matrix independent variables
Fam Fin Gov Ind Inst <25 25>50 >50 Size Vol Lev B/M AUM InstO NetShort Conver
Fam 0 0 0 0 -0,68 0,01 0,07 -0,01 -0,04 -0,03 0 0,08 -0,13 0 0,06 Fin 0 0 0 0 0,13 -0,08 -0,07 0 0,04 0,05 0,09 -0,04 -0,01 -0,02 -0,11 Gov 0 0 0 0 -0,08 0,14 -0,04 0 0 -0,09 -0,03 0,11 -0,13 0,15 0,06 Ind 0 0 0 0 -0,26 0,1 0,2233 0,07 0 0,06 -0,06 -0,08 -0,3 -0,08 0,05 Inst 0 0 0 0 0,23 -0,12 -0,17 -0,05 0 -0,03 -0,02 0 0,5 0 0 <25 -0,68 0,13 -0,08 -0,26 0,23 0 0 -0,02 -0,07 0,07 0,05 0 0,28 0 -0,05 25>50 0,01 -0,08 0,14 0,1 -0,12 0 0 -0,02 0,02 -0,06 -0,05 -0,03 -0,08 0,02 0,03 >50 0,07 -0,07 -0,04 0,2233 -0,17 0 0 0,03 0,06 -0,02 -0,01 0,03 -0,03 -0,03 0,03 Size -0,01 0 0 0,07 -0,05 -0,02 -0,02 0,03 -0,02 -0,02 -0,01 -0,03 -0,09 -0,02 -0,03 Vol -0,04 0,04 0 0 0 -0,07 0,02 0,06 -0,02 0,04 -0,01 0,02 -0,13 0,05 0,2 Lev -0,03 0,05 -0,09 0,06 -0,03 0,07 -0,06 -0,02 -0,02 0,04 0,04 -0,14 -0,02 0,03 0 B/M 0 0,09 -0,03 -0,06 -0,02 0,05 -0,05 -0,01 -0,01 -0,01 0,04 0,01 0,06 0 -0,05 AUM 0,08 -0,04 0,11 -0,08 0 0 -0,03 0,03 -0,03 0,02 -0,14 0,01 0 -0,05 0,16 InstO -0,13 -0,01 -0,13 -0,3 0,5 0,28 -0,08 -0,03 -0,09 -0,13 -0,02 0,06 0 0,05 -0,11 NetShort 0 -0,02 0,15 -0,08 0 0 0,02 -0,03 -0,02 0,05 0,03 0 -0,05 0,05 0 Conver 0,06 -0,11 0,06 0,05 0 -0,05 0,03 0,03 -0,03 0,2 0 -0,05 0,16 -0,11 0
This table reports the Pearson correlation matrix between the ownership type, ownership concentration and the control variables
The correlation coefficients in bold are significant at a 5% significant level
21 5 Results
5.1 Cumulative abnormal returns
Table 4 reports the cumulative abnormal returns for the short sale announcements. 3 different event windows are tested to evaluate the effect. The event windows vary from -1 to-1, -2 to 2 and -3 to 3 respectively. I will further denote the cumulative abnormal returns with respect to the event windows as CAR(-1,1), CAR(-2,2) and CAR(-3,3). When looking at panel A of table 4 there is a significant positive cumulative abnormal return found across the whole sample for all the event windows tested. For the CAR (-1,1) and CAR(-2,2) the abnormal return is 0.6% and significant at a 5% level. The CAR (-3,3) is 0.5% and significant at a 5% level. This goes against the main findings of prior literature where a negative abnormal return is reported (Diamond and Verrecchia, 1987), (Desai, Ramesh and Thiagarajan, 2002), (Asquith, Pathak and Ritter, 2005). The 𝑯𝟎(1) where there is a negative abnormal return surrounding short sale announcement is not rejected. But there is enough evidence at a 5% significance level that the short sale announcements have a positive abnormal effect on the stock price. The positive abnormal return may indicate that the market does not find the short sale announcement credible and associates the short sale announcement with an increased demand in the future. Better known as a short squeeze (Dechow, Hutton, Meulbroek and Sloan, 2001). Because there is a time lag of maximum one day between the actual short sale and the short sale announcement, another explanation for the positive abnormal returns is due to momentum. Stocks tend to overreact to information (Jegadeesh and Titman, 1993). For example, if a company announces something to which the market reacts overly positive in the eyes of the short sellers. The short sellers may short the stock. Given that the announcement of the short sale comes after the actual sorts sell the positive cars may be attributed to the fact that the stock has an upwards momentum. Panel A also further examines the short sale announcements. A high variation in cumulative abnormal return is found across the whole sample and for all the event windows. CAR (-1,1), for instance, has a standard deviation of 1%, a minimum abnormal return of 87% and a maximum cumulative abnormal return of 61%. Also the median values found are below average indicating a slight bias upward. This high variation in CAR reveals that also negative abnormal returns are present in the sample and, for the various events, the market assesses and acts differently. Panel B and C of table 4 further looks at the CAR of each event window tested. Panel B of table 4 divides the sample in the subgroups of the different ownership types. Again, except for financial owned firms, all CARs(-1,1) are positive and significant. For family and industrial owned firms this is at a 10% significance level and for governmental and institutional owned firms this is significant at a 5% level. Financial owned firms report a negative CAR across all event windows tested at a 5% significance level.
