Short selling by institutional investors as an instrument for
shareholder activism: a study in Dutch listed firms
Msc Business Economics Student: Tony Arkes Studentnumber: 6058329 Field: Corporate Goverannce Supervisor: Dr. Z. Sautner
University of Amsterdam, Amsterdam Business School January 2014
Abstract
This paper examines significant short positions in Dutch listed firms, disclosed by institutional investors. The market reacts negatively to short selling activity, indicated by a significant price decline in the short-term. Cumulative abnormal returns range from -2.12% for the 11-days event window down to -7.69% for the 41-days event window. The price effect is permanent, which suggest that short selling registers are not sensitive to price manipulation. Analysing targets’ firm characteristics in the pre-target period reveal that the negative market reaction is largely driven by the characteristics Cash, ROA, Ownership, and Board Size. Furthermore, findings show that these variables have a significant impact on the probability of being targeted by short sellers. This paper gives more insight in short selling and the market reaction underlines the potential of short selling to become an effective instrument for shareholder activism.
2
Table of contents
Introduction ... 3
I. Institutional background and literature review ... 5
A. Types of shareholder activism ... 6
B. Motives for shareholder activism ... 7
C. Effectiveness of shareholder activism ... 7
D. Short selling as an instrument for shareholder activism ...10
E. European legislation regarding short selling ...11
II. Hypotheses ...13
III. Research design ...15
A. Data sample ...15
B. Two sample examples of recent short positions ...16
B.1. Imtech ...16
B.2. SNS Reaal ...16
C. Descriptive statistics ...17
IV. Research methodology ...19
A. Market reaction ...19
B. Firm characteristics ...21
V. Research results ...23
A. Market reaction ...23
B. Underlying determinants of the market reaction ...24
C. Difference target firms and peer groups ...26
C1. Firm characteristics targets and matched firms ...26
C2. Firm Characteristics targets and other Dutch listed firms ...27
D. Probability of being targeted ...28
D1. Matched firms ...28
D2. Dutch listed firms...29
E. Summary results ...30
VI. Conclusion and discussion ...31
References ...32
3
It has been widely investigated to what extend shareholder activism result in improvements in target firms' performance, capital structures, operations, and governance mechanisms. It is concluded that the effectiveness of shareholder activism depends on the type of institutional investors and the form of activism (Karpoff, 1997). The most popular forms include the ‘Wall Street Walk’, also known as activism through ‘exit’, and shareholder activism through ‘voice’ (McChahery, Sautner and Starks, 2010). Both forms have a non-hostile character, but proved to have a disciplinary effect on managers [Admati and Pfleiderer, 2005; Karpoff, 1997). Hedge funds are most successful in achieving their goals associated with shareholder activism. Intervention by hedge funds improves overall performance of target firms, which results in a significant price run-up in the short-term. Shareholder activism arises from institutional investor who has taken a long position in the target firm, with the expectation that the stock price will rise in value.
This research investigates to what extent short selling could become an effective instrument for shareholder activism. Researchers already suggested short selling as a form of shareholder activism, however there is little empirical evidence that supports these suggestions. An effective form of shareholder activism has disciplinary effects on the target’s management. Before such time, it requires a major shift in belief systems and culture of most boards and managements teams (Jensen, 2008).
In response to the financial crisis the European Securities and Markets Autority introduced changes in European legislation regarding short selling. European financial regulators are required to disclose significant short selling positions. Short selling is highly risky and has a high potential for manipulating stock prices (Jones et al. 2012). Since November 2012 the Dutch Authority of Financial Markets (AFM) publishes a short selling register on its public website. Institutional investors are obligated to disclose significant short positions that exceed a threshold of 0.5% in issued capital of Dutch listed firms. The main purpose of such a uniform European policy is improving market efficiency and increasing financial stability. Another important goal of short selling registers is preventing price manipulation, like the Herbalife casus. In 2013 the stock price of Herbalife has become a so-called battleground; two major US hedge funds revealed different perceptions about the future of the firm. Bill Ackman took a large short position of 10% in the firm, worth 1 billion dollar, and believed that the firm would suffer from a pyramid scheme. After the accusation the stock price dropped by 40%. Daniel Loeb, who believed that Ackman tried to manipulate the stock price, decided to take a large long position in Herbalife. After the announcement of the long position the stock price increased rapidly. (The Economist, 2013).
The research question is: To what extend does short selling have the potential to become an effective instrument for shareholder activism? This empirical research formulates an answer to this question by focuses on two aspects: the market reaction on disclosed short positions and the firm characteristics of target firms. A significant decrease in the stock price indicates that short selling could have a disciplinary impact on target’s management. However, it requires a permanent price
4
decrease to exclude the existence of price manipulation. Examining firm characteristics of target firms gives more insight in why firms get targeted. Short sellers are characterized as investors who position themselves in overvalued firms. It is unclear which other covariates are important factors in explaining short selling activity.
The AFM provides data of disclosed short positions, which are used in the first part of this empirical research. Datastream provides daily stock price and market returns, which are used to perform an event study to measure the market reaction. The second part of this empirical research compares the firm characteristics between target firms and non-target firms. Panel data for firm characteristics are both retrieved from Datastream and hand-collected. Probit regression models determine the marginal probability of firm characteristics on the probability of being targeted by institutional investors.
This paper is structured as follows. The next section serves as a literature review. Background information, types, motives, and the effectiveness of shareholder activism are discussed. Furthermore, this section provides information about changes in European regulation regarding short selling. Section II describes several hypotheses that are tested in this paper. Section III and IV cover the research design. The data sample and descriptive statistics are described and discussed in section III. Section IV describes the research methodology. The research results are analysed in section V. Finally, a conclusion and an answer to the central research question are formulated in section VI.
5
I. Institutional background and literature review
This section serves as a literature review and starts with background information about institutional investors that engage in short selling activism. Subsequently, the shareholder activism forms and motives are described. Third, short selling in relation to shareholder activism is discussed. Finally, the recent European legislation regarding short selling is described and discussed.
Institutional investors that are not satisfied with aspects of the firm’s management, operations or governance structures can try to bring changes by engaging in shareholder activism, which can be done without a change of control. The SEC describes in Rule 13-F that institutional investors are described as entities such as bank trusts, insurance companies, hedge funds, mutual funds, and pension funds that invest on the behalf of others and manage at least $100 million in equity (Bushee, 1998).
