The relationship between corporate governance quality and the probability of being a target of a large short position : an analysis of the Dutch stock market.

(1)

The relationship between corporate

governance quality and the probability of

being a target of a large short position:

An analysis of the Dutch stock market.

-Faculty of economics and business-

Author: Martine Scheerder Student number: 5978092

MSc Business Economics, Finance track Master Thesis

Completion: 2014-04

(2)

Abstract

In this study the corporate governance quality of Dutch firms that are target of a large short position is compared to the corporate governance quality of comparable Dutch firms that are not targeted. Besides, the influence of corporate governance quality on the probability of being a target of a short position is examined. There are two main findings in this paper. First, the governance index scores of

the untargeted firms are significantly better than the governance index scores of the target firms. Target firms score on average, only better on the audit category. On all other categories (i.e. board, anti-takeover and compensation & ownership) the untargeted firms have a better score. Second, the probability of being a target of a short sale position increases if the governance index score of the

(3)

1. Introduction

Institutional investors have become increasingly important as equity holders in financial markets (Smith, 1996). When one of the firms of which the institutional investor holds shares, is

underperforming, the institutional investor has to decide what to do with his holdings in the firm. Although an institutional investor could simply sell his holdings in the underperforming company, often the amount of the holdings is of such large size that the shares cannot be sold without driving the stock prices down and suffering further losses (Gillian and Starks, 2000). Another limitation that institutional investors face when they want to sell part of their portfolio is that their holdings are often indexed. Due to this indexation they cannot sell their shares in the underperforming firm, without selling the rest of the indexed portfolio. (Gillian and Starks, 2000).

When an institutional investor chooses to hold the shares in the underperforming company he can choose to undertake nothing or he can express his dissatisfaction through a form of shareholder activism (Aggarwal, Erel, Ferreira and Matros, 2011). If there exists managerial misbehaviour in the underperforming firm, an action that the institutional investor can undertake is monitoring the

management (Massa et al., 2013). The big disadvantage of directly monitoring the management is that this activity is very time consuming. The threat of a possible short position might be a good

alternative for monitoring the management (Massa et al., 2013). Massa et al., (2013) find that the potential of being a target of a short position disciplines managers and simultaneously reduces their incentives to manipulate.

Since November 2012 the ESMA (European Securities and Markets Commission) implemented new rules regarding the notification of short positions in listed companies. The main new requirements are: An investor that is holding a net short position in the issued capital of a company shall issue a

notification on each event that the position reaches the set threshold of 0.5% of the total issued capital of the company in consideration. When a significant net short position goes below the 0.5% threshold, the notification will still continue to be shown in the notifications part of the register for one more business day. The notification will thereafter be moved to the short sale archive. (ESMA, 2012). The purposes that the ESMA wants to achieve by means of these new regulations are: preventing market fragmentation (thereby increasing the efficiency of the internal market), reducing systemic risk, reducing the risks that can violate financial stability, reducing risks to market integrity arising from short selling and reducing the scope for regulatory arbitrage and compliance costs (ESMA, 2012). By means of this study I try to contribute to the existing literature on the target process of short sellers. This paper focuses on the differences in governance characteristics of firms that are target of a large short sale position and firms that are not. Besides, the relationship between the potential of being a target of a short position and the corporate governance quality of a firm is examined. In this paper a large short position is a position that reaches the threshold of 0.5% of the issued capital of the

(5)

company. The reason for setting this threshold is that only positions of this size and larger are

published in the online short sale register that is maintained by AFM (Autoriteit Financiële Markten). All firms included in this research are Dutch firms that were listed on Euronext Amsterdam during 2012. The Netherlands is chosen as research area since the Netherlands is part of the area where the new short sale notification rules of the ESMA are implemented. Furthermore, I have easy access to the data that need to be collected for the computation of the governance index. Besides, the

Netherlands is a very interesting corporate governance research area since companies can for example choose if they want to create a one-tier or a two-tier board. In this research the short sale target firms, which are collected from the AFM website, are matched to comparable untargeted Dutch firms. The matching procedure is based on industry, total assets and Tobin’s Q. By using this matching method the comparison between the two groups becomes fairer, since the governance scores of similar firms are compared.

The question that I try to answer through this research is as follows: “Has the overall governance quality of a firm any influence on the probability of being a target of a short position?” The research question is answered in two parts. In the first part the governance quality of target firms and their untargeted matching partners are displayed and discussed. In the second part, a probit regression is performed. The dependent variable in this regression is the target variable, which has a value of one in the case the firm is targeted by a short seller and a value of zero otherwise. The governance variable and control variables that might have power in explaining the target process of short sellers are added to the regression. The output of this regression shows if the governance index variable has any influence on the probability of being a target of a short position.

This research is valuable since it incorporates recently introduced regulation. Besides, there is little known about the characteristics of shorted stocks. Dechow et al., (2001) point out that there might be a relationship between fundamental value to market value ratios and future stock returns. Brent et al., (1990) find that stocks that have high betas and for which options or convertible securities are traded, tend to have higher levels of short interest. Angel et al., (2003) find that shares that are actively traded are more often targeted by a short position than shares that are not actively traded. Diether et al., (2009) find that short-selling activity is higher for large stocks with high capitalization, growth stocks, stocks for which the level of institutional ownership is high, expensive stocks, and stocks with

actively traded put options.

Furthermore, a very extensive governance index is hand collected for the aim of this research, which may give some new insights in the overall governance quality or shortcomings of Dutch firms. Besides, newspapers are reporting regulatory about short selling. For example: Banks are more sensitive to short selling than other firms, so it might be a good idea to ban shorting on banks in times

(6)

of financial distress (Ballegeer, 2013, het Financiele Dagblad). American and British hedge funds have earned more than €100 million by gambling on stock price decreases (shorting the stock) of Dutch listed companies (Cohen and Kakebeeke, 2013, het Financiele Dagblad). Hedge funds are making millions by targeting stocks. What is their strategy? Is short selling phenomenon for the overall market? Which stocks do short sellers pick? Unfortunately, little information can be found in previous researches concerning the targeting process of short sellers. Through this study, I try to take away part of the opaqueness around the characteristics of shorted stocks and the targeting variables of short sellers.