22
This is consistent with the findings of (Diamond and Verrecchia, 1987), (Desai, Ramesh and Thiagarajan, 2002), (Asquith, Pathak and Ritter, 2005). Panel C of table 4 divides the sample into the different kinds of ownership concentrations at the firms. Widely held firms have an average CAR(-1,1) of 1.4% and for CAR(-2,2) and CAR(-3,3) an average of 1.2% significant at a 5% level. Firms with an ownership concentration between 25 and 50 percent have an average CAR around 0%. This is significant at a 10% level. Closely held firms find a negative average CAR across all the event windows. CAR(-1,1) finds an average -0.9% significant at 5%, CAR(-2,2) find an average of -0.08% significant at 5% and CAR(-3,3) an average of -1.2% significant at 10%.
Table 4
cumulative abnormal returns of short sale announcements
Panel A: Average CAR for the whole sample (n=398)
Mean Median Max Min Std.Dev
CAR(-1,1) 0,006** 0,002 0,61 -0,87 0,01
CAR(-2,2) 0,006** 0 0,42 -0,3 0,06
CAR(-3,3) 0,005** 0,001 0,83 -0,34 0,08
Panel B: Average CAR for the subgroup ownership types
car11 car22 car33
Family (n=47) 0,0078* 0,0029** 0,0046**
Financial (n=124) -0,0082** -0,0037** -0,0044**
Government (n=34) 0,017** 0,011** 0,02**
Industrial (n=95) 0,017* 0,0052 0,015*
Institutional (n=97) 0,011** 0,01* 0,006*
Panel C: Average CAR for the subgroup ownership concentrations
car11 car22 car33
<25 (n=236) 0,014** 0,012** 0,012**
25>50 (n=79) 0,0003* 0,0004* -0,005*
>50 (n=79) -0,009** -0,008** -0,012*
This table reports the event study results of short sale announcements on European listed firms from the period 11/01/2012 till 5/30/2013; 3 different event windows are tested referred to as
CAR(-t,t), where -t is the total days prior to the event and t is the total days after the event
For a definition of the ownership types and the ownership concentrations see appendix A
*,**,*** show the significance level for 10 , 5 and 1 percent
5.2 Regression analysis
To examine how the different kind of ownership structures and ownership concentrations have influence on the cumulative abnormal return of short sale announcement a regression analysis is conducted. The cumulative abnormal return (CAR(-1,1) of short sale announcements is estimated for a 3 day window with one day before the announcement and one day after the announcement. The standard errors of the regression are robust and clustered at ownership type and ownership concentration respectively to account for any
23
variation between the groups. The period of the estimation is from 11/01/2012 till 05/30/2012. All the continuous variables are in log form. Table 5 A shows the regression of the CAR of short sale announcements against the ownership types with the control variables: Size, Volume, Leverage, Book-to-market, Assets under management, Institutional ownership, net short position and the dummy convertible bonds outstanding. The family owner type yield insignificant result and is slightly negative. Regression I shows that the dummy the financial owner type is highly significant at a 1% level, t-stat -25.92 > -2.57. If a firm’s largest owner is a financial company the short sale announcement return drops with 1.8%. The market knows that, on average, the financial company is better informed and thus has less informational asymmetries (Thomsen and Pedersen, 2000). If there is less informational asymmetries the share price of a firm reflects the true value better. With more information available for the market, it perceives the short sale announcement as an unknown shock and acts accordingly. Regression IV shows that industrial