Shareholder activism surfaced in the 1980s, when large block holders intervened in highly diversified firms that reported poor performance. These activist block holders included insurance companies, pension funds, money managers, banks, and strategic investors such as conglomerates. Holding a significant share in the target allowed activists to demand corporate changes that resulted in assets divestures and a decreasing budget for mergers and acquisitions. (Brav et al., 2008)
Proxy proposals were introduced in the late 1980s and are recognized as the first formal mechanism for shareholder activism. The first proxy proposals focused mainly on corporate
governance issues (Gillan and Starks, 1998). These first proxy proposals proved to be effective and resulted on average in an increasing stock price and improved performance in the post-submission period. (Bethel, Liebeskind, and Opler, 1998).
During the 1990s the role of shareholder activism became more important as institutional ownership increased. The proportion of institutional investors investing in equities increased to almost 50% in 1994, compared to 24.2% in 1980 (Sias and Starks, 1998). The increase can be explained by the significant growth of pension fund assets, which started to mimic stock index returns (Karpoff, 1997). Subsequently, a large proportion of institutional investors that were passively managing their portfolio shifted towards a more active approach. Minow and Monks (1991) explain that shareholder activism also served as an alternative for selling underperforming stocks. Pension funds increased equity ownership to diversify their portfolios, selling underperforming stocks could increase the risk of an undiversified portfolio, which could have a negative impact on the portfolio return. In addition, during the 1990s the number of institutional investors actively seeking for underperforming target firms increased (Gillan and Starks, 1998).
Smith (1996) explains the strong increase in shareholder activism during the 1990s by a less active the takeover market. Shareholder activism became an alternative for a less disciplining market of corporate control. The threat of a takeover has a disciplining effect on managers to maximize
6
shareholder value. The takeover market was significantly less active from 1989-1993 in comparison to 1984-1988. Mikkelson and Partch, (1997) find a significant lower level of CEO turnover during the period 1989-1993, which might be explained a lower threat-level of being acquired. The decreasing turnover-level can also be explained by the strong increase in takeover defences during the late 1980s (Bethel et al. 1998).
A. Types of shareholder activism
There are several methods for institutional investors to express their dissatisfaction with the management and firm performance. The existing literature makes a distinction between hostile and non-hostile forms of shareholder activism. Non-hostile events are the most common, since they require less time and fewer resources. In addition to this, regulation restricts pension and mutual funds in the level of resources that can be spend on shareholder activism. Hedge funds have more freedom in capital allocation and show more aggressive behaviour compared to other institutional investors. However, hedge funds show aggressive behaviour in less than 30% of all events (Brav et al. 2008).
The most popular and easiest form of activism is selling shares of underperforming firms, also known as the ‘Wall Street Walk’ (McChahery, Sautner and Starks, 2010). This form is without juridical costs and is not time consuming. However, selling large blocks of shares can drive the share price down, which would lead to further losses (Gillan and Starks, 2000). On the other hand, the threat of selling shares has a disciplining effect on managers who receive bonuses that are related to the stock price. Therefore, the ‘Wall Street Walk’ can be seen as a governance mechanism itself (Admati and Pfleiderer, 2005).
Other popular methods among institutional investors include the submission of proxy proposals, voting against the company at the annual meeting, conducting discussions with executives and demanding a change in corporate control (McChahery, Sautner and Starks, 2010).
Hedge funds’ most popular method, which occurs in almost 50% of the cases is communicating with the board or management on a regular basis with the intention to increase shareholder value. Hostile forms include threatening a proxy fight, publicly criticizing the target and demanding changes, suing the target and launching proxy contests with the goal to replace the current board. In the most rigorous case, which occurs in 4.2% of the events, the hedge funds tries to take control of the target, for instance through a hostile takeover bid.
7 B. Motives for shareholder activism
The motives for institutional investors to engage in shareholder activism are divergent. As described earlier, selling significant equity holdings could drive the share price down. Furthermore, shareholder activism could be a solution when portfolio managers are restricted to a yearly turnover-ratio. Many public pension funds manage indexed portfolios, which restrict the fund of selling all shares of underperforming firms in their portfolio.
Shareholder activism can solve agency problems between managers and shareholders, which have evolved from managers who have the incentive to act out of self-interest at the expense of the shareholder (Agrawl and Knoeber, 1996). For instance, management decisions focusing on the short-term firm performance could decrease long-short-term shareholder value. A higher level of institutional ownership makes it more attractive and profitable to monitor managers of firms in the current portfolio. An increase in shareholder’s value probably outweighs monitoring costs (Gillan and Starks, 2000).
Brav et al. (2008) state that the objectives of hedge funds can be divided into five major categories, each containing multiple subcategories. The first category includes events in which the institutional investor believes that the firm is undervalued and/or that the fund can help managers increasing shareholder value. This category represents almost 50% of the cases. This category has a non-hostile character, since this form only involves communication with the management.
In case of the second category, the institutional investor demands changes in the capital structure of the target. Hedge funds demand changes regarding reduction of excess cash, an increase in leverage or higher payout ratios to shareholders using stock repurchases or dividend payouts. The hedge fund could also suggest an equity issuance and reduced seasoned equity offerings. In the third case, firms are targeted because the institutional investor wants to change the current business strategy. The institutional investor makes proposals regarding: operational efficiency, tax efficiency, division spin offs, mergers or acquisitions, and growth strategies. The last category includes activism with the goal to bring changes in corporate governance. Subgroups include firing the CEO or chairman, challenging board independence, demanding a higher level of information disclosure and questioning potential fraud, challenging the level or pay-for-performance sensitivity of executive compensation and decreasing the number and quality of takeover defences.
C. Effectiveness of shareholder activism
The effectiveness of shareholder activism is widely investigated and researchers seem to disagree to what extent shareholder activism is successful and results in improvements in target firms’ values,
8
earnings, operations or governance structure. The effectiveness of shareholder activism differs between the types of institutional investors and the activism forms.
Karphoff (2001) analyses the mixed results published by the existing literature and concludes that the consensus between researchers is larger than it seems at first sight. Differences often arise when the ‘success’ term is defined. Increasing shareholder value and increasing firm performance in the post target period are common indicators of success. Another criterion focus on the number of supporting votes on proxy proposals submitted by the institutional investor. It can be concluded that caution is required when comparing empirical results. Karphoff (2001) also emphasizes that success differs between different types of institutional investors and forms of activism. When comparing results, a clear description of the sample is required.
