The Effect of the Short Selling Ban in 2008-09
on Market Quality
Author: Julian Sie Student Number: 10373381
Thesis Supervisor: Jan Lemmen Course: BSc ECB
Abstract
This paper studies the effects of the short selling ban of 2008-09, as imposed on the U.K. stock market by the Financial Services Authority (“FSA”), on intraday volatility and liquidity. It also investigates whether an upward and downward bias in stock prices exists after imposing and lifting the ban. The results show that the intraday volatility increased significantly during the period of the ban. However, this could be the result of the financial crisis instead of the short selling ban itself. After the ban was lifted, intraday volatility remained substantially higher than in the pre-ban period. Liquidity decreased significantly in the ban period as well. Contrary to what academic literature suggests, this paper finds that liquidity for banned stocks did not decrease more relative to the control group stocks. Furthermore, while an upward bias in share price is observed when the ban is imposed, there is no indication of a downward bias upon lifting the ban. However, a more specific analysis of banned stocks with a large market capitalisation shows both an upward and downward bias in share prices. Overall, this paper finds that market quality during the ban period was severely reduced, yet this cannot be fully attributed to the short selling ban solely.
Keywords: Short Selling, Volatility, Liquidity, Market Quality, Financial Crisis JEL Classification: G01, G14, G18
1
Statement of Originality
This document is written by Student [Julian J.Y. Sie] who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2 1. Introduction
Amid one of the most intense periods of the crisis and a few days after the fall of Lehman Brothers, the FSA introduced a short selling ban which initially targeted 29 financial firms on the London Stock Exchange and prohibited short selling and trading in futures and options of these stocks. Although the ban was motivated by the FSA’s desire to protect the market quality and bring a halt to further instability in the financial sector, constraining short selling negatively affects market efficiency and security pricing (Boehmer, Jones and Zhang, 2013). This seemingly contradictory move makes research into this topic of social relevance.
This paper will provide an overview of existing literature and aims to empirically assess the effects of the U.K. short selling ban on the relevant stocks’ stability and market quality.
The remainder of this paper is structured as follows: Section 2 provides the theoretical framework, comprising of an in-depth explanation of short selling, a time line of regulatory actions against short selling in 2008-09, as well as an overview of the existing academic literature on this topic. The hypotheses of this paper can be found in section 3, which is followed by an explanation of the methodology used in section 4. Section 5 deals with the results of the statistical tests performed and in section 6 the main conclusions of this paper will be drawn.
2. Theoretical Framework 2.1 Short Selling
Short selling refers to the sale of a security that is not owned by the seller, but that is promised to be delivered. Investors engage in this type of transaction when they believe the price of a security will decline, after which they will buy back the security at a lower price to generate a profit. Commonly, this practice is either driven by speculation or with the intention to hedge a particular long position. Although short selling is sometimes viewed with hostility by company executives, many academic researchers, as summarised in the literature review below, have shown that it does have a social function through providing market liquidity and contributing to the efficiency in pricing securities.
3
A short sale is in fact a loan, because a stock is borrowed but paid for at a later point in time. A covered short sale implies that the short-seller must purchase the stock at the later date in order to replace the stock that was borrowed. In a naked short sale however, investors sell shares that have not yet been borrowed, but assume that they can provide these shares at any delivery deadline. The FSA's ban in 2008-09 banned all forms of short selling, with an exemption for registered market makers. Existing short positions that accounted for more than 0.25 percent of a company's asset value had to be disclosed. Prior to the ban in 2008, both naked and covered short sales were allowed in the U.K.
2.2 Literature Review
There is general consensus among academics that the presence of short sellers is beneficial for the stock market for multiple reasons. First, it is claimed by many researchers that short-sellers effectively enhance price formation. In a world where traders have rational expectations, short selling constraints should not lead to a long-term bias in security prices (Diamond and Verrechia, 1987). However, if agents have less than fully rational expectations there will be a divergence of opinion as described by Miller (1977). Short selling constraints will in this case result in an upward bias on prices, because pessimistic investors no longer have the ability to short sell securities and thus influence the market.