owners have positive significant effect of 1.4% on the CAR. The t-stat 2.14>1.96 is significant at a 5% level. This increase in the CAR is because industrial owned companies are undervalued on average because of the conglomerate discount (Lins and Servaes, 1999). Implicating that the goal of shareholder wealth maximization is not always met and the risk of the company is diversified. Institutional owned companies yield an increase in the CAR of 2.5% and are, with a t-stat of 1.62, not significant. The control variables size, volume and leverage are all slightly negative but not significant across all ownership types. The book to market ratio is positively correlated with the CAR. All t-stat are above the critical value of t=2.57 and are thus significant at a 1% level. This corresponds with the findings of Barber and Lyon (1997) and Fama and French (1992). A 1% increase in the book-to-market ratio yields in a 1% increase in the CAR of short sale announcements. The total asset under management of the position holder was added to control for the reputation on the market (Asquith, Pathak and Ritter, 2005). For all ownership types an 1 % increase in the AUM yields in a 0.5% decrease in the CAR of short sale announcements. This is significant across all ownership types for a 1% level with none of the t-stats above -2.57. The market seems to follow firms that initiate a short sale with a good reputation on the market.
24
Table 5 A
Dependent: cumulative abnormal return of short sale announcement
(I)
(II)
(III)
(IV)
(V)
Family
Financial
Government
Industrial
Institutional
Owner- type
-0,006
-0,018***
0,024
0,014**
0,025
[-0,42]
[-25,92]
[1.58]
[2,14]
[1,62]
Size
-0,004
-0,004
-0,004
-0,003
-0,004
[-0,69]
[-0,73]
[-0,76]
[-0,67]
[-0,7]
Vol
-0,04
-0,004
-0,004
-0,04
-0,004
[-1,28]
[0-1,07]
[-1,19]
[-1,11]
[-1,2]
Lev
-0,03
-0,003
-0,003
-0,003
-0,003
[-1,25]
[-1,08]
[-1,19]
[-1,08]
[-1,14]
B/M
0,01***
0,01***
0,011***
0,011***
0,01***
[6,45]
[7,5]
[5,85]
[9,5]
[7,48]
AUM
-0,005***
-0,005***
-0,005***
-0,005***
-0,005***
[-3,69]
[-3,48]
[-3,88]
[-3,34]
[-3,7]
InstO
-0,0001
-0,0001
-0,0021
-2,10E-05
-0,0001
[-0,45]
[-0,5]
[-0,36]
[-0,08]
[-0,52]
NetShort
0,0039
0,0034
0,0021
0,0044
0,004
[0.77]
[-0,76]
[0,39]
[0,92]
[0,77]
Conver
0,01
0,006
0,009
0,009
0,009
[1]
[-0,65]
[0,95]
[0,91]
[0,95]
R^2
0,043
0,05
0,047
0,047
0,043
Obs #
397
397
397
397
397
This table looks at the cumulative abnormal returns of short sell announcement regressed against ownership type dummies and the various control variables. The period is from 11/01/2012 till 5/30/2012. The continuous variables are in log form. For a description of the variables see appendix A. The constant is not shown in the regression *,**,*** show the significance level for 10 , 5 and 1 percent respectively
Table 5 B shows the regression of the CAR of short sale announcements against the ownership concentration levels with the control variables: Size, Volume, Leverage, Book-to-market, Assets under management, Institutional ownership, net short position and the dummy convertible bonds outstanding. Regression VI shows the CAR of short sale announcements on a dummy with an ownership level below 25%. If the firm is widely held the CAR increases, after inclusion of the control variables, with 1.3% and is significant at a 5 % level. T-stat 1.99>1.96. Widely held firms are an indication of better corporate governance mechanism (Clark and Wojcik, 2005). This implicates that these firms, on average, are better able to deal with the short sale announcement. The market seems to value this positively.