Gillan and Starks (2000) examined the impact of more than 2000 shareholder proposals submitted by both institutional and individual investors. The stock market reaction and shareholder voting depends on the issues addressed by the proposal and the identity of the proposal sponsor. Institutional investors like pension funds and mutual funds receive more votes than individuals, but have a small negative impact on stock prices.
Del Guercio and Hawkins (1998) investigated the impact of shareholder proposals submitted by the largest pension funds between 1987 and 1993. Compared to other institutional investors, pension funds have different investment strategies, which result in different activism objects, tactics, and impact on targets. The paper concludes that submitted shareholder proposals increase corporate governance activity and corporate changes such as asset divestures and business restructurings. However, the proposals have no significant impact on the stock price and accounting measures of performance in the three year post-targeting period.
Karphoff (2001) summarizes the findings from multiple researches and concludes that shareholder activism by pension and mutual funds often result in changing corporate governance structures. However, in the short-run the market doesn’t respond to pension and mutual funds participating in shareholder activism. Therefore, shareholders of target firms don’t benefit significantly in the short-run (Karpoff, 2001).
While multiple researchers question the effectiveness of pension and mutual funds, there is consensus on the effectiveness of hedge funds engaging in shareholder activism. Hedge funds are more successful in achieving their activism goals when compared to institutional investors. Hedge funds are less regulated compared to pension and mutual funds which is reflected in the different organizational structure. While pension funds are limited to a certain risk level, tax laws disable mutual funds taking a concentrated share in a particular company. However, hedge fund managers manage large unregulated funds and are more incentivized because they often participate in the fund with private capital. In addition, regulation allows hedge funds to use large amounts of debt and hold significant stakes in firms. Furthermore, derivatives enable hedge funds to minimize risk on the downside (Brav et al. 2008). As a result, hedge funds are able to influence corporate boards and
9
managements of target firms more effectively than pension- and mutual funds. Therefore, it can be concluded that hedge funds are better positioned to act as informed monitors than other institutional investors.
Brav et al. (2008) investigated the impact of multiple forms of shareholder activism by hedge funds. In 67% of the cases hedge funds succeed fully or partially in achieving their goals. Markets respond positively on the announcement of shareholder activism by hedge funds. Targets receive, on average, a positive abnormal return of 7% in the short-term. The presence of hedge funds and their opportunities to intervene management of undervalued firms results in disciplinary pressure on the management of public firms.
Kein and Zur (2009) also find a positive market reaction if a hedge fund engages in shareholder activism. Shareholders of target firms receive a positive abnormal return of 10.2% on average. Firms that are targeted by private investors experience an abnormal return of 5.1% on average. Both hedge funds and private investors are extremely successful in implementing their objectives.
Brav et al. (2010) give another explanation for high abnormal returns around the announcement day. Value creation includes a temporary price impact caused by buying pressure from the initial hedge fund or its followers. If the price impact is purely temporary and reflects a trading friction, rather than information about potential value changes, the market corrects the price, indicated by negative abnormal returns shortly after the event.
Brav et al. (2008) also investigated the firm characteristics of the target firms. They find evidence that target firms tend to be low growth firms, but are significantly more profitable compared to their matched peers. Regarding the capital structure, Brav et al. (2008) conclude that targets firms have more leverage, a lower cash-to-asset ratio and pay their shareholders fewer dividends compared to the comparable firms. The research also focuses on the quality of the target’s corporate governance by using a corporate governance index. The resulting conclusion is that that target firms have more takeover defences. The presence of many takeover defences could lead to bad corporate governance and agency problems. Managers have less incentives to focus on increasing shareholder’s value, since the pressure of a takeover is much smaller, which could lead to bad managerial behaviour.
The general conclusion is that shareholder activism improves overall performance of target firms. Brav et al (2008) find evidence for increasing payout ratios, improving debt, and return on assets in the post-target period. In addition to this, activism by funds result in a reduction of free cash flow related agency problems.
10
D. Short selling as an instrument for shareholder activism
The existing literature doesn’t provide conclusive empirical evidence that supports or rejects the theory that short selling could become an effective form of shareholder activism. However, like the ‘Wall Street Walk’, it could also become an effective instrument when it creates disciplinary pressure on the target’s firm management.
Short sellers don’t actually own, but borrow a stock which is perceived to be overvalued from a broker, a large institutional investor or another broker-dealer1. The borrowed stock is sold in the market for the current market price. The position is closed when the stock is bought back in the market at a later point in time and is returned to the lender. Short sellers can profit from a decrease in the stock price. The risk-return profile for a short position differs from the profile of a long position. A short-sellers' maximum gain is the sale price of the stock (if the stock price falls to zero), while the loss is potentially unlimited (if the stock price rises) (Dechow et al., 1997).
Short sellers are characterized as sophisticated investors who incur relatively large transaction costs attempting to short-sell and subsequently repurchase temporarily overpriced securities. In addition, Asquith and Meulbroek (1996) find that short-sellers, as a group, successfully identify securities that underperform the market. In general, short sellers maximize their investments returns by following three strategies. First, short sellers avoid securities that incur high transactions costs. Second, investors use information in addition to fundamental-to-price ratios that has predictive power with respect to future returns. Dechow et al. (1997) find a strong relation between the trading strategies of short-sellers and ratios of fundamentals to market prices. Third, short sellers avoid securities with low fundamental-to-price ratios when the low ratios are attributable to temporarily low fundamentals.
Because short selling is more expensive and riskier than taking a long position, it is suggested that short-sellers will not trade unless they expect the price to fall enough to compensate them for the additional costs and risk. This could indicate that short selling is a valuable signal to the market that the company is overvalued. The method becomes effective when the management bring changes regarding corporate policies or overall firm performance in the post-target period. However, short sellers also have the reputation of being price manipulators.
Massa, Zhang and Zhang (2012) state that short sellers provide an external governance mechanism to discipline managerial incentives. Jensen (2008) analysed and defined the agency costs of overvalued firms and supports these findings. Substantially overvalued equity result in a set of organizational forces that are extremely difficult to manage. He also concludes that governance systems have failed to eliminate agency problems. Because short sellers are an obvious source of potentially valuable information for the governance system, boards should communicate with them.
