The effect of short-sale restrictions during the financial crisis

The effect of short-sale restrictions

during the financial crisis

A study into the abnormal returns of financial stocks

Abstract: We investigate the effects of short-sale bans on abnormal returns of financial stocks during the 2008 financial crisis. Using a sample of 12 small- to mid-sized European stock markets with diverging rules on short-selling, we find negative effects of short-selling restrictions on abnormal returns. Instead of the intended support to prices and volatility, restrictions on naked and covered short-selling lead to worsened performance.

Keywords: short-selling restrictions, naked, covered, abnormal returns, volatility, financial crisis, price support.

Master thesis

The effect of short-sale restrictions during the financial crisis

Table of Contents

1. Introduction... 4

1.1 The short-selling mechanism... 4

1.2 The short-selling debate... 5

2. Literature ... 8

3. Methodology ... 15

3.5.1 Paired sample T-test ... 20

1. Introduction

At the height of the financial crisis in September 2008, many stock market regulators imposed short-selling restrictions to support stock prices and to reduce the heavy volatility that accompanied the market crash. Some banned short-selling on all listed stocks, some were more selective and only chose financials, and some did not even insert a shorting ban at all. The variety in scope and insertion dates allows for a cross-country comparison of stock returns on small- to medium-sized European markets during a period of financial decline. Although the intentions of the regulators were obvious, literature suggests that short-selling restrictions can have adverse effects on stock prices. In this study we investigate whether the short-selling bans in the financial crisis of 2008 were irreproachable.

Short-selling has existed as long as stock exchanges have. Only recently has a certain level of consensus been established in literature regarding short-selling and its effect on stock returns. Yet, the discussions on short-selling of stocks – borrowing a stock, subsequently selling it and hoping for a

decrease in price by the time the stock needs to be returned – have resurfaced amongst the turmoil of the current financial crisis. In a knee-jerk reaction to collapsing market prices, financial regulators adjusted legislation regarding the shorting of shares. Where predominant stock markets such as the New York Stock Exchange and the NASDAQ tightened regulations, few publicly announced their support for short-selling. In Europe, consensus was not reached when the Committee of European Securities Regulators (CESR) appealed to regulators to instate a common ban on short-selling (Autoriteit Financiële Markten, 2008). The diversity of rules regarding short-selling in different European countries makes the European Union a suitable geographical area to conduct research on the effects of short-sales.

1.1 The short-selling mechanism

promise to the buyer of the short to deliver in the future. Most lenders are intermediaries that act on behalf of institutional investors. Brokerage firms, on the other hand, usually have access to an internal supply of stocks. Once a lender has been located, the borrowing transaction will start with the placement of collateral with the owner and a lending fee paid to the owner of the stock. The level of the lending fee is determined by supply and demand on the stock loan market (Lamont, 2004), and is paid over agreed upon time periods. Collateral generally consists of cash and treasury securities (DAvolio, 2002) and generates interest for the lender. In case of sizable amounts, the borrower of the stock (i.e. the party who puts up the collateral), might demand a rebate of the earned interest (i.e. the short stock rebate rate) (The Street.com, 2010). The transaction does not affect legal ownership, so any dividend or distribution rights for the stock remain with the original owner. Furthermore, the lender has the right to recall the stock at any time. In case of a covered short transaction, the sale is only possible after the stock has been borrowed. The short seller realizes a profit when, at a later point in time, he is able to buy the stock at a lower price to return it to the lender. A numerical example of a short-sale transaction is shown in Table 1 in

the Appendix. Apart from speculating reasons, motives to sell a security short include hedging and balancing of buy and sell orders. Market makers need to balance buy orders with sells, and shorting is key component in strategies that involve offsetting the risk of a long position (DAvolio, 2002). Risks for the short seller include involuntary closing of a short position due to a recall, and the implicit risk of the possibility of an increase in share price. The upside risk for the short-seller is limited to the price of the share; indeed the price can only

drop as much as it is worth.

Contrastingly, the downside risk for a short-seller is unlimited, since there is no maximum price for which he will have to buy the share back. In case of a naked short-sale, there is no risk of involuntary closing for the selling party, as there is no position to
close; instead the buying party runs the risk of delivery failure.

1.2 The short-selling debate

However, taken together, conclusions seem to contradict (Charoenrook & Daouk, 2005). Although this study does not strive to solve this contradiction, a mere step towards consensus could pose fruitful to all parties involved, including academics, regulators, investors, and investees.

Senchack and Starks (1993) investigate the effects of short-selling at the individual stock level and find empirical evidence to support the findings of Diamond and Verrechia: stock returns indeed have a negative reaction to short-sale constraints. Additionally, several researchers have investigated the effects of short-selling combined with crashes. Allen and Gale (1991) argue that short-selling may potentially lead to a crash. On a highly related note, Endo and Rhee (2006), argue that short-sales hypothetically enable a quicker recovery from a crash. The proposed arguments by Allen and Gale on the one hand and Endo and Rhee on the other are rather paradoxical. It raises the question whether short-sales should only be allowed after a crash to stimulate recovery, or, perhaps not at all in order to prevent one. Christophe et al. (2004) also question the timing of short-sale restrictions.

By looking at the behaviour of various stock markets in world, one is unable to identify a pattern in terms of regulatory decision-making. During the turmoil of the European debt crisis, the German councillor Angela Merkel introduced a ban on naked short-selling in May of 2010, for example (Kirschbaum, 2010). The Japanese Financial Services Agency furthermore, has recently announced an extension of the ban on short-selling (White, 2009), while the Italian market watchdog Consobo has just announced a lift of the ban on short-selling that was instated during the beginning of the crisis (Forden, 2009).

restrictions are dispersed. At this point, most large stock markets have inserted bans on short-selling. Inclusion of a large influential stock market could lead to a biased sample, which in turn could cause problems in terms of external validity. We have therefore selected a number of smaller European stock markets to investigate the possible effects of short-sale constraints. The selection of European stock market is shown in Table 2.

The aforementioned lack of consensus is witnessed within the selected sample of countries. The Dutch and Swedish authorities, for example, took divergent actions regarding short-sales. The Dutch financial markets authority (AFM) instated a ban during the early stages of the start of the global financial turmoil in 2008. A temporary ban was instated on September 22nd and was supposed to take effect during a three-month period. The ban turned out to be extended a number of times and was only abolished in June of 2009. Certain aspects of the regulation are still in effect – public disclosure, for example, is still required if the sale involves more than 0.25% of the investee
s securities (Autoriteit Financiële Markten, 2008).

Sweden, however, allowed short-selling before the financial crisis and as one of

the few, continued to do so during the financial crisis. Even after other countries imposed bans on short-selling, Sweden
s Finansinspektionen (the financial market regulator) deemed such measures unnecessary and did not consider implementing them either (Reuters, 2008). Furthermore, legislation was already in place that prohibited “using short-selling to unlawfully influence the price of a security” (Strömqvist, 2009). At a later stage, the Finansinspektionen announced that it had started to monitor short-selling and increased its market surveillance of trading in financial companies (Hedval, 2008a). In spite of those measures, the discussions to ban short-selling in Sweden dwindled (Dagens Industri, 2008). Closer inspection of short ban policies in other countries did not result in a change of course for the Finansinspektionen, as these actions were seen as not effective (Hedval, 2008b).

terms of returns since the beginning of the financial crisis. In order to investigate the possible effects of short-sale restrictions, this study focuses on the returns of financial institutions across multiple countries in Europe. The research question that therefore has arisen can be stated as follows:

Do short-sale regulations influence abnormal returns of financial stocks during a financial crisis?