A study by Boehmer, Jones and Zhang (2008) finds that heavily shorted stocks slightly underperform lightly shorted stocks, which implies that short-sellers are on average better informed about the true value of stocks. A security that is heavily shorted could therefore indicate that the stock is currently overvalued. This is an example of short-sellers reducing mispricing in securities. This theory is supported by Ofek, Richardson and Whitelaw (2004), who show on the basis of empirical evidence that there is a clear relationship between short selling constraints and the level of mispricing.
Overall, the presence of short selling constraints reduces market quality, not only because of an increase in the level of mispricing but also because of a decrease in liquidity for stocks in which short selling is constrained as well as reduced market efficiency as measured by the bid-ask spread (Boehmer, Jones and Zhang, 2013).
4
In light of such a large number of existing academic literature pointing out the inefficiencies regarding short selling constraints, the question arises as to why the FSA carried out the ban in 2008. The FSA justified their decision at the time by claiming that the market for financial stocks was disrupted by extreme circumstances. A large number of short-sellers active in for instance a stock of a bank could imply an expected decrease of the bank's stock price and this uncertainty could then result in a bank run. A short selling ban would prevent such a scenario from occurring. The ban was an attempt to bring short-term stability in an unstable market.
2.3 Regulatory Actions Taken in 2008-2009
Temporary bans and other regulatory efforts that either prohibit or make it more costly and difficult to sell short have been applied on numerous occasions before. Jones (2012) provides an in-depth analysis of many regulations regarding short selling in the U.S. starting as early as the 1930's. The bans in 2008 imposed by the FSA in the U.K., the Securities and Exchange Commission (“SEC”) in the U.S and bans in many other countries however, were the broadest and most unexpected bans to date (Boehmer, Jones and Zhang, 2013). The last time shorting was banned was in 1931.
The first regulatory action taken in this period was a restriction on naked short selling in 19 financial stocks in July 2008 in the United States. This temporary ban lasted until mid-August. On September 17 however, the SEC once again imposed a temporary ban, this time prohibiting naked short selling in all U.S. stocks. One day later, the FSA announced a temporary short selling ban in 29 financial stocks, which after revisions were to include 34 stocks. Existing short positions in these stocks that accounted for more than 0.25 percent of the company's asset value also had to be disclosed.
3. Hypotheses
The presence of many short sellers in a stock indicates that the stock is currently overvalued (Boehmer, Jones and Zhang, 2008). A stock that is heavily shorted is therefore expected to decrease in the future. A short selling ban would imply that there should be an upward bias in stock prices. This is in line with Miller's (1977) theory of heterogeneous expectations, which
5
suggests that there is an upward bias associated with short selling constraints. These theories lead to the first hypothesis of this paper:
Hypothesis 1: Banned stocks experience an upward bias in price at the moment the short
selling ban is imposed, as well as a downward bias the moment the ban is lifted. This paper also aims to research the effect of the short selling ban on intraday volatility. While academic literature so far does not clearly state a significant effect of short selling bans on volatility, we would expect intraday volatility to reduce slightly as a result of pessimists being shut out of the stock market, following the line of reasoning of Miller (1977). From this follows the second hypothesis:
Hypothesis 2: Intraday volatility decreased more for banned stocks than for non-banned
stocks.
As discussed in the literature review, market quality as well as market efficiency are expected to deteriorate after the introduction of the ban. In order to investigate the effect of the short selling ban on market efficiency we will calculate the bid-ask spread. A larger spread indicates that a stock is less liquid. This paper's third hypothesis will therefore be:
Hypothesis 3: Liquidity, as measured by the size of the bid-ask spread, decreased more for
banned stocks than for non-banned stocks.
4. Methodology 4.1 Data
The data used for this paper consists of daily data for all constituents of the U.K. FTSE 350, as well as the group of financial firms that were banned during the period of 2008-2009. The sample includes thirty-four financial firms, most of which were constituents of the FTSE 350 at the time of the ban. Five of these banned firms defaulted over the course of the observation period and are therefore not included in the various analyses in this paper. This paper acknowledges therefore the existence of a survivorship bias.
6
The ban was implemented by the FSA on September 19, 2008. Some firms were later added to this original list, resulting in the list of firms released on September 30, 2008. This paper uses the list released on September 30, comprising of 32 firms, but assumes for simplicity's sake that all firms were banned starting from September 19.
The time span of the data starts two months ahead of the ban period (September 2008 - January 2009) up to two months after. All data is on a daily basis and includes the following variables for all firms, when possible: stock price, market capitalisation, daily high and low prices, as well as daily bid and ask prices at market close.
4.2 Investigating the Upward and Downward Bias in Share Prices
An upward or downward bias on share prices is a short-term effect. It is therefore important to use an appropriate time period. This paper therefore uses the time period by Hansson and Fors (2009), which includes ten trading days before and after the ban was implemented and ten trading days before and after the ban was lifted. I will run a set of paired t-tests to compare the ten days after the ban with the ten days preceding the ban. The same will be done for the two periods surrounding the lifting of the ban. The t-statistics of the banned group of stocks can then be compared to t-statistics of control group stocks. This paper will also graphically show the movements of share prices of banned stocks and control group stocks during the entire period.
4.3 Calculating Intraday Volatility
This paper uses the difference between the daily high and daily low values of stock prices to calculate the intraday volatility, a method first used by Parkinson (1980). The formula for this method is as follows:
Intraday Volatility = (Day High - Day Low) / Volume Weighted Average Share Price
In this formula "Day High" and "Day Low" represent the highest and lowest share price of a certain day, respectively. This paper does not have access to the necessary data to compute the
7
Volume Weighted Average Share Price and will instead use the Closing Share Price, thus changing the formula to:
Intraday Volatility = (Day High - Day Low) / Closing Share Price
The total period will be divided into four sub-periods: Pre-ban, the first half of the ban, the second half of the ban and post-ban. The control group will consist of the non-banned stocks that are constituents of the FTSE 350. To investigate whether there is a difference in the intraday volatility movements of banned and control group stocks, I will use paired t-tests, as seen in Hansson and Fors (2009). A graph depicting the movements of intraday volatility will also be included.
4.4 Measuring Liquidity
As a measure of liquidity, I will use the daily bid-ask spread. This spread is calculated as follows: Spread = (Ask Price - Bid Price) / Share Price * 100%
In this formula "Ask Price" represents the lowest price a seller of a security is willing to accept, whereas "Bid Price" represents the highest price a buyer is willing to accept. "Share Price" is the daily closing share price for a security.
The level of spread over time for banned and control group stocks will be shown graphically. Higher levels of spread would suggest that the market is less liquid at that time and the market is less efficient. As described in the previous sections, I will run paired t-tests to measure the differences between sub-periods. The sub-periods are identical to those that will be used in the analysis of intraday volatility.
4.5 Additional Analysis of a Specific Group of Banned Stocks
Most of the existing literature compares the entire group of banned stocks with an entire group of non-banned stocks on the same market. This paper does so as well, as explained above, but also aims to provide an example of an analysis of a select group of comparable banned stocks with comparable non-banned stocks. Ideally one would create a control group with stocks that
8
operate in the same sector to eliminate sector-specific factors from influencing results. For the U.K. short selling ban in 2008-09 this is not possible however, as the ban specifically targeted firms in the financial sector.
Instead, this paper will create a control group based on stocks that are comparable in market capitalisation (MCAP). The select group of banned stocks consists of the seven largest companies in terms of MCAP. The control group consists of 17 stocks with a comparable value for MCAP. The list of companies in both groups is shown in the following table:
Banned Stocks
Control Group (based
on MCAP)
Company Name MCAP
(in 1,000)
Company Name MCAP
(in 1,000)
Aviva plc 10,404 GlaxoSmithKline plc 61,651
Barclays PLC 14,717 Unilever plc 41,107
HSBC Holdings plc 83,297 AstraZeneca PLC 36,818
Lloyds Banking Group plc 11,898 Rio Tinto plc 35,523
Standard Chartered PLC 14,965 BG Group plc 32,759
The Royal Bank of Scotland Group plc 18,999 Tesco PLC 27,594
Prudential plc 9,806 Diageo plc 23,158
Average 23,441 Anglo American plc 22,649
Reckitt Benckiser Group plc 18,944
SABMiller plc 16,753
National Grid plc 16,431
Imperial Tobacco Group plc 14,874
BAE Systems plc 13,700 Carnival plc 12,184 Centrica plc 11,926 SSE plc 10,951 BT Group plc 10,621 Average 23,979
I will investigate for this list of companies the upward and downward bias in stock price, intraday volatility and the bid-ask spread. The methodology used will be the same as the one described in the other sections above.
9 5. Results
5.1 Upward and Downward Bias in Stock Prices
As described in the methodology, this paper has used four sub-periods of ten days each. The ten-day period after the ban was imposed (lifted) has been compared with the ten-day period preceding the imposing (lifting) of the ban.
Share Price - Group of
Banned Stocks Period 10 Days Before Start Ban 10 Days After Start Ban 10 Days Before End Ban 10 Days After End Ban
Period After Start -/- Before Start After End -/- Before End Means 5.518 5.534 4.134 3.919 Difference 0.29% -5.20% St. Dev. 0.091 0.073 0.042 0.047 T-statistic 0.207 -2.681
Share Price - Group of
Control Stocks Period 10 Days Before Start Ban 10 Days After Start Ban 10 Days Before End Ban 10 Days After End Ban
Period After Start - Before Start After End - Before End Means 6.967 6.620 5.485 5.008 Difference -4.98% -8.70% St. Dev. 0.077 0.078 0.058 0.050 T-statistic -6.429 -5.552
The four sub-period paired t-test results are listed above. The top and bottom halves of the table represent results for banned and control stocks respectively. For all ten-day sub-periods I have calculated the means and standard deviations. Defaulted stocks or stocks for which not enough information was available have been excluded. On the right hand side of the table, the differences between the sub-periods are shown. The t-statistics have a 95% confidence interval. For the first period (imposing of the ban), I find that the t-statistic of the banned group of stocks (0.207) does not show a significant difference, whereas the t-statistic of the control group
(-6.429) shows a significant decrease. This would imply an upward bias.
10
significant decrease, whereas the t-statistic of the control group (-5.552) shows an even larger decrease. This is not in line with my expectation of a downward bias.
The movements of the share prices for both the banned group of stocks and the control group over the entire period is shown graphically above. The vertical axis contains share price, while the horizontal axis depicts time (in trading days). I have also included two vertical lines in this graph. The first vertical line indicates the implementation of the ban (September 19, 2008), whereas the second vertical line indicates the day the ban was lifted (January 16, 2009). The graph starts at July 21, 2008 (t = 0) and ends at March 19, 2009 (t = 170).
5.2 Intraday Volatility
The intraday volatility for the banned and control group of stocks has been plot and the result of this plot can be seen below. Intraday volatility and time are on the vertical and horizontal axis, respectively. The two vertical lines represent the days the ban was imposed and lifted.
11
In the graph we can see that the volatility increased for both banned and non-banned stocks at the moment the short-selling ban was imposed. The increase in volatility for banned stocks outweighs that of the control group however. During the second half of the ban, the volatility for both groups decreases again, but remains above the pre-ban level. When the ban was lifted volatility spiked for banned stocks, but remained stable for control group stocks. After the spike, volatility for banned stocks remains at mostly the same level as the second half of the ban period.
One could look at the graph and conclude that the short selling ban increased volatility for banned stocks, because the volatility for banned stocks increased more than the volatility for the control group. In the second half of the ban however, volatility (for both groups) decreased again. This would imply that the short-selling ban caused a short-term increase in volatility.
However, this is not in line with any of the theories discussed in the literature review. Instead, we would expect a short selling ban to exclude pessimists from the stock market, thus
12
partially relieving banned stocks of downward pressure. This would result in less volatility (less spread in stock prices).
The result is clouded, because the short selling ban was sector-specific, as it only targeted financial firms. Sector-specific factors could therefore influence the results. A possible explanatory factor could be the fact that the crisis in 2008-09 was mostly a financial crisis, and thus affected financial firms more than regular firms. This would imply that the banned group of stocks would be more volatile in this period than the control group stocks. The large increase in that specific period for both groups of stocks might be appointed to the unstable market conditions that prevailed. Only a few days before the short selling ban was implemented the American bank Lehman Brothers defaulted, which had significant effects on global stock markets. This event could cause the higher volatility during that time.
Immediately after the ban was lifted, we observe a spike in intraday volatility for the banned group of stocks. This is potentially caused by a large number of short selling orders on the first day the ban was lifted. This would indicate that the market is still trying to correct the stock prices of the banned stocks, as it was unable to do so when the ban was still in place. For the calculation of t-statistics, I consider it an outlier and therefore do not include this observation.
After the ban is lifted the intraday volatility does not return to its lower pre-ban level, but instead remains at a higher level. I believe this to be the result of ongoing effects of the financial crisis at that time.
13 Intraday Volatility - Group
of Banned Stocks Period Pre-Ban 1st Half Ban 2nd Half Ban Post-Ban
Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.047 0.100 0.071 0.084 0.075 Difference 115.67% -28.89% 16.99% St. Dev. 0.019 0.027 0.015 0.031 0.031 T-statistic 11.793 -5.777 *2.609
Intraday Volatility - Group
of Control Stocks Period Pre-Ban 1st Half Ban 2nd Half Ban Post-Ban
Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.042 0.074 0.060 0.051 0.057 Difference 74.32% -19.13% -14.44% St. Dev. 0.010 0.017 0.012 0.006 0.016 T-statistic 10.901 -4.033 -4.195
*One observation in the post-ban time period was considered an outlier and therefore not included in the calculation of the t-statistic.
The table above shows the results of the paired t-tests for intraday volatility. The top and bottom halves of the table represent results for banned and control stocks respectively. I have included means and standard deviations for the four sub-periods, as well for the total period. Defaulted stocks or stocks for which not enough information was available have been excluded. On the right hand side of the table, the differences between the sub-periods are shown. The t-statistics have a 95% confidence interval.
5.3 Liquidity
The spreads of both the banned and control group of stocks have been calculated. For all sub-periods, as well as the total period, mean spreads and standard deviations are shown in the table below. The differences between two periods and the results of the paired t-tests are also displayed. T-tests have been taken with a 95% confidence interval. Stocks that defaulted in the observed period or stocks for which not enough data was available have been excluded from the analysis.
14 Spread - Group of Banned
Stocks Period Pre-Ban 1st Half Ban 2nd Half Ban Post-Ban
Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.23% 0.28% 0.28% 0.31% 0.28% Difference 23.84% -2.83% 11.79% St. Dev. 0.05% 0.07% 0.08% 0.02% 0.06% T-statistic 4.694 -0.465 2.419
Spread - Group of Control
Stocks Period Pre-Ban 1st Half Ban 2nd Half Ban Post-Ban
Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.27% 0.33% 0.36% 0.37% 0.33% Difference 20.73% 10.73% 2.27% St. Dev. 0.02% 0.03% 0.05% 0.01% 0.05% T-statistic 8.065 3.253 0.800
The analysis of the bid-ask spreads leads us to the following results: On average, banned group stocks have a smaller bid-ask spread compared to control group stocks. This would suggest that the banned stocks are on average more liquid. This result can most likely be attributed to the fact that the banned group of stocks are all financials firms.
Another observation that can be made is that bid-ask spreads increase significantly for both group of stocks during the ban period. These spreads are also very volatile for both groups in the ban period, as shown by the higher values for standard deviation in the table above. In the post-ban period the bid-ask spreads remain high, but less volatile. Because bid-ask spreads remain high even after the ban was lifted, the observed results indicate that bid-ask spreads increased not necessarily as a result of the ban itself. It is likely that the higher bid-ask spreads were instead caused by the prevailing financial crisis at that time.
The hypothesis regarding liquidity stated the expectation that banned stocks have relatively higher bid-ask spreads as a result of the ban. This paper cannot prove this hypothesis on the basis of the observed results of the paired t-tests. Instead, we find that the increase in bid-ask spreads of non-banned stocks outweighed that of banned stocks in the ban period, as shown by the higher t-value for the control group in the table. One possible explanation for this result could be that the paired t-test might not provide accurate results in this particular analysis. The
15
bid-ask spreads for banned stocks showed a relatively high standard deviation during the entire period, which diminishes the power of the paired t-test.
A graphical representation of the movements of bid-ask spreads over the course of the entire period is shown above. On the vertical axis the value of the bid-ask spread is shown, whereas the horizontal axis indicates time. The two vertical lines represent the days the ban was imposed and lifted.
5.4 Results of the Large MCAP Group Analysis
5.4.1 Upward and Downward Bias in the Large MCAP Group
The results of the specific control group analysis indicate that both an upward as well a
downward bias were present after the imposing and lifting of the ban. For the first period, the t-statistic for the banned group 0.849) is larger than the t-statistic for the control group (-4.824), indicating an upward bias in share price when the ban was imposed. For the second period, the t-statistic for the banned group (-6.724) is smaller than the t-statistic for the control
16
group (-2.017), indicating a downward bias in share price when the ban was lifted. These results can be verified in the table below. T-statistics were calculated with a 95% confidence interval.
Share Price - Banned Group of Large MCAP Stocks
Period 10 Days Before Start Ban 10 Days After Start Ban 10 Days Before End Ban 10 Days After End Ban Period After Start -/- Before Start After End -/- Before End Means 8.635 8.489 4.387 3.208 Difference -1.69% -26.87% St. Dev. 0.701 0.492 0.310 0.290 T-statistic -0.849 -6.724
Share Price - Control Group of Large MCAP Stocks
Period 10 Days Before Start Ban 10 Days After Start Ban 10 Days Before End Ban 10 Days After End Ban Period After Start - Before Start After End - Before End Means 14.728 14.043 12.317 12.001 Difference -4.65% -2.57% St. Dev. 0.423 0.479 0.402 0.164 T-statistic -4.824 -2.017
I have also included a graph depicting the share price over the course of the entire period. On the vertical axis the share price is shown, whereas the horizontal axis indicates time. The two vertical lines represent the days the ban was imposed and lifted.
17 5.4.2 Intraday Volatility of the Large MCAP Group
A graphical representation of the intraday volatility of the "Large MCAP" group can be found below. On the vertical axis the intraday volatility is shown, whereas the horizontal axis indicates time. The two vertical lines represent the days the ban was imposed and lifted.
The results of the "Large MCAP" group do not differ substantially from the results of the entire banned group analysis of volatility. Volatility for both the banned and control group stocks increased significantly in the ban period. When the ban was lifted volatility remained higher than the pre-ban level. The spike in volatility on the day the ban was lifted is also observed here.
The results of the paired t-tests are on the next page. I consider the observed spike in volatility an outlier and therefore exclude it from the calculations. All t-statistics were calculated with a 95% confidence interval.
18
Intraday Volatility - Banned Group of Large MCAP Stocks
Period Pre-Ban 1st Half Ban
2nd Half Ban
Post-Ban Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.048 0.118 0.084 0.133 0.097 Difference 146.14% -29.01% 58.90% St. Dev. 0.016 0.046 0.028 0.088 0.061 T-statistic 8.853 -3.881 4.552
Intraday Volatility - Control Group of Large MCAP Stocks
Period Pre-Ban 1st Half Ban
2nd Half Ban
Post-Ban Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.033 0.065 0.051 0.041 0.047 Difference 97.88% -22.36% -20.12% St. Dev. 0.006 0.019 0.014 0.007 0.017 T-statistic 9.470 -3.586 -4.244
*One observation in the post-ban time period was considered an outlier and therefore not included in the calculation of the t-statistic. 5.4.3 Liquidity of the Large MCAP Group
The table below shows means, standard deviations and t-statistics for the bid-ask spread of the "Large MCAP" group. T-statistics were calculated with a 95% confidence interval.
Spread - Banned Group of Large MCAP Stocks
Period Pre-Ban 1st Half Ban
2nd Half Ban
Post-Ban Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.15% 0.16% 0.17% 0.31% 0.20% Difference 8.84% 6.25% 82.35% St. Dev. 0.01% 0.06% 0.05% 0.05% 0.08% T-statistic 0.878 1.150 12.417
Spread - Control Group of Large MCAP Stocks
Period Pre-Ban 1st Half Ban
2nd Half Ban
Post-Ban Total Period 1st Half Ban -/- Pre-Ban 2nd Half Ban -/- 1st Half Ban Post-Ban -/- 2nd Half Ban Means 0.11% 0.12% 0.13% 0.13% 0.12% Difference 8.18% 7.98% 3.81% St. Dev. 0.01% 0.01% 0.01% 0.01% 0.01% T-statistic 5.411 4.607 2.634
The hypothesis regarding liquidity states that I expect the banned group of stocks to become relatively less liquid as a result of the short selling ban. To proof that liquidity has decreased more for the banned group than for the control group, the bid-ask spreads of banned stocks must have widened when the ban was imposed. However the observed t-statistics do not
19
indicate that this is the case. This is potentially caused by high standard deviations in the spread of the banned group.
When compared to the previous analysis of the entire group of banned stocks, these results exhibit more or less the same pattern. The spreads of the banned group are substantially more volatile than those of the control group, which prevents the paired t-tests to yield any significant results.
6. Conclusions
Overall this paper observes that market conditions worsened for both banned stocks and non-banned stocks during the ban period. Both the bid-ask spread and intraday volatility increased considerably. These results can however not necessarily be attributed to the short selling ban, as the entire ban and post-ban period coincided with the financial crisis. It is therefore difficult to make a clear distinction between effects caused by the short selling ban and effects caused by the financial crisis.
This paper finds evidence suggesting an upward bias in share prices, but cannot significantly proof that a downward bias occurred. Hansson and Fors (2009) and Boehmer, Jones and Zhang (2013) also find an increase in share prices in the days following the ban. They conclude however that these results are not caused by an upward bias as described by Miller (1977), but rather by signalling effects and government support programmes.
Both the bid-ask spreads and intraday volatility increased significantly for both banned stocks and non-banned stocks during the short selling ban. This is in line with results from Hansson and Fors (2009) and Marsh and Payne (2012). This paper is not able to validate however that the ban worsened market quality more for banned stocks in comparison to other stocks.
The analysis of the "Large MCAP" group indicates that both an upward and downward bias was present when the ban was imposed and lifted. For this group, the intraday volatility and bid-ask spreads show mostly the same pattern as those of the entire group of banned stocks.
20
Because the upward and downward biases in share price are of such a short-term nature, data with a daily time interval might not be sufficient to accurately analyse the bias effect. A future study using for instance hourly or semi-daily data could provide more insight in the short-term movements of share prices and yield more significant results.
This paper has shown an example of the effects on a specific group of banned stocks with a control group matched by their value of market capitalisation. A suggestion for future research could be to investigate the effects of short selling bans on banned stocks with control groups matched on other variables, such as for instance company-specific betas.
21 Bibliography
Boehmer, E., Jones C.M., Zhang, X., (2008). Which shorts are informed? Journal of Finance 63,
491-527.
Boehmer, E., Jones C.M., Zhang, X., (2013). Shackling short sellers: the 2008 shorting ban. Diamond, D.W. & Verrecchia, R.E., (1987). Constraints on short selling and asset price adjustment to private information. Journal of Financial Economics 18, 277-311.
Hansson, F. & Fors, E.R., (2009). Get Shorty? Market impact of the 2008-09 U.K. short selling ban. Working Papers in Economics, no. 365.
Jones, C.M., 2012. Shorting restrictions: Revisiting the 1930s. The Financial Review 47, 1-35. Marsh, I.W. & Payne, R., (2012). Banning short sales and market quality: The UK's experience.
Journal of Banking & Finance 36, 1975-1986.
Miller, E., (1977). Risk, uncertainty, and divergence of opinion. Journal of Finance 32 (4),
1151-1168.
Nagel, S., (2004). Short sales, institutional investors and the cross-section of stock returns.
Journal of Financial Economics 78, 277-309.
Ofek, E., Richardson, M. & Whitelaw, R.F., (2004). Limited arbitrage and short sales restrictions: Evidence from the options markets. Journal of Financial Economics 78, 305-342.
Parkinson, M., (1980). The Extreme Value Method for Estimating the Variance of the Rate of Return. The Journal of Business 53 (1), 61-65.