1
Borrowing stocks from a pension fund or other large institutional investors is only an opportunity for professional investors, who pay a premium to the lender. Private investors can only profit from decreasing stock prices by using options, turbo´s or sprinters.
11
Jensen (2008) believes that short selling could become an effective tool for shareholder activism, but it requires a major shift in belief systems and culture of most boards and managements teams.
Jones et al. (2012) investigated the impact of short selling disclosures in the French, UK, and Spanish market and believe that short selling will not become an effective form of shareholder activism. They find that disclosures have a small direct effect on the targets’ stock prices. Only in the two days event window [0, 2] they calculated a significant cumulative abnormal return of -1.78%. The decrease is permanent, which indicates that short sellers are well informed but don’t manipulate the market.
E. European legislation regarding short selling
Short selling is heavily regulated or forbidden in many stock markets. For instance, short selling constraints in the United States were developed from beliefs that short-sellers cause follow-on short positions and may cause stock prices to spiral downward. High risk and a high potential for manipulating the stock prices resulted in a heavily regulated U.S. stock market. Therefore, US investors are often restricted in the size of short positions compared to the overall size of their managed portfolios (Jones et al. 2012).
Christophe et al. (2004) investigated the level of short selling in the five days prior to the earnings announcement of Nasdaq-listed firms. For a subset of companies, short selling activity is significant higher in the days prior to a negative earnings announcement. They find evidence that underline the existence of insider trading among short-sellers. Disclosing short positions improves market efficiency and the variance of changes in stock prices decreases, as information becomes publicly available.
In response to the financial crisis of 2008, France, Spain, and the United Kingdom introduced short selling registers that require market participants to disclose large short positions in equities. By increasing transparency, European financial regulators aim to prevent a further increase in financial instability. Policies between the countries differ in terms of the disclosure threshold and to what extent information about the positions becomes publicly available. Under the UK short selling policy, investors are required to disclose their position when exceeding the threshold of 0.25% of the target’s issued capital. France and Spain adopt a threshold of 0.5%. In July 2012 the Spanish regulator Comisión Nacional del Mercado de Valores (CNMV) even prohibited short selling activities for a period of three months, in an attempt to reduce volatility in the equity markets. In March 2012, the European Parlement and Council approved a uniform short selling legislation, which applies to all European Member States. The new legislation, established by the European Securities and Market
12
Authority, ends the fragmented situation, which could hinder the functioning of the internal European financial market (Jones et al., 2012).
Jones et al. (2012) find evidence for significant follow-on short selling activity after the first disclosure. However, follow-on short selling doesn’t result in a significant increase in the net short positions, which seems odd. This contradiction can be explained by the existence of short positions in the target firm on the day of the initial disclosure, which are below the threshold. Based on the findings by Jones et al. (2012) it can be concluded that the short selling register doesn’t signal other investors and doesn’t drive down the stock price. This underlines the potential of the new disclosure legislation for other European member states.
Since November 2012, in response to the new European Legislation, the Dutch Authority for financial markets (AFM) has disclosed information about significant short positions. The online disclosure register is used as a starting point for the empirical section of this research.
13
II. Hypotheses
This section develops multiple hypotheses that are tested in the empirical section. Testing hypotheses help formulating an answer to the central research question. The hypotheses are based on empirical findings described by the existing literature.
The first hypothesis focuses on the short-term market reaction to the disclosure of short positions. Jones et al. (2012) investigated the market reaction on disclosed short positions and did not find significant abnormal returns in the days around the disclosure. However, the first hypothesis is based on empirical evidence that underline the effectiveness of hedge funds that engage in shareholder activism. Hedge funds are quite successful in achieving their objectives when they intervene in underperforming firms. In two-third of the cases their goals are fully or partially in achieved (Brav et al. 2008). Markets react positively to the announcement by hedge funds to engage in shareholder activism, which enforces the disciplinary effect on managers. Appendix C gives an overview of institutional investors that disclosed significant short positions in Dutch listed firms between November 2012 and August 2013. The majority of the events are disclosed by hedge funds. The first hypotheses don’t follow the results by Jones et al. (2012), since hedge funds act as well informed monitors, which are managed by highly informed managers. In general, short sellers position themselves in overvalued target firms and will not trade unless they expect a significant price fall, which compensates for a high exposure to risk. Based on these findings, the first subset of hypotheses is as follows:
Hypothesis 1a: Disclosed short positions result in a significant stock price decline.
Hypothesis 1b: Target firms are more overvalued in comparison to non-target firms, which is indicated by a significantly higher Tobin’s Q and Market-to-Book ratio.
Hypothesis 1c: The presence of hedge funds decreases cumulative abnormal returns further.
The second category of hypotheses focuses on firm performance and the the cash position. Brav et al. (2008) find that hedge funds take long positions in target firms that show higher operational performance in terms of ROA and free cash flows. Because short-selling works the other way around, it is expected that target firms show lower operational performance in the pre-target period. In addition to this, it is expected that target firms have a lower cash position, measured by net cash relative to total assets. The amount of net cash evaluates the company’s cash flow and is commonly used among investors as an indicator for the attractiveness of the company’s stock. A positive net cash
14
position indicates that cash is available to finance projects or pay out dividends in the nearby future. Therefore the hypotheses are as follows:
Hypothesis 2a: Target firms show lower performance in the pre-target period compared to non-target firms; higher ROA decreases the probability of being targeted.
Hypothesis 2b: Target firms show a lower net cash position compared to non-target firms; improving the net cash position decreases the probability of being targeted.
The third category focuses on the impact of corporate governance structures on the probability of being targeted. Shareholder activism often results from dissatisfaction with the current corporate governance mechanisms. Brav et al. (2008) find that target firms of shareholder activism show lower quality of corporate governance structures. It is unclear to what extend short sellers take corporate governance variables in to account when they determine their targets. Therefore, this research examines the impact of closely-held shares and board size on the market reaction and the probability of being target.
Closely-held shares are shares in a publicly traded firm, held by managers, insiders, and large block holders. Holdernes et al. (1999) concludes that large block holders and inside ownership minimalizes agency problems, because managers are more monitored and have more personal incentives. Furthermore, companies with higher percentages of closely-held shares may be more stable, since the stock price are less determined by the investors’ sentiment.