The answer is both relevant for the assessment of past restrictions and for future actions. The direction and size of the abnormal returns can indicate whether the shorting bans were an irreproachable and sensible action during the market crash. Projecting this to future situations can show whether shorting bans are effective in the support of stock prices, or whether they have adverse effects.

Although both findings are important, the latter could have the more far stretching implications. From an academic perspective it can be said that outcomes may help build consensus, or at least, more clarification on the subject by testing assumptions and hypotheses with a tailored conceptual model. From a business perspective, the answer is useful for regulators, financials, and investors and investees. Regulators may use the information to ensure that

markets perform better. At the same time, the information is useful to financials as opinions towards the costs and benefits of short-sales are diverging. Finally, possible outcomes may help (institutional) investors in their decision-making with regards to corporate social responsibility and the ethical dilemma of making profit from a bear market.

2. Literature

In an efficient market, security prices reflect all information available to investors, with the assumption that information and trading cost are zero. However, assessing the more realistic situation that these prices are not zero, the efficient market hypothesis can be reformulated as: “prices reflect information to the point where the marginal benefits of acting on information do not exceed the marginal costs“ (Fama, 1991). Furthermore, in its strongest form, a single investor with superior information - e.g. insider trading (Jaffe, 1974), cannot profit from this information since the price would immediately reflect this (Ross, Westerfield, Jaffe, & Jordan, 2008).

possible reasons for a security to be mispriced. In case investors deem a stock as overpriced, they may opt for selling it short. The underlying argument being that, over time, the market will recognize overpricing, leading to prices to decrease. By closing out the short position at a later point in time when prices have declined, short sellers are able to realize profits. In essence, the actions of short sellers are comparable to those of arbitrageurs, who also see opportunities for capturing value in case of mispricing of securities (Jones & Lamont, 2002). Apart from arbitrage purposes, the mechanics of short-selling are a key component in hedging strategies. Selling stocks short can be used to hedge the market risk (beta risk) of a portfolio and is generally used to hedge against market-wide decline. Equity managers may hence employ a short-sale to reduce the net market exposure of a portfolio (Aima, 2007).

Miller (1977) is among the most influential researchers to investigate the implications of short-sale restrictions on market returns. Miller sets out that the ability to short a stock will incorporate the adverse opinions of investors, therefore leading to prices that are in line with the efficient market hypothesis. An impediment to short-selling would consequently imply that these adverse

short-selling can be imposed on different levels: a complete prohibition of short-sales including both naked and covered shorting, a partial prohibition on naked shorting, and a partial prohibition on shorting of shares in particular industries (e.g. a prohibition of naked short-selling in financial shares). Restrictive factors attributable to costs include the stock-loan costs and opportunity costs that arise when the proceeds from the short-sale are not directly available for investment (Figlewski, 1981). Senchack and Starks (1993) attempt to test the theory developed by Diamond and Verrecchia using cross-sectional research on individual stock returns. They find that the possibility of short-selling enhances the distillation of information into stock returns.

Short-sale constraints imposed by regulators are directly observable. However, this is not always the case for other types of constraints. Boehme et al. (2009) describe three types of proxies for short-sale constraints: short-stock-rebate rates, the relative short interest level, and the presence of exchange traded options. Firstly, the short-stock-rebate rates are known to be the optimal proxy, but are also problematic as the required information is not publicly available (Jones & Lamont, 2002).
Relative short

interest (RSI)1, secondly, is said to be the most commonly used proxy. The RSI is a measure that is calculated by dividing the short interest – the number of shares sold short – by the number of outstanding shares, therefore representing the percentage of the shares that are held short for a particular company. The RSI is an indicator of the level of difficulty to take a short position in a security: a low RSI rate implies that there are relatively few short positions for a certain stock, therefore indicating that the stock is difficult to short. The RSI alone might not fully explain the degree to which a stock is being shorted. As restrictions become more severe, it will become increasingly more difficult to short a stock – a cause for the RSI to decrease simultaneously. Thirdly, the presence of tradable options plays an important role in measuring the impact of constraints as well. Should a short position be unavailable to an investor, the investor may opt for a so-called
synthetic short position, where the short position is realized by buying put options.

Short-sale proxies are a sensible approach when the amount of short-sale constraint is not of a regulatory nature and thus hard to identify. However, they

1

are still an indirect way of measuring the concept (Chang, Yu, & Complex, 2004). Instead of proxies, a direct and easily quantifiable method is made possible by the use of the lists of prohibited stocks. Financial regulators specify which stocks are affected by its regulations. In case of a short-selling ban, a list of unshortable stock is therefore available. This research will focus on the use of these publicized lists in order to use the most direct approach possible with the available data.

2.1 Relationships

2.1.1 Short-sale constraints

The majority of the extensive literature described above has findings consonant with those of Miller, where a short-sale ban results in overvaluation of the respective stocks, because negative information is harder to price into the market. However, some other authors found the opposite. As mentioned, Diamond and Verrechia (1987) argue that short-sale constraints could reduce the speed of information adjustment, hence not necessarily leading to overvaluation. Bai, Chang and Wang (2006) set out that overvaluation could occur in markets where the information asymmetry between informed and uninformed investors is relatively low. Because short-sales constraints suppress some investors of trading on

negative information, prices only reflect positive information. This is the so-called allocational effect. The situation is different for markets or periods where the asymmetry between investors is high - without short-sales constraints, prices will reflect all available information, and uninformed investors know that private information of the informed investors is priced into the market. However, if restrictions make it more difficult to short a stock, uninformed investors will know that not all information is priced in the market; it only reflects the positive news. This decrease in informativeness of the market prices increases the uncertainty and risk, lowering demand and consequently the price of the stock - i.e. the informational effect. In all, less information therefore contributes to a decline instead. Here the informational effect outweighs the allocational effect that could have caused overvaluation. This is an important finding, because it specifically aims at declining markets, where a high volatility and a sudden increase in uncertainty can cause stock markets to crash.

information about a stock and short-sales restrictions allows them only to have positive quantities of stock, the negative information does not become immediately clear in the price of the stock. Because the negative information is not absorbed in the price yet, some information might be hidden, since it is not revealed through short positions. Thus when a market is declining and no buying demand for the stock appears, the signal is given that the hidden information is more negative then expected. The decline is amplified, because the information is only revealed when the stock prices are falling. This again is in line with the above: crashes and large declines are not so much caused by significant news events but by a large increase in uncertainty and hidden information that is revealed in a declining market. In both cases negative information was not translated into the price, because of the presence of short-sales restrictions.

The research focus on the financial crisis and the contemporaneous stock market crash makes both studies by Hong and Stein (2003), and Bai et al. (2006), on information asymmetry particularly relevant. This however, would mean that the short-sales bans had adverse effects, instead of providing the necessary support for the stock prices.

