Cigi Paper 15

7/28/2019 Cigi Paper 15

1/22

CIGI PaPers

no. 15 aPrIl 2013

Are Short SellerSPoSitive FeedbAcktrAderS? evidenceFrom the GlobAlFinAnciAl criSiS

MartIn t. Bohl, arne C. KleIn andPIerre l. sIKlos

7/28/2019 Cigi Paper 15

2/22

7/28/2019 Cigi Paper 15

3/22

Are Short SellerS PoSitive

FeedbAck trAderS? evidence

From the GlobAl FinAnciAl criSiS

Marti T. Bhl, Ar C. Kli a Pirr L. Sikls

7/28/2019 Cigi Paper 15

4/22

Copyright 2013 by The Centre or International Governance Innovation.

The opinions expressed in this publication are those o the authors and do not

necessarily reect the views o The Centre or International Governance Innovationor its Operating Board o Directors or International Board o Governors.

This work is licensed under a Creative Commons Attribution Non-commercial No Derivatives License. To view this license, visit (www.creativecommons.org/licenses/by-nc-nd/3.0/). For re-use or distribution, please include this copyrightnotice.

ACKnowLedgeMenTS

We are indebted to Christian A. Salm or helpul comments and suggestions. Theauthors would like to thank the participants o the International Finance and BankingSociety Conerence (Rome), the seminar o the Financial Services Research Centre atWilrid Laurier University (Waterloo), the 15th Conerence o the Swiss Society orfnancial market research (Zurich), the 19th Annual Conerence o the MultinationalFinance Society (Krakow), the Workshop on Reorming Finance: Balancing Domesticand International Agendas (Ljubljana) as well as two anonymous reerees. Pierre L.Siklos would like to thank The Centre or International Governance Innovation (CIGI)or fnancial support.

57 Erb Street WestWaterloo, Ontario N2L 6C2Canadatel +1 519 885 2444 ax + 1 519 885 5450www.cigionline.org

7/28/2019 Cigi Paper 15

5/22

tAble oF contentS

4 About the Authors

4 Executive Summary

4 Introduction

6 Banned Stocks, Construction o Control Groups, andData

8 Methodology

10 Empirical Results

14 Conclusion

15 Works Cited

18 About CIGI

18 CIGI Masthead

7/28/2019 Cigi Paper 15

6/22

CIGI PaPers no. 15 aPrIl 2013

4 The CenTre for InTernaTIonal GovernanCe InnovaTIon

executive SummAry

Short sellers are routinely blamed or destabilizing stockmarkets by exacerbating deviations rom undamentalvalues. In response, regulators periodically impose short-sale constraints aimed at preventing excessive stock marketdeclines. One explanation is that policy makers regard shortsellers as behaving like positive eedback traders. Relying

on the theoretical model put orward by Sentana andWadhwani (1992), which stresses the conditional natureo returns persistence, bans on selected fnancial stocks insix countries during the 2008-2009 global fnancial crisisare examined. These provided a setting to analyze theimpact o short-sale restrictions on eedback trading. Thefndings suggest that, in the majority o markets examined,restrictions o this kind ampliy positive eedback tradingduring periods o high volatility and, hence, contributeto stock market downturns. On balance, thereore, short-selling bans do not contribute to enhancing fnancialstability.

introduction

During the recent global fnancial crisis in 2008-2009,regulators, politicians and high-profle media coverage

blamed short sellers or ampliying stock marketdownturns. In this spirit, regulatory authorities aroundthe world imposed bans on short sales with the hope ostabilizing stock markets, thereby preventing excessiveprice declines. For instance, in the announcement o the

July/August 2008 ban on naked shorts, the US Securitiesand Exchange Commission (SEC) declared that there waspanic selling: As a result, the prices o securities may

artifcially and unnecessarily decline well below the pricelevel that would have resulted rom the normal pricediscovery process (SEC, 2008). A potential rationalizationor this kind o behaviour is to view short sellers as akinto positive eedback traders who ampliy deviations romundamental values.

Feedback trading is a well-known phenomenon duringtimes o fnancial turmoil (see: Sentana and Wadhwani,1992; LeBaron, 1992; Koutmos, 1997; Kaminsky andSchmukler, 1999; Karolyi, 2002; Kaminsky, Lyons andSchmukler, 2004; and Salm and Schuppli, 2010). However,the literature is almost completely silent about the impact

o short-sale constraints on institutional investorseedback trading behaviour. This paper aims to fll this gapby investigating the short-selling regimes in the UnitedStates, the United Kingdom, Germany, France, South Koreaand Australia during the recent global fnancial crisis. Itcontributes to the literature by providing evidence againstthe stabilizing eects o short-sale constraints, which mayexacerbate positive eedback trading rather than mitigateit. Thus, banning short sellers may actually ampliy marketdownturns rather than attenuate them.

About the AuthorS

Martin T. Bohl is proessor o economics,Centre or Quantitative Economics, WestphalianWilhelminian University o Mnster. From 1999to 2006, he was a proessor o fnance and capital

markets at the European University ViadrinaFrankurt (Oder). His research ocuses onmonetary theory and policy as well as fnancialmarket research.

Arne C. Klein is an assistant lecturer inDepartment o Economics at the WestphalianWilhelminian University o Mnster. From July toOctober 2011, he was a visiting scholar at WilridLaurier University, Waterloo, Canada.

Pierre L. Siklos, a CIGI senior ellow, is thedirector o the Viessmann European ResearchCentre at Wilrid Laurier University, and a researchassociate at Australian National University'sCentre or Macroeconomic Analysis. His researchinterests are in applied time series analysis andmonetary policy, with a ocus on ination andfnancial markets.

