University of Amsterdam, Amsterdam Business School MSc Business Economics, Finance track
Master Thesis
The effect of the 2008 naked short sale restriction on firm returns
Joost Reus 5933706
January 2014
Supervisor: Dhr. J.E. Ligterink
Abstract
This thesis looks at the effect of the Securities and Exchanges Commissions short sale
restriction of July 2008 on the returns of restricted stocks. An event study shows high positive abnormal returns after the announcement of the regulation. During the ban, stocks hardly outperformed the market and returns turn negative around the removal. These findings are in line with the overpricing hypothesis of Miller (1977). A difference-in-difference regression shows no significant difference in daily returns for restricted stocks during the restriction period, including the days after the announcement. Only the returns for restricted stocks are significantly higher in the days after the announcement. Regulators can use a restriction to calm down the market temporarily, but not as solution for downward pressure on the long run.
Table of contents
1. Introduction 4
2. Literature review 7
2.1 Short Selling 7
2.2 The short sale restrictions of 2008 8
2.3 Theories of short sales and empirical findings 9
2.3.1 The SEC thoughts 10
2.3.2 Overpricing 10
2.3.3 Unbiased stock prices 12
2.3.4 Downward pressure 13
3. Hypothesis 15
4. Methodology 17
5. Data and descriptive statistics 22
6. Empirical results 24
6.1 Results Event Study 24
6.2 Results Difference-in-Difference Regression 27
7. Conclusion 31
References 33
Appendix 35
1. Introduction
By the end of 2008, almost all financial markets in the world turned into a crisis. National governments tried with anything in their power to calm down the markets. One way was with restrictions on short selling. The goal of these restrictions was to protect firms from
speculating investors, who bet on a stock price decline that firms were already experiencing (Boulton and Braga-Alves, 2010). Speculation puts more downward pressure on the stock price, which can cause a downward spiral. The first regulation in the United States on short selling came from the Security and Exchange Commission (SEC) in July 2008. The markets acted agitated and after a couple of incidents, intervention was in their eyes necessary. On July 15th 2008, the SEC announced a restriction on naked short selling. The restriction was intended to restore confidence in the US markets after a period of false rumors spreading (SEC Emergency Order, 2008). For investors it was no longer possible to naked short sell publicly traded stocks for nineteen substantial financial firms until August 12th. A second restriction was introduced in September 2008. This restriction had the same purpose, but this time all financial stocks were included in the ban. Not only naked short selling was
prohibited, also covered short selling was not allowed in this period.
Short selling is selling a stock you do not own. Most short sales are covered, meaning that the stock sold is borrowed from for example a broker. Naked short selling is not covering the sale. Traders short sell stocks to speculate on a decrease in stock value in the future, so they can buy it back cheaper at a later time. More short sales on a stock create more
downward pressure on the stock price. This can be followed with a loss in confidence in one particular stock. The Security and Exchange Commission (SEC) tried to prevent this with the emergency order in July 2008.
On the 15th of July the SEC announced that the restriction on naked short selling was going to be implemented five days later, on the 21st of July, The restriction lasted for seventeen
trading days and was lifted on the 12th of August. The restriction is a natural experiment to investigate the consequences on the stock returns of the restricted firms and to see how the market reacted on this interference of a regulator. There are many articles about the effects of the short sale restrictions of 2008 on stock prices. Most authors, like Autore et al. (2011), Beber and Pagano (2011) and Boehmer et al. (2009), lay focus on the second restriction for the US. Only Boulton and Braga-Alves (2010) and Kolasinki et al. (2009) studied the first restriction. Thus, this field is still interesting to investigate. Also using the
difference method of March and Payne (2012), complements the existing literature. My research question in this article is as follows: What is the effect of a restriction on naked short selling for returns on the stock market? I will focus my research on the first restriction on short selling in 2008 for the US market. This study is relevant for regulators and policy makers in order to know what the consequences of a restriction are on naked short sales in turbulent times. It is also interesting for traders to know how the market reacts and how they should respond with their trades.
The interventions of the regulators are not fully supported by the findings and theories of academic research. Miller (1977) predicts that restrictions on short selling can lead to overpriced securities due to an absence of pessimistic traders. On the other hand, Diamond and Verrecchia (1987) argue that stock prices are not biased during the restriction, because investors take the absence of the pessimistic investors into account. Bai et al. (2006) add a risk factor to the model of Diamond and Verrecchia. Bai et al. argue that there is more uncertainty about the stock price and that it may lead to a stock price decrease. Boulton and Braga-Alves (2010) investigated the first restriction in combination with stock returns. Their findings are in line with the overpricing hypothesis of Miller. The abnormal returns of the restricted stocks went up after the announcement of the ban and dropped after lifting the ban. Because this is the only study for the July restriction focusing on stock returns, it is of interest to confirm the conclusions using a different approach. First, like Boulton and Braga-Alves (2010), I use an event study with abnormal returns to analyze the reaction of restricted stocks. I use the market model for calculating the returns and control the findings with the Fama-French three factor model. The control group is based on the NAICS codes which sort firms for a high level of comparability in business statistics. The advantage above other
classifications systems, like the SIC, is the specification and the number of subcategories in an industry. The control group is further deducted by excluding firms with a beta that is not near the beta of the restricted firms. I expect that the remaining firms are subject to the same market shocks and react similar as the restricted firms in normal circumstances due to these criteria.
The second part of my research contains a difference-in-difference regression to measure significant differences in stock returns over several time periods. Marsh and Payne (2012) used this method when focusing on the UK stock market. Daily returns of the period before the announcement of the restriction, the period between the announcement and the
introduction, the period during the restriction and the period after lifting the ban will be compared to a benchmark period to find significant differences in returns.
My findings for the event study are roughly in line with the finding of Boulton and Braga-Alves. I find a positive cumulative abnormal return of 17.35% in the period between the announcement and the introduction of the restriction. Boulton and Braga-Alves found a CAR of 12.90% in the same period, almost similar (13.04%) as my findings when I used the Fama-French three factor model. During the restriction, the restricted stocks hardly outperform the market. The findings confirm the overpricing hypothesis of Miller (1977). In the days around the removal, the abnormal returns turned negative. Investors anticipated the removal and the stock prices dropped to the level from before the restriction.
The difference-in-difference outcome shows that the returns of the restricted stocks during the restriction period, including the days after the announcement, are not significantly different compared to the control group. However, the overall effect for the whole portfolio is positive. The daily returns are in higher compared to the benchmark period. In the period after the announcement, stock returns of the restricted stocks are significantly higher. The effect of the restriction is temporarily, because the returns drop after lifting the ban. The results are partially in line with March and Payne (2012). They found no evidence for return differences for the financial and control group once the ban was introduced and when the ban was lifted. However their study was focused on the UK market and there restriction went immediately into effect, without an announcement period. Regulators can use this kind of intervention as a way to calm down the markets for a short period, but not to solve downward pressure on stock prices in the long run.
The remainder of this paper is structured as follows. Section two gives an overview of the existing literature, consisting of academic theories and empirical research. Section three describes the development of the hypothesis, followed by the methodology in section four. Section five shows the statistics of the descriptive variables and section six presents the results of the event study and the difference-in-difference regression. Finally, section seven gives the conclusion.