In a situation without short-sale bans, the stock price would reflect all available information, thus the price is in that sense efficient, as it is fully informative to all investors. However, in a situation with a short-sale ban this does no longer hold, because the negative information is not priced in the market, resulting in information asymmetry. This information asymmetry causes a decline in prices of the banned stocks, leading to negative abnormal returns of those stocks. Using the informational effect of short-sale restrictions in our framework, we have therefore arrived at our first hypothesis:

Hypothesis 1

A short-sale restriction will cause negative abnormal returns of the restricted stocks during a declining stock market

2.1.2 Option status

possibility to take a synthetic short position. This gives rise to hypothesis 2a.

Hypothesis 2a

Effects of a short-sale restriction on abnormal returns are moderated by the possibility to take a synthetic short position.

2.1.3 Firm size

There is reason to expect that short-selling restrictions have differing effects for firms with varying market values. As such, size effect might come into play (Dimson & Marsh, 1986). These suspicions stem from previous research that has shown that stocks with smaller market capitalization levels tend to be less desirable for market makers. In an investigation on the market for stock borrowing, D
Avolio (2002) points out that stock with small market values are generally not shorted at all.

Furthermore, market makers such as institutional investors are the main suppliers of lendable stock. This could mean a bias towards stocks that are generally held in those investors
portfolios. Again, those securities tend to be large cap stocks (D
Avolio, 2002). Having said that, firm size could exhibit a moderating effect on the relationship between short-sale restrictions and abnormal returns. Incorporating the firm size aspect as a moderating effect

between short-sale restrictions and abnormal returns gives us hypothesis 2b.

Hypothesis 2b

Effects of a short-sale restriction on abnormal returns are moderated by firm size.

2.1.4 Volatility

Next to these control variables a third factor is expected to influence the abnormal returns, the volatility of a stock. However instead of providing another exogenous control variable, the volatility is also influenced by the shorting ban. Ho (1996) finds evidence that the volatility of stock returns is larger for stocks with a short-selling ban, which is supported by Boehmer, Zhang and Jones (2008) in their research on the US stock market.

Hypothesis 3a

A short-sale restriction will cause larger volatility in returns of the restricted stocks during a declining stock market

Hypothesis 3b

Restricted stocks with a larger volatility will have larger negative abnormal returns than those with lower volatility and no restriction, during a declining stock market.

2.2 Context

As previous sections described the research field of short-selling, the hypotheses are principally based on the broad academic context of short-selling. This paper also aims at providing a specific insight in the past financial crisis, projecting the already known in an empirical yet contemporary setting.

In the past financial crisis, regulators did not act uniformly - some imposed bans or constraints on short-selling, whereas other did not. For those who did, the aim was to protect the stocks, which were mostly financials, from excessive shorting in anticipation to asset depreciations and even possible bankruptcies. The existing literature shows similarly divergent conclusions. Combined with the 2004 article by Lamont on the battle between short-sellers and firms, and a study of a class action lawsuit of general investors versus short-sellers by Apfel et. al. (2001), the effects of short-sales can be rather controversial. This could explain the divergence of regulations on short-sales on various stock markets. In further support of short-sale restrictions,

Haruvy and Nousair (2006) set out that short-selling regulation might have strong implications for the formation of bubbles. They argue that short-sales influence the supply and demand, by increasing the supply when prices fall. In support of relieving short-sale restrictions, Endo and Rhee (2006) set out that two forces for improved demand during a declining might arise due to short-selling. The first originating from buying back shares by the short sellers covering their short positions and the second caused by the reinvestment of the profits of short sellers.

Research on US stock markets by Boehmer, Jones and Zhang (2009) reveals that short-sales during the ban from September till early October, did not evaporate to zero, but remained at 1,96%, where non banned stocks have 12.46% of shorted stocks to total. Several market makers in the US were exempted from the ban, so they could keep hedging their positions. Despite that, the shorting ban led to a significant large decrease in the amount of stock sold short.

short-sales decreased the bid-ask spread. As the bid-ask spread increases, liquidity decreases as trading on negative information becomes more costly.

In line with the model of Diamond and Verrecchia (1987) and other previous research by Bris, Goetzman and Zhu (2007), Saffi and Sigurdsson (Sigurdsson & Caffi, 2007), and Boehme and Wu (2007), short-selling bans influence the amount of information encapsulated in share prices due to the exclusion of trade decisions based on negative information. Beber and Pagano (2009) use the data from January 2008 till June 2009, hence widely including the crash and the subsequent recovery of the stock market indices. The findings of the crisis do not only support the previous literature, but also find that the price discovery is significantly smaller when the markets were in decline.

Perhaps most relevant to our research is their third finding: instead of supporting market prices during the crash, abnormal returns in most markets (except for the US) were negative as compared to an equally-weighted market index in decline. Evidence therefore suggests the short-sales ban worked adverse. Other effects were found for the US, where the stocks affected by the ban had cumulative positive returns as opposed to their control group, the non-banned stocks (Boehmer, Jones, & Zhang,

2009). Despite these positive returns in the US market, the loss of liquidity, unintended market consequences, and the lack of support of the ban to stock prices made SEC
s chairman say that with the current knowledge, “the [SEC] would not do it again” (Younglai, 2008).

3. Methodology

3.1 Event definition

the financial crisis are yet to fade off, and only few regulators have lifted bans.

3.2 Time period

This study focuses on trade regulations regarding short-sale restrictions starting September 15th 2008 (from various newspaper sources it is the observable start of the perils of the financial meltdown as Lehman Brothers Holdings Inc filed for bankruptcy (Wilchins, 2008), and up to the point of world market recovery in financial markets on March 2009. Figure 1 at the top of this page depicts the timeline of this study. The timeline of events and specific list of financials that result from this setup provide for a unique sample that is suited for testing the effects of short-sale restrictions during a period of economic downturn.

3.3 Model

The framework we utilize in this study compares the actual return (ex post) of a portfolio or individual security to the normal return in order to find abnormal returns. The normal return is specified as

the return of the security had the event not taken place. In order to find the normal return, conditioning information is included into the equation. Two types of models can be used for conditioning expected normal returns (Wong, 2002). The first, the constant mean return model, assumes a constant return of the security. The model we will employ, however, is the market model, which relates the return of a security to the return of the market portfolio (MacKinlay, 1997). The use of the market model with the given data sample of financials is likely to give more meaningful results (Dyckman, Philbrick, & Stephan, 1984), as it will incorporate the overall effect of the financial crisis during the time window of this research. In other event study-based research, we find similar ways of approaches when the timeline involves extreme market-wide trends (Netter & Mitchell, 1989) and (Kho, Lee, & Stulz, 2000).

3.3.1 Market Model

Abnormal returns are measured using the equation below, simply stating that the abnormal returns (AR) for security i at time t are equal to the actual return of security i, minus the expected return. Xt

depicts conditioning information, excluding the event in question.