To my knowledge, Massa et al., (2013) were the first to examine the relationship between corporate governance and short selling. They examined what effect an increase in short selling potential has on the internal corporate governance quality of a firm. The main proxy that Massa et al., (2013) use for their short selling potential (SSP) variable is the part of company shares available to be lent to short sellers. They find that an increase in the potential of being a short sale target increases the internal corporate governance quality of a firm. However they did not investigate if corporate governance quality is an explanatory variable in the targeting process of short sellers. This study provides a supplement to the existing literature, since this study does investigate the following question: Is the corporate governance quality of a firm an explanatory variable in the targeting process of short sellers? The main contribution of this paper is to gather more information about the targeting process of short sellers and the characteristics of shorted stocks. Besides this research tries to complete the research of Massa et al., (2013), since I try to confirm that the action of investors to establish better internal governance when a firm is threated by a potential short sale (Massa et al., 2013) is a scientifically supported action to undertake.

The structure of this paper is as follows: In section 2, a broad overview of the most important literature concerning short selling is presented. In section 3, the hypothesis is constructed and explained. Next, in section 4, the methodology is revealed. Then, in section 5, the data are described, the variables are defined, and the summary statistics are displayed. In section 6 the results of this research are discussed. Next, section 7 shows three important additional tests on the data and the model. Thereafter, section 8 provides a discussion towards the limitations of this research. Finally, in section 9 the conclusions and the recommendations for future research in this area are revealed.

(7)

2. Literature Review

In this chapter the most important literature with regard to short selling is discussed. In section one the short selling concept is explained and the market players are revealed. In section two the three most important reasons for investors to go short are mentioned. Next, in section three a review of the findings of existing papers on characteristics of target firms is given. Furthermore, section four provides a discussion about short sale regulations. Finally, in section five the link between short selling and corporate governance quality is explained, and the connection between previous research and this research is explained.

2.1. Explaining the concept

How is a short sale created? An investor borrows stock from another investor’s margin account to subsequently sell the borrowed stock to a third party (Brent et al., 1990). This transaction usually happens through a broker. To make sure the borrower pays back the shares, he provides the lender a collateral. The lender pays the short seller interest on this collateral, which is called rebate rate (Diether et al., 2009). This interest rate is based on loan supply and demand (Diether et al., 2009). If the number of shares that is available to borrow is higher than the number of shares demanded to short, then the rebate rate will be almost equal to the risk free interest rate (Asquith et al., 2005). All rights declared or dividends paid during the period the stock is shorted have to be paid to the lender of the shares, since the short seller only borrows the shares and does not own them (Nagel, 2005). The investor who shorted the stock will only be able to make a profit when the stock price of the shorted stock sufficiently declines. Since there is no limit on the increase of the stock price, the potential losses for the short seller are infinite. On the other hand, the stock price cannot be lower than zero, so the profit is limited. Thus, an investor can lose more than he initially invested, but the highest profit he can make is a 100% gain if a company goes bankrupt. This is the reason why brokers often advise inexperienced, risk-adverse investors to purchase put options instead of shorting stock (Brent et al., 1990).

There are two types of short positions: covered positions and naked positions. When a short seller has a naked position he is not borrowing any stock to cover his position within the standard settlement period. Whereas when an investor holds a covered position he managed to borrow the shorted stock from a lender (Beber and Pagano, 2013).

In the United States, the big custody banks are the largest lenders of shares (D’Avolio, 2002). These banks lend as agents on behalf of large institutional owners such as large index funds, pension funds, public retirement funds, mutual funds and endowments. Among these institutional investors, the passive index funds participate most extensively in the lending process (D’Avolio, 2002). The short sellers are mostly sophisticated private investors, hedge funds and institutional investors (Nagel, 2005). As argued before, the risk of short selling is infinite. Since the short selling process is very

(8)

risky investors have to be well informed and experienced traders who are wealthy enough to survive the potential losses.

The majority of the short-sellers in the Dutch stock market are large U.K. hedge and investment funds like Blackrock, Marshall Wace LLP, The Children’s Investment Fund Management (UK) LLP (AFM Register). Unlike the regulation in some other European countries, the Dutch regulators allow naked short positions. Private investors can hold a short position in the Netherlands, however the

requirements for private investors to acquire a short position are very strict and most banks do not allow their customers to go short.

2.2. Motives behind short selling

There are three main motives for investors to go short, investors can use a short position for hedging arbitrage, besides they can use a short position to speculate or to delay tax recognition. First, the hedging motive is discussed. A perfect hedge, a long and a short position in the same stock that offset each other, cannot provide any profits to the security holder (Brent et al., 1990). To gain profit on this position the security holder needs to obtain an additional product, which is linked to the value of the obtained stock (Brent et al., 1990). An example of such a product is a convertible security. In the last two decades, the most successful hedging strategy among big hedge funds has been convertible arbitrage (Loncarski et al., 2009). The aim of this technique is to take advantage of the underpricing of convertible bonds by simultaneously taking a long position in a convertible, and a short position in the underlying asset (Loncarski et al., 2009). The total number of shares that has to be sold short is a function of the following three elements: the conversion ratio, the sensitivity of the conversion option’s value to shifts in the price of the underlying equity, and the sensitivity of the delta to shifts in the price of the underlying (Loncarski et al., 2009). Hedge funds that use convertible arbitrage

strategies select underlying stocks that pay low dividends, are undervalued, are liquid, and have no constraints to be sold short (Loncarski et al., 2009). If, at the time the strategy is setup, the convertible

bonds are underpriced, investors can make use of this arbitrage to gain profits. (Loncarski et al., 2009)

Short positions may also be used to speculate in the stock market. This motif is according to McDonald and Baron (1973) in two-thirds of the cases the reason for investors to go short (Brent et al., 1990). Bearish speculators, who expect stock prices to decline, may use short positions to express their feelings (Randall and Dickson, 1994). These bearish speculators believe that the existing prices do not cover all information available. They believe that the true stock price is lower than the existing one (Brent et al., (1990)). The speculators try to make a profit by betting on stock price decreases of the, in their eyes, overvalued stocks.

(9)

A third reason why investors may go short on a stock is to delay tax recognition (Brent et al., 1990). An opportunity to delay the capital gain taxes can be created by simultaneously holding a long and short position in the same stock. By going short in the stock that is also held long, an investor can lock in a profit, but postpone the recognition of a capital gain (Brent et al., 1990). This may be an event of serious impact when the investor, due to this strategy, will be taxed at a lower interest rate in the upcoming tax period (Brent et al., 1990). Besides postponing capital gains, it might also be interesting for investors to postpone the recognition of a loss (Brent et al., 1990). Tax rates may rise next year, which makes it more valuable for an investor to postpone the recognition of a loss to next year (Brent et al., 1990).