About the Project

This publication emerges rom a project calledEssays in Financial Governance: PromotingCooperation in Financial Regulation and Policies.The project is supported by a 2011-2012 CIGI

Collaborative Research Award held by Martin T.Bohl, Badye Essid, Arne Christian Klein, PierreL. Siklos and Patrick Stephan. In this project,researchers investigate empirically policy makersreactions to an unolding fnancial crisis and thenegative externalities that emerge in the ormo poorly unctioning fnancial markets. At themacro level, the project investigates whetherthe bond and equity markets in the throes o afnancial crisis can be linked to overall economicperormance. Ultimately, the aim is to proposepolicy responses leading to improved fnancialgovernance.

7/28/2019 Cigi Paper 15

7/22

are ShorT SellerS PoSITIve feedbaCk TraderS? evIdenCe from The Global fInanCIal CrIS

marTIn T. bohl, arne C. kleIn and PIerre l. SIkloS

Unlike the literature that reports unconditionalautocorrelations (Beber and Pagano, 2013), this paper is thefrst to highlight the conditional nature o return persistencestemming rom eedback trading, a point emphasized bySentana and Wadhwanis (1992) seminal article. Notably,the interaction between institutional investors eedbacktrading and volatility is addressed.1 This is o particularinterest to regulators, as the bans were designed or times

o market turmoil rather than or tranquil trading periods.

The debate on short selling is not new to academics (see:Boehmer, Huszar and Jordan, 2010; and Bris, Goetzmannand Zhu, 2007 or reviews). Empirical research on shortsellers investment strategies ocuses on their ability toidentiy overvalued stocks. There is widespread evidencesupporting the view that high short interest predictsnegative subsequent returns (see: Seneca, 1967; Figlewski,1981; Senchack and Starks, 1993, Aitken et al., 1998, Desaiet al., 2002; Asquith, Pathak and Ritter, 2005; and Boehme,Danielsen and Sorescu, 2006), although there are dissenters(see: Hurtado-Sanchez, 1978; Dickinson and Woolridge,

1994; and Huszar and Qian, 2011).

Direct evidence on short sellers trading strategies isscarcer. Dechow et al. (2001) document short sellers abilityto exploit inormation rom undamental-to-price ratios togenerate positive abnormal returns. Drawing on daily NewYork Stock Exchange (NYSE) order ow data, Boehmer,

Jones and Zhang (2008) show that heavily shorted stockssignifcantly underperorm relative to lightly shortedones. Similar fndings or the Nasdaq stocks are reportedin Diether, Lee and Werner (2009). Blau et al. (2010)study short sellers trading behaviour during periods ostrong market movements. In particular, they examine

the tendency o these investors to ollow the crowd whenthe market is heating. Their results suggest that whileshort sellers act as contrarian investors on average, theseinvestors are prone to herd behaviour during extreme upor down swings in the market.

Another perspective argues that short sales may be dueto arbitrage, hedging or tax-related trades and, thus, thistype o activity does not necessarily contain inormationabout uture perormance (Brent, Morse and Stice, 1990).Boehmer and Wus (2013) recent evidence lends credenceto the notion that the ability to sell short signifcantly addsto the accuracy o stock prices. In particular, a higher short

1 Measured in terms o trading volume or asset holdings, institutionalinvestors have been playing a dominant role in mature stock markets ormany years. By 2007, fnancial assets o institutional investors as a percento GDP exceeded 200 percent in the case o the United States and theUnited Kingdom, and were well over 100 percent in the other countriesconsidered in this study, with the exception o Korea, where the value isaround 90 percent o GDP (Gonnard, Kim and Ynesta, 2008). Moreover,short sales are mainly used by institutional investors whereas individualinvestors play only a minor role. Consequently, this investigation o theimpact o short-selling constraints amounts to an analysis o institutionalinvestors eedback trading behaviour.

order ow tends to increase the inormational efciencythe pricing process.

Paralleling regulators reaction to the global fnanccrisis, academic interest dealing with the impact o shosale constraints has once again experienced a revivAnalyzing the ban in the United States in July and Augu2008, Bris (2008) and Boulton and Braga-Alves (201

report evidence o negative eects on market liquidisuch as rising bid-ask spreads, lower trading volumand reductions in pricing efciency. Additionally, tresults o Harris, Namvar and Phillips (2009) and Boultand Braga-Alves (2010) lend urther support to Mille(1977) overvaluation hypothesis. The US short-selliregime in September and October 2008 is the subject Boehmer, Huszar and Jordan (2011). Their results confrdeteriorations in market quality but cannot corroboraovervaluation. Based on the same ban period, GrundLim and Verwijmeren (2012), as well as Battalio anSchultz (2011), provide evidence stressing the thesis thoptions represent a substitute to short sales. Eviden

or both US short-sale regimes given in Kolasinski, Reand Thornock (2013) supports Diamond and Verrecchi(1987) prediction that negative eects on market qualare stronger or stocks with listed options.

Marsh and Payne (2012) report lower trading volumes aorder book liquidity in the case o the United KingdoFocusing on Australia, Helmes, Henker and Henk(2011) document reduced trading activity and increas

bid-ask spreads or stocks excluded rom short sellinConsidering restrictions in 30 countries, Beber and Paga(2013) add to the evidence that short-sale constraints entreductions in market liquidity. In addition, they analy

the residuals o market model regressions reportiincreased autocorrelation or banned stocks.