AR_it = Rit
E R

[

]

E[e

] = 0

Var[e

] =

In the market model, normal returns are measured with the next equation:

(2) Using the equation for normal returns (2), the abnormal returns equation (1) can be rewritten as:

(3)

Where Rit is the return of the security on

period t and is equal to the expected return of the security and the abnormal returns for the security. The expected return is denoted by return on the market, Rmt and the parameters
and ,

and AR denotes the excess abnormal return. We follow the methodology of MacKinlay (1997) in our estimation of the market model. The returns in the market model can be estimated with the observations from the estimation window, expressed as:

(4.1)

These can be found with matrices, where is a vector of

estimation-window returns and an

matrix with a vector of ones in the first column and the vector of market return observations

in the second column and is

the parameter vector. The

parameters can then be estimated using ordinary least squares. Formally using matrix algebra, the parameter estimation is shown in 4.2 to 4.5 below. In 4.2, the parameter vector that contains the alphas and betas is found my multiplying

X

is then multiplied with the transpose of

X

and the vector with estimation-window returns. 4.3 and 4.5 depict the standard deviation and variance of respectively the abnormal returns, and market model parameters. The calculation for abnormal returns is then carried out in 4.4 by subtracting the market model returns (

X

R

(4.2)

(4.3)

(4.4)

(4.5)

in the market, hence making it well qualified in measuring the returns in the financial crisis. Because the market returns are estimated by implementation in matrices, after which the abnormal returns can be calculated, the above means:

(3)

Rearranging by moving AR to the left and Rit to the right, we arrive at (6) which

equals the computation of abnormal returns using matrices (7).

(6)

(7)

Aggregation of the abnormal returns is needed to make conclusions about the event, both in time and across the securities. Data clustering due to event date overlap is a typical problem in event studies (Cambell, Lo, & MacKinlay, 1997) (Lee & Varela, 1997), and may cause difficulties in statistical analysis (Brown & Warner, 1985). The design of the time window for this study, however, should prevent such extreme correlations due to clustering. The event dates are rather spread throughout the time window of the financial crisis and should not cause concern for extreme clustering of events.

3.3.2 Volatility measure

The volatility is measured with the intraday measure. First, this is supported by the methodology of Boehmer, Zhang and Jones (2009) who also use the intraday measure. Second, other measures of the volatility such as 20 or 30 day standard deviations are too long for our 106-day period to be effective. Hence, the intraday measure is the most suitable one.

For each trading day, intraday high and intraday low are used. The volatility is then calculated by subtracting the intraday low from the intraday high, resulting in the daily stock price movement. Standardizing these with the daily arithmetic average (Jacquier, Kane, & Marcus, 2003) results in a measure that can be compared across companies.

3.4 Sample design

that show more trading activity, and will therefore have a smaller probably of non-trading. Information on short-selling is gathered from the financial authorities and central banks in respective countries, which combined with releases from Reuters and Data Explorers, gives a detailed picture of the measures in each country. Three countries in Eastern Europe – Poland, Slovenia and the Czech Republic – are removed because short-selling is not practiced. This results into an initial selection of 13 countries.

The criteria for inclusion of a firm in the sample follow the stock exchanges categorization: a list of financials is compiled by including stocks that are labelled as
financial by the operator of the stock exchange and/or regulator. We filter for firms in banking, life and non-life insurance. Large, mid- and small-cap stocks are all included, as long as they are categorized as
financial, however, only firms that have had stocks publicly available for trading throughout the entire event window are included. Possible reasons for missing trade data could be an IPO, withdrawal due to bankruptcy or non-trading. We investigate local news sources for such causes, should such exclusion occur. Shares that exhibit non-trading over most of the estimation and event period are removed from the sample, to avoid a non-trading bias.

Publicly traded investment funds such as those common in Sweden are removed, because they are used for private investments instead of providing financial services.

Having constructed a set of firms, we proceed in collecting daily trade for each stock. Trade data is retrieved from Datastream (Thomson Reuters). The index used in the market model is the MSCI Barra country index for each country. This index is available for all countries, except for Luxembourg. Unavailability of an alternative index equal to the method used by MSCI Barra, has led to excluding Luxembourg. The final sample consists of 117 companies in 12 countries (Table 2, Appendix).

for testing the effects of short-sale restrictions during a period of economic downturn.

3.5 Statistical tests

3.5.1 Paired sample T-test

We conduct tests over different post-event periods in order to distinguish between an immediate and overall impact on stock returns in a longer time frame (i.e. during the period of economic downturn). Refer to Figure 1 on page 16 for a graphical overview of the timeline and its segments. Immediate effects are investigated by comparing the 7-day comparison period to the 7-day post-event time period. A paired sample T-test is employed to T-test for these possible immediate effects of the shorting ban.

The event window starts one day before the event and ends one day after the event, effectively ranging from t-1 to t+1.

Using information from announcements made by financial regulators, we are able to pinpoint the exact event dates. In order to allow for possible information leakages or lagging effects, a three-day event window is deemed appropriate. The comparison period covers the 7 days prior to t-1 over which an average

cumulative abnormal return (aCAR) is computed. The T-test is used to test the null hypothesis that there are no

significant differences between the aCAR and the CARs of each of the seven days after the event.

3.5.2 Non-parametric test

Although the T-test should provide robust results even in case of slight non-normality of the sample (Henderson Jr, 1990), we conduct a non-parametric sign-ranked test in order to detect any case of unacceptable levels of non-normality. Again we proceed in testing the null hypothesis that there is no significant difference between the abnormal returns at t-1 and t+1 for the

returns of stocks that face a shorting ban. Due to the possibility of clustering in the sample, and possible non-normality of the distribution due to overlapping event dates, we opt for a Wilcoxon sign-ranked test. Said test is deemed appropriate over a simple signed rank test in that it incorporates information about the magnitude of difference, whereas the simple signed rank test does not. The abnormal returns found by computation are prepared such that matched pairs of dates can be formed for each of every security (i.e. the matched pair for testing H0 for security X would be

the abnormal returns at t-1 and at t1).

statistic T that is to be tested against the Wilcoxon statistic T for large samples2:

(8)

Where the variance of T equals:

(9)

Since we test for a difference in abnormal returns, without specifying any direction (i.e. negative or positive), we use the two-sided decision rule to reject the null hypothesis if:

(9)

where the significance level is
.

3.5.3 OLS regression

After having tested for immediate effects, we carry on with investigating effects over the period of decline. In order to investigate the interaction between abnormal returns and shorting bans, while taking into account the different types of bans and firm-specific characteristics, we employ regression analysis. Using OLS we test the 106-day post-event window for effects during the financial crisis.3 The dependent variable is the abnormal stock return per firm cumulated over the post-event window

2

3 _{The number of days between the last event and}

the day the overall market index decline ended is 106

as computed with the market model. The independent variables in the model are the dummies for bans, option status and firm size. In order to distinguish between the different types of short-bans, dummy variables are assigned to each observation. The dummy for a naked ban equals 1 if naked short-sales are restrained, but covered short-sales are still possible, and equals 0 otherwise. The dummy variable for a covered ban equals 1 if neither naked nor covered short bans are possible, and is 0 otherwise.

Each stock receives a dummy variable for its option status. Unity is assigned to stocks with traded options and zero otherwise. Option status is therefore the second independent variable in our model.