2.3. Characteristics of shorted stocks

There has been executed quite some research on the fundamentals of firms that are targeted with short positions. There exists evidence that indicates that ratios of fundamental value to market value often forecast future stock returns (Dechow et al., 2001). As mentioned before one of the main motives for investors to go short is to speculate on stock prices. Given the proven forecasting capability of the fundamental value to market value ratios, those ratios provide a decent starting point for further research to find out more about the target process of short sellers (Dechow et al., 2001).

Brent et al., (1990) find that stocks that have high betas and for which options or convertible securities are traded, tend to have higher levels of short interest. This finding is consistent with the previously mentioned popular investment strategy among hedge funds, the convertible arbitrage strategy. High beta stocks are attractive for arbitrageurs and investors that want to hedge their portfolios, since they tend to be highly correlated with the market (Brent et al., 1990).

Angel et al., (2003) find that shares that are actively traded are more often targeted by a short position than shares that are not actively traded. Furthermore, they find that short selling is mostly occurring among the very volatile stocks and happens much less frequently among stocks with relatively low price variability (Angel et al., 2003).

Like this study, Jones et al., (2013) made use of the new regulations introduced by the ESMA, concerning the public notification of large short position disclosures. In their research they use data from Spain, the United Kingdom, and France. They find the following: When the stock of company X is shorted and this short position is disclosed to the public, the probability that company X will be shorted again, within a month by a different investor, increases (Jones et al., 2013). They call this type of short selling “follow-on shorting”. The explanation that Jones et al., (2013) give for follow-on-shorting is that it just reflects the unrelated acquisition of correlated signals by a number of asset managers. Furthermore, they find that follow-on shorting is more likely to occur when the size of the initial short position is large, and when follow-on short sellers are located geographically in the neighbourhood of the initial short seller (Jones et al., 2013). However, they do not find any evidence

(10)

to prove that the total amount of short interest increases after disclosing a short position. Finally they conclude that short disclosures are not used as a signal for the market participants to short the same stock and to manipulatively drive the stock price down (Jones et al., 2013).

Diether et al., (2009) find that short-selling activity is higher for large stocks with high capitalization, growth stocks, stocks for which the level of institutional ownership is high, high price stocks, and stocks with actively traded put options. They argue that those observations are in line with the, according to them, most popular trading strategies among short sellers: trading on short-term overreaction, opportunistically providing risk-bearing services and acting as a voluntary liquidity provider (Diether et al., 2009). Besides those findings, they also find evidence which shows that short sellers increase their short-selling activity after periods of positive returns, on days with significant high buying pressure, and on days with high levels of asymmetric information (Diether et al., 2009). Engelberg et al., (2012) investigated how short sellers become informed traders. They do not find any evidence that might suggest that short sellers can anticipate news events better than the market. Nevertheless, they do find that the trade timing of short sellers differs from the overall market. Actually, the volume of short sales increases after news events. This finding suggest that short sellers are acting on publicly available information and do not significantly anticipate information before it becomes public (Engelberg et al, 2012). In short, the evidence of Engelberg et al., (2012) suggests that short sellers gain an information advantage, which they create by fast and superior processing of publicly available news sources.

The results of Karpoff and Lou (2010) imply that short sellers own skills to identify financial misrepresentation before the general investors do. Like Karpoff and Lou, Desai et al., (2006) examined whether the motive for short selling is related to questionable financial reporting. Their results show that short sellers disproportionately target firms that show extremely high accruals in their financial reports. Desai et al., (2006) explain that short sellers are trying to target overpriced companies. The finding of Dechow et al., (1996) supports the underlying argument of shorting stocks with high accruals since they show that accounted high accruals often turn out to be overvalued. Secondly, Desai et al., (2006) findings suggest that short sellers can identify suspect financial reporting prior to public announcement.

2.4. Short selling regulations

Nowadays, the regulation of short selling is a popular debate topic. Governmental regulators are still struggling to find the best way to deal with short selling. As mentioned before the ESMA introduced new regulations (November 2012) for short selling in Europe. Short selling has on one hand very

(11)

positive effects on the market and the market participants. Whereas, on the other hand, there are a couple of possible issues that go together with the trading concept.

What are the main drawbacks of short selling? Short selling can disturb markets and increase the trading volatility (Payne, 2012). When markets are healthy, this is not a major problem. But, when short selling is applied in an unstable market, it may push share prices of weak stocks even further down (Payne, 2012). Furthermore, investors may use short selling to manipulate the stock market or to act profitably on inside trader information (Payne, 2012). Market participants who want to

manipulate the stock market can use a short position to display their untrue worries about a particular stock. This action may lead to an increase of the amount of shorted stock in that particular company. Other investors are now trading on this fake rumour, which was introduced by the person who wanted to manipulate the market (Payne, 2012). Finally, short sellers face settlement risk (Payne, 2012). If the lending market for shares becomes illiquid, the investor who shorted stock might be unable to borrow the shares that he sold short. This results in situations where the short seller is not able to fulfil his settlement obligations (Payne, 2012).

What are the main benefits of short selling? Short selling can ease price revision in situations where the market is overvaluing securities (Payne, 2012). When new valuable information becomes available to a select number of market participants they can reveal this information to the rest of the market by shorting stocks, which are overvalued and take a long position in securities, which are undervalued (Payne, 2012). If short selling would not be allowed, only the prices of securities that are undervalued could be directly corrected (Payne, 2012). Besides, short selling can facilitate liquidity and can create new trading opportunities (Payne, 2012). Short sellers are providing liquidity to the market, and they will be rewarded for this. After the liquidity injection, prices will again return to their true value. (Payne, 2012).

The best way to regulate short selling still needs to be found. Karpoff and Lou (2010) find that shorting stocks contributes to more efficient price discovery. If the finding of Karpoff and Lou is correct, than limitations on short selling may hinder the communication of negative information into prices (Payne, 2012). The benefits that go together with the regulations have to be bigger or at least equal to the losses obtained by the less efficient price discovery. It will be very hard to measure the benefits and losses of new regulations. This is why Payne (2012) suggests to conduct more research into effects of specific regulation plans of for example the ESMA.

2.5. Short selling and corporate governance

As mentioned before, one of the main objectives of short sellers is to pick stocks of which they believe the share price will drop. If the share price drops, they will make a profit, but when it increases they will make a loss. This study focuses on the targeting process of short sellers. More specifically, this study is trying to analyse the relationship between corporate governance quality and

(12)

the probability of being a target of a short position. Since short sellers bet on future share price decreases, which are obviously connected to the firm valuation of the overall market, the existing literature about the relationship between corporate governance quality and firm value is reviewed. The influence of corporate governance quality on firm value has been examined in several studies. La Porta et al. (2001) find that poor shareholder protection is punished by the stock market, by giving those firms lower valuations. Brown and Caylor (2004) find that good-governed firms are more profitable, pay out more dividends to their shareholders, and are more valuable than poorly governed firms. While Bhagat and Bolton (2008) find that none of their corporate governance measures are correlated with future stock market performance.