Yermack (1994) find evidence for an inverse relation between board size and firm value. His results support the existing theory that small boards are more effective in decision making. As a result, smaller boards make it more easy to dismiss the CEO, which has a strong disciplinary effect. Hermalin and Weisbach (2003) support these results and conclude that large board sizes are less effective than small boards, because they suffer from agency problems and free rider problems. Therefore, the corporate governance related hypotheses are as follows:
Hypothesis 3a: The percentage of closely-held shares to target firms is smaller compared to non-target firms; the percentage of closely-held shares is negatively related to the probability of being targeted by short sellers.
Hypothesis 3b: Target firms have larger board sizes compared non-target firms; board size is positively related to the probability of being targeted by short sellers.
15
III. Research design
This section describes how the research is designed. Section A describes the sample selection and the databases. Section B describes two recent cases of short selling in the Dutch market, which provide more insight in short selling in the Dutch market. Section C describes and analyses the descriptive statistics.
A. Data sample
This research analyses the impact of large short positions in Dutch listed companies, held by institutional investors in the period November 2012 – August 2013. Since 1 November 2012, the Dutch Authority Financial Markets (AFM) publishes an online short selling register including notifications of investors holding significant net short positions in Dutch listed firms. Investors are required to report their position to the AFM when the size of the position exceeds 0.2% of the issued capital. The position is disclosed in the register if it exceeds a pre-determined threshold of 0.5% of the issued share capital of the target concerned. In addition to this, every 0.1% increase or decrease above the threshold must be disclosed, which gives more insight in the development of net short positions. Furthermore, the list contains the names of the market participant holding the net short position, the stock issuer and disclosure dates. Once a significant net short position falls below the 0.5% threshold, the position will remain visible in the active register for one more day before it is moved to the archive overview.
The market reaction can be determined by calculating cumulative abnormal returns on the days around the disclosure dates. Determining the market reaction requires daily stock returns of the target firms and daily index returns, the latter is also recognized as the benchmark or normal return. Daily returns are retrieved from Thomson Reuters Datastream, a financial database containing global company data, stock returns, and macro-economic data. The AEX index, Amsterdam Mid Cap index (AMX) and Amsterdam Small Cap index (AScX) are used as benchmark returns. Datastream also provides panel data concerning firm characteristics of target firms and its peers, which are used to measure the underlying determinants of the market reaction2.
2
Variables Board Size and R&D expenses are mainly collected by hand from annual reports, as Datastream frequently reports missing values for both variables.
16 B. Two sample examples of recent short positions
In order to provide more insight into why Dutch firms are targeted by institutional investors, two recent cases are described. Information about the positions is retrieved from ‘Het Financieele Dagblad’, a daily Dutch financial newspaper. Both examples are included in the research data sample.
B.1. Imtech
Royal Imtech N.V. is a technical services provider in the European market of electrical engineering, ICT and mechanical engineering. During the period 1 November 2012 – 1 August 2013 the firm has been targeted by 15 institutitional investors, which resulted in 208 disclosured net short positions.
In February 2013 Imtech announced a depreciation of more than EUR 100 million on Polish projects, including the construction of a large amusement park. This park, Adventure World Warsaw, was estimated to be worth EUR 680 million. It was the largest order in the company’s history and Imtech would be responsible for ICT, security and power generation.
The company did not receive money for the services it provided which resulted in the large depreciation on its Polish projects. It remains unclear whether the fraud has been performed by Imtech’s local management or by its Polish client. Due to the high depreciation, Imtech was unable to meet covenants that have been agreed with the bank. Despite the fact that the company claimed to have sufficient liquidity to stay out of the danger zone, the stock price dropped by 47% on the day of the announcement. Profit for short sellers are estimated to be EUR 170 million (Cohen and Kakebeeke, 2013).
B.2. SNS Reaal
SNS Reaal N.V. is a Dutch bank-insurer and provides products and services in the fields of insurance, property finance, mortgages, savings, pensions and investments. Clients include retailers and SMEs that are mainly active in the Dutch market.
In 2006 the company bought the Property Finance portfolio from ABN AMRO, which included loans to risky Real Estate companies. After the financial crisis of 2008, the majority of the borrowers were not able to fully repay their debts. SNS Reaal incurred significant losses, which could not be compensated by profitable banking and insurance activities. The bank did not have enough financial capital to continue its activities. Savers extracted money from their deposit accounts, sometimes up to EUR 500 million per day. The Dutch state nationalized the bank-insurer on 1 February 2013.
17
Between 1 November 2012 and 1 February 2013, SNS Reaal was targeted by four large hedge funds. With the nationalization of the company, shareholders of the company are expropriated and the obligation for short sellers to return the borrowed shares expired. Put differently, short sellers earned a profit of 100%, which is equal to SNS Reaal’s stock price at the moment of borrowing and selling. The profit for the hedge funds is estimated to be EUR 26 million (Bos, 2013).
C. Descriptive statistics
An overview of the AFM short selling register regarding the period November 2012 – August 2013 is presented in Table I. The period of interest includes 27 Dutch listed firms, targeted by 57 institutional investors. Short positions have a minimum value of at least 0.5% of target’s issued capital. A disclosure is the result of an increasing or decreasing short position by at least 0.1%. Changes in net short positions are only disclosed in the current register if the current short position exceeds the threshold of 0.5%. The institutional investors together have 100 short positions in the 27 target firms and the total number of disclosed positions is 797. The average number of disclosures per short position is 7.97, which indicates that institutional investors change a single short position by almost eight times on average before it closes the position. The average number of disclosed positions per target firm is 29.52; the average with respect to the position holders is 13.98. The average net short position is 0.95%, the median short position 0.72%, and the maximum net short position reaches an interest of 5.07%.
Table II gives an overview of the industries in which the target firms operate, and the number of disclosures per industry. Most disclosures are assigned to companies that produce or trade Industrial Goods, this industry accounts for 41% of total disclosures. The large majority can be explained by the presence of Royal Imtech N.V., which accounts for 208 disclosed net short positions. Appendix B gives a more detailed overview of the disclosures per target firm.
Appendix C gives information about institutional investors. The Appendix reports the name of the position holder, country of origin, the number of disclosures per position holder, and the percentage of total disclosures. Furthermore, the Appendix makes a distinction between hedge funds and other asset managers. Of 57 institutional investors 15 are recognized as a hedge fund. The percentage of position disclosed by hedge funds equals 69%. The majority of the institutional investors are US- and UK-based. Institutional investors from the US disclosed 61.3% of total disclosures; UK investors disclosed 35.8%; 2.9% include investors from other countries, which are mostly European-based.