Firm size in terms of market value is accounted for by grouping our sample into quartiles and assigning labels accordingly (i.e.
1,
2,
3, or
4). To test whether firms with high market value experience different effects from a shorting ban than small market value firms, dummy variables for association with the fourth quartile are assigned. Firms that are in the fourth quartile receive a dummy variable of 1, all others 0.

estimated. The first equation is depicted as follows:

CARt106 = a + b1NAKED BAN + b2OPTION

+ b3SIZE (10)

where the dependent variable CART106

depicts cumulative abnormal returns over the 106-day post-event window; NAKED BAN is the dummy variable for a naked ban; OPTION is the dummy variable for option status; SIZE is the dummy variable for inclusion in the fourth quartile; and
1,
2, and
3 are the

standardized corresponding coefficients.

The OLS regression is repeated over a similar equation for testing the effects of a covered ban:

CARt106 = a + b1COVERED BAN + b2

OPTION + b3SIZE (11)

Here, the first dependent variable, COVERED BAN and its corresponding coefficient represent the dummy variable for a covered ban. With the first proposed relationship in mind – shorting bans cause negative abnormal returns – we are particularly interested in the coefficient of the first independent variable, the dummy for a short-sale ban (be it naked or covered). The second independent variable will help in explaining the second proposed relationship – option status as a moderating variable. Finally, the third independent variable will be used in

testing for any firm size moderating effects. More specifically, it will tell us whether firms with high market value experience different effects from a shorting ban.

The impact of volatility is measured using linear regression, where both hypotheses will be tested separately. The hypothesis 3a) will be regressed using the following equation for a naked short-selling ban:

VOL T106 = a + b1NAKED BAN (12)

where the dependent variable VOLT106

depicts cumulative abnormal returns over the 106-day post-event window; NAKED BAN is the dummy variable for a naked ban; and
1 is the standardized

corresponding coefficients. The covered shorting ban is measured with the same equation, but instead the covered ban is the independent variable:

VOL T106 = a + b1COVERED BAN (13)

Hypothesis 3b) can only be tested if the results of the previous tests are significant, that is, if there is a relation between the bans and the volatility. The second hypothesis is then calculated using equation 14 for the naked ban and equation 15 for the covered ban

CAR T106 [E COVERED]= a + VOL T106 (15) 4. Results

The impact of the shorting ban is tested over the short and longer time periods. The short period is used to assess the immediate impact of the announcement of the shorting ban, whereas the longer period is used to compare the overall impact of the shorting ban of stock performance during financial downturn. We first offer the results of the tests for the immediate impact and will then proceed with the results using the longer time frame.

4.1 Abnormal returns

The market model is used for the calculation of the abnormal returns. A 400-day time window prior to the event is employed to estimate the normal returns for each stock. Because the comparison window for the first test begins at t-8, the estimation window ranges from 9 to t-409 for each stock. The resulting alpha and beta are then used to calculate the expected normal returns in the event windows. Subtracting the normal returns from these raw returns then gives the residuals, or abnormal returns.

4.2 Normality

computation of abnormal returns, logarithmic raw returns, and logarithmic abnormal returns are calculated. Logarithmic returns exhibit larger normality and are therefore frequently used in stock return studies (Tsay, 2005). Normality is formally tested using the Kolmogorov-Smirnov (K-S) statistical test.

The results of the K-S test in Table 3 for the 17-day sample show some dispersed results. Most values are not significant at the 5 percent level and hence normal distributed, however the important values of t+1 and t+2 are significant, and negatively skewed and leptokurtic (i.e. it has a greater kurtosis than the normal distribution). Because these values are high, some levels could be unacceptable. Therefore it cannot be confirmed that the observations for all days are significantly normally distributed. Both parametric and non-parametric tests are performed to test the sample. The latter is a paired sample T-test that is rather robust. Slightly decreased normality should therefore not pose any serious problems.

The normality of the 106-day sample is confirmed with a K-S test, which is not significant at the 5 percent level, although the data is slightly negatively skewed and leptokurtic, which is

expected, because some companies experienced large stock price losses.

4.3 Regression

The test results from the paired sample T-test suggest that the event that takes place between t-1 and t+1 leads to a

change in abnormal returns. As can be seen in Table 4 in the Appendix, the difference between average abnormal returns during the comparison window and the returns on the day of the ban and the two consecutive thereafter are significant at the 5 percent level for stocks that are subject to a naked ban.

The notion that the market could react to rumours before the event date – that is, on t-1 - is not confirmed by the test

results. However, since the impact of a ban on abnormal returns during the 3-day event window is significant, the null hypothesis that abnormal returns are not different after a shorting ban can be rejected. The results from the T-test are confirmed when conducting the non-parametric test over the same sample. A graphical representation of this period is depicted in Figure 2 in the Appendix. The graph shows a reaction on and after the event date.

has yet to be described. Therefore the entire sample is used, which includes all companies and countries. This allows the comparison between companies that were impacted by the ban and those that were not. Although the collinearity diagnostics for the model show little sign of interaction between the ban and option availability, another variable is computed combining both. Simultaneously measuring both could increase the explanation of the model, hence:

NAKED_NOOPTION = NAKED_BAN x

NOOPTION (13)

COVERED_NOOPTION = COVERED_BAN

x NOOPTION (14)

Reversing the OPTION variable into NOOPTION allows the combination with the NAKED BAN and COVERED BAN, otherwise the 0 and 1 would have ambiguous meanings. The regression models as stated and explained in the methodology are:

CARt106 = a + b1NAKED BAN + b2OPTION

+ b3NAKED_NOOPTION (15)

CARt106 = a + b1COVERED BAN +

b2OPTION + b3COVERED_NOOPTION

(16)

The coefficients are estimated using multiple linear regression analysis. At first, inclusion of all variables does not seem to yield significant values for each coefficient in both the naked and covered ban equations. To exclude variables which have no impact or overlap with others, stepwise regression is used, of which the results are shown in Tables 5 and 6 in the Appendix. Again, these are depicted in Figure 3 in the Appendix, where stocks with a shorting ban show negative abnormal returns.

4.3.1 Naked ban

Inserting the unstandardized b-coefficients from Table 4 in the regression model for the naked ban, yields:

CAR106 = -0,254 -0,307 NAKED BAN +

0,281NOOPTION

4.3.2 Covered ban

Results for the test on covered bans are presented in Table 5 of the Appendix. When the unstandardized b-coefficients are inserted in the regression model for the covered ban, the regression is: CAR106 = -0,473 + 0,464NOOPTION-

0,289 COVERED_NOOPTION

This differs from the regression for the naked ban, because the combined dummy of covered ban and option availability is included (significant at 5 percent) whereas the covered ban is not. Again, the option availability is found to have a significant impact on the regression results, at 5 percent.

4.3.3 Volatility

The impact of the shorting ban on stock return volatility is tested with linear regression, of which the results are shown in Table 6 of the Appendix.

As can be seen, the influence of the covered ban has no influence of a significant level. However, there is an influence of the naked ban. Although the model is limited, because of the number of variables, there is a significant influence of the naked ban on volatility. When the unstandardized b-coefficients are inserted in the regression model, this yields:

VOL T106 = 0,045 + 0,015 NAKED BAN

Because the covered ban has no influence on volatility, the impact of volatility on stock returns is only tested for the naked ban. The results are shown in Tables 7 and 8 of the Appendix.