Beber and Paganos (2013) paper comes closest to tstudy described in this paper. Their fndings, howevonly shed light on unconditional autocorrelationUnconditional autocorrelations in single stock aportolio returns is a well-known phenomenon (see, example: Lo and MacKinlay, 1988; Conrad and Kaul, 198Lo and MacKinlay, 1990; Mech, 1993; Chang, McQueand Pinegar, 1999; and Bris, 2008). Theoretical explanatioor unconditional serial correlation ocus on the speo price discovery, nonsynchronous trading, transacti

costs and market microstructure issues, but do not dewith institutional investors trading patterns. By contraSentana and Wadhwani (1992) show that eedback tradi

behaviour can lead to autocorrelations that are conditionon volatility. Specifcally, their model is able to explathe stylized act that increased volatility is known

be accompanied by serial correlations, which are monegative and higher in absolute values than during perioo low volatility. A version o the Sentana and Wadhwa(1992) model is used to shed light on a short-sale banimpact on institutional investors eedback trading.

7/28/2019 Cigi Paper 15

8/22



Large parts o the literature on short-sale restrictions dealwith overvaluation as suggested by Miller (1977), whoargues that excluding pessimists rom the market resultsin an upward bias in stock prices. In contrast, the presentstudy deals with eedback trading behaviour, which maydrive prices temporarily below their undamental values.Like Miller (1977), this paper adopts the approach that theintroduction o short-sale constraints leads to a change

in the investor structure in the market. In the case oMiller (1977), the ratio o optimists is increased, whereasthis paper examines a shit rom classical mean-varianceinvestors towards eedback traders.2

Focusing on bans that aect only selected fnancialstocks means the eects o the restrictions and the crisisper se can be disentangled by creating matched controlgroups rom a sample o unbanned frms. This enablesthe contrast o changes in the extent o eedback tradingin restricted stocks with those or unrestricted onesduring the time period the ban is in place. The evidencesuggests that autocorrelations in daily returns dier

signifcantly between crisis and non-crisis periods onthe one hand and between ban and non-ban periods onthe other. First, as predicted by Sentana and Wadhwani(1992), autocorrelations become much stronger in absoluteterms during the crisis. Second, and more important,there is evidence that, in the majority o countries, short-selling constraints urther intensiy institutional investorspositive eedback trading behaviour. The latter clearlycontrasts with regulators view that short-sale constraintsmay constitute a tool to stabilize stock markets duringperiods o turmoil.

The structure o the paper is as ollows. In the next section,

the timeline o the short-selling bans and the constructiono control groups is sketched. The third section outlinesthe eedback trader model and urther econometricmethodology. The ourth section discusses the empiricalresults and the fnal section is the conclusion.

bAnned StockS, conStruction

oF control GrouPS And dAtA

In many countries, short-selling bans were part o the frstregulatory changes intended as countermeasures againstalling stock market prices during the fnancial crisis

o 2008-2009. On July 15, 2008, the SEC announced anemergency order banning naked short selling in the stockso 19 large fnancial frms, which came into orce on July 21.On July 29, the day the ban was originally meant to expire,the SEC issued an extension, with the ban remaining in

2 We thank an anonymous reeree or pointing out this distinction.

orce until August 12.3 These early restrictions were onlyoreplay on September 17, the SEC imposed a ban onnaked shorting in all stocks, which came into orce at 12:00a.m. the next day. Late on September 18, ater the closingo the market session, regulators prohibited all short salesin nearly 800 fnancial stocks, eective immediately. OnOctober 2, regulators announced an extension o the banor up to 30 days beyond September 17. The ban expired

at midnight on October 8, three days ater the adoption othe so-called Troubled Asset Relie Program. This secondUS ban is not included in the analysis, since it lasted oronly 14 days and does not provide a sufcient number oobservations to consistently estimate changes in eedbacktrading.

On September 18, 2008, the Financial Services Authority(FSA) in the United Kingdom established the strongestversion o the short-selling bans considered in this study,which came into orce the next day. A prohibition to createa net short position using any instrument (includingderivatives with an exemption or market makers and

specialists), it aected 34 fnancial frms. The rule waslimited until January 16, 2009 and expired on schedule.

The German Bundesanstalt rFinanzdienstleistungsausicht preerred a relatively longleash or short sellers, only orbidding naked short sales in11 large fnancial frms. Announced on September 19, andestablished the next trading day, the ban was extendedthree times in 2008 and 2009, and fnally phased out on

January 31, 2010. In France, the Autorit des marchsfnanciers ollowed the same time schedule as in Germany,and placed limits to short selling in 15 fnancial institutions.

On September 30, 2008, the South Korean FinancialSupervisory Commission imposed a ban on all short salesin all South Korean stocks, which was justifed on thegrounds o malignant rumors in the market. On May 20,2009, it was announced that the ban would be lited ornon-fnancial stocks eective June 2009. As this rameworkremained unchanged, the analysis or South Korea wasrun until the end o 2010.

On September 22, 2008, the Australian Securities &Investments Commission prohibited naked short salesor all frms listed at the Australian Securities Exchange(ASX)and established a reporting regime or covered short

sales. In eect rom November 19, 2008, this ban was litedor all stocks, with the exception o fnancial stocks in theStandard & Poor's (S&P)/ASX 200 plus fve other stocksthat were part o businesses regulated by the Australian

3 Normally, a short seller must borrow or expect to borrow theunderlying security in question. Naked shorting implies that the sellerdoes not obtain the security. Since not all bans considered here appliedto naked shorting, it is doubtul that our results are inuenced by thisdistinction.

7/28/2019 Cigi Paper 15

9/22



Prudential Regulation Authority. This ban expired on May24, 2009.

For the United States, the United Kingdom, Germany,France and Australia, most banned stocks can be includedin the analysis.4 For South Korea, the analysis is limitedto the fnancial stocks in the KOSPI 100 due to liquidityconcerns.