Table III reports the firm characteristics for the 27 Dutch target firms. The variables of interest are described in Appendix A. More than 50% of the target firms are listed on the AMX. Firm size is indicated by the variables MV and EV. The average (median) market capitalization for target
18
firms is EUR 2.963 million (EUR 1.201 million). The sample is relatively small, which results in a relatively large standard deviation. The smallest firm targeted has a market value of EUR 126 million, while the largest firm has a market value of EUR 13.512 million. The average (median) enterprise value is EUR 5.933 million (EUR 2.424 million). EV also reveals a large dispersion in minimum and maximum value; EV ranges from EUR 324 million to EUR 30.9000 million.
The average Tobin’s Q for the target firms is 1.45, which indicates that the average market value is 1.45 times higher than the book value. In general a value higher than 1 means that a firm is overvalued. The median value is 1.07, indicating a slightly overvalued target sample. Market-to-book, which represents the market value of equity divided over the book value of equity, has an average (median) value of 1.55 (1.16). Both valuation indicators conclude that the average and median target firm is overvalued.
The average and median Cash ratio are slightly negative, which indicates that the average and median net cash position is negative. Net cash refers to the amount of cash after subtracting interest-bearing liabilities. The ratio evaluates the company’s cash flow and is a common indicator for the attractiveness of the company’s stock. A positive net cash position indicates that cash is available to finance projects in the nearby future. Leverage, defined by net debt relative to total asset, has an average and median value of 0.26. Summary statistics show a minimum value of -0.13, which can be explained by a cash position that exceeds interest-bearing liabilities.
Target firms report an average (median) ROA of 9% (6%). The worst performing target has a negative ROA of -7% in the pre-target period.
Capex and R&D give more insight in the investment side of target firms. Capex, defined by capital expenditures relative to sales, has an average (median) value of 0.18 (0.04). On average, target firms allocate a relatively small proportion to R&D expenses. The average (median) value of R&D expenses relative to sales is 0.02 (0.00).
Target firms show an average (median) dividend yield of 4.52% (3.64%). Payout, defined by the percentage of net income before extraordinary items that is paid as dividend, has an average (median) value of 39.25% (41.32%). Binckbank N.V. reports the largest payout ratio (98.64%), however its dividend yield is approximately 5%.
Ownership and Board Size examine the quality of corporate governance structures. Ownership represents the percentage of closely held shares by managers, relatives of the founders and large blockholders. Ownership has an average (median) value of 20.89% (10.57%). The average (median) Board Size, including both the management and executive board member, consists of 7.44 (7) members.
19
IV. Research methodology
This research investigates the market reaction and the underlying determinants of disclosed short positions in Dutch listed firms. This sections starts describing the methodology that calculates the market reaction. Secondly, the methodology regarding the underlying determinants is mentioned. Finally, the models that calculate the target probability are mentioned.
A. Market Reaction
Event studies are commonly used in Finance and are used to examine the market reaction on the days around a particular event. In this case the market return is expressed by the cumulative abnormal return, which is the sum of the daily abnormal returns on the days around the position date. The abnormal return is the difference between the expected stock return and observed stock return:
ARjt = Rjt – E(Rjt), (1)
Where,
ARjt = abnormal return for for stock j Rjt = observed stock return for a for stock j
E(Rjt) = excess return, also recognized as normal return or benchmark return
The single-index market model estimates the benchmark return, which is the return as if there would not be any influence of an announced event. The formula for the benchmark return is expressed as follows:
E(Rjt) = αj + βjRMt + εjt (2)
Where:
E(Rjt) = benchmark return for a given target firm aj = constant
βj = beta of the target firm’s stock relative to the market return Rmt = market return or index return
ԑit = residual returns, assumed to be normally distributed with a mean of zero and standard deviation equal to σi
20
The market model is based on the method suggested by Brawn and Warner (1985), who estimate the coefficients by an OLS-regression model. The benchmark return depends on the alpha coefficient, the beta coefficient, the market return, and the error term. In order to eliminate any influences of the announcement on the return, alpha and beta are estimated by an OLS-regression for the time window [-252; -30]. The software program Stata 12 performs the OLS-regressions and corrects the parameters for heteroscedasticity.
Three index returns are retrieved from Datastream, since the target firms are listed on different exchange caps. The AEX index return is used for large-cap target firms, the AMX return for mid-cap firms, and the AScX for small-cap firms. Abnormal returns are calculated on the days around the positions dates. The event windows of interest include: [-1, 1], [-2, 2], [-5, 5], [-10,10], [-15, 15], [-20, 20], and [-25, 25]. Including larger event windows show whether the price impact is permanent or temporary.
The market response can be expressed by the change in shareholder value, which is calculated by the cumulative abnormal return (CAR). CAR is the sum of daily abnormal returns and are calculated by the following formulas:
CARjτ = ∑ (3)
CAARjτ = 1/n ∑ (4)
The sample might contain follow-on short selling activity, which means that the initial disclosure of by an institutional investor, in a particular target firm, is followed by disclosed positions in the same firm by other institutional investors. In addition to this, a particular investor might increase or decrease its net short position within a short time window. Follow-on activity and multiple disclosures within a short time window might bias cumulative abnormal returns, as event-windows contain the effect of multiple disclosures.
The presence of multiple events in a single event window results in increasing variances, which could induce misspecification when testing the null hypothesis. This research follows the method provided by Khotari and Warner (2006),who correct standard errors for multiple events that occur in a particular event window. Portfolios are constructed to capture correlations between stock returns in the 100 days event window. The portfolio standard deviation for the portfolio return accounts for cross-sectional dependence between the stock returns. The standard deviation is used to test cumulative average abnormal returns on their significance and is used for multiple event-windows.