5. Analysis

5.1.1 Research question

Before the instatement of a naked ban, data on naked shorting activity is largely unavailable. As soon as the ban is instated, however, all investors know that there is zero naked short-selling activity. This loss of information causes even more severe uncertainty for less informed investors and might trigger relatively more sell activity. The result is a decline in stock returns that is more severe than it would be in the case of a covered ban. The negative effect of a naked shorting ban becomes even more pronounced when taking into account the availability of a traded option: stocks without traded options exhibit larger negative abnormal returns than stocks with traded options.

Relating our finding to the theoretical framework, we find congruence with the informational theory of stock prices (Senchack Jr & Starks, 1993); removing the possibility of shorting deteriorates the information encapsulated by the share price as can be witnessed by an increase in volatility and abnormal returns.

5.1.2 Hypothesis 1

The first hypothesis is confirmed with our data; a short-sale restriction does indeed cause negative abnormal returns of restricted stocks during a bearish market. In a market that is showing signs of it entering a period of strong decline,

or even a crash, share prices do not benefit from being protected from short sellers.

In fact, Bai Chang and Wang (2006) are supported in that information asymmetry causes short-sale bans to have an adverse effect on share prices. Rather than supporting prices, restrictions on short-selling remove a significant informational feature of the price, and lead to higher uncertainty.

The significant difference between the abnormal returns during the comparison window (the 7 days leading up to the event) and the abnormal returns on the day of the event and the two days afterwards, imply that the effects of a shorting ban are immediately reflected in share prices: they exhibit stronger negative abnormal returns.

This finding is consistent with Diamond and Verrecchia (1987) in that short-sale restrictions do not uniformly lead to overvaluation.

5.1.3 Hypothesis 2

inclusion of the interaction variable (Table 4). Including option status in the test for covered short-sale restrictions, however, leads to stronger effects of the moderating variable (Table 5). Option status therefore becomes particularly important from the moment options are the only possibility to achieve a short position, which is true in case of a covered shorting restriction. With regards to short sellers, this finding implies that the availability of traded options seems to play a significant role only when neither naked nor covered short positions are possible. An economic reason could be that synthetic short positions are less desirable compared to naked or covered short positions due to transaction costs, and superior flexibility of naked and covered short transactions. The findings of Danielsen and Sorescu (2009) imply that stocks with traded options are less affected by short-sale restrictions. Our findings confirm that this is indeed the case for stocks that have a restriction for covered short-sales. No such evidence was found for the interaction term for naked bans and option status. In that respect the inclusion of option status is a valuable addition to the research of the impact of the financial crisis. Exclusion of the variable was a limitation of previous research by Boehmer, Zhang and Jones (2009).

5.1.4 Hypothesis 2b

The impact of firm size on the shorting bans is more difficult to assess. Dividing the sample in quartiles according to their size immediately shows that most companies in the upper quartile have option availability, whereas the other quartiles have not. As option status is strongly related to firm size, testing all quartiles separately is troublesome. Furthermore, division into quartile renders sample sizes too small. Hypothesis 2b, the hypothesis that firm size is a moderating variable in measuring the effects of short-sale restrictions on abnormal returns is therefore neither accepted nor rejected.

5.1.5 Hypotheses 3a and 3b

Another impact of the shorting ban is measured in the volatility of stock returns. There is a significant relation between the naked ban and volatility, where stocks with a naked ban have 19,8 percent higher volatility than non-banned stocks (Table 8), but no significant relation between the covered ban and volatility. Thus, hypothesis 3a is confirmed for the naked ban. Again, our finding is contradictory to the aim of the financial authorities. Instead of decreasing volatility and heavy movements in the stock prices, the naked ban had an adverse effect.

The impact of volatility on the returns of restricted stocks is also found for naked stocks. An obvious explanation for the relationship could be the risk-return relationship, where a higher return is expected by investors for a stock with larger risk, measured as the volatility. The negative coefficient is pointing to another situation. Although most stocks were already in decline, restricted stocks that displayed large volatility were also the stocks that had the most negative cumulative abnormal returns. Hence, the impact of the ban was the strongest for stocks with large volatility.

A conceivable explanation is the loss of information due to the exclusion of naked short sellers in the market, as explained in Bai, Chang and Wang (2006). The increase of uncertainty about the information available might cause larger movements in stock returns. The market was in decline and stocks were already volatile with large amounts of information and new events emerging. In such crashing stock markets and turbulent economic developments it is more difficult to assess whether all information is available or which information is missing, because it is not traded upon by naked short sellers. Hence, intraday volatility becomes larger when the investors become more uncertain about

the future and when this uncertainty peaks during the day.

5.2 Implications

The implications of our findings can be explained on different scales – there are direct implications for short-sellers, the general investing public and investees, regulatory bodies, and finally society as a whole.

The direct implications of short-sale restrictions are clear-cut for short-sellers. A restriction will cause a short-seller to have to cease activities, or resort to other means, such as using options. A more indirect implication could be the stigma created by media coverage that is mainly negative. Firms could pick up on this
fear, and might try to prevent investors from shorting their stock. Lamont (2004) explains a number of ways in which firms may raise the difficulty of shorting.

Little is known about the reaction of firms to short-selling behavior. Although Lamont (2004) sets out ways in which firms may react, little empirical evidence is available. Having said that, it is not unlikely for firms to be concerned about share value being deteriorated by market regulations, and for them to take action on it. On the other hand, increased negative abnormal returns due to short-sale restrictions, as shown in the empirical evidence, are neither in the best interest of firms.

Although the motives for introducing short-selling bans are increased stability or prevention of market crashes, the empirical evidence from our sample shows contrasting results. Finally, ethical considerations of short-selling and the consequences of its effects on a greater scale can be pointed out. Although speculation is not the only motive for selling stock short, it is the motive that gains the lions share of attention in newspaper articles and governmental publications. This is not without reason, since the profits of short-sellers indeed depend on the losses of firms. The main problem in most arguments against short-selling is, however, the implicit reasoning that says short-selling is the actual cause of the downfall. Short-selling might put a downward pressure on prices, but for it to be the cause of a bankruptcy remains debatable. Similarly,

one can argue against short-selling restrictions as they often stir up the market mechanism in that they pose as reasons for more uncertainty. Indeed, investors might think along the lines of:
if the government thinks certain stocks need protection, then surely there must be something wrong with them?

5.3 Limitations

Although the design of our research stems from confidence and ongoing discoveries in the subject of short-selling, there are limitations to our research as well.

stock markets in the world, because situations can be different. As the past financial crisis showed, the effect of the shorting bans was different from Europe and the United States (Boehmer, Jones, & Zhang, 2009). But then this was beyond the scope of our research.

A first limitation to our research is a consequence of a lack of data. Our study would have benefitted from a larger and more extensive database, however to our regret this was not available. The inclusion of more control variables, such as institutional ownership and cross-listings, would have made the results stronger and would have given more precise results. Because these variables are a worthwhile contribution to future research we describe them in section 5.4.

Another variable that would have made a valuable contribution is the amount of short-sales. This is related to the implications for the governmental bodies and regulators in section 5.2. Because we have no data on the amount of short-selling, we do not know the difference before and after the ban. Short-selling is practiced in all countries in our sample, as mentioned in section 3.4. But whether the ban was necessary in the first place cannot be researched, there could have been no short-selling in some companies

before the ban. Although this does not directly alter the results of our hypothesis, it is an interesting question that gets more weight because of our results. Our results that the shorting bans are associated with negative returns could make the insertion of the ban even more controversial or doubtful then it was, maybe it was not even necessary at all.

The opposite can be said as well. Although we found significant negative abnormal returns the ban could have supported the stock price of some companies. Researching every single case in our sample is an enormous task, requires far more data and time, and was not our research topic. Still, some companies under great pressure from short sellers could have benefited, or even prevented from going bankrupt, from the ban. We do acknowledge the fact that the market crash was as strong as it was unusual, and the insertion of the shorting ban was in some cases an act of last resort for regulators and governments that time. However, with hindsight their decision might have been different.

5.4 Future research

5.4.1 Institutional ownership

short a stock. This supply of stocks intended for loans is mainly provided by institutional investors (D
Avolio, 2002). As a result, the degree to which a stock is owned by institutional investors, i.e. the breadth of ownership of a share, is likely to moderate the possible effects of a short-sale constraint. One would expect stocks with significant institutional ownership to be more readily available for shorting than stocks with mainly individual owners. Nagel (2005) is indeed able to explain abnormal returns with the use of institutional ownership as a proxy. D
Avolio (2002) similarly concludes that institutional ownership is the main determinant for the loan supply of stock. Due to unavailability of information, the moderating effects of institutional ownership are a subject for future research.

5.4.2 Cross-listing

A cross-listed stock that is constrained on home market X, might not be constrained on market Y. A ban on the home market affects liquidity on both home and foreign market; vice versa however, a ban on the foreign market might not influence liquidity on the home market (Beber & Pagano, 2009). For this reason, the occurrence of cross-listing could play a role in curbing the effects of a short-sale constraint, given that this constraint is only of effect in one of the

countries in which a stock is listed. Due to sample size limitations, the existence of cross-listing effects is a subject for further research. It would most likely require an investigation into market-wide short-selling bans across all types of firms, in order to obtain usable sample sizes.

6. Conclusion

Bibliography

Aima. (2007). An Overview of Short Stock Selling. Aima Canada.

Aitken, M., Frino, A., McCorry, M., & Swan, P. (1998). Short sales are almost

instantaneously bad news: Evidence from the Australian Stock Exchange. Journal of

Finance , 2205-2223.

Alexander, G., & Peterson, M. (1999). Short Selling on the New York Stock Exchange and the Effects of the Uptick Rule. Journal of Financial Intermediation , 8 (1-2), 90-116. Allen, F., & Gale, D. (1991). Arbitrage, short sales, and financial innovation. Journal of the

Econometric Society .

Apfel, R., Parsons, J., Schwert, G., & Stewart, G. (2001). Short Sales, Damages and Class Certification in 10b-5 Actions. NBER Working Paper .

Autoriteit Financiële Markten. (2008). Jaarverslag 2008. The Hague: DeltaHage. Bai, Y., Chang, E., & Wang, J. (2006). Asset prices under short-sale constraints.

University of Hong Kong and MIT Working Paper .

Beber, A., & Pagano, M. (2009). Short-Selling Bans around the World: Evidence from the 2007-09 Crisis. CEPR Discussion Papers .

Binder, J. (1998). The event study methodology since 1969. Review of Quantitative

Finance and Accounting , 11 (2), 111-137.

Boehme, R., Danielsen, B., & Sorescu, S. (2009). Short-sale constraints, differences of opinion, and overvaluation. Journal of Financial and Quantitative Analysis , 41 (02), 455-487.

Boehmer, E., & Wu, J. (2007). Short Selling and the Price Discovery Process. Working

Paper Series .

Boehmer, E., Jones, C., & Zhang, X. (2009). Shackling short sellers: The 2008 shorting ban. Johnson School Research Paper Series , 34 (9).

Bris, A., Goetzmann, W., & Zhu, N. (2007). Efficiency and the Bear: Short Sales and Markets Around the World. The Journal of Finance , 62 (3), 1029-1079.

Brown, S., & Warner, J. (1985). Using daily stock returns. Journal of financial economics ,

14 (1), 3-31.

Cambell, J., Lo, A., & MacKinlay, C. (1997). Event Study Analysis. In The Econometrics

of Financial Markets (2nd Edition ed., pp. 149-180). Princeton: Princeton University

Press.

Chang, E., Yu, Y., & Complex, M. (2004). Short-sales constraints and price discovery: evidence from the Hong Kong market. papers.ssrn.com .

Charoenrook, A., & Daouk, H. (2005). A Study of Market-wide Short-selling restrictions.

SSRN .

Christophe, S., Ferri, M., & Angel, J. (2004). Short-selling prior to earnings announcements. Journal of Finance , 59 (4), 1845-1875.

Dagens Industri. (2008, 23 September). Inget blankningsförbud i Sverige. Retrieved 2010 
 1-March from Dagens Industri:

http://di.se/Default.aspx?pid=50690__ArticlePageProvider

Danielsen, B., & Sorescu, S. (2009). Why do option introductions depress stock prices? A study of diminishing short sale constraints. Journal of Financial and Quantitative Analysis

, 36 (04), 451-484.

DAvolio, G. (2002). The market for borrowing stock. Journal of Financial Economics , 66 (2-3), 271-306.

Dimson, E., & Marsh, P. (1986). Event study methodologies and the size effect: The case of UK press recommendations. Journal of Financial Economics , 17 (1), 113-142.

Dyckman, T., Philbrick, D., & Stephan, J. (1984). A comparison of event study

methodologies using daily stock returns: A simulation approach. Journal of Accounting

Research , 1-30.

Endo, T., & Rhee, S. (2006). Purchases and Short Sales in Emerging Markets: Their Rationales and Design Variables. World Bank Financial Sector Paper Series . Fama, E. (1991). Efficient capital markets: II. Journal of Finance , 46 (5), 1575-1617. Figlewski, S. (1981). The informational effects of restrictions on short sales: some empirical evidence. Journal of Financial and Quantitative Analysis , 16 (4), 1-15. Forden, S. G. (2009, 30 July). Italian Regulator Lifts Short-Selling Ban on Most

Companies. Retrieved 2010, 1 March from Bloomberg:

http://www.bloomberg.com/apps/news?pid=newsarchive&sid=a9ivw6fvfb.A

Glover, J. (2010, 20 May). Borrowing Costs Rise as German Ban Saps Confidence . Retrieved 2010, 24 May from Bloomberg:

http://www.bloomberg.com/apps/news?pid=newsarchive&sid=aGt9NTtcx8mc Haruvy, E., & Noussair, C. (2006, 1 Jan). The effect of short selling on bubbles and crashes in experimental spot asset markets. The Journal of Finance .

