Short Selling Bans and Institutional Investors' Herding Behavior: Evidence from the Global Financial Crisis Martin T. Bohla, Arne C. Kleina and Pierre L. Siklosb a Department of Economics, Westphalian Wilhelminian University of Münster, Germany b Department of Economics, Wilfrid Laurier University and Viessmann European Research Centre, Canada Testable Hypotheses • Does herding become more relevant during a financial crisis? In other words, are regulators desired to displace SS during a crisis because herding is exacerbated during falling markets? – YES? Herding implies investors’ difference of opinion is relatively small – NO? Divergences of opinion increase during a crisis. Therefore, adverse herding is a possibility • SENTIMENT plays a role • Is the evidence for/against herding similar across countries? Contribution to the literature • Do short sales constraints (SSC) have a significant impact on herding behavior? Contribution to the literature • Do short sales constraints (SSC) have a significant impact on herding behavior? • The answer, in turn, entails important information for stock market regulators Contribution to the literature • Do short sales constraints (SSC) have a significant impact on herding behavior? • The answer, in turn, entails important information for stock market regulators • and deepens insights into institutional investors’ trading behavior Markets under consideration Setting and Data: Short Sale Constraints in The United States • 07/15/2008 – 08/12/2008 Naked short sales in (18) selected stocks United Kingdom • 09/19/2008 - 01/16/2009 All economic short positions in (32) selected stocks Germany • 09/22/2008 – 01/31/2010 Naked short sales in (10) selected stocks Markets under consideration France • 09/22/2008 – 01/31/2010 Short Sales in (12) selected stocks South Korea • 09/30/2008 All short sales • 06/01/2009 Lifted for non-financials Markets under consideration Australia • 09/22/2008 Naked short sales • 11/19/2008 Lifted for non-financials being member of the S&P/ASX 200 and APRA-regulated business • 05/24/2009 Ban expires Markets under consideration US UK GER FR ROK AUS No. Stocks banned (N) 18 29 10 12 16 44 No. Stock Control group 18 29 10 12 16 44 T 17 83 343 347 317 127 Markets under consideration Institutional investors holdings GDP • US & UK > 200% • GER, FR, AUS > 100% • ROK ≈ 90% In 2007 (Gonnard (2008)) Literature Review • Miller (1977), Diamond and Verrecchia (1987) • Short selling bans Miller (1977) • Divergence of opinion Miller (1977) • Divergence of opinion • SSC deter pessimists from expressing their beliefs Miller (1977) • Divergence of opinion • SSC deter pessimists from expressing their beliefs • therefore, market prices are build upon optimists’ valuation Miller (1977) • Divergence of opinion • SSC deter pessimists from expressing their beliefs • therefore, market prices are build upon optimists’ valuation Overvaluation Diamond and Verrecchia (1987) SSC reduce informational efficiency: new information is impounded into prices with a delay Diamond and Verrecchia (1987) SSC reduce informational efficiency: new information is impounded into prices with a delay • this holds for both positive and negative news Diamond and Verrecchia (1987) SSC reduce informational efficiency: new information is impounded into prices with a delay • this holds for both positive and negative news • but the effect is stronger for negative information Crisis related Bans Previous literature on the short selling bans reports strong evidence for deteriorations in market quality and liquidity • Bris (2008), Boulton and Braga-Alves (2010) and Kolanski et al. (2010) for the July/August ban in the US • Boehmer et al. (2009) and Kolanski et al. (2010) for the September/October ban in the US Crisis related Bans • Marsh and Payne (2010) for the UK • Helmes et al. (2010) for Australia • A broad international perspective incl. 30 countries is given in Beber and Pagano (2011) Beber and Pagano (2011) • Their results for 30 countries underscore negative effects on market liquidity Beber and Pagano (2011) • Their results for 30 countries underscore negative effects on market liquidity • In addition, they find increased autocorrelations in the residuals of market model regressions Empirical Approach • We aim at identifying the impact of short sale constraints on herding behavior Empirical Approach • We aim at identifying the impact of short sale constraints on herding behavior 1. A measure of herding is needed Empirical Approach • We aim at identifying the impact of short sale constraints on herding behavior 1. A measure of herding is needed 2. Control for the effects of the crisis per se is needed Empirical Approach • We aim at identifying the impact of short sale constraints on herding behavior 1. A measure of herding is needed 2. Control for the effects of the crisis per se is needed 3. Robust inference based on small/medium size samples Measure of Dispersion Measure of Dispersion as an input to evaluate Herding: details Measure of Dispersion • St = dispersion: captures a key characteristics of herd behavior – N = number of stocks, – T = number of observations – rit = return, stock i, time t; – rmt = cross-sectional weighted average of returns in a ‘portfolio’ of N stocks NOT an E(r) Average deviation of a stock from the market proxies how investors discriminate between stocks Christie and Huang (1995) Rational Pricing Herding Adverse Herding Methodology: Regression Form (2) > 0 under rational Asset pricing; {e.g., changing may be one reason} BANNED CONTROL IMPLIES SSR have an effect 0 means deviation from rational Asset pricing Proxies variance since 2 2 2 mt E (rmt ) E (rmt )2 E (rmt ) Chang et al. (2000) Autocorrelation: Schwert’s criterion From max to min, use a 10% criterion Matching • Matching variables: Market capitalization, trading volume and market beta (all standardized) Matching • • Matching variables: Market capitalization, trading volume and market beta (all standardized) Matching metric: Sum of squared differences in those three variables (Euclidean distance) Interpretation In general, evidence supporting an effect of short sale constraints is found if the estimate for significantly differs between test and control groups Interpretation In particular, support for regulators’ point of view is given in case of a dampening effect of SSC on herding which, in turn, is found if is significantly negative for the control group while being equal to zero for the banned stocks. Control 0 Test 0 Interpretation By contrast, evidence in line with a amplifying effect of SSC on herding, is found if is negative for the test group but equal to zero for the control stocks. Control 0 Test 0 Bootstrap A bootstrap algorithm • enables us to draw reliable inference from small and medium samples Bootstrap A bootstrap algorithm • enables us to draw reliable inference from small and medium samples • allows us to directly test the H0 of Rational Asset Pricing (i.e., CAPM-type) Bootstrap We generate data by the following processes 1. r r i ,t i i m ,t i ,t Bootstrap We generate data by the following processes r r 1. 2. i ,t i i m ,t i ,t r r SMB HML i ,t i i m ,t i i i ,t Further empirical issues • Persistently rising vs falling markets may make a difference: sort St, by length of runs l {1,2} Further empirical issues • Persistently rising vs falling markets may make a difference: sort St, by length of runs l {1,2} • Threshold effects? Further empirical issues • Persistently rising vs falling markets may make a difference: sort St, by length of runs l {1,2} • Threshold effects? • Small cap versus large cap: former exhibit more herding than latter; former lag latter in terms of correlation of returns Empirical Results Recall that we bootstrap deviations from Rational Asset Pricing Empirical Results Recall that we bootstrap deviations from Rational Asset Pricing • Significance does not mean significantly different from zero Empirical Results Recall that we bootstrap deviations from Rational Asset Pricing • Significance does not mean significantly different from zero • but significantly different from the value implied by the asset pricing model Empirical Results Adverse Herding! Herding Empirical Results Empirical Results • Almost no herding (either adverse or regular) in case of unbanned stocks Empirical Results • Almost no herding (either adverse or regular) in case of unbanned stocks • strong evidence for adverse herding for the stocks subject to the constraints for some countries Interpretation • It is well known in the literature that short sale constraints create uncertainty about fundamental asset values Interpretation • It is well known in the literature that short sale constraints create uncertainty about fundamental asset values • The work of Hwang and Salmon (2004, 2009) suggests that during such turmoils investors loose trust in the market consensus and come back to fundamental based pricing Interpretation • It is well known in the literature that short sale constraints create uncertainty about fundamental asset values • The work of Hwang and Salmon (2004, 2009) suggests that during such turmoils investors loose trust in the market consensus and come back to fundamental based pricing • This may show up in adverse herding, via an increased dispersion of returns Thank you for your attention!