To my knowledge Massa et al., (2013) were the first to explore a relationship between corporate governance and short selling. They examined what effect an increase in short selling potential has on internal corporate governance quality. Their research sample includes data between 2003 and 2009, they use all publicly listed companies for which data were available in Datastream. They find a strong positive relationship between the short selling potential (SSP) of a firm and the firms’ ISS index (governance index used). Investors require an increase in the internal corporate governance quality of a firm when the potential of being a short sale target rises. The ISS index consists of 64 governance attributes for U.S. firms and 55 attributes for firms that are not U.S. firms and comes from

RiskMetrics/Institutional Shareholder Services (ISS).The positive relationship between the short

selling potential of a firm and the firms’ ISS index tends to be stronger for firms that have investors who have a shorter investment horizon, for companies that rely more on equity financing, and for firms that are operating in less developed areas (Massa et al., 2013).

Massa et al., (2013) contribute to the existing literature on internal corporate governance. This paper will also contribute on the literature on corporate governance but in a slightly different way. Like Massa et al, this research will try to find a relationship between short selling and corporate

governance quality. Massa et al., (2013) researched the effect that an increase in short selling potential has on the internal corporate governance quality of a firm. Exactly the other way around, this study will look at the relationship of internal corporate governance quality on the potential of being a target of a short position. This research tries to confirm that the action of investors to establish better internal governance when a firm is threated by a potential short sale (Massa et al., 2013) is a scientifically substantiated action to undertake. Besides, this study tries to take away part of the opaqueness around the characteristics of shorted stocks and the targeting variables of short sellers.

(13)

3. Constructing hypotheses

Dechow et al., (2001) find that stocks picked by short-sellers have relative bad fundamental value to market value ratios. Since it is likely that there are a bunch of stocks with bad fundamental to market value ratios, the short-sellers have to choose between those bad performing stocks based on an addition reason. If a firm is poorly governed, processes of improving fundamental value to market value ratios will probably be less quick and efficient. The probability of a stock price increase for the short term will be less likely. This is attractive for short sellers, who are mainly betting on stock price decreases (Nagel, (2005)).

As argued before, La Porta et al. (2001) find that poor shareholder protection is punished with lower firm valuations. Brown and Caylor (2004) find that good-governed firms are more profitable, pay out more dividends to their shareholders, and are overall more valuable than poor-governed firms. Firm valuation and future share price are highly correlated, since the share price reflects the firm valuation made by the market. This is why I expect firms that are target of a short position to have lower governance scores than untargeted firms.

The first hypothesis that is tested is the following:

Hypothesis 1: Firms that are target of a large short position have a lower governance score than

untargeted firms.

In the second part of this paper I am interested in finding the relationship between the probability of being a short sale target and the overall governance quality of a firm. I try to find an answer the following question: “What is the influence of corporate governance quality on the probability of being a target of a short position?” As argued before, in poorly governed firms it is very likely that

processes of improving fundamental value to market value ratios will be less quick and efficient, and thus the probability of a stock price increase for the short term is less likely. By means of the above argumentation I have constructed the second hypothesis.

Hypothesis 2: Firms with lower corporate governance quality are more likely to be a target of a short

(14)

4. Methodology

To test the first hypothesis introduced in the previous paragraph, the target firms are matched to control firms. The initial database consists of all firms that were listed on Euronext Amsterdam during 2012. Since the matching firms should be comparable firms, the database is split into six different sections, based on industry. In each industry the target firms are matched to the firms that best match their total assets and Tobin’s Q values.

Why are the target firms matched to firms with comparable total assets and Tobin’s Q values? Big firms tend to have different optimal governance guidelines than small firms. Small firms may face resource constraints that enclose their choice to optimize their corporate policies (Yermack, 1996). To control for size, the logarithm of total assets is used. As argued before, short-sellers are mainly targeting firms of which they expect share price decreases (Nagel, 2005). To control for firm performance Tobin’s Q is used. The matching is executed by hand, through scrolling in the excel sheet.

After the matching has occurred, the corporate governance score difference between the two groups is tested using a t-test for matched pairs. The t-test is executed on the mean and on the median of the two sample groups. Next, the answer to the first hypothesis is formed. This answer is constructed by a comparison between the governance scores of the targets and their matching partners. In this comparison, not only the differences between governance scores, but also the differences in the subsections (board, audit, anti-take-over, and compensation & ownership) of the governance scores are compared and discussed.

To answer the second hypothesis the following probit regression is performed:

𝐏𝐫 𝐘 = 𝟏 𝐗_𝟏, 𝐗_𝟐,𝐗_𝟑,𝐗_𝟒,𝐗_𝟓, 𝐗_𝟔,𝐗_𝟕, 𝐃_{𝐢!𝟏^𝟓} = 𝚽(𝛃_𝟎+ 𝛃_𝟏𝐗_𝟏+ 𝛃_𝟐𝐗_𝟐+ 𝛃_𝟑𝐗_𝟑+ 𝛃_𝟒𝐗_𝟒+ 𝛃_𝟓𝐗_𝟓+ 𝛃_𝟔𝐗_𝟔+ 𝛄_𝐢𝐃_𝐢)

Where: Y: Target (y=1), Untargeted (Y=0)

Φ : is the cumulative normal distribution. X1: Log (Total Assets)

X2: Governance index X3: Tobin’s Q

X4: Log (Cash)

X5: Log (Trading Volume)

X6: High Beta (30% highest betas) (X6=1), Low Beta (X6=0)

(15)

The control variables used in this regression are total assets, Tobin’s Q, cash, the beta of the stock, trading volume of the stock, and several industry dummies. Why are those variables selected? The logarithm of total assets is selected to control for differences in firm size. As argued before, small firms may face resource constraints that enclose their choice to optimize their corporate policies (Yermack, 1996). I expect the logarithm of total assets to have a positive effect on the Target variable. My expectations are based on the findings of Diether et al., (2009) who find that short-selling activity is higher for large stocks with high capitalization.