For a given set o stocks and a given period, the authorsaim at comparing the return dynamics under short-saleconstraints with an unobservable hypothetical processwithout restrictions on shorting. This requires the creationo a control group with similar characteristics. In order todo so, the matching techniques also employed by Boehmer,

Jones and Zhang (2011) and Beber and Pagano (2013) areollowed. In many o the countries included in this study,all or at least all important fnancial stocks are subject tothe ban; thereore, it is necessary to match these controlgroups mainly rom non-fnancial frms.5

The matching variables have to be careully selected tobuild a reliable match on the stocks available. FollowingBoehmer, Jones and Zhang (2011), market capitalizationand trading volume are used. Capitalization controls orfrm size, while trading volume is included in order torule out an inuence o liquidity concerns on institutionalinvestors eedback trading. Furthermore, since the testgroups consist o fnancial stocks, which are, in general,characterized by above-average exposure towards themarket, the market beta calculated or the respectivecommon market index is used as an additional matchingvariable.

Similar to Boehmer, Jones and Zhang (2011), mean values othese variables are calculated or the period rom January2008 until the introduction o the ban in the case o theUnited States, the United Kingdom, Germany, France andAustralia. For South Korea, the period rom September2008 until the end o the ban on non-fnancial stocks onMay 31, 2009 is used. The aim is to choose the matchingpartners such that they reect, as closely as possible,the characteristics o the banned stocks. Thereore,Beber and Pagano (2013) are ollowed and the matching

4 In the United States, Merrill Lynch is not included as there is no

longer sufcient data. In Germany, Hypo Real Estate is excluded romthe analysis since it was nationalized and delisted during the ban. In theUnited Kingdom, Bradord & Bingley and Tawa were dropped, as thefrst was announced to be partly nationalized on September 29, 2008 andthe second was hardly traded during the ban period. In France, Dexiaand Allianz are not included in the sample, as their notations in Pariswere delisted during the ban. Data is no longer available or Paris Re.In Australia, Macquarie DDR Trust and Challenger Financial ServicesGroup are excluded because no data is available.

5 An exemption is the July/August 2008 ban in the United States,where the control group includes a lot o fnancials, or instance AmericanInternational Group. In the United Kingdom, almost 20 percent o thecontrol stocks are fnancials.

partner that minimizes the sum o squared dierencesthe mean values o the matching variables are selectwith replacement. Note that replacement is advisable prevent the composition o the control groups rom beiinuenced by the order in which we match frms to our tegroups. As the beta, volume and capitalization strongdier with respect to mean value and standard deviatiostandardized variables are used. This ensures that equ

weights are assigned to each matching variable, in tsense that the selection o control stocks depends equaon market sensitivity, trading volume and capitalization

The datasets consist o daily total returns, markcapitalizations and trading volumes o the stocks subjeto the ban as well as those in the indices used or matchicontrol groups, where the index composition as it wthe day beore the introduction o the short-sale banused. The S&P 100 (United States), the FTSE 100 (UnitKingdom), the DAX and MDAX (Germany), the CA40 and the French stocks in the Next CAC 100 (Francthe KOSPI 100 (South Korea) and the S&P/ASX 1

(Australia) are used. To estimate the models describedthe third section, value-weighted return indices rom tstocks in the test and control groups are calculated usilog returns. This avoids the necessity o estimating timseries models rom noisy single stock data.

To consistently and robustly estimate the eedback tradmodel, a relatively long sample period is preerreHowever, ewer stocks in the test and control samplare available or calculating the return indices when tperiod extends to well beore the ban. To cope with boaspects, the period rom January 2003 until December 20is used.7

Thus, or each sample, there are eight years o daily dawhich ought to be adequate to draw reliable conclusion

All time series are obtained rom Thomson ReuteDatastream. The historical constituents o the indices weprovided by S&Ps, the FTSE Group, the Deutsche BrGroup, NYSE Euronext and the Korea Exchange. Tduration o short-sale constraints in days is: 402 (SouKorea), 347 (France), 343 (Germany), 127 (Australia), (United Kingdom) and 17 (United States). The number stocks in the test and control groups are: 44 (Australia),(United Kingdom), 18 (United States), 16 (South Korea),

(France) and 10 (Germany). Table 1 provides a summarythe time schedules and key eatures o the six short-selliregimes together with some descriptive statistics or treturn indices o the stocks in the test and control group

6 A list o the stocks in the test and control groups is available uprequest.

7 This period is selected to ensure that or the countries with smallest number o stocks, Germany and France, returns o at least nstocks are available at every point in time.

7/28/2019 Cigi Paper 15

10/22



Tabl 1: ovrvi abut th Bas a dscriptiv Statistics

Ban Period Type of Ban Mean SD Ex. Kurtosis

United States

Test Group07/15/200808/12/2008 naked short sales

0.817 3.642 1.112

Control Group 0.222 3.116 1.178

United Kingdom

Test Group09/19/200801/16/2009

all economic shortpositions

0.435 4.686 3.364

Control Group 0.048 5.150 0.499

Germany


0.020 3.278 4.420

Control Group 0.030 3.261 3.542

France

Test Group09/22/200801/31/2010 all short sales

0.010 3.332 2.556

Control Group 0.038 2.299 6.053

South Korea

Test Group06/01/2009 all short sales

0.099 1.819 1.008

Control Group 0.195 1.521 0.375

Australia


0.133 2.136 0.423

Control Group 0.185 2.968 2.490

Notes: Mean, SD, and Ex. Kurtosis reer to the mean, standard deviation, and excess kurtosis o the respective market return during the ban periodwhere the market return is expressed in percentage points. In South Korea, the ban started on September 30, 2008 but with eect rom June 2009 the banwas lited or non-fnancials. In Australia, the ban started on September 22, 2008 but with eect rom November 19, 2008 it was lited or non-fnancials.

methodoloGy

Relying on the previous work o Shiller (1984), De Longet al. (1990) and Cutler, Poterba and Summers (1991),

Sentana and Wadhwani (1992) put orward a modelbased on the behaviour o two heterogeneous groups oinvestors, namely undamentalists and eedback traders.The frst group, also called smart money traders, makesits investment decisions within a rational mean-varianceramework. In particular, its relative stock holding is given

by

St=

Et1

rt

, (1)

whereEt1

rtdenotes the expectation on the stock return in

period tand the risk-ree rate. t is a positive unction

o the conditional variance 2

,t =(2

), and accounts ora risk premium in the spirit o capital asset pricing typemodels. Thus, undamentalists demand increases withthe expected excess returnE

t1r

t and decreases with 2.