2. Literature review
To understand what effect a restriction has on the stock market, we first need to understand the process of short selling. Next, I briefly explain the restriction SEC introduced, followed by the theoretical frameworks and existing studies.
2.1 Short selling
Short selling a stock is selling a stock that the seller does not own, but is promised to be delivered by the seller. When a stock is sold short, the seller borrows the stock from a broker. The stock comes from the inventory of the broker. The short seller receives an amount from selling the stock, but has an open position to close by buying back the same number of shares borrowed at a later time and returning them to the broker. Because the seller does not own the share, it must pay dividend to the lender. If the stock price drops, the seller can buy the stocks back at a lower price and make a profit. If the price rises, the seller has to buy the shares back at a higher price and make a loss. Most of the time it is possible to hold the short position as long as desired, but when the lender wants the borrowed stock back and the broker has no other shares in his inventory, you can be forced to close the position. This is possible when many investors are short on a particular security (Brent, Morse and Stice, 1990).
There are two types of short selling, covered short selling and naked short selling. Covered short selling is discussed above: a share is borrowed from a broker and is used as collateral. In a naked short sale, the seller does not borrow or arrange to borrow the securities in time to make a delivery to the buyer within the standard three-day settlement period (SEC, 2005). The seller fails to deliver the stock to the buyer, also known as fail-to-deliver. This can happen with a delivery of, for example, few traded, illiquid stocks. Naked short selling is not necessarily a violation of the securities law (SEC, 2005). Sometimes, market makers agree to sell a security immediately to maintain market liquidity, even if they can’t deliver the shares to the buyer at that time. Usually the security will be delivered a couple days too late, so it is a temporary failure of deliver. The most concern is caused by manipulative naked short selling. Traders can overflow the market with sales of a company’s shares. This creates downward pressure on the stock prices. Traders can cover their short position at a profitable price after the share price drops (Christian, Shapiro and Whalen, 2006).
A short sale creates an extra stock in a company. An investor agrees to pay the owner of the existing stock any dividend and gives the borrowed shares back when the investor closes his position. This created stock is equal to the original stock, except for voting power. The lender of the stock will normally not notice that his stock is lent to someone else, because his broker arranges this process. The effect of short sales is comparable to the effect of a bank on the supply of money (Miller, 1977). People deposit money in a bank, which they can withdraw at any possible time. However, the bank has the opportunity to lend the deposited money to a third party. The supply of money increases, when the bank does this. The depositors can still withdraw their money upon demand. Investors, like depositors, do not know and do not mind that their share is used for other purposes, as long as they can close their position anytime. Thus, short selling provides extra liquidity in the stock market.
2.2 The Short Sale Restrictions of 2008
During the crisis of 2008, two restrictions were implemented regarding short selling in the US. The government institution with the power to introduce this is the Securities and Exchange Commission (SEC). The SEC announced the first restriction on July 15th, which took effect on July 21st. Nineteen ‘substantial financial firms’ were subject to the ban. The ban was intended to be terminated on July 29th. But on 29 July, the SEC extended the rule to August 12th and on August 12th the restriction was lifted. During this ban, it was not allowed to naked short sell a stock.
In September, the SEC introduced again a restriction on naked short selling. This restriction was informed in the night of September 17th and became effective the day after on September 18th. However, in the evening the SEC surprised the market with a temporary ban on short sales for financial stocks. This included the covered short sales. This article focuses on the first restriction of the SEC. Although there are many articles about the effects of the short sale restrictions of 2008 on stock prices, most authors, like Autore et al. (2011), Beber and Pagano (2011) and Boehmer et al. (2009), lay focus on the second restriction for the US. The second restriction had more influence on the financial market as it restricted all short sales for financial firms. The duration of the ban was also longer than the first restriction, which make the result more reliable. Boulton and Braga-Alves (2010) study the first restriction and Kolasinki et al. (2009) do both to investigate the difference between
prohibition and constraints on short sales. The effect of the first restriction on stock prices is somewhat under-expressed and therefore interesting to investigate further.
On July 15th, 2008, the Securities and Exchange Commission (SEC) announced an
emergency order pursuant to section 12(k)(2) of the securities exchange act of 1934. This act gives the SEC the power to take temporary action to respond to market developments. In the months before the announcement, the SEC had taken several actions to limit consequences of spreading false rumors (SEC, 2008). In March 2008, the rumor spread about liquidity
problems of Bear Stearns. Investors lost confidence in the firm and the stock price dropped. In addition, counterparties were concerned and were unwilling to secure the funding of Bear Stearns. The Federal Reserve stepped in before this could influence the whole market. In April 2008, the SEC charged a trader who spread false rumors about Blackstone Group’s acquisition of Alliance Data Systems. Due to this negative events in the months before and the worsening market circumstances, the SEC concluded that it was “in public interest” to put a restriction on naked short selling for nineteen financial firms (SEC, 2008). A short sale was only possible if the person had borrowed or arranged to borrow the security before an order for a short sale. On July 18th, the SEC announced to exclude certain entities from the
restrictions. Registered market makers, block positioners and other market makers who were selling short as part of market making and hedging activities were excluded.
2.3 Theories on short sales and empirical findings
First, I give an overview of the theories, followed by a more extensive explanation and corresponding empirical findings. There are two leading theories with regard to restrictions on short selling. Miller (1977) expects that a restriction leads to overpricing, because the more pessimistic investors are excluded and only the view of more optimistic investors is incorporated in the stock price. Diamond and Verrecchia (1987) argue that stock prices are not biased during a restriction. Informed traders are driven out of the market, which leads to less informativeness of the short sales. Uninformed traders take this into consideration and prevent that the stock price is biased. Bai, Chang and Wang (2006) add a risk factor to the model of Diamond and Verrecchia. Investors experience more uncertainty about the stock price due to a decrease of informativeness. This uncertainty leads to a decrease in demand if the investors is risk averse and results in a lower equilibrium price.
2.3.1 The SEC thoughts
The SEC emergency order was in their eyes a necessary consequence of spreading false rumors that led to a loss of confidence in the market. Naked short sales could put further pressure on this lack of confidence and an unnecessarily decline in stock prices. The SEC was afraid that if financial institutions were involved, the stability of the US markets could be disrupted (SEC, 2008).
Goldstein and Guembel (2008) support the thoughts that a restriction helps to stabilize the market. They investigated the behavior of investors using a trading model, where an investor is either informed or not informed about the state of the market. In their model, they added a ‘feedback effect’, where managers learn from the stock price when making an investment decision. In the model with feedback effect, it turned out that speculators have an incentive to sell, even when they are uninformed. A strategy with a short position in a stock and with more sell orders, will eventually lead to an actual decline in the share price and not to a temporarily price decline due short sales. The manager may see the share price decline as negative information about the investment decision and turns the investment project down. On average, this is not the best decision and will lead to a loss in firm value. This strategy of taking a short position in a company is called manipulation and the speculator can make a profit with this strategy. Goldstein and Guembel advice to discourage manipulation to increase further the efficiency of the market, through for example restrictions on short sales.