The next variable added is the logarithm of cash. This variable is included since it may give us important information about recent firm performance (Masulic et al., 2007). When the amount of the cash variable is high this may indicate recent good firm performance. If a firm is currently performing very well, this good performance may last for some time. This is why I expect the logarithm of cash to have a negative effect on the Target variable. The next variable added to the regression is Tobin’s Q. Dechow et al., (2001) find that fundamental value to market value ratio’s regularly forecast future stock returns. Due to this, I expect Tobin’s Q to have a positive effect on the Target variable. The final two variables added to the regression are the beta of the stock and the logarithm of the trading volume. The beta variable is included since Brent et al. (1990) find that stocks with high betas for which options or convertible securities are traded, tend to have higher levels of short interest. Unfortunately no information concerning convertible securities or options could be found for the sample stocks. Based on previous findings (Brent et al., 1990) and (McDonald and Baron, 1973) I expect the beta variable to have a positive effect on the Target variable. The economic reasoning behind this expectation is that stocks with high betas are very attractive stocks to use for arbitrageurs or hedgers because they tend to be highly correlated with the market (McDonald and Baron, 1973). The logarithm of the trading volume is added to the regression because Angel et al., (2003) find that shares that are actively traded in the market are more often target of a short position than inactively traded stocks. Stocks that are actively traded offer more liquidity, which reduces the probability for a “short squeeze” (Angel et al., 2003). A short squeeze arises when short sellers, as they cover their positions, face rising share prices because there is a lack of supply and an excess of demand for the shares of the company concerned (Angel et al., 2003). Based on previous research of Angel et al., (2003) I expect the trading volume variable to have a positive effect on the Target variable.

(16)

5. Data and descriptive statistics

5.1. Data Collecting

The initial database consists of all Dutch firms that were listed on Euronext Amsterdam during 2012. Data on net short positions are collected from the Netherlands Authority for the Financial Markets (AFM) website. In the Netherlands, AFM is appointed to track the register of short sales. On their website you can download the short selling register in excel format. The register is divided into two different sections. One section regarding the current positions and the other section the past positions (archive). In the register the position holder, the name of the issuer, the ISIN number, the size of the net short position and the position date are reported. The firms included as target firms in this research are all companies that were reported to have a net short position between the first of November 2012 and the first of July 2013.

All information needed to construct the governance index is hand collected. The sources used to collect this information are annual reports and firm websites. The financial firm information is collected from DataStream. After the target firms are matched to comparable companies, only those comparable companies and the selected target firms are used in this research, all other remaining companies that were located in the initial database are dropped. All variables are collected over the year 2012. The reason for this is that the governance policies of companies do not change regulatory and often when they change, those changes are only reported in the annual reports. In September, when I was collecting the governance index the annual reports of 2013 were not yet available. The financial variables are collected as an average over the year 2012. My motivation for this choice is that in this research only Dutch companies are used, some of those companies are very small and do not provide the specific financial firm information that is needed for this research, on day-to-day or monthly basis.

5.2. Defining the variables

The governance index that is used in this research is a self-created measure following the approach of Aggarwal et al., (2007). This index is widely used among researchers since it is a very comprehensive governance measure consisting of more than forty attributes. The governance attributes are organized in four subcategories: board (twenty-five attributes), audit (three attributes), anti-takeover provisions (six attributes), compensation and ownership (ten attributes) (Aggarwal et al., 2007).

The board attributes are trying to give an overview of the overall functioning of the company’s board of directors. The twenty-five board attributes relate to independence, composition, size, transparency, and the way of conducting work (Aggarwal et al., 2007). The audit section reflects the independence of the audit committee and the ratification of the auditor’s report. The anti-takeover category overlays the following subjects: share class structures, role of shareholders, blank check preferences and the

(17)

existence of poison pills. In the latter category the executive compensation and ownership structure of the firm are reviewed. The attributes consist of subjects among executive compensation and

monitoring, and identifying compensation structures. (Aggarwal et al., 2007).

To compute the governance index, the governance dummy attributes are used. The attributes are all dummy variables. The attribute has a value of one if the answer to the attribute question is “yes”, when the answer to the attribute question is “no” the value assigned to the attribute is zero. All attributes are determined by using hand-collected information from annual reports and firm websites. The governance index is expressed as a percentage, if a firm satisfies all acquirements the governance index will be 100% (Aggarwal et al., 2007). (For more detailed information about the Governance index see table 1, appendix 1)

The industry dummy that is used in this research is collected from Datastream. The variable is called general industry classification and splits the Dutch market into six categories. These six categories are: industrial, utility, transportation, banking, insurance, and other financial company. Tobin’s Q is calculated by using the next formula (Equity Market Value+Liabilities Book Value)/(Equity Book Value+Liabilities Book Value). The variable used for the total assets values during 2012 is the total assets world scope data variable, this variable is also collected from Datastream. The variable that is used for the beta during 2012 is collected from Datastream and is expressed as a dummy variable in the research model, the top thirty percent highest beta companies get a value of one assigned, and the remaining companies receive a value of zero. The variable that is used to measure the trading volume is the turnover by volume variable and the variable that is used for measuring the firms’ cash amounts is the worldscope cash variable, both are collected from Datastream. (For more detailed information on data definitions and descriptions see appendix 2).

5.3. Summary statistics

In this paragraph the summary statistics are displayed. First the total sample summary statistics are discussed. Second, a discussion is provided, in which the sample is divided into a control and a target group. The summary statistics for those two groups are displayed separately. Finally, the pearson correlation matrix is revealed and discussed. Table 3, shows the summary statistics for the total sample, table 4 show the summary statistics for the target group, and table 5 for the control group. The mean governance index for the total sample is 69%, whereas when the control and target group are separated, the mean governance index is 66% for the target group and 74% for the control group. This finding is in line with my previously mentioned expectation that target have, on average, corporate governance of a lower quality than untargeted firms.

As can be found in table 4 and 5, the mean number of the various types of industries, the total assets and Tobin’s q are approximately the same between the target group and the untargeted group. This is

(18)

the logical consequence of the matching method, since the matching was based on industry, total assets and Tobin’s Q. The mean cash levels differ between the two groups. The mean Log(cash) level of the target group is 5.16, for the control group this is 4.75. In contrast to the comparison between the average governance indexes of the two sample groups, the average cash levels differ from my

expectations. I expected target firms to have lower average cash levels than untargeted firms. When looking at the average beta of the two groups, the target group has an average score of 0.33 and the control group has an average beta of 0.24. Since the beta in this model is a dummy variable (the top thirty percent highest beta companies received value 1, the remaining companies received value 0) the only conclusion that I can make based on this average is that, on average, more target firms are included in the top thirty percent highest beta companies. This outcome is in line with my

expectations. Besides this finding is in line with the finding of Brent et al., (1990), stocks that have high betas tend to have higher levels of short interest. Finally, one can observe a difference in the average trading volumes between the target and untargeted group. The average trading volume for target firms is 5.14 whereas the average for the untargeted group is 4.26. I infer from this that, on average, the stocks of target firms are traded more often than the stocks of untargeted firms. This outcome is in line with my expectations. This outcome is also in line with the findings of Angel et al., (2003), shares that are actively traded in the market are more often target of a short position, since they offer more liquidity.