Hedval, L. (2008a, 24 September). FI monitors short selling of shares. Retrieved 2010 
 1-March from Finansinspektionen: http://www.fi.se/Templates/Page____10667.aspx Hedval, L. (2008b, 14 October). Investigation of share loans and price trends. From Finansinspektionen: http://www.fi.se/Templates/Page____10799.aspx

Henderson Jr, G. (1990). Problems and solutions in conducting event studies. Journal of

Risk and Insurance , 57 (2), 282-306.

Ho, K. (1996). Short-sales restrictions and volatility The case of the Stock Exchange of Singapore. Pacific-Basin Finance Journal .

Hong, H. (2003). Differences of Opinion, Short-Sales Constraints, and Market Crashes.

Review of Financial Studies , 16 (2), 487-525.

Jacquier, E., Kane, A., & Marcus, A. (2003). Geometric or arithmetic mean: a reconsideration. Financial Analysts Journal , 59 (6), 46-53.

Jaffe, J. (1974). Special information and insider trading. Journal of Business , 47 (3), 410-428.

Jones, C., & Lamont, O. (2002). Short-sale constraints and stock returns. Journal of

Financial Economics .

Kho, B., Lee, D., & Stulz, R. (2000). US banks, crises, and bailouts: From Mexico to LTCM. American Economic Review , 90 (2), 28-31.

Kirschbaum, E. (2010, 18 May). Germany announces short-selling ban. Retrieved 2010 
 18-May from Reuters: http://www.reuters.com/article/idUSBAT00546720100518 Lamont, O. (2004, 1 Jan). Go Down Fighting: Short Sellers vs. Firms. NBER Working

Paper .

Lee, S., & Varela, O. (1997). An Investigation of Event Study Methodologies with Clustered Events and Event Day Uncertainty. Review of Quantitative Finance and

Accounting , 8 (3), 211-228.

MacKinlay, A. (1997). Event studies in economics and finance. Journal of Economic

Literature , 35 (1), 13-39.

Miller, E. (1977). Risk, uncertainty, and divergence of opinion. Journal of Finance , 32 (4), 1151-1168.

Netter, J., & Mitchell, M. (1989). Stock-repurchase announcements and insider

transactions after the October 1987 stock market crash. Financial Management , 18 (3), 84-96.

Reuters. (2008 
 19-September). Swedish watchdog says not considering shorting

ban. Retrieved 2010, 06 February from Reuters:

http://www.reuters.com/article/idUSSAT00556520080919

Ross, S., Westerfield, R., Jaffe, J., & Jordan, B. (2008). Modern Financial Management (8th Edition ed.). New York: McGraw-Hill.

Senchack Jr, A., & Starks, L. (1993). Short-sale restrictions and market reaction to short-interest announcements. Journal of Financial and Quantitative Analysis , 28 (2), 177-194. Sigurdsson, P., & Caffi, P. (2007). Price Efficiency and Short Selling. Social Science

Research Network .

Strömqvist, M. (2009, 24 June). How does shortselling affect financial markets?

Economic Commentary , pp. 1-8.

Tsay, R. S. (2005). Analysis of financial time series (2nd Edition ed.). New Jersey: John Wiley and Sons.

White, S. (2009, 24 July). Japan extends stock short-selling curbs to Oct -FSA. Retrieved 2010, 1 March from Reuters: http://www.reuters.com/article/idUST314120090724

Wilchins, D. (2008, 15 September). Lehman files for bankruptcy, plans to sell units. Retrieved 2010, 05 February from Reuters:

http://www.reuters.com/article/idUSN0927996520080915

Wong, H. (2002). A comparison of event study methods for foreign firms listed on the US stock exchanges. Journal of International Accounting Research , 1 (75).

Younglai, R. (2008, 31 December). SEC chief has regrets over short-selling ban. Retrieved 2010, 20 February from Reuters:

Appendix

Table 1: Example of a short position

In this numerical example, the mechanism of the short-sale trade is illustrated using historical prices of Swedbank (SWED A) shares.

The short seller engages in a short position in Swedbank A shares on 1/1/2007 and closes out his position one year later. The investor effectively shorts 100,000 of Swedbank A shares at SEK 205.55 per share. After 1 year he closes out his position by buying back 100,000 shares at SEK 151.37 per share.

This table illustrates the mechanism of the trade. Collateral for the 100,000 shares is placed with the lender at the start of the period. Returns are calculated at the end of the period when the position is closed.

Percentages for the interest rebate, borrowing costs and shorting margin are based on industry averages (Aima, 2007).

Action Collateral (mSEK) Exit proceeds (mSEK) Return (mSEK) Comments 1. Short position: SWED A 20.56 SEK 15.14 5.42 100,000 shares shorted @ SEK 205.55

2. Short rebate 0.72 3.5% interest rebated

on placed collateral of mSEK 20.56

3. Cost of borrowing shares

(0.51) 0.25% paid over borrow value of initial stock loan 4. Dividend payment on short position (0.90) Dividends of SEK 9.00/share received 5. Margin Costs on short position (0.21) (20.56) x 140% short margin = (28.78) + 20.56

Net margin used = (8.22) x 2.5% rate

Total return 4.52

Table 2: Sample structure

Scope: “NK” denotes a ban on naked shorting. “CV” denotes a ban on covered shorting.

Country Scope Start of ban Number financials

Austria NK 22/09/2008 6 Belgium NK 22/09/2008 6 Denmark CV 13/10/2008 32 Finland - - 4 Greece CV 10/10/2008 16 Hungary - - 2 Ireland CV 19/09/2008 4 Netherlands NK 19/09/2008 7 Norway CV 08/10/2008 17 Portugal NK 22/09/2008 4 Spain CV 24/09/2008 12 Sweden - - 7 Total 117

Table 3: Results from K-S test

17-day window

Kolmogorov-Smirnov

Sig. Skewness SD Kurtosis SD

t-8 1,166 0,132 0,396 0,251 1,240 0,498 t-7 0,927 0,357 -0,188 0,251 1,817 0,498 t-6 1,096 0,181 0,429 0,251 2,410 0,498 t-5 1,050 0,220 -0,870 0,251 1,921 0,498 t-4 0,940 0,340 -1,019 0,251 4,512 0,498 t-3 0,892 0,404 -0,487 0,251 0,699 0,498 t-2 1,116 0,166 0,255 0,251 3,415 0,498 t-1 1,257 0,085 0,124 0,251 1,759 0,498 t0 1,092 0,184 1,636 0,251 5,667 0,498 t+1 1,454 0,29 -1,256 0,251 9,039 0,498 t+2 1,261 0,83 -1,342 0,251 8,056 0,498 t+3 0,848 0,469 -0,807 0,251 1,103 0,498 t+4 1,537 0,018 1,376 0,251 11,177 0,498 t+5 1,592 0,013 -1,902 0,251 5,235 0,498 t+6 1,628 0,010 -3,810 0,251 23,405 0,498 t+7 1,485 0,024 1,612 0,251 8,496 0,498 t+8 1,075 0,198 -0,576 0,251 3,552 0,498 AVt-8-2 0,815 0,519 -0,689 0,251 1,076 0,498 106-day window Kolmogorov-Smirnov

Sig. Skewness SD Kurtosis SD