2.3.2 Overpricing
As mentioned earlier, the amount of a short position is the amount of increase in supply of stock on the market. An increase in the supply lowers the price of the stock, simply by the effect of supply and demand. Thus, the presence of adverse opinions affects the value of a stock. Miller (1977) predicts that restrictions on short selling leads to overpriced securities, because only the investors with a positive view on the security remain in the market. The more pessimistic traders are excluded, so not all information is incorporated in the stock price. During the ban, the price exceeds the equilibrium level, but as soon as the restriction is lifted, the prices should decline.
Boulton and Braga-Alves (2010) took the July 2008 ban to test Miller’s conjecture that short sale bans lead to overpriced securities and low subsequent returns. They used the restriction as a natural experiment by studying the impact on the stock prices and the quality of the market. Their sample was extracted from the CRSP database. Boulton and Braga-Alves compared the stock return of the restricted firms with a comparable group of unrestricted firms based on the same SIC code in the 6000-6799 (financial firms) range. For each
restricted stock, they found one stock that minimizes the difference in market value, closing stock price, return volatility and daily turnover. The Fama and French’s three-factor model for abnormal returns was used to control for size and book-to-market factors. The authors found, consistent with Miller’s overpricing hypothesis, a positive market reaction to the announcement. The cumulative abnormal return (CAR) was 12.9% in the three days before the implementation. During the restricted period, the abnormal returns of both the restricted as the control group did not differ significantly from the market return. Only the amount of naked short sales increased dramatically for the comparable unrestricted firms. After lifting the ban, the restricted stocks had a decline in abnormal returns, also consistent with Millers hypothesis. In the seven days around the expiration the CAR was -10.9%.
Autore et al. (2011) studied the September ban and support Millers overpricing hypothesis with data from CRSP. The abnormal returns are calculated with the Fama and French three-factor model. The authors create portfolios of stocks based on size, beta and book-to-market ratio. The banned stocks had a positive abnormal return at the initiation of the ban. Especially the stocks with a greater short interest perform better. After the ban is lifted, stocks with initially higher abnormal returns, have lower abnormal returns and vice versa. During the ban, the restricted stocks perform better. These results show up even when controlled for the Troubled Asset Relief Program (TARP). The TARP was a device of the United States government to address the subprime mortgage crisis.
The study by Boehmer, Jones and Zhang (2009) support partly the overpricing hypothesis of Miller (1977). Boehmer et al. found large increases in abnormal return after the
announcement of the September ban, but during the ban stock prices decreased gradually. Overall, the prices of the stock increased, but the authors were not convinced that the ban was solely responsible for this increase. At the same time, TARP and other initiatives were
announced. Next to these bank bailout interventions, they find no positive share price effects in stocks added later to the ban list. These stocks were added after interventions of the
government and should therefore not be influenced. Boehmer et al. believe that the bank bailout interventions leaded to the price increase rather than the short sale restriction.
The study by Beber and Pagano (2011) supports the thought of Boehmer et al. about the influence of the bailout interventions in the US and does not fully support the overpricing hypothesis. The study investigated every country in the world where the ban was introduced. The ban did not lead to higher stock prices in most countries, except for the US. It turned out that bans on covered short sales are correlated with significantly lower returns relative to unrestricted stocks. Bans on naked short sales do not have any significant correlation with returns. When looking at bans on financial stocks, only the US September restriction has a correlation with positive returns. Beber and Pagano (2011) also dedicate the effect of the positive return on the bail-out interventions. This is in line with the conclusion of Boehmer et al. (2009).
2.3.3 Unbiased stock prices
Diamond and Verrecchia (1987) use a rational expectations framework that does not match with the prediction of Miller (1977). Diamond and Verrecchia (1987) argue that two effects arise from restrictions on short selling. The first effect is referred to as the short-prohibition effect, which eliminates investors who want to short a stock but are prohibited to do so. A restriction will prevent informed investors to trade, when considering informational trades, because these trades are driven by the information of more informed traders. This leads to a decrease of information in the stock price. However, the less informed traders account for the information role of the stock price, so the stock price is not biased.
The second effect is called the short-restriction effect, which eliminates traders due to additional costs of short selling. Only an investor who is convinced that in the future a price decline occurs that exceeds the additional transaction costs is willing to short and pay the increasing costs (Diamond and Verrecchia, 1987). The informed traders are more likely the ones who are willing to pay these additional costs. The consequence is a change in the composition of the remaining short sellers. The less informed traders are more likely to drive out the market compared to the informed traders. Diamond and Verrecchia argue that this increases the informativeness of the short sales. The share price will not be biased during the restriction period.
Kolasinki et al. (2009) studied both restrictions with private short selling data and share price data from CRSP. They tested the theoretical predictions of Diamond and Verrecchia. They focused on the short selling activity with intra-daily short sales data. The costs of borrowing a stock during the emergency order increased significantly for the restricted stocks with on average seven times compared to before the restriction. The volume of short sales decreased with approximately fifty percent for the affected firms.
Kolasinki et al. tested the prediction of Diamond and Verrecchia that the
informativeness of the short sales will change during the restriction, but they failed to find evidence that the first restriction had any impact. They think the small size of the restricted group (17 firms) and the short time of the restriction mainly contributes to this finding. Testing the same prediction for the second restriction, the authors did find support for an increased informativeness of short sales.
Marsh and Payne (2012) use a different method to study the effect of a restriction. They used a difference-in-difference regression for daily returns to compare the effect of a restriction with a benchmark period. Their study focused on the UK market and they found a significant deterioration in returns the period before introducing a restriction for UK stocks. During the ban and after lifting the ban they could not find any significant difference in stocks prices between restricted stocks and matched stocks compared to the benchmark period. However, during the ban the stocks returns turned positive. They concluded that the goal of the UK government succeeded because the market calmed down.
2.3.4 Downward pressure
Most studies focus on how the investor’s expectation of future payoffs affects the share price and what role a short sale restriction plays in this expectation. Most analyses, like Diamond and Verrecchia (1987), Scheinkman and Xiong (2002) and Hong and Stein (2003), assumed that investors are risk neutral. Bai, Chang and Wang (2006) disagree and say that investors are risk averse. Bai et al. argue that risk influence the price of an asset during short sale restrictions. When better informed investors are excluded from the market by short sale restrictions, share prices become less informative. This is in line with the short-prohibition effect of Diamond and Verrecchia (1987). However, the less informed traders are less certain if the share price is correct and this increases the risk. When investors are risk averse,
investors reduce their demand for the asset due to an increase in uncertainty. The consequence is a decrease in the equilibrium price.
In a perfect efficient financial market, prices of financial assets reflect all available
information, no matter what sentiment the information is, negative or positive. The reaction to new information of buyers and sellers is reflected in market prices. This process is called price discovery (Brenner and Subrahmanyam, 2009). Restrictions have both influence on potential sellers and buyers. Sellers are constrained and excluded by a restriction. For buyers it is more difficult to find products to protect themselves from downside losses, because short sales are not permitted. The remaining supply of products that limits the downside risk, like put options, is used by the sellers to hedge their own exposure (Brenner and Subrahmanyam, 2009). A restriction on short selling leads to less activity in the stock market, which delays price discovery and decreases liquidity. Benner and Subrahmanyam argue that this causes stocks prices to fall.