Table 6 displays the correlation coefficients between the variables, those results show the directions and significance levels of the various correlations. First of all, I only discuss the correlations that might be interesting for this study. According to table 6, there exists a positive significant correlation between the Log(Total Assets) variable and the Log(Cash), Log(Trading volume) and Industry Dummy 5 variables. When a firm increases in size, on average, the total cash holdings of the firm increases. Thus, cash holdings/amounts are correlated with firm size. Besides firm size and trading volume tend to be correlated. This correlation might be explained by the following: bigger firms might have more outstanding shares or at least have a higher amount of total outstanding capital than smaller firms and thus might have automatically a higher trading volume than smaller firms. The result that firms who are active in industry 5, which is the dummy variable for the insurance industry, are showing a positive significant correlation with the Log(Total Assets) variable is an interesting finding. On average, and in this sample, firms in the insurance industry are bigger than firms that are

active in the other industries. Furthermore, the Log(Cash) variable is showing a positive significant

correlation with the Log(Trading volume) variable and the industry dummy 5 (insurance industry) variable. Firms with more cash have a higher trading volume, probably because those firms have a bigger size than firms with lower cash holdings, and as argued before bigger firms might have more outstanding shares or at least have a higher amount of total outstanding capital than smaller firms and thus might have a higher trading volume than smaller firms. As mentioned before the

(19)

Log(TotalAssets) variable is also correlated with the insurance industry dummy (dummy 5), so again the correlation between the Log(Cash) and industry dummy 5 might be explained by the on average bigger firm size of industry 5 firms.

(20)

6. Empirical results

6.1 Governance index

In table 2 the target firms and their matching partners are displayed. The matching is based on industry, total assets, and Tobin’s Q. The number of target firms (N=27) is larger than the number of matching partners (N=21). This is due to the fact that several target firms are matched to the same control firm. The following matching partners have been used twice in this research: LBI

International, Kardan, Prologis Euro Prop, Reed Elsevier, Hal trust, and Hes Beheer. I was forced to use these matching partners again, because there are only a few companies listed on Euronext Amsterdam. Furthermore the matching is based on industry, size, and profitability. So when there only exist three listed firms for a particular industry, of which two were a target of a short position, both targets have to be matched to the same matching firm.

When I mention meeting the attribute, I mean that the attribute question can be answered with “yes” and thus a value of one is assigned to that firm for that particular governance attribute. Table 8 displays the percentages of Target firms and matching partners that received a “yes” (value of one) on the different governance attribute questions. The attributes that are answered with “yes” by all firms in the sample are: performance of the board is reviewed regulatory; does not ignore shareholder proposal; auditors are ratified at most recent annual meeting; company is not authorized to issue blank check preferred; no interlocks among compensation committee members; and all stock-incentive plans adopted with shareholder approval. The target firms score slightly better on the following attributes: board size is greater than 5 but less than 16; chairman and CEO are separated or there is a lead director governance committee; audit committee comprised solely of independent directors; shareholders may call special meetings and officers and directors own at least 1% but not more than 30% of total share capital.

On the thirty-one remaining attributes, the percentage matching firms that meet the attribute is higher than the percentage target firms that meet the attribute. The most outstanding differences lie in the following attributes: shareholders vote on directors selected to fill vacancies; policy exist on outside directorships (5 or fewer); shareholder approval is required to increase/decrease board size; board has the express authority to hire its own advisors; board approved succession plan in place for CEO; directors are required to submit resignation upon a change in job and single class is common. When dividing the categories into board audit anti-takeover and compensation & ownership, the target firms are on average only performing better in the audit category. In the remaining categories, the matching partners are on average outperforming the target firms. This finding is in line with my previously introduced hypothesis, which stated: “Firms that are target of a large short position have a lower governance score than the firms that are not picked.”

(21)

In table 7 the governance scores between the target firms and their matching partners are compared. Post NL is the only target firm that is outperforming its matching partner. Accell Group, NSI, and SBM score exactly the same as their matching firm. The biggest difference between governance index scores is found in the comparison of Akzo Nobel and Philips.

The t-test for matched pairs is provided in table 9. The mean governance score of the target firms is 0.66 (standard deviation 0.06) whereas the mean score for the matching partners is 0.74 (standard deviation 0.06). The student t-statistic for matched pairs is -5.45. This score is significant at α 0.05%. Furthermore, the difference in median is tested; both the Pearson chi square and the continuity corrected Pearson chi square statistics indicate that the medians of the two samples differ. Those results provide evidence that the governance scores of the target firms in this sample significantly differ from the scores of the matching partners (at α=0.0005).

Table 10 displays a comparison between the outcomes of this study and the outcomes of Aggarwal et al., (2007). Row one displays the percentage of target firms that meet the different governance attributes, the second row displays the same for the matching firms, the third row displays the

percentage of foreign firms meeting the different governance attribute questions, these “foreign firm” data are collected by Aggarwal et al., (2007). With foreign firms Aggarwal et al., (2007) refer to firms in countries outside the U.S. It is interesting to compare my results to the results of Aggarwal et al., (2007) since now I can compare the governance quality of Dutch targeted and untargeted firms to the overall corporate governance quality of “foreign firms”.

First, on average the Dutch target and untargeted firms have both a better governance score than the foreign firms. It is notable that the Dutch firms are excessively outperforming the foreign firms on eight out of the twenty-five board category attributes. The most striking result in the board category is that on average Dutch firms tend to be controlled by a majority of independent outside directors. The majority of the Dutch nominating, compensation and audit committees are ruled by more than fifty percent outside directors whereas this is not the case for the majority of the foreign firms.

Another striking result is that on average 93.2% of the foreign firm has one share class whereas 42.9% of the Dutch matching firms and only 7.4% of the Dutch target firms have one share class. When collecting the governance index I noticed that in many of the Dutch firms that didn’t meet the requirement of the existence of only one share class, there existed a preferred share class. Obviously, it is not optimal to compare the average governance scores of Dutch and “foreign firms”, since we are comparing firms that differ in industry and size. However, these results do indicate that the global corporate governance quality and composition vary. A possible explanation for this observation is the global difference in government policy on corporate governance.

(22)

The most important finding of this part of the research is that the governance index scores of the target firms significantly differ from the scores of the matching partners. Looking more carefully into the data it can be found that, on average, target firms are only scoring better than matching partners on the audit category part of the governance index. The target firms score, on average, worse than the matching partners on the other three categories (board, anti-takeover, and compensation &

ownership). These findings are in line with my hypothesis and can be explained as follows. If a firm is poorly governed, processes of improving fundamental value to market value ratios will be less quick and efficient. The probability of a stock price increase for the short term will be less likely. This is attractive for short sellers, who are mainly betting on stock price decreases (Nagel, 2005). Those findings might also scientifically approve the action of investors that in case of the threat of short selling establish better internal governance (Massa et al, 2013). Since, indeed corporate governance quality might be an explanatory variable in the targeting process of short sellers. In short, the findings show that firms that are target of a short position have, on average, a lower governance quality than comparable firms, which are not targeted. This might suggest that firms with lower governance quality have a higher probability of being a target of a large short position. In the next section this is tested.

6.2 Probit regression

In this section the previously introduced probit regression is performed and discussed. The dependent variable of this binary model shows whether or not a firm is target of a large short position (target=1, untargeted=0). The variable of interest is the Govindex (the previously explained and computed governance index), The control variables are: Log(TotalAssets), Toibin’s Q, Log(Cash), Beta, Log(Trading volume) and Industry dummy 1 till 5. Table 11 reports the binary model output. The variable Log(Trading Volume) is the only variable that is significantly different from zero, at α=0.05. When increasing alpha to 0.10, Tobin’s Q, Log(Cash), Govindex, Beta and Industry dummy 1 are also significantly different from zero.

The results can be interpreted as follows: an increase in Tobin’s Q, Cash, Govindex or Beta makes the outcome of y=1 less likely (at α=0.10), an increase in Industry dummy 1 makes the outcome of y=1 more likely (at α=0.10). An increase in trading volume makes the outcome of y=1 more likely (at α=0.05). Concerning the other control variables included in the model, nothing can be concluded since their P-values are too high to safely interpret the results. The pseudo R-squared of the model is 0.81, which is relatively high.

The results of the probit model output cannot be interpreted as straightforward as the regular linear regression outputs (Ai and Norton, 2003). Since the magnitude of the probit model coefficients cannot be interpreted, the average marginal effects are calculated, the results are shown in table 12. The variables Tobin’s Q, Log(Cash), Govindex, Beta, Log(Trading volume), Industry dummy 1 and

(23)

Industry dummy 4 are significantly different from zero, at α=0.01. Therefore, there can only be assigned value to the coefficients of those variables. The results can be explained as follows: for each additional 0.01 increase in Tobin’s Q, firms are 0.012% less likely to be a target of a short position. For every one percent increase in cash, firms are 0.0046% less likely to be a target of a short position. For each additional 1% increase in the governance index, firm are 5.63% less likely to be a target of a short position. A firm that is ranked in the top thirty percent beta category is 51% less likely to be a target of a short position than a firm that is not ranked. For every one percent increase in trading volume, firms are 0.0082% more likely to be a target of a short position. A firm that is active in industry 1 (industrial) is 61% more likely to be a target of a short position than a firm that is not active in this industry. A firm that is active in industry 4 (banking) is 46% more likely to be a target of a short position than a firm that is not active in this industry.

When comparing my findings with the findings of Brent et al., (1990) it is remarkable that my research shows that high beta stocks are less likely to be a target of a short position than low beta stocks, whereas Brent et al., (1990) find exactly the opposite. My findings do support the findings of Angel et al., (2003), who find that shares that are actively traded in the market are more often target of a short position. The most important finding of my research is that firms with low corporate

governance quality are more likely to be a target of a short position. My findings actually confirm that the action of investors to establish better internal governance when a firm is threated with a potential short sale (Massa et al., 2013) is a scientifically supported action to undertake.

The reported results confirm my second hypothesis: “Firms with lower corporate governance quality are more likely to be a target of a short position.” The governance quality of a firm does clearly matter in the picking process of a short seller. Short sellers prefer to pick stocks of firms that are governed less sound. To check the data on multicollinearity and correctly specified data and model, a goodness of fit test, a multicollinearity test and a linktest are presented in the next section.

(24)

7. Additional tests

In this chapter, checking the model and the dataset play a central role. One of the most important shortcomings of the probit model is the lack of diagnostic measures relative to the OLS regression

(Hagle and Mitchell, 1992). The very familiar coefficient of multiple determinations, the R!_{, is absent}

in the probit model regression output (Hagle and Mitchell, 1992). Since the R!_{of the model is}

missing, we should find another way to test the actual fit and explanatory power of the model. First, the goodness of fit test is executed on the data. It is important to check if the model actually ‘fits’ the data. By fitting the actual data is meant: what percentage of the firms is correctly specified as being a target/untargeted firm? To measure the fit of the binary model to the actual data the goodness of fit test is performed. Table 13 represents the goodness of fit measure outcomes. The model clarifies 22 out of the 24 target firms correctly, and 16 out of the 17 untargeted firms. Consequently, the model clarifies 92.68% of the data correctly, which makes this model a decent fit.

The second additional test that has to be executed is a test for multicollinearity. Multicollinearity exists when two or more independent variables in the model are highly correlated. When there exists collinearity between two or more variables, the model can estimate parameters with an incorrect sign and of implausible magnitude (O’Brien, 2007). This makes the model and the model output useless. To test for multicollinearity, the variance inflation factors (VIF) method is used. The VIF provides an

indication of how much the approximated variance of the I!" regression coefficient has risen above

the expected value that it would reach when R!_{would be equal to zero (O’Brien, 2007). Table 14}

shows the results of the multicollinearity test. As can be obtained from table 14, the mean VIF score is 2.22 and none of the individual VIF scores of the variables is greater or equal to 10, which means that there does not exist multicollinearity in this model.

Finally, the linktest is carried out. The linktest tests the model on overall specification. Table 15 displays the results of the linktest. Through this test I can conclude that I cannot detect a specification error, since the variable _hat is significant but the variable _hatsq is not. The linktest is not significant, so no model misspecification can be detected.

(25)

8. Discussion

In this chapter the limitations of this research are discussed. First I want to mention the most

important limitation of this study, namely the size of the used sample. Since the short-selling register of AFM is only active since November 2012, and only large short-sale positions are recorded, there are only twenty-seven target firms included in this research. Given that some companies had to be matched to the same matching partner (due to the limited number of comparable listed Dutch companies per industry) the matching was not optimal and the final database consisted of only forty-eight firms. Since the final database is very small, it might be doubtful to apply the conclusions of this research to all Dutch listed firms.

Second, the results of this research can only be applied to Dutch listed companies. Almost every country in Europe has its own corporate governance code. In the Netherlands for example, companies can choose between one and two tier boards. The same holds for the differences in short-selling regulations. If one is interested to establish a link between corporate governance and short selling to countries outside the Netherlands, extensive additional research has to be executed.

Finally, the last limitation of this study is the endogeneity problem. Endogeneity issues can be split into three main causes: omitted variables, simultaneity bias and measurement error (Robberts and Whited, 2011). An omitted variable refers to a variable that should actually be included in the model, but is not included. This variable should be added since it has explanatory power on the dependent variable. Simultaneity bias refers to the problem where one or more dependent variables and the independent variable are determined in equilibrium so that it can likely be reasoned that the dependent variable causes the independent variable or that the independent variable causes the dependent

variable (Robberts and Whited, 2011). Measurement error refers to the problem that occurs when the researcher interprets or collects information incorrectly. Measurement error can also occur when there exist conceptual inequalities between proxies and their unseen counterparties (Robberts and Whited, 2011).

As argued before there is only little known about the variables that influence the targeting process of short sellers, I tried to add all explanatory control variables, but there exist a reasonable chance that not all explanatory variables are added. There might exist a simultaneity bias between corporate governance quality and short sale potential, since Massa et al., (2013) find that the threat of short selling activates investors to establish better internal governance. There might also exist a

measurement error. Before collecting the governance index of the firms, I knew which firms were target of a large short position, and which firms were not. Another point that should be mentioned about the governance index collecting process is that firms that reported their governance information in unusual chapters or in unclear language might also have lower governance index scores, since some

(26)

governance questions could not be correctly answered. In my opinion there was no big difference between reporting techniques of targeted and untargeted firms.

(27)

9. Conclusion

In this paper, I have examined the differences in corporate governance quality between shorted Dutch firms and comparable Dutch firms that were no target of a short position. Besides, I examined the following question: “Has the corporate governance quality of a firm any impact on the probability of being a target of a large short position?” My main findings, suggest two conclusions. First, the governance index scores of the untargeted firms are significantly better than the governance index scores of the target firms. The target firms score, on average, only better on the audit category. On all other categories (board, anti-takeover, and compensation & ownership) the matching partners have a better average score.

Second, the probability of being a target of a short sale increases if the governance index score of the firm decreases. The governance quality of a firm does clearly matter in the picking process of a short seller. Short sellers prefer to pick stocks of firms that have a lower overall governance quality. I cannot compare my main findings to previous research since a comparable research has, to my knowledge, never been executed before. Nevertheless, I can compare the influence of some of the control variables on the probability of being a target of a short sale with previous studies.

When comparing my findings with the findings of Brent et al., (1990) it is remarkable that my results indicate that high beta stocks are less likely to be a target of a short position than low beta stocks, whereas Brent et al., (1990) find exactly the opposite. My findings do support the findings of Angel et al., (2003), who find that shares that are actively traded in the market are more often target of a short position. The most important finding of my research is that firms with low corporate governance quality are more likely to be a target of a short position. My findings actually confirm that the action of investors to establish better internal governance when a firm is threated with a potential short sale (Massa et al., 2013) is a scientifically supported action to undertake.

As reported in the discussion part of this thesis, the main limitations of this paper are: the small sample size, the application of the findings is restricted to the Dutch market and the endogeneity issue. Taking into account these limitations I am still convinced that my findings add an important value to the literature on the short selling target process, namely that short sellers prefer to target stocks with lower governance quality.

For future research I recommend to generate a database of a bigger size, to use more control variables, and to control better for endogeniety issues. It would also be interesting to look at countries other than The Netherlands. Furthermore it would be interesting to investigate which variables within the

governance index are the most influential in explaining the differences in the probability of being a target of a large short sale. Besides, the extensive governance index, that I hand collected, can be used to answer the following question: “What is the influence of a large short position on the corporate governance quality of a firm?”

(28)

10. Literature

Ai, C., Norton, C. (2003), Interaction terms in logit and probit models, Economics Letters 80, 123-129.

Aggarwal, R., Erel, I., Stulz, R and Williamson, R. (2007), Do U.S. firms have the best corporate governance? A cross-country examination of the relation between corporate governance and shareholder wealth, NBER Working paper Series.

Aggarawal, R., Erel, I., Ferreira, M. and Matos, P. (2011), Does governance travel around the world? Evidence from institutional investors, Journal of financial economics 100, 154-181.

Angel, J., Christophe, S. and Ferri, M. (2003). A close look at short selling on Nasdaq, Financial

Analyst Journal 59(6), 66-74.

Asquith, P., Parag, A. and Ritter, J. (2005). Short interest, institutional ownership, and stock returns,

Journal of Financial Economics 78, 243-276.

Beber, A., Pagano, M. (2013). Short-selling bans around the world: evidence from the 2007-09 crisis,

The Journal of Finance 68 (1), 343-381.

Brent, A., Morse, D. and Stice, E. (1990). Short interest: explanations and tests, Journal of financial

and quantitative analysis 25(2), 273-289.

Bhagat, S. and Bolton, B. (2008). Corporate governance and firm performance, Journal of Corporate

Finance 14, 257-273.

Brown, L and Caylor, M. (2004). Corporate governance and firm performance, SSRN paper. D’Avolio, G. (2002). The market for borrowing stock, Journal of financial economics 66, 271-306. Dechow, P., Hutton, A., Meulbroek, L. and Sloan, R. (2001), Short-sellers, Fundamental analysis, and stock returns, Journal of Financial Economics 61, 77-106.

Dechow, P., Sloan, R. and Sweeney, A. (1996). Causes and consequences of earnings manipulations: An analysis of firms subject to enforcement actions by the SEC, Contemporary Accounting Research 13, 1-36.

Desai, H., Krishnamurthy, S. and Venkataraman, K. (2006), Do short sellers target firms with poor earnings quality? Evidence from earnings restatements. Review of accounting studies 11(1), 71-90. Diether, K., Lee,K. and Werner, I. (2009). Short-sale strategies and return predictability, The review

The relationship between corporate governance quality and the probability of being a target of a large short position : an analysis of the Dutch stock market.