Feedback traders relative holdings are determined in theollowing manner

Ft= r

t1, (2)

where captures the type and degree o eedback trading.The case o > 0 reers to positive eedback trading. Thismeans buying ater price increases and selling ater pricedeclines. Such a behaviour can be caused by stop-loss

orders, portolio insurance or trend chasing. In contrast,negative eedback trading, < 0, is in line with commonbuy low, sell high strategies.

Market clearing requires that all stocks are held so thatS

t+ F

t= 1. Together with (1) and (2) this implies

Et1

rt = (2 (2)r

t1. (3)

Note that in the absence o eedback trading, Ft = 0, (3)

collapses to the classical capital asset pricing model (seeMerton, 1973) where stock returns do not display auto-correlation. By contrast, the presence o eedback traders,

Ft 0, implies frst order serial correlation in stock returns.Relying on a linearized ormulation or the risk premium,(2) = + 2, and assuming rational expectations,r

t= E

t1r

t+

t, leads to the ollowing testable equation

rt= +2 (

0+

12) r

t1+

t, (4)

where = + , 0

= and 1

= . Given a positive risk-return relationship, > 0, positive eedback trading,

t

t t

t

t t

t t

t t

7/28/2019 Cigi Paper 15

11/22



> 0, induces negative conditional autocorrelations in stockreturns as

1= > 0. This eect increases with conditional

variance 2. By contrast, negative eedback trading, < 0,leads to positive conditional autocorrelations as

1= < 0.

By means o the term 0r

t1, (4) is able to capture

unconditional autocorrelations in stock returnsdierent rom 0 induced by eedback trading; however,

eedback trading is not the only theoretical rationale orautocorrelations in daily stock index returns. The mostcommon alternative explanations are nonsynchronoustrading, transaction costs and time-varying expectedreturns. Nonsynchronous trading, as proposed by Lo andMacKinlay (1988, 1990), rests on the assumption that somestocks in a portolio are traded at time twhile others aretraded at t+ 1. I new inormation arrives in the market inperiod t, the frst group o stocks will react to this news intwhile the inormation is impounded into the prices o thesecond group only at time t+ 1. Thereore, the returns othe overall portolio will be serially correlated.

Mechs (1993) transaction cost approach recognizes that,due to market making, bid and ask prices dier. Intuitively,inormed investors only trade i their estimate or the truestock value lies outside o the bounds defned by the bid andask quotes, that is to say, higher (lower) than the ask (bid)price. When new inormation enters the market, aectingthe undamental values o the stocks in a portolio, it maymove investors valuation or some o these securitiesoutside o these bounds. However, the news may not besignifcant enough to move investors assessment o thetrue stock value beyond the bid or ask price in the caseo other stocks. For the second group o stocks, there may

be no change until later on, when additional inormation

arrives or simply because noise trading changes the bidand ask quotes. Similar to the nonsynchronous tradinghypothesis, auto-correlated portolio returns may be theconsequence.

By contrast, the time-varying expected returns modelproposed by Conrad and Kaul (1988) is a purely empiricalapproach that is not limited to portolio return but alsoapplies to single stocks. Conrad and Kaul (1988) assumethat expected returns are driven by an autoregressive(AR)(1) or random walk process and use Kalman fltertechniques to extract these returns. Although they are ableto reject the random walk, the economic determinants

behind such an AR(1) process remain unknown.

Empirical evidence, however, suggests that the observedautocorrelations are too large to trace back to theseexplanations. For instance, the results o Mech (1993) andBoudoukh, Richardson and Whitelaw (1994) lend littlesupport to Lo and MacKinlays (1988, 1990) nonsynchronoustrading hypothesis. Similarly, the transaction costs modelput orward by Mech (1993) and the time-varying expectedreturns proposed by Conrad and Kaul (1988) ail to explaina large portion o the serial correlation in index returns

(see McQueen, Pinegar and Thorley, 1996 and Ogde1997). Moreover, it should be stressed that the hypothesoutlined above reer to unconditional autocorrelatioand, hence, are unable to explain the empirical observatithat return autocorrelations turn negative during timeshigh volatility.

To shed light on potential changes in eedback tradin

behaviour due to short-selling restrictions, (4) is extendin the ollowing way

rt=+2

(

0+

12+

2ISSR2) r

t1+

t.

The dummy variable ISSR is equal to 1 i short sarestrictions are in place and 0 otherwise. In the presepaper, positive eedback trading is the ocus o intere

because it may ampliy stock market downturns adeviations rom undamental values in times o hiconditional variance, 2, during periods o short-selliconstraints. Thus, the parameters

1and

2are o particu

importance where 2

accounts or potential changes the extent o eedback trading when the ban is in placGiven positive eedback trading, so that

1> 0, a parame

2 = 0 indicates unchanged positive eedback tradi

during the period when the constraints are in eeIntensifed positive eedback trading is ound in the cawhere

2> 0 since the coefcient on 2r

t1rises to

1+

as long as the restrictions are in orce. By contrast, fndithat

2< 0 is evidence or a moderation o positive eedba

trading during the ban.