Boehmer and Wu (2012) studied this topic. They use daily short selling flows and relate it to variables that measure efficiency of prices. Their key result is that short sellers play an important role in price discovery and that their trading activity contributes to more informational efficient prices. More shorting flow leads to less deviation of transaction prices from a random walk, fewer price delays and not to large negative price shocks. All these points suggest that short sellers improve price efficiency.
3. Hypothesis development
The SEC imposed two restrictions in 2008, the naked short sale ban of 15 July for nineteen financial stocks and the restriction on short sales for all financial stocks on 18 September. The focus in this thesis is on the first restriction, because this restriction is, as earlier mentioned, somewhat underexposed. The main difference is the kind of restriction on short sales. However, there is also a difference in time between the announcement and the introduction of the rules. After the announcement of July 15th, traders had the time to cover their naked short position by borrowing or arrange to borrow the share.
The literature about the consequences of short sale restrictions contains a number of theories mentioned earlier. Miller (1977) predicts an increase in share prices due to the fact that the pessimistic traders are excluded. Diamond and Verrecchia (1987) have a
contradictory theory. They argue that the composition of informed and uninformed traders changes after introducing a restriction on short sales. With a prohibition, it excludes mainly informed traders, but informed traders account for this. Uninformed traders are more likely to be driven out of the market then informed traders when selling a stock short is more difficult or limited by a restriction. This increases the informativeness of short sales. Therefore, stock prices are in both situations not systematically overpriced. Bai, Chang and Wang (2006) add the factor risk to the model of Diamond and Verrecchia. Uninformed traders have more uncertainty about the price and therefore more risk. Risk averse traders reduce the demand due the increased uncertainty, which decreases the price.
This event is a natural experiment to investigate the consequences of a restriction. Based on the theory of Miller (1977), I expect that after introducing the restriction stock prices will increase due to a decrease in short sellers. The decrease in short sellers is a
consequence of the restriction, which makes it less attractive to sell short a stock. More effort to borrow or arrange to borrow a share and higher costs to borrow due to an increase in demand makes it less attractive. The high value of the stock prices will stay until the rule is suspended. After the suspension the prices will fall, because more pessimistic investors can enter the market than before the restriction. This is in line with the theory of Miller (1977). Several empirical finding support this theory. Boulton and Braga-Alves (2010) and Autore et al. (2011) fully support the theory. Boehmer, Jones and Zhang (2009) and Beber and Pagano (2011) party support the theory because they argue that also other government interventions influenced the returns during the restriction, like the TARP.
The theory of Miller does not take into account that the restriction is announced before the implementation. The SEC introduced the restriction four trading days after the notification. Investors were given time to cover the naked short position. Market makers were aware of an increase in demand and could increase the cost of borrowing. For short sellers it already became less attractive before the implementation to short a stock. Therefore, my expectation is that after the announcement the stock prices of the later restricted stocks will increase. Boulton and Braga-Alves (2010) support this expectation with their empirical findings.
Hypothesis 1a: After announcing the restriction, stock returns of the restricted stocks increases and stay at high level during the restriction period compared to the market return
Hypothesis 1b: Returns of restricted stocks will decline compared to the market return as soon as the restriction is lifted
Marsh and Payne (2012) used in their study the difference-in-difference regression with regard to the effect of short sell restrictions. They investigated the restriction for the UK market. A diff-in-diff regression is used to measure the effect of a regulatory change. This method is not used before in studies focusing on the first restriction of the SEC. Therefore, it is interesting to see what outcome the diff-in-diff regression give.
The diff-in-diff regression shows the effect of the intervention of the SEC for different periods. Using the theory of Miller (1977), I expect that the restricted stocks perform better in the period between the announcement and the expiration compared to the period before the restriction. Next to that, I expect that the restricted stocks perform better than the control group, because the control group is not exposed to the ban. This expectation should also be in line with the expectation of the SEC. Their goal was to calm down the markets after a couple of incidents (SEC, 2008).
Hypothesis 2: Restricted stocks perform significantly better in the period between the announcement and the expiration of the restriction compared to the benchmark period and control group.
4. Methodology
Data is required to examine the hypotheses. Data will be extracted from the database Center for Research in Security Prices (CRSP), a part of the Wharton Research Data Services (WRDS). I will use Eventus to perform an event study. Eventus subtract data directly from the CRSP database. The focus of this study is on the first regulation of 2008, a restriction on naked short selling. This restriction was announced on the 15th of July and took effect on the 21th of July. The ban was intended to be terminated on July 29th. But on 29 July, the SEC extended the rule to August 12th. On august 12 the SEC not postponed it again and ended the restriction.
The SEC Emergency Order affected the nineteen largest firms which were publicly traded. Two of the firms, BNP Paribas Securities Corp. and Daiwa Securities Group Inc., trade over the counter. For these two firms no information is available because CRSP does not track firms that trade over the counter. The remaining seventeen firms are all operating in the financial sector. The Emergency Order only applied to nineteen financial firms and left many of other financial firms unprotected. This gives the possibility to create a matched group with the same characteristics and industry as the restricted firms.
Kolasinki, Reed and Thornock (2009) used as a control group financial firms only based on the same GICS industry subgroup. Their control group consisted of 66 firms from four GICS subgroups from Compustat: Diversified Banks, Investment Banking and Brokerage,
Diversified Capital Markets, and Other Diversified Financial Services. They do not control for other variables and admit this limitation in the conclusion: “…, the firms subject to the rules are not well matched to a control group of firms not subject to the rules, which limits the ability of the study to use relevant control firms.” Boulton and Braga-Alves (2010) specify their control group not only on an industry code. First they match firms based on the SIC Code in the 6000-6799 (financial firms) range from CRSP. Next, they calculate the mean market value of shares outstanding, closing stock price, return volatility and daily turnovers for 2007 and match each restricted stock to a stock that minimizes the difference in the mentioned characteristics. Therefore, each restricted stock has one equivalent stock. This is tricky, because one abnormality of a control stock during the ban period has huge impact on the results. Besides, there are only nineteen financial stocks that are affected by the
restriction, leaving a lot of financial firms unaffected. Therefore it is possible to search for 17
more than one suitable control firm. The SEC restricted the nineteen most substantial financial firms with transactions in publicly traded securities (SEC, 2008). Therefore, it is difficult to find financial firms with the same size in the control group.
I will take a different approach compared to the existing research. My control group is based on The North American Industry Classification System (NAICS). The NAICS is the standard used by Federal statistical agencies and is adopted in 1997 to replace the SIC system. It is developed to allow for a high level of comparability in business statistics among the North American countries. I match firms based on the six digit code beginning with 52 (Finance and Insurance). For each stock, I match the industry and the NAICS industry. For example for Goldman Sachs Group Inc I use the NAICS code 523110 (Investment Banking and Securities Dealing) to find matched stocks. The advantage of using NAICS over SIC is that NAICS is more specified and symbol more subcategories. I further decrease the list of matched
companies by excluding firms with a beta that is not near the beta of the restricted firm. The beta is a factor of the market model that explains the movement of a firm stock compared to a market index. The market model is explained later. The beta is based on past performance to predict the movement of a stock in the future. By matching firms on basis of nearly the same beta, you can expect the same return of a stock. The control group consists of 32 firms and is presented in table 1 together with the restricted firms. Unfortunately, both Boulton and Braga-Alves (2010) and Kolasinki et al. (2009) did not publish their control group. Boulton and Braga-Alves used a control group of 17 firms and Kolasinki et al. 66 firms based only on the GICS industry code.
