The Impact of Short Sales Restrictions
Ian W Marsh and Norman Niemer*, Cass Business School
Restrictions on short selling financial sector stocks were introduced in the UK on 19 September 2008. Several countries followed this move, though to different extents. In this report we analyse the impact these restrictions had on the behaviour of stock returns. We characterise stock returns using standard moment definitions (mean, median, standard deviation, skewness and kurtosis).
The literature on the impact of constraints on short selling would suggest that we might initially expect to see higher skewness in the returns of stocks subject to the restrictions, probably in conjunction with lower kurtosis. This would result from the exclusion of pessimistic investors from the market. Other things equal, mean returns should be higher and standard deviations should be lower. Given the extraordinary nature of economic and financial news during this period, the other things equal assumption is even more heroic than usual and we do not focus extensively on these two moments.
We also consider first order autocorrelation coefficients and the strength of the correlation between individual stock returns and the market index return (a goodness of fit statistic). If the imposition of restrictions excludes a class of participants with relevant information, it might be expected that the efficiency of the market declines. This would be reflected in an increase in the magnitude of autocorrelation coefficients and, more controversially, by a rise in the goodness of fit statistics.
Our main findings can be summarised as follows:
1. We find no strong evidence that the imposition of restrictions on short selling in the UK or elsewhere changed the behaviour of stock returns. Stocks subject to the restrictions behave very similarly both to how they behaved before the imposition of restrictions and to how stocks not subject to the restrictions behave.
2. Further, comparing behaviour across countries where the nature of the restrictions differs, we fail to find systematic patterns consistent with the expected effect of the new regulations.
3. We also find no sign of the expected detrimental impact of constraints. Autocorrelation coefficients and goodness of fit statistics are if anything slightly lower in the post restrictions period.
4. Regression analysis suggests that any change in the key statistics is mainly driven by sector-wide influences rather than the restrictions on short selling. That is, some systematic changes in the behaviour of financial sector stocks could be discerned but no strong evidence of a systematic impact of the restrictions could be identified.
One major caveat to our analysis is worth mentioning. The post restrictions period analysed is short in most countries (typically 30 or 31 days) and in some countries is extremely short (12 days in the US and Canada). Placing great emphasis on a statistical analysis of such short samples is foolish.
Many of the statistics presented in this report are very sensitive to the addition or subtraction of one or two days in the sample. However, we feel that the main thrust of the findings appears to be robust, even if the exact numbers are not.
To view the paper, please click here (244KB)
*: This independent analysis was commissioned and funded by the International Securities Lending Association (ISLA) the Alternative Investment Management Association (AIMA) and London Investment Banking Association (LIBA). However, the results and interpretations are the independent work of the cited authors and are not necessarily those of any of the commissioning and funding organisations.
The literature on the impact of constraints on short selling would suggest that we might initially expect to see higher skewness in the returns of stocks subject to the restrictions, probably in conjunction with lower kurtosis. This would result from the exclusion of pessimistic investors from the market. Other things equal, mean returns should be higher and standard deviations should be lower. Given the extraordinary nature of economic and financial news during this period, the other things equal assumption is even more heroic than usual and we do not focus extensively on these two moments.
We also consider first order autocorrelation coefficients and the strength of the correlation between individual stock returns and the market index return (a goodness of fit statistic). If the imposition of restrictions excludes a class of participants with relevant information, it might be expected that the efficiency of the market declines. This would be reflected in an increase in the magnitude of autocorrelation coefficients and, more controversially, by a rise in the goodness of fit statistics.
Our main findings can be summarised as follows:
1. We find no strong evidence that the imposition of restrictions on short selling in the UK or elsewhere changed the behaviour of stock returns. Stocks subject to the restrictions behave very similarly both to how they behaved before the imposition of restrictions and to how stocks not subject to the restrictions behave.
2. Further, comparing behaviour across countries where the nature of the restrictions differs, we fail to find systematic patterns consistent with the expected effect of the new regulations.
3. We also find no sign of the expected detrimental impact of constraints. Autocorrelation coefficients and goodness of fit statistics are if anything slightly lower in the post restrictions period.
4. Regression analysis suggests that any change in the key statistics is mainly driven by sector-wide influences rather than the restrictions on short selling. That is, some systematic changes in the behaviour of financial sector stocks could be discerned but no strong evidence of a systematic impact of the restrictions could be identified.
One major caveat to our analysis is worth mentioning. The post restrictions period analysed is short in most countries (typically 30 or 31 days) and in some countries is extremely short (12 days in the US and Canada). Placing great emphasis on a statistical analysis of such short samples is foolish.
Many of the statistics presented in this report are very sensitive to the addition or subtraction of one or two days in the sample. However, we feel that the main thrust of the findings appears to be robust, even if the exact numbers are not.
To view the paper, please click here (244KB)
*: This independent analysis was commissioned and funded by the International Securities Lending Association (ISLA) the Alternative Investment Management Association (AIMA) and London Investment Banking Association (LIBA). However, the results and interpretations are the independent work of the cited authors and are not necessarily those of any of the commissioning and funding organisations.