One might argue that changes in positive eedbacktrading patterns might be explained by fnancialturmoil rather than by the short-selling constraints perse. Thereore, to disentangle the eects o the short-selling ban and the crisis, the results or the bannedstocks are contrasted against those in the unrestrictedcontrol groups. There are three possible parameterconstellations. First, i the parameter

2does not dier

between the stocks in the test and control groups, short-sale constraints do not aect eedback trading. Second,i the test groups parameter, Test is greater than the oneound or the group o unrestricted stocks, Control, the baamplifes positive eedback trading. I this is the case, adisproportionately high share o positive eedback tradein the market sell ater past price declines irrespective

o undamental values, exacerbating fnancial distress.This would be evidence or the destabilizing eectso short-sale constraints during stock market turmoil.Third, a value or

2being lower or banned stocks

compared to unconstrained ones indicates a dampeningeect on positive eedback strategies and, thus, supportregulators point o view that short-selling bans stabilizestock markets during crises. Concisely stated, whenassuming positive eed back trading, that is,

1> 0, a

destabilizing eect is ound iTest Control is positive,

t

t t t t

t

t

t

2

2

2 2

7/28/2019 Cigi Paper 15

12/22



whereas a negative dierence is in line with a stabilizingimpact o the constraints.

To take into account volatility clustering and ARCH eects,(5) is jointly estimated with Bollerslevs (1987) GARCH (1,1) approach

2= +0

t1

+1

t1

, (6)

where the parameter restrictions ,0,

1> 0 and

0+

1< 1

apply. Finally, t-tests are perormed on the signifcance odierences in

2between test and control groups. To check

or robustness, the eedback trader model is re-estimated(5) using the T-GARCH specifcation proposed by Glosten,

Jagannathan and Runkle (1993). The models are estimatedby quasi maximum likelihood, with standard errorscorrected as proposed by Bollerslev and Wooldridge(1992).

emPiricAlreSultS

The empirical approach to measuring the inuenceshort-sale constraints exert on eedback trading relies onmatched control samples. Thereore, assessing the qualityo these matches is important. Since autocorrelated stockreturns are being dealt with, the autocorrelation unctionsor up to 25 lags or the test and control groups in a given

country are compared, where the period when the ban is ineect is excluded. The results, displayed in Figure 1, showthat the serial dependence in returns is relatively similaramong test and control stocks. In addition to the visualcomparison, the correlation between the autocorrelationunctions o the test and control groups or a given countryis also considered. In the case o Australia, there is amoderate correlation o 0.273.

For all other markets, however, the correlation coefcientslie between 0.593 (United Kingdom) and 0.850 (UnitedStates).

Fiur 1: Autcrrlati Fuctis fr Six Cutris

0.15

0.1

0.05

0

0.05

0.1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

USTest USControl

0.1

0.08

0.06

0.04

0.02

0

0.02

0.04

0.060.08

0.1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

UKTest UKControl

0.1

0.08

0.06

0.04

0.02

0

0.02

0.04

0.06

0.08

0.1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

GermanyTest GermanyControl

0.1

0.08

0.06

0.04

0.02

0

0.02

0.04

0.06

0.08

0.1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

FranceTest FranceControl

0.1

0.08

0.06

0.04

0.02

0

0.02

0.04

0.06

0.08

0.1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

S.KoreaTest S.KoreaControl

0.1

0.08

0.06

0.04

0.02

0

0.02

0.04

0.06

0.08

0.1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

AustraliaTest AustraliaControl

Notes: Autocorrelation unctions with 25 lags or daily value-weighted return indices or the test and control groups or the United States, the UnitedKingdom, Germany, France, South Korea and Australia.

t

2 2

7/28/2019 Cigi Paper 15

13/22



Table 2 provides average values or market capitalizations,trading volumes and market betas, that is, the threematching variables or each test and control group, to veriythe absence o a systematic bias in the matched samples.In most cases, the values or all three variables matchrelatively closely between the test and control groups. The

market capitalization is, on average, two percent highor the control stocks while their trading volume is abo6.5 percent lower. The sensitivity towards the market is, average, fve percent lower or the control groups. Henthese relatively small dierences suggest that our contstocks are not subject to any signifcant systematic bias.

Tabl 2: Matchi Statistics

Market Capitalization Trading Volume

United States

Test Group 81895 675438 1.949

Control Group 62490 671904 1.747

United Kingdom

Test Group 9275 6304323 1.140

Control Group 12331 7102369 1.102

Germany

Test Group 16630 1218 1.139

Control Group 15319 1056 1.066

FranceTest Group 28666 120472 1.241

Control Group 22143 77008 1.087

South Korea

Test Group 4414913 52424645 1.225

Control Group 4476268 51265244 1.228

Australia

Test Group 8907 45516 1.063

Control Group 11691 45682 1.082

Notes: Average values or market capitalizations, trading volumes and market betas or test and control groups over the nine months preceding respective ban. Trading volume is expressed in thousand units o home currency, while market capitalization reers to a million units o home curren

The parameter estimates or the baseline model given in (5)and (6) are reported in Table 3. As with most daily fnancialtime series, strong ARCH eects and volatility clustering,measured by

0and

1, are present. The stationarity

conditions or the parameters o the conditional variance

equation are met in all cases. Turning to the meequation, all stock return indices display unconditionautocorrelations dierent rom 0 as all

0parameters a

statistically signifcant at the one percent level.