21 B. Firm characteristics
In order to determine the underlying determinants of the market reaction, the cumulative abnormal returns are regressed on a set of firm characteristics. Firm characteristics that have a significant impact on the cumulative abnormal returns might explain why firms are targeted. The regression equation is as defined follows:
CARjτ = α + β1 Short Position + β2 ln(MV) + β3 Tobin’s Q + β4 Mtb+ (5) β5 Cash + β6 Leverage + β7 ROA + β8 ln( Capex) + β9 ln (R&D) +
β10 Dividend + β11 Payout + β12 Ownership + β13 Board size + D1 Hedge Fund
Targets’ firm characteristics are compared to two peer groups, using different matching approaches. The first approach compares target firms to firms operating in the European market, same industry segment, with a comparable firm size in terms market capitalization. This approach is based on the criteria proposed by Brav et al (2008). Firms operating in the European market are subject to the same for forms of European regulation. Firms operating in the same industry segment are exposed to similar risk-levels and growth opportunities, since they operate in the same stage of the industry life cycle. Information about the target’s industry segment is retrieved from Infinancials, a database that provides data and analytics on more than 80.000 listed firms. Appendix B reports the industry segment for the 27 target companies. Of the European competitors four companies are selected with a comparable firm size, in terms of market capitalization3.
The second approach compares the target firms to the remaining Dutch listed companies on Euronext Amsterdam. Comparing the results between different matching approaches might reveal consistent differences between target and non-target firms.
First, the average differences in firm characteristics between the targeted companies and both peer groups are compared. The differences are calculated by the following equation:
Difj = Xi – 1/n ∑ (6) Where:
Xi = firm characteristic target firm i N = the number of matched firms Xj = firm characteristic matched firm j
3 In case of missing comparable firms in terms of market capitalization, the target firm is matched to a Dutch listed firm with
22
Second, probit regression models are run to determine the marginal probability of firm characteristics on the probability of being targeted. The dependent variable is a dummy variable, which equals one if the firm is registered in the AFM short selling register between 1 November 2012 and 1 August 2013. The same independent variables are included in the model as described in equation 5, except for the variables Short Position and the dummy Hedge Fund. The first probit model includes the set of target firms and market/industry/size matched firms. The second model includes on all Dutch listed firms. The equation model is as follows:
Pr ( Targeti = 1) = α + β1 ln(MV) + β2 Tobin’s Q + β3 Mtb + β4 Cash + (7) β5 Leverage + β6 ROA + β7 ln (Capex) + β8 ln (R&D) +
23
V. Research results
This section describes and analyses the empirical findings of this research. First, the regression results regarding the market reaction are examined. Second, the underlying determinants of the cumulative abnormal returns are investigated. Subsequently, differences in firm characteristics between target and non-target firms are analysed. The second ends by summarizing consistent results reported by the different research models.
A. Market Reaction
The market reaction is determined by the cumulative abnormal returns on the days around the disclosure of the short positions. Table IV reports the cumulative average abnormal returns (CAARs) for two categories during multiple event windows. The first category measures the impact of the first disclosed short position, held by a single investor in a single target firm. The first disclosure of a short position is also known as the initial disclosure. The second category measures the total effect of both initial disclosure and any following disclosures. In addition to this, the second category also incorporates the effect of changes in the net short positions. For the convenience, the second category is called the ‘total disclosure effect’4
.
The market responds negatively to initial short positions during all examined event windows. CAARs are significant at the 1% level for the 11, 21, 31, 41, and 51-days event windows. CAAR decreases as the event window becomes larger, and reaches a minimum value of -6.90% in the 41 days event window, after which it increases to -4.93% in the 51-days event window.
The total disclosure effect also has a negative impact on the target’s stock price.
The results show a similar trend compared to the initial disclosures: the minimum value for CAAR is reached during the 41-days event window, after which the CAAR increases. The values for the 41 and 51-days event windows are respectively -7.69%, and -7.30%. The 11- days event window is statistically significant at the 5% level, the larger event windows are significant at the 1% level. The findings support Hypothesis 1a: The Dutch market reacts negatively to the disclosure of significant short positions.
A large proportion of the total disclosure effect results from the initial disclosure effect, which is 78% on average. Furthermore, the development of CAAR over the multiple event windows shows a permanent price decrease. These results conclude that there is no stock price manipulation associated with the short selling register. The results support the findings by Jones et al. (2012).
4
The total number of dsiclsoures is lower than 797, which can be explained by the nationalization of SNS Reaal. As a result,
24
Figure 2 plots the average daily return and abnormal trading volume for the 51 days event window. Abnormal returns are decreasing in the pre-disclosure period.
After the disclosure the abnormal return shows an slightly upward trend. The abnormal trading volume shows an upward trend. Trading volume is highest at the day of the disclosure, after which it stabilizes. The upward trends support earlier findings that short sellers act as well informed traders.
B. Underlying determinants of the market reaction
In order to determine the underlying determinants of the market reaction on disclosed positions, firm characteristics are regressed on the cumulative abnormal returns. The OLS-regression results are presented in Table V. Firm fixed effects control for firm specific characteristics that are constant over time. Including firm fixed effects in the regression model lowers the probability of omitted variables bias and serves as a robustness check. Firm fixed effects increase the explanatory power of the regression model, which is indicated by a higher R-squared. The OLS regression model is not able to regress the variables Payout, Ownership and Board Size, when including firm fixed effects. These variables capture a substantial part of the firm fixed effects; including these variables and firm fixed effects causes omitted variable bias. For instance, Board Size is relatively stable over time and differs between firms. The standard errors are clustered at firm level.
The impact of Position, defined by the size of a short position, on cumulative abnormal returns is not as large as expected. Smaller event windows report a negative relation between Position and CAR, however the coefficients are insignificant. The coefficient sign becomes positive for the larger event-windows, but remain insignificant. Only the 31-days event window shows a significant coefficient at the 10% level equal to 2.04. Increasing the short position by 1% results in a 2.04% higher cumulative abnormal return, which seems obvious.
The results show a positive relation between firm size and CAR. MV is significant at the 1% and 10% level for the 31 and 41-days event window. Increasing MV by 1% leads to a 24.68% higher CAR for the 31-days event windows. Such a large impact seems improbable and might be explained by the small sample size. As discussed in section III, the small sample results in a large dispersion in market capitalization between the target firms5.
The regression results show conflicting results between Tobin’s Q and Market-to-Book. Tobins’ Q has a negative impact on CAR, which is significant for every event window. The coefficient is equal to -4.32 in the 3-days event window, which suggest that an increase by one-unit in Tobins’ Q results in a -4.32% lower CAR. Market-to-book has a positive impact on CAR, which is significant for all event windows. Including firm fixed effects doesn’t reveal the same results; variables are
5 EV is excluded from the regression models, since the variable is highly correlated to other firm characteristics.
25
insignificant or show the opposite coefficient sign. It is possible that firm specific effects influence both variables, which causes a variable bias.
Cash is positively related to CAR, and significant at the 1% level for every event window. A 1% in the net cash relative to total assets results in a 41.74% increase in CAR for the 11-days event window. The positive relation supports Hypothesis 2b, however the large impact seems improbable. Leverage has a small impact on CAR. Only the 3-days event window reports a significant coefficient of -2.68 at the 10% level.
The impact of firm performance on CARs is measured by ROA, which is an important factor in explaining the market reaction. ROA is significant at the 1% level for the 3, 5, 11, and 21-days event window. The impact of a changing firm performance is smaller for the 31, 41, and 51-days event windows, but remains significant at respectively the 5% and 10% level. At first sight, the negative relation seems obvious and the impact improbable. A 1% increase in ROA results in a -36.92% decrease in CAR for the 11-days event window, which seems improbable. The negative relation seems obvious at first sight, but might be explained by increasing agency problems associated with free cash flows. Nevertheless, Hypothesis 2a is rejected.
Capex has a positive impact on CAR, however the relation is mainly insignificant. Capex is only significant at the 10% level for the 5-days event window, when firm fixed effects are included. A 1% increase in capital expenditures relative to assets, results in a 5.21% increase in CAR. R&D reveals a negative relation between R&D expenses and CAR, which is insignificant for every window. However, including firm fixed effects shows a significant positive relation. Both Capex and R&D seem to be influenced by firm fixed effects. Therefore these variables are no important factors in explaining the market reaction on disclosed positions. The same conclusion seems to be valid for the variables Dividend and Payout.
The regression results show a positive relation between the corporate governance related variables and the CARs. An increase in Ownership, defined by the percentage of closely-held shares, results in a higher CAR. The impact is significant for every event-window and supports Hypothesis 3a. An increasing number of board members results in increasing CARs, the results are significant at the 10% for the 3, 5, and 11-days event window. The positive coefficient doesn’t support Hypothesis 3b, which is therefore rejected.
The impact of hedge funds on the market reaction is measured by the dummy Hedge Fund. The dummy variable is not an important factor in explaining the negative market reaction. Therefore, Hypothesis 1c is rejected.
26 C. Difference target firms and peer groups
This subsection investigates the main differences between target firms and peer groups and gives more insight in why firms are targeted. First, the mean differences in firm characteristics between targets and non-targets are described. Second, the probit regression results are described.
C1. Firm characteristics targets and matched firms
Table III gives more insight why firms are targeted by institutional investors. Column six and seven report the mean differences between firm characteristics of target firms and their matched comparables. Each target firm is matched with four European firms that are active in the same industry segment, with a comparable firm size in terms of market capitalization. Differences are tested for their significance, which is indicated by the t-values. Appendix D gives an overview of the target firms and their matched peer groups.
Comparable firms have a larger firm size in terms of both market capitalization and enterprise value. Matched firms have an average MV amounting to EUR 3.139 million, which is EUR 176 million higher than the average target firm. Matched firms have an average EV amounting to EUR 6.112 million, which is EUR 179 million higher than the average target firm. Differences are insignificant, since market capitalization accounts as a matching criterion.
Tobins’ Q and Market-to-Book are analysed to determine whether target firms are more overvalued than non-target firms. The mean difference in Tobins’ Q indicates that target firms are slightly more overvalued compared to their matched comparables. This result support Hypothesis 1b; short sellers position themselves in overvalued stocks. However, the insignificant difference of 0.02 indicates that the average comparable firm is overvalued as well. The difference in Market-to-Book shows different outcome. The results conclude that target firms are less overvalued, which seems obvious. However, the relatively small difference of -0.01 is statistically insignificant. The valuation variables conclude that both the target sample and the matched peer group are overvalued. Therefore, the mean differences between both variables don’t have explanatory power in determining why firms are targeted.
Capital structures of targets differ significantly compared to matched firms. Target firms have smaller cash positions, the average net cash position relative to total assets is equal to zero. The net cash position for matched firms equals 5% of total assets. The mean difference between both groups is significant at the 5% level. Target firms are more levered than their comparables. The net debt position relative to total assets is 14% higher for target firms. The difference is significant at the 1% level.
27
As expected, target firms show lower firm performance than their matched competitors. The matched peer group shows a ROA of 10% on average, 1% higher than the target sample. However, the difference is insignificant.
The investment side reveals no significant differences in terms of capital expenditures and expenses in R&D. Capex, defined by the capital expenditures relative to sales, equals 14% for matched firms, which is 4% higher compared to target firms. R&D expenses equal 3% of total sales for matched firms, which is 1% higher than the R&D level for target firms.
In the pre-year target period the average dividend yield is 0.93% higher for target firms, the difference is significant at the 5% level. The matched peer group shows a higher payout ratio of 4.72% on average. Brav et al. (, 2008) states that lower payout ratios could lead to agency problems associated with free cash flows. However, the difference is statistically insignificant.
Closely-held shares equal 20.98% for target firms, compared to 30.32% for matched firms. The difference of 9.43% is significant at the 5% level and indicates that stocks of target firms are less liquid. Matched firms have 9.89 members on average, which is 1.45 more than board of target firms. The difference is significant at the 5% level. However, Board Size might better be compared to other Dutch listed firms, since European countries may show different trends in board structures. For instance, in the Netherlands, the two-tier board is commonly implemented.
C2. Firm Characteristics targets and other Dutch listed firms
The second matching approach compares the targets to other Dutch listed firms. Consistent differences in firm characteristics between both matching approaches help explaining why firms get targeted.
Differences in MV and EV equal respectively EUR -3.487 million and EUR -7.810 million, indicating that targets firms have smaller firm sizes in comparison to other Dutch listed firms. However the difference are insignificant, which can be explained by high standard errors resulting from a small sample size.
The valuation measures Tobins’ Q and Market-to-Book reveal the same results: the Dutch peer group has a lower Tobins’ Q, but a higher Market-to-Book ratio compared to target firms. The remaining Dutch listed firms have an average Tobin’s Q of 1.38 and a Market-to-Book ratio of 1.55. The mean differences between both groups are statistically insignificant. Based on the results it can be concluded that the average Dutch listed firm was overvalued in the period November 2011 – November 2012.
The capital structure variables reports a better net cash position and a lower net debt level relative to total assets for the remaining Dutch listed firms. The peer group has an average net cash