The focus of the study lies on the stock return reaction around the Emergency Order. Abnormal returns are used to investigate how the stocks responded to the restriction. The first part of the study is done by an event study, which examines the reaction of the restricted firms. To investigate the returns I use the Eventus software available in WRDS, which subtracts data from CRSP. I use daily data with the market model as a benchmark for estimation of the abnormal returns. The market model is a model that relates the return of a security with the return of the market portfolio. The model is linear, because it is assumed that asset returns have a joint normal distribution. The market model is specified as follows
𝑅𝑖,𝑡 = ∝𝑖+ 𝛽𝑖𝑅𝑚,𝑡 + 𝜀𝑖,𝑡 with 𝐸(𝜀𝑖,𝑡) = 0 and 𝑣𝑎𝑟(𝜀𝑖,𝑡) = 𝜎𝜀2
Ri,t and Rm,t are the period returns on security i and the market portfolio and εi,t is zero mean
disturbance term (MacKinlay, 1997). The model assumes that the expected value of that term is zero and has no influence on the equation. That means that the return of a security at time t is based on ∝𝑖, a constant term, 𝛽𝑖, beta of the security and 𝑅𝑚,𝑡 , the return of the market at the same t. The parameters ai and Bi are the ordinary least squares (OLS) estimates of the
intercept and the slope of the market model regression.
The assumption of the disturbance term does not hold in practice. There is a
difference in the actual return (Ri,t ) and the expected return (∝𝑖+ 𝛽𝑖𝑅𝑚,𝑡). In that case is εi,t
the difference, also called the abnormal return or unexpected return.
To investigate the whole sample of abnormal returns a cross-sectional aggregation is used (Kothari and Warner, 2006). The method compares the distribution of the actual returns with the distribution of the predicted returns. The focus is on the mean of the distribution. For a sample of N stocks, the cross-sectional mean abnormal return (AR) for any period t is:
𝐴𝑅𝑡= 𝑁 � 𝜀1 𝑖,𝑡 𝑁
𝑖=1
Time-series aggregation is used to examine whether mean abnormal returns for periods around the event are equal to zero. I use the cumulative abnormal return (CAR) method, which is defined as:
𝐶𝐴𝑅(𝑡1, 𝑡2) = � 𝐴𝑅𝑡 𝑡2
𝑡=𝑡1
The Eventus software uses the CRSP database to calculate these cumulative abnormal returns for given periods. I use two groups as input for Eventus, the restricted group of 17 firms and the control group of 32 firms. The CRSP Value Weighted market index is used as market indices and the market model as benchmark. The estimation period for calculating the alpha and the beta, where the market model is based on, ends 50 days before the event date, with a minimum length of 100 and a maximum of 365 days. The estimate method is Ordinary Least Squares (OLS).
The event date is July 15th, the date of the announcement of the Emergency Order. 19
That is the zero point. I have chosen for six windows for the event study to make a distinction between several occurrences. The first period is several days before the announcement, to see if investors anticipated this and how the market sentiment was. The second period is after the announcement until the day before the implementation of the restriction. In this period, we can see how the market acts after hearing the extra rules. The third period, the first five days after the introduction, and the fourth period, the remaining days until the end of the
restriction, we see how the market behaves during the Emergency Order. The fifth period is the period after lifting the ban and the last period is a couple of days thereafter.
I use a difference-in-difference regression for daily returns to model the behavior of the stock price. A diff-in-diff regression is used to measure the effect of a treatment to a treated group compared to a control group. The regression specification includes robust standard errors and is also used by Marsh and Payne (2012).
𝑅𝑖,𝑡 = ∝ +𝛽1𝑅𝑠𝑑𝑖,𝑡+ 𝛽2𝑃𝑟𝑒𝑖,𝑡+ 𝛽3𝑃𝑟𝑒𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽4𝐵𝑎𝑛𝑖,𝑡+ 𝛽5𝐵𝑎𝑛𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽6𝑃𝑜𝑠𝑡𝑖,𝑡
+ 𝛽7𝑃𝑜𝑠𝑡𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽8𝑀𝑟𝑘𝑡𝐶𝑎𝑝𝑖+ 𝛽9𝐿𝑒𝑣𝑖+ 𝜀𝑖,𝑡
On the left-hand side is Ri,t the daily return of firm i on time t. The right-hand side contains a constant α and a dummy variable Rsdi,t to select the stocks restricted by the emergency order. In the regression, I use three periods that will be compared to the benchmark period. Like Marsh and Payne (2012) I took as benchmark period two months before the pre ban period. This benchmark period is from begin May until the end of June. The pre ban period (Pre) have a duration of twee weeks and is the period before the announcement. The ban period (Ban) is the period between the announcement and the expiration of the ban. The days after the announcement is included because investors had already the possibility to cover their short position and had time to respond on the situation. This period is the most important. The results of this period will show if the restricted firms perform significantly better compared to the benchmark period and compared to the control firms. The post ban period (Post) have a duration of one week and is the period after lifting the restriction.
The difference-in-difference terms, which give the effect of the treatment in the different periods, are the interaction variables of the time dummies in combination with the dummy for restricted stocks. If the coefficients are significantly different from zero I can conclude a difference in market reaction between the portfolio of restricted stocks and the portfolio of matched stocks for a particular period compared to the benchmark period.
The model is not completely similar to the model Marsh and Payne (2012) uses. Marsh and Payne add a volatility control variable to control for market wide changes and spreads. This study doesn’t investigate spreads and volume, so the volatility variable is deleted in the model. Marsh and Payne suggest to add control variables using previous work. I add two variables to control for the abnormal return, market capitalization and leverage ratio. These variables are also used in the other papers, like Boulton and Braga-Alves (2010), Kolasinki et al. (2009), Harris et al. (2009) and Autore et al. (2011). Market capitalization is the market value of companies’ outstanding shares. Short sellers want to put downward pressure on the stock price of firms with as minimal effort as possible. Relatively, the impact of one short sale is higher for firms with a low market capitalization. Therefore, the
relationship between the daily return and market capitalization is expected to be negative. The second control variable is the leverage ratio of a firm. The ratio is calculated by dividing the liabilities by the stockholders equity. A firm with high leverage is more likely to have financial distress compared to a firm with no leverage. It is more likely that a highly levered firm is to be sold short. Thus, the relationship between the daily return and leverage is expected to be positive.