Tabl 3: gARCH estimati Rsults fr th Fback Trar Ml

0

1

2

0

1

t-test

United States

Test Group 0.053*** 0.007*** 0.027*** 0.002 0.001 0.006*** 0.055*** 0.944*** 0.622

Control Group 0.000 0.024*** 0.011*** 0.010*** 0.000 0.009*** 0.072*** 0.923***

United KingdomTest Group 0.032*** 0.005*** 0.035*** 0.002*** 0.005*** 0.005* 0.096*** 0.901*** 6.588***

Control Group 0.058** 0.017** 0.022*** 0.009*** 0.004** 0.003* 0.067*** 0.933***

Germany

Test Group 0.040*** 0.021*** 0.052*** 0.002** 0.002*** 0.043*** 0.111*** 0.874*** 2.712***

Control Group 0.000*** 0.004*** 0.003*** 0.007*** 0.000 0.002*** 0.100*** 0.883***

France

Test Group 0.046*** 0.006*** 0.011*** 0.001*** 0.002*** 0.0152*** 0.109*** 0.891*** 0.259

Control Group 0.003 0.020 0.049*** 0.001 0.003 0.022*** 0.092*** 0.901***

South Korea

7/28/2019 Cigi Paper 15

14/22



0

1

2

0

1

t-test

Test Group 0.055*** 0.000 0.010*** 0.007*** 0.002*** 0.042*** 0.104*** 0.891*** 3.509***

Control Group 0.051*** 0.025*** 0.056*** 0.005* 0.001 0.043*** 0.068*** 0.919***

Australia

Test Group 0.029*** 0.014** 0.004*** 0.003 0.013*** 0.002*** 0.116*** 0.857*** 4.389***

Control Group 0.005*** 0.012*** 0.011*** 0.002* 0.001 0.006*** 0.081*** 0.919***

Notes: The estimates are based on the ollowing mean equation, rt

= +2 (0

+ 1

2 + 2

ISSR2)rt1

+ t

where the conditional variance is modelled by2 = +

02 +

12

1. When assuming positive eedback trading, i.e.,

1> 0, a destabilizing eect is ound iTest Control is positive, whereas a negative

dierence is in line with a stabilizing impact o the constraints. t-test indicates the t-value or the test o the signifcance in dierences in 2. ***, **, and *

denote statistical signifcance at the one percent, fve percent and 10 percent level, respectively.

Recall that this parameter is designed to capture theimpact o the explanations or unconditional eedbacktrading discussed in the section on methodology. Loand MacKinlays (1988, 1990) nonsynchronous tradinghypothesis is unlikely to be relevant in this case, sincethe test and control groups consist o large cap stocks,which are heavily traded each day. Similarly, time-varyingexpected returns are relatively unlikely to be the root causeo unconditional return autocorrelations since Conrad and

Kaul (1988) demonstrate that the explanatory power othis hypothesis is inversely related to frm size. For theportolio ormed rom the stocks with the largest size, theirmodel explains only one percent o the variation in returns.Thereore, the observed unconditional serial correlationsmight be a consequence o transaction costs as proposed

by Mech (1993).

The estimates or the parameter capturing the interactionbetween conditional variance and autocorrelation,

1,

is ound to be signifcant and positive or nine out o12 samples indicating positive eedback trading. Nowattention is turned to

2, the estimates or the parameter

capturing changes in eedback trading during the periodwhen short selling is constrained. For all test groups,except the United States, the estimates are positive andsignifcant, indicating higher conditional autocorrelationand, thus, increased positive eedback trading when theconstraints are in place. For the control groups, insignifcantparameters

2are observed in the majority o cases, except

or the United Kingdom, where a negative and signifcantestimate is reported. t-tests on the signifcance indierences suggest that or the United Kingdom, Germany,South Korea and Australia,

2is signifcantly higher or the

stocks acing short-sale restrictions than or the unbannedones. In all our cases, this result holds at the one percent

level. Thus, in these stock markets, displacing short sellersleads to more pronounced eedback trading.

Interestingly, there are markets where an amplifcation opositive eedback trading as a consequence o both banningonly naked shorts (Germany and Australia) and banningall shorts but leaving derivative trading unaected (SouthKorea) is observed. However, the eect is the strongestin the United Kingdom, where regulators imposed aninsurmountable hurdle or pessimists not owning a stock,that is, a prohibition to establish any kind o economic short

position including derivatives. The fnding o intensifedpositive eedback trading among dierent kinds oinstitutional short-sale regimes can be interpreted as akind o robustness check. However, a signifcant impacto short-selling restrictions on institutional investorseedback trading is not ound in the case o the UnitedStates and France.

The United States diers to some extent rom the other

markets under consideration, since there are liquidand advanced derivative markets, which may provideinvestors with substitutes or short sales. Empiricalevidence, however, does not lend much support to thesubstitutability hypothesis. Dealing with the case o the USshort-sale regime in September and October 2008, Battalioand Schultz (2011) provide evidence avouring the notionthat single stock options constitute only a partial substituteor short sales. In particular, they show that banning shortsales leads to a dramatic increase in trading costs in termso wider bid-ask spreads. As a result, the use o thesederivatives becomes unattractive to pessimists. Grundy,Lim and Verwijmeren (2012) show that short sellers do not

switch to single stock utures, either.

At frst glance, the fnding that banning short salesmakes investors more prone to positive eedback tradingcontrasts with the tendency o short sellers to ollow thecrowd on trading days when the absolute value o themarket returns exceeds a certain limit as reported in Blauet al. (2010). However, the authors only investigate USstock market data over the sample period January 2005 toDecember 2006.8 This period was characterized by tranquilstock trading and economic expansion while dealing witha period o an extraordinarily severe crisis and marketuncertainty. The literature claims behavioural eects like

eedback trading and herding to be closely related toinvestors sentiment (see Shiller, 1984; Lee, Shleier andThaler, 1991; Devenow and Welch, 1996), which, in turn,is time-varying (see Lee, Jiang and Indro, 2002; Bakerand Wurgler, 2006). Thereore, it seems unlikely thatinstitutional investors behaviour could be expected to be

8 Consequently, their sample does not contain extended periods omarket downturn. In particular, their data set only contains 12 days oextreme negative returns

t t t tt t1 t 2 2

7/28/2019 Cigi Paper 15

15/22



similar between tranquil periods and times o fnancialcrisis.