25
Table 5 B
Dependent: cumulative abnormal return of short sale announcement
( VI )
( VII )
( VIII )
<25
25<50
<50
Ownership -
0,013**
-0,0007
-0,018**
concentration
[1,99]
[-1,17]
[-2,01]
Size
-0,003
-0,003
-0,004
[-1,06]
[-1,19]
[1,29]
Vol
-0,004**
-0,004**
-0,004**
[-2,22]
[-2,09]
[-2,31]
Lev
-0,003**
-0,003**
-0,003**
[-2,57]
[-2,53]
[2,06]
B/M
0,01
0,01
0,01*
[1,73]
[1,87]
[1,84]
AUM
-0,005*
-0,005*
-0,005*
[-1,87]
[-1,91]
[-1,84]
InstO
-0,0001
-0,0001
-0,0002*
[-1,55]
[-0,91]
[-1,66]
NetShort
0,0038
0,0039
0,0035
[0,7]
[0,73]
[0,69]
Conver
0,01
0,009
0,008
[0,51]
[0,52]
[0,48]
R^2
0,046
0,043
0,047
Obs #
397
397
397
This table looks at the cumulative abnormal returns of short sell announcement regressed against ownership concentration dummies and the various control-
variables. The period is from 11/01/2012 till 5/30/2012. The continuous variables are in log Form. For a description of the variables see appendix A The constant is not shown in the regression *,**,*** show the significance level for 10 , 5 and 1 percent
Regression VIII analysis the effect of a firm that is closely held, share ownership above 50%, and finds a significant negative relationship at 5% with the CAR of short sale announcements. The T-stat is 2.01>.196. When a firm is closely held the CAR decreases with 1.8%. This finding also corresponds with the main findings of Clark and Wojcik (2005) and Slovin and Sushka (1993). A negative relationship is found when the ownership concentration at a subsequent firm increases. The control variables have the same slopes as in Table 5B but differ in their significance. The variable volume is significant at a 5% level across all ownership concentration levels with none of the T-stats above -1.95. A higher trading volume serves as a proxy for the degree of informational asymmetries (Kalay and Kronlund, 2009). When the volume increases with 1% the CAR decreases with 0.4%. When there is more information available to the market the short sale
26
announcement is a bigger shock to the market. The variable leverage is significant at a 5% level across all degrees of ownership concentration. None of the T-stats is below 1.96. An increase in leverage of 1% decreases the CAR of short sale announcements with -0.3%. The market thinks that with more leverage the risk of going bankrupt is more severe (Altman, 1968). For this reason the market reaction is bigger. The variable asset under management of the short position holder has also a negative impact on the CAR of the short sale announcement. But this is only significant at a 10% level with none of the T-stats above -1.64. A 1% increase in the total asset under management decreases the CAR with -0.05%. Notice that this is the same inference as the variable AUM found in table 7 A. The variable institutional ownership is only significant at a 10% level for closely held firms. The T-stat - 1.66 <-1.64. A 1% increase in the degree institutional ownership decreases the CAR of short sale announcements with -0.02%
5.3 Analysis of variance
Table 6 reports the analysis of variance and analysis of covariance with the CAR(-1,1) as the dependent variables on the variables ownership type and ownership concentration with the various control variables. The continuous variables are transformed into logarithmic because outliers are problematic for the ANCOVA analysis (Berry, 1987). Panel A reports the two way ANOVA table for the ownership type and the ownership concentration factors and the interaction between two variables. Levene’s test is performed to see if the variances are not different. With and P value of 0.54<0.05 the null hypothesis is not rejected and the variances are homogeneous. The adjusted-R of the model is 0.0093 and with an F statistic of 1.62<2.10 for a significance of 5% the null hypothesis cannot be rejected that both the categories have influence on the short sale announcement. The ownership type yields insignificant results. With an F statistic of 1.39<2.37 the null hypothesis cannot be rejected at a significance level of 10% that the ownership type has influence on the short sale announcement. The ownership concentration factors yield significant results with an F statistic of 2.46>2.30 at a 10% level. But is not significant for a significance level of 5% (F-stat 2.46<3). There is, at a significance level of 10%, enough evidence to believe that at least one of the ownership concentration factor means differentiate from one another. The interaction variable is added to check if the ownership type and the ownership concentration have influence on each other. If this is violated, hence a significance correlation, the two ways ANOVA cannot be computed and the interpretation of the categorical variables are meaningless (Keller, 2008).
27 Table 6
Dependent variable: Cumulative abnormal return of short sale announcement Panel A: ANOVA table for ownership type and ownership
concentration
df F prob>F Adj-R
0.0093
Model 6 1.62 0.1409
Ownership type 4 1.39 0.18 Obs #
Ownership concentration 2 2.46* 0.087 397
Interaction 8 0.45 0.9
Panel B: ANCOVA table for ownership type and ownership
concentration
df F prob>F Adj-R
Model 11 2.24 2.24 0.034
Ownership type 4 1.63 0.16
Ownership concentration 2 2.45* 0.087 Obs #
Size 1 1.47 0.22 397
Vol 1 2.47 0.12
Leverage 1 0.3 0.59
B/M 1 4.29** 0.03
AUM 1 4.93** 0.04
Panel A reports the two way anova table of the factors ownership type and the ownership concentration. Panel B reports the ancova table of the factors ownership- type and ownership concentration. Continuous variables are in log form
For a description of the variables see appendix A
*,**,*** show the significance level for 10 , 5 and 1 percent
The interaction term, ownership type x ownership concentration, is not significant for all levels (F-stat 0.45<2.51) meaning that the interpretations of the independent variables tested are meaningful. Panel B of table 6 reports the two ways ANCOVA. The model in itself, with an F-stat 2.24>1.79, becomes significant at a 5% level. This implicates that the independent variables are significantly different from zero. The adjusted-R is 0.034. The factor ownership type is still insignificant after the inclusion of the covariates. The F-stat is 1.63<1.94 at a 10% significance level. The ownership concentration factor remains the same after the inclusion of the covariates. The F-stat for the ownership concentration is 2.45>2.30 and significant for a 10% level. The post-hoc test of the ANCOVA in table 7 panel B is examined to see how the ownership concentration factors differ. The covariates in Table 7 Panel B are significant for the book-to-market ratio and the AUM at a 1% level. The covariates size, volume and the leverage ratio are insignificant. Notice that the covariates institutional ownership, net short position and the
convertible bonds outstanding
28
are left out because they correlate unevenly with the categorical variables ownership type and
ownership concentration. The assumption of the homogeneity of the regression is thus satisfied.
Table 7 reports the difference in means of the cumulative abnormal return of short sale announcement with respect to the ownership types and the ownership concentrations. A post hoc Bonferroni adjustment to the least significant difference fisher method is conducted because this is the best method when analyzing small number of factors within each category (Keller, 2008. P.537).This test adjusts the t value by dividing with the number of pair wise comparisons. This is calculated as follows,number of pair wise comparisons = factor (factor – 1) 2
The factor is the number of levels for each category. For the ownership type category this is 5 and the ownership concentration this is 3, see table 1. Thus the number of pair wise comparisons for ownership type is 10 and the number of pair wise comparisons for the ownership concentration is 3. T-statistics for ownership type: For a 5% significance level the T-stat (0.05/10,∞) ≈ 2.58 and for a 10% significance level the T-stat (0.1/10,∞) ≈ 2.33. T-statistics for ownership concentration: For a 5% significance level the T-stat (0.05/3,∞) ≈ 2.33and for a 10% significance level the T-stat (0.1/3,∞) ≈ 1.645
Panel A of table 7 reports the absolute differences in means for the types of owners. No type of owner has a significant different mean from each other. All T statistics are below the critical value T<2.33 at a 10% level. From this, the hypothesis H (2), There is a significant difference in return when the owner is family or government owned cannot be accepted. The assumption that every investor differs in their mix between wealth, risk aversion and the emphasis they put on shareholder value (Shleifer and Vishny, 1997) does not alter the cumulative abnormal return of short sale announcements with a 3 day event window. Panel B of table 7 reports the absolute differences in mean for the different degrees in ownership concentration. Firms with an ownership concentration below 25% compared to firms that have an ownership concentration between 25% and 50% have a difference in mean of 13% significant at 5% with an T-stat 1.99>1.96. The difference in mean between firms that have a concentration below 25% and firms that have a concentration greater than 50% is 23%. This is significant at 5% with a T-stat of 2.21>1.96. The difference between firms that have an ownership concentration between 25 and 50 % and firms with an ownership concentration above 50% is 0.9% but yields insignificant results. The degree of ownership concentration at firms seems to have significant influence on cumulative abnormal return of the short sale announcement. The hypothesis (3) that the degree of ownership concentration has a negative impact on the abnormal return is accepted at a 5% significance level. The F-stat 2.43>2.3 from table 5 panel b shows that, at an 10% significance level, that the means of the