5. Data and Descriptive Statistics
The stock price data is used from CRSP with the Eventus software of WRDS to calculate abnormal returns. In Eventus I make a distinction between two portfolios, one portfolio of stocks that are subject to the Emergency Order and one portfolio of stocks that have similar characteristics as the restricted group to control for the effect of the restriction. As mentioned earlier firms in the control group are matched on basis of NAICS codes. The codes are divided by specified industries where the companies have their main activity in. Because the emergency order affected only 19 financial firms, many firms with similar characteristics are available for the control group. Next to matching firms on basis of NAICS codes, firms are matched on basis of their betas. The beta is a coefficient related to the market return and is based on past performance. Firms with similar betas have the same stock movements. Matching firms betas, shows the difference in reaction of both portfolios on the Emergency Order. Therefore, the control portfolio consists of companies with near similar betas in the same sub industry based on the NAICS codes.
The restricted group consists of 17 firms, excluding BNP Paribas Securities Corp and Daiwa Securities Group Inc. Those two firms are not recorded in CRSP because they trade over the counter. The control group consists of 32 firms, based on NAICS codes and matched values of the firm’s beta. In table 1 the restricted group and control group are presented.
Compustat and CRSP, both parts of WRDS, are used to extract data for the diff-in-diff regression. CRSP provides the stock price and total shares outstanding. The period of the restriction is used for total shares outstanding. Compustat provides the liabilities and stockholders’ equity. I used the end balance of 2007 for the values of the liabilities and stockholders’ equity. The market capitalization and leverage ratio are calculated with this information.
Table 2 presents the descriptive statistics of the variables mentioned above. The average beta of the restricted group is 1.67 and a median of 1.70. For the control group the average beta is 1.47 and a median of 1.44. The difference is caused by the control companies of Lehman Brothers (NAICS 523930) and Fanny Mae and Freddy Mac (NAICS 522294). There were not control companies with the same high beta as the restricted firms for these NAICS codes. Next, there is a huge difference in market capitalization. The mean for the restricted stocks is about 37 billion, against approximately 12 billion for the control group. This is not surprising,
since the SEC constrained the nineteen biggest financial firms. There is almost no difference in leverage ratio. The restricted firms have a mean of 30.30 and the control firms a mean of 29.73. This high leverage ratio is typical for the financial industry.
6. Empirical Results
The focus on the result consists of reporting the reacting of the stock market on the
announcement of the SEC. First, I examine the abnormal returns of the restricted and control portfolio during several periods of the Emergency Order (EO). Second, I report the results of an analysis of daily returns around the announcement and expiration period with a
difference-in-difference model.
6.1 Results Event Study
The results in table 3 show that in the period before the announcement the restricted shares decreased in value compared to the other shares. In the ten days before the announcement the stocks of the restricted portfolio had a cumulative abnormal return of -12.63%, whereas the control portfolio reacted more in line with the predicted market model and had an abnormal return of -4.60%. Both results are significant for a confidence interval of 95% and 99% respectively. In the period between the announcement of the SEC and the introduction of the new rules, the restricted shares reacted significant positive, for a significance level of 0.01, with an abnormal return of 17.35%. This reaction is the announcement effect of the
restriction. All the shares in that portfolio have a positive return during that period. Also the controlled portfolio reacted significant positive (9.73%) for a significance level of 0.01, but that was not as extreme as the restricted shares. Boulton and Braga-Alves (2010) are the only authors who also did an event study. Their results are in line with my results, but differ somewhat. They find a positive CAR of 12.9% after the announcement, which is a difference of 4.45%. This could be explained by the calculation method of the abnormal returns.
Boulton and Braga-Alves used the Fama-French three-factor model, where I used the market model. The Fama-French model controls for the size of the firm and the book-to-market ratio, the market model does not.
After the introduction of the Emergency Order, the restricted portfolio did not outperform the market. It performed in the first four days 0.18% worse and in the remaining days -3.20%. Both periods are not significant different from zero. The matched sample did outperform the market in this period with respectively 7.06% and 5.01%. The findings confirm hypothesis 1a. The restricted stocks significantly outperform the market after the announcement of the restriction and the return did not drop during this period.
After lifting the restriction, the former restricted shares fell with 10.55%. This result is significant for a significance level of 0.01. All the shares in this group performed worse than the market in the five days after the removal. The matched group did not significantly
outperform the market. The last period in the sample is a week after the removal. Both groups remain close to the market level. These findings are in line with hypothesis 1b. Returns drops directly after the removal of the restriction. The decrease in abnormal return almost
compensated the increase of returns in the announcement period, as shown in table 3.
As mentioned earlier, the market model does not control for the firm size and book-to-market ratio. The Fama-French three-factor model does control for these factors. To verify the results I use the Fama-French model. The model is:
𝑅𝑖,𝑡 = ∝𝑖+ 𝛽𝑖𝑅𝑚,𝑡 + 𝑠𝑗𝑆𝑀𝐵𝑡+ ℎ𝑗𝐻𝑀𝐿𝑡+ 𝜀𝑖,𝑡
The model is similar to the market model, but it adds two control variables. SMBt is the average return on small market-capitalization portfolios minus the average return on three large market-capitalization portfolios. HMLt is the average return on two high book-to-market equity portfolios minus the average return on two low book-to-book-to-market equity portfolios.
The results of the event study using the Fama-French model are shown in table 4. The results of the restricted group are, as expected, in line with the finding of Boulton and Braga-Alves. They used the same model and therefore the results are different in the announcement period compared to the market model. In the announcement period the CAR is 13.04%. Boulton and Braga-Alves calculated a CAR of 12.9%. The differences with the market model for the other periods are small and negligible. During the ban the CAR is -5.76%, against a CAR of -3.38% using the market model. After lifting the ban, the CAR is -8.35% using the Fama-French model and -10.55% using the market model. However, the abnormal returns of the control group are significant different. The abnormal returns of the control group are lower (2.22%) in the announcement period compared to the earlier findings with the market model (9.73%), but it is significant for a significance level of 0.05. Overall, the results with the Fama-French model show the same movements in abnormal return as with the market model, only less extreme.
The findings of the event study are in line with the theoretical predictions of Miller (1977). Miller (1977) predicted that when short sales are restricted, the more pessimistic investors are excluded from trading and therefore the market price only reflects the optimistic investors. Therefore, the share price will increase after the implementation of the restriction. In this case investors knew after the announcement on 15 July the upcoming restriction of 21 July.
Investors already reacted in the period between the announcement and the implementation (17.39%). During the restriction period the prices remain on the same, overpriced level, according to Miller. The results confirm this assumption. In this period the portfolio of restricted stocks hardly outperform the market.
The overpricing theory and the behavior of stocks after removing the restriction are also consistent with each other. In the period after lifting the ban, the abnormal returns turn negative for more than ten per cent. After the removal it was possible again to naked short sell a stock. Investors with a more negative view on the market had again easily access to the stock market. This reflects in the negative returns after lifting the ban. Also in the days before the expiration the market turned negative. In the four days before, the restricted portfolio drops respectively 3.51%, 1.94%, 1.44% and 1.10%. In total a decline of 7.99%, against a small plus of 0.76% of the matched portfolio. Originally, the order should be terminated on 29 July at the end of the day, but the SEC announced on 29 July an extension until 12 August. Investors could anticipate the removal of the restriction and did so.
The period around the removal of the Emergency order almost completely
compensates the upward movement the restricted portfolio made. From the announcement date until five days after removing the ban, the value of the stocks has nearly the same value. The SEC’s Emergency Order was a temporary and artificial move to stop the downward pressure on the restricted stocks and even increase the value during this period. The reason of the SEC was to prevent a disruption in the market based on false rumors. Those false rumors caused prices to fall unnecessarily. The emergency order succeeded in plan, but only for seventeen trading days.
6.2 Results Difference-in-Difference Regression
To answer the second hypothesis I use a diff-in-diff regression to see if the daily returns are significantly higher during the restriction compared to the control group. The event study already shows that the abnormal returns of restricted stocks outperforms the market in the days after the announcement. During the period of restriction the abnormal returns are more closely to the market return and after lifting the restriction the returns drop. This panel regression gives us with another method and another measure of return more certainty about the significance of implementing a restriction. The model used is:
𝑅𝑖,𝑡 = ∝ +𝛽1𝑅𝑠𝑑𝑖,𝑡+ 𝛽2𝑃𝑟𝑒𝑖,𝑡+ 𝛽3𝑃𝑟𝑒𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽4𝐵𝑎𝑛𝑖,𝑡+ 𝛽5𝐵𝑎𝑛𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽6𝑃𝑜𝑠𝑡𝑖,𝑡
+ 𝛽7𝑃𝑜𝑠𝑡𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽8𝑀𝑟𝑘𝑡𝐶𝑎𝑝𝑖+ 𝛽9𝐿𝑒𝑣𝑖+ 𝜀𝑖,𝑡
To be clear the ban period in this regression is the period between the announcement and the expiration of the ban. The results are shown in table 5. The first regression is a regression of Marsh and Payne (2012), only now the first restriction of the US market is studied instead of the restriction in the UK. During the benchmark period restricted stocks performed 0.37% worse than the firm of the control group. The average daily return of the restricted stocks is -0.83% and significant for a significance level of five per cent. During the pre-ban period the returns turned even more negative, for both the control group and restricted group. The restricted firms return is -0.525% worse compared to the matched firms, but this is not statistical significant. The overall return is 0.70% lower than the benchmark period and significant for a p-value of 0.05. The result show that in the period before the announcement return turned negative, especially for the later restricted firms. The returns do not explain precisely why the SEC stepped in, but it justify their action. Marsh and Payne (2012) also found negative returns in their benchmark period and pre-ban period and suggest that this was probably one of the key factors for the UK government to introduce a restriction.
During the ban period, stocks performed 1.85% significantly better for a significance level of 0.01. The returns of the restricted stocks performed even 0.03% better, but this value is not statistically significant. Also based on the results of the event study, the reason could be that investors anticipated the introduction and the removal. Because only in the days after the announcement the restricted stocks reacted strongly positive and the days before the
expiration negative. The investors were aware of the date of introduction and the date of expiring. Between these days the abnormal return was not different from the market return.
The control group reflects during the ban period the market return and the restricted group returns not differ from these market return. After removing the restriction the daily returns turn negative again. The return of the restricted stocks is 1.11% worse compared to the benchmark period and is significant for a pvalue of 0.05. The average restricted return is -1.73% and is approximately the same level as the restricted return of the ban period (1.87%), except this return was positive. That means that the upward movement during the ban is almost completely vanished in the days after the expiration. These results are also in line with the earlier finding of the event study. Marsh and Payne (2012) could not find any return differences during the restriction period and after the expiration for both the restricted group as the control group. But they concluded that the goal of the UK authority was to calm disorderly market and that this succeeded, because the return turned positive during the ban.
The second part of the regression, including two logarithm control variables, neither gives significant result for the interaction variables. Marsh and Payne (2012) add only a volatility control variable to control for changes in the information environment. This is useful by testing spreads and volume, as did Beber and Pagano (2009). Autore et al. (2011), Boulton and Braga-Alves (2010) and Harris et al. (2009) use the log of market capitalization to control for returns. I add leverage to the regression because theory predicts that highly levered firms are often target for short sellers. The chance that highly levered firms are getting in financial trouble is higher compared to firms with no leverage (Massoud et al, 2011).
The results of the second regression are also shown in table 5. The dummy variables are the same to the first regression, except for the constant and restricted factor. The constant changes to -0.18% and variable restricted to -0.33%, meaning that the return is dependent on the market capitalization and leverage ratio. The market capitalization has a negative
influence on the daily return. If the capitalization increases with one percentage, the daily return decreases with 0.02%. The leverage ratio, the amount of liabilities compared to stockholders equity, has a positive effect. An increase in the leverage ratio of one per cent, leads to an increase of 0.01% in daily return. Both the influence of the market capitalization and the leverage ratio is too small to be significant.
Based on the results shown in table 5, I do not find statistical evidence to conclude that the restricted stocks during the ban period perform statistically better compared to the control group. Therefore, I have to reject hypothesis 2.
Following the result of the event study, I expect a significant effect in the period between the announcement and the introduction. In the first regression, this period is combined with the ban period to show the overall effect during the restriction. It is interesting to split the period into an announcement period and the actual ban period. The following model is used:
𝑅𝑖,𝑡 = ∝ +𝛽1𝑅𝑠𝑑𝑖,𝑡+ 𝛽2𝑃𝑟𝑒𝑖,𝑡+ 𝛽3𝑃𝑟𝑒𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽4𝐴𝑛𝑛𝑖,𝑡+ 𝛽5𝐴𝑛𝑛𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽6𝐵𝑎𝑛𝑖,𝑡+
𝛽7𝐵𝑎𝑛𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽8𝑃𝑜𝑠𝑡𝑖,𝑡+ 𝛽9𝑃𝑜𝑠𝑡𝑖,𝑡∗ 𝑅𝑠𝑑𝑖,𝑡+ 𝛽10𝑀𝑟𝑘𝑡𝐶𝑎𝑝𝑖+ 𝛽11𝐿𝑒𝑣𝑖+ 𝜀𝑖,𝑡
Table 6 shows the results of a regression with four periods instead of three. It is important to see that in the regression without the control variables, the values of the coefficients of the unchanged variables have the same values as in the first regression. This is expected, since only one dummy variable is divided in two variables.
In the days after announcing the emergency order until the introduction, the daily returns of the both the restricted as the control portfolio are higher. On average the restricted stocks outperform the market significantly with 2.41% for a significance level of 0.1. This is above the 3.62% higher return of the whole group compared to the benchmark period. That value is significant for a p-value of 0.01. After introducing the restriction the portfolio reacted logically less positive than in the first equation, but still statistically significant for a p-value of 0.01. Remarkably, the restricted stocks had on average a negative return of 0.57% during the restriction compared to the control group. These results are in line with the event study, where the CAR was also negative during the restriction period. Overall the daily return of the restricted stocks during this period is 0.00%. Even though the restricted stocks performed worse than the control group during the restriction, the average return not turned negative.
The effect of the restriction, based on the event study, was in line with my hypothesis. After the announcement stock returns of the restricted stocks increased and stayed on that high level. After lifting the ban, stocks return immediately declined. This confirmed the theory of Miller (1977) and support the finding of Boulton and Braga-Alvas (2010). The overall effect of the restriction was small and almost disappeared when the ban was lifted.
Based on the difference-in-difference regression the returns of the restricted firms was not significantly different compared to the control group during the period of the restriction including the announcement period. Therefore, hypotheses 2 is rejected. These finding are in line with the results of Marsh and Payne (2012).
The stock market did react, but only on two time intervals. The first reaction of the market is in the days after the announcement and before the implementation. The second reaction is in the days around the removal of the restriction. Investors knew about both the implementation and the expiration date and probably they toke this into account. The goal of the SEC was to calm down the markets (SEC, 2008) and they succeeded temporary by introducing the restriction. But after the expiration the market returned to same level as before the interference.
7. Conclusion
In this thesis, I examine the effect of the short sale restriction on stock returns of the July SEC emergency order in 2008. During this restriction, 19 stocks of financial firms were prohibited from naked short selling. I used an event study and a difference-in-difference regression to investigate the reaction of stock returns on this restriction in several periods. A control group of stocks is selected on basis of the NAICS code and comparable betas of firms to compare the reaction of the restricted stocks with unrestricted stocks.
The event study shows a huge increase in abnormal return of 17.39% for restricted stocks in the days after the announcement of the SEC and before the implementation. During the ban these stocks hardly outperform the market. In the six days after the removal, the cumulative abnormal return is 10.55%. The restricted stocks underperformed compared to the market. The abnormal returns turned not only negative after the removal, also in the four days before lifting the restriction. The negative abnormal return is 7.99% in this period. Boulton and Braga-Alves (2010) roughly find the same results.
The findings are in line with the theoretical predictions of Miller (1977). Miller predicts an increase in share prices after introducing a restriction on short sales. During the restriction prices are overpriced and after lifting the restriction, stocks prices will drop. The results support Millers overpricing hypothesis. Investors knew after the announcement when the restriction took effect. The abnormal return of 17.39% after the announcement and before the introduction show that investors reacted after the announcement instead of after the introduction as Miller suggested. The results also show that investors anticipated the removal. This was possible because the SEC announced the removal date.
Also Beber and Pagano (2011), Boehmer et al. (2009) and Autore et al. (2011) support Millers overpricing hypothesis for the US market. But they investigated the second restriction on also covered short sales. However, it supports that government intervention with restrictions on short sales is a temporary solution.
The difference-in-difference regression was intended to find significant evidence for a higher return for restricted stocks during the restriction using the model of Marsh and Payne (2012). I could not find statistical evidence during the ban period, even when I included the
announcement period. Only the announcement period showed significantly positive daily returns. These results are in line with the results of the event study.
Some factors may bias the results. The regulation not banned all short sales. The emergency order only prohibited naked short selling. Covered short sales were still possible in this period. On the other hand, recent studies of Beber and Pagano (2011), Boehmer et al. (2009) and Autore et al. (2011) shows similar results for covered short sales restrictions. Next, the SEC restricted the nineteen largest financial firms. Therefore, it is hard to control for the firm size using other financial firms in the control group. Another biased factor that influenced the results is that bona fide market makers were excluded from the restriction to facilitate
customer orders to avoid delay. At last, the half of the nineteen firms affected by the
restriction, have international listings that are not subject to the ban. The SEC only has power to regulate stocks in the US market. The restriction has less effected for that firms than for the firms only operating in the United States. These factors are likely to decrease the effect of the regulation.
Both the event study and diff-in-diff regression show that the effect of the restriction period is temporary. The stocks did react, but mostly not during the restriction. The stocks
outperformed the market after the announcement and underperformed around the removal. I can conclude that the restriction of naked short sales is a temporarily and artificial instrument of the SEC to slow down the downward movement of stocks. The goal of the SEC’s
restriction, to prevent a disruption in the market based on false rumors, succeeded only during this restriction. Regulators can use this kind of instrument for a short-term solution and influence the market behavior for a short and temporary period. In the long term, the effect of the restriction is small. Therefore, it should not be used to solve downward pressure on stock prices.
References
Autore, D.M., Billingsley, R.S., Kovacs, T., 2011, The 2008 short sale ban: Liquidity, dispersion of opinion, and the cross-section of returns of US financial stocks, Journal of Banking and Finance 35, 2252-2266.
Bai, Y., Chang, E.C., Wang J., 2006, Asset prices under short sales, working paper, University of Hong Kong.
Beber, A., Pagano, M., 2012, Short selling bans around the world: evidence from 2007-2009 crisis, Journal of Finance 68, 343-381.
Boehmer, E., Jones, C.M., Zhang, X., 2009, Shackling short sellers: The 2008 shorting ban, Review of Financial Studies 26, 1363-1400.
Boehmer, E., Wu, J., 2013, Short selling and the price discovery process, Review of Financial Studies 26, 287-322.
Boulton, T.J., Braga-Alves, M.V., 2010, The skinny on the 2008 naked short-sale restrictions, Journal of Financial Markets 13, 397-421.
Brenner, M., Subrahmanyam, M.G., 2009, Restoring financial stability: how to repair a failed system, chapter 12 short selling, New York University Stern School of Business.
Brent, A., Morse, D., Stice, K.E., 1990, Short interest: Explanations and Tests, The Journal of Financial and Quantitative Analyses 25, 273-289.
Christian, J.W., Shapiro, R., Whalen, J.P., 2006, Naked Short Selling: How Exposed are Investors, Houston Law Review 43, 1033-1090.
Diamond, D.W., Verrecchia, R.E., 1987, Constraints on short selling and asset price adjustment to private information, Journal of Financial Economics 18, 277-312.
Goldstein, I., Guembel, A., 2008, Manipulation and the allocational role of prices, The Review of Economics Studies 75, 133-164.
Harris, L.E., Namvar, E., Phillips, B., 2009, Price inflation and wealth transfer during the 2008 SEC short-sale ban, working paper, University of Southern California.
Hong, H., Stein, J.C., 2003, Difference of Opinion, Short-Sales constraints and market crashes, Review of Financial Studies 16, 487-525.
Kothari, S.P., Warner, J.B., 2007, Handbook of Corporate Finance: Empirical corporate finance Volume 1, Chapter 1 Econometrics of Event studies, Elsevier.
Kolasinksi, A.C., Reed, A.V., Thornock, J.R., 2009, Prohibtions versus constraints: The 2008 short sales regulations, working paper, University of Washington.
MacKinlay, A.C., 1997, Event studies in economics and finance, Journal of Economic Literature 35, 13-39.
Marsh, I.W., Payne, R., 2012, Banning short sales and Market Quality: The UK’s experience, Journal of Banking and Finance 36, 1975-7986.
Massoud, N., Nandy, D., Saunders, A., Song, K., 2011, Do hedge funds trade on private information? Evidence syndicated lending and short-selling, Journal of Financial Economics 99, 477-499.
Miller, E.M., 1977, Risk, uncertainty and divergence of opinion, Journal of Finance 32, 1151-1168.
Scheinkman, J., Xiong, W., 2003, Overconfidence and speculative bubbles, Journal of Political Economy 111, 1183-1219.
Security and Exchange Commission 2008, Statement of Securities and Exchange Commission concerning short selling and issuer stock repurchases.
http://www.sec.gov/news/press/2008/2008-235.htm 34