Conditional return autocorrelation coefcients, (0

+ 12

+ 2ISSR2, are plotted in Figure 2. For all samples, these

coefcients decline sharply during the global fnancial crisiso 2008-2009. This fnding is in line with the phenomenon

o intensifed positive eedback trading during perioo fnancial turmoil reported in a large body o empiricliterature (see, or example, Sentana and Wadhwani, 199LeBaron, 1992; Koutmos, 1997; Kaminsky and Schmukl1999; Karolyi, 2002; Kaminsky, Lyons and Schmukler 20and Salm and Schuppli, 2010).

Fiur 2: Citial Crrlatis fr Six Cutris

0.18

0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0

0.02

FranceTest FranceControl

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0

0.05

0.1

S.KoreaTest S.KoreaControl

0.2

0.15

0.1

0.05

0

0.05

AustraliaTest AustraliaControl

0.5

0.45

0.4

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0

USTest USControl

0.3

0.25

0.2

0.15

0.1

0.05

0

0.05

UKTest UKControl

0.45

0.4

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0

0.05

0.1

GermanyTest GermanyControl

Notes: Conditional correlation coefcients or the test and control groups or the United States, the United Kingdom, Germany, France, South Korea aAustralia based on the specifcation given in (5) and (6).

It is well known that the volatility process o fnancial

returns oten exhibit asymmetries. As outlined above, theseeects can be studied using the Glosten, Jagannathan andRunkle (1993 T-GARCH model. Table 4 shows the results.This robustness check broadly confrms the main results,as the signifcance in dierences in

2remains unchanged.

t

t t

7/28/2019 Cigi Paper 15

16/22



Tabl 4: T-gARCH estimati Rsults fr th Fback Trar Ml

0

1

2

0

1

2

t-test

United States

Test Group 0.059*** 0.000 0.029*** 0.003*** 0.002*** 0.007*** 0.005*** 0.959*** 0.060*** 0.916

Control Group 0.031*** 0.008*** 0.028*** 0.003 0.000 0.009*** 0.059*** 0.886*** 0.089***

United Kingdom

Test Group 0.033*** 0.001*** 0.024v 0.002** 0.000 0.006*** 0.0987*** 0.899*** 0.004*** 5.321***Control Group 0.059*** 0.008*** 0.010*** 0.007*** 0.003*** 0.010*** 0.097*** 0.876*** 0.053***

Germany

Test Group 0.024*** 0.003** 0.004*** 0.001*** 0.003*** 0.032*** 0.009*** 0.922*** 0.104*** 4.214***

Control Group 0.000 0.004*** 0.003*** 0.007*** 0.000 0.002*** 0.100*** 0.883*** 0.010***

France

Test Group 0.026*** 0.003** 0.002 0.006*** 0.007 0.014 0.115*** 0.881*** 0.007 0.686

Control Group 0.000 0.023*** 0.030*** 0.000 0.002 0.034*** 0.026*** 0.898*** 0.111***

South Korea

Test Group 0.063*** 0.000 0.000 0.007*** 0.002*** 0.042*** 0.095*** 0.912*** 0.019*** 1.743*

Control Group 0.040*** 0.022*** -0.035*** 0.007***

0.002 0.005*** 0.069*** 0.920***

0.002***Australia

Test Group 0.040*** 0.002 -0.024* 0.002 0.021** 0.007*** 0.014*** 0.926*** 0.110*** 2.326***

Control Group 0.001*** 0.005*** -0.000 0.003*** 0.001 0.011*** 0.060*** 0.911*** 0.054***

Notes: The estimates are based on the ollowing mean equation, rt= +2 (

0+

12 +

2ISSR2)r

t1+

twhere the conditional variance is modelled by

2 = +02

1+

12

1+

2It1 < 02

1withIt1 < 0 being equal to 1 i the lagged error,

t1, is negative and equal to zero otherwise. When assuming positive

eedback trading, i.e., 1

> 0, a destabilizing eect is ound iTest Control is positive, whereas a negative dierence is in line with a stabilizing impact othe constraints. ***, **, and * denote statistical signifcance at the one percent, fve percent and 10 percent level, respectively.

2reers to the asymmetry

parameter in the Glosten, Jagannathan and Runkle (1993) T-GARCH model. t-test indicates the t-value or the test o the signifcance in dierences in 2.

To sum up, the evidence suggests that, in the majority omarkets under consideration, short-selling bans ampliypositive eedback trading. Thus, contrary to regulators

expectations, these constraints do not stabilize stockmarkets in times o fnancial distress but can actually leadto additional selling during market downturns.

concluSion

In the recent fnancial crisis, politicians, regulators andhigh-profle media coverage blamed short sellers orexacerbating stock market downturns. Institutional shortsellers adhering to positive eedback trading strategiesare a potential justifcation or this allegation. The extantliterature underscores the negative eects o short saleconstraints on inormational efciency and liquidity butis silent about their impact on positive eedback tradingduring fnancial crises. The aim o this paper is to fll thisgap. Insights into this topic are o great interest to stockmarket regulators, enabling evaluations o the efciencyo short-sale constraints in keeping prices closer to theirundamental values. Positive eedback trading can ampliystock market downturns in times o fnancial turmoil.Given that short sellers ollow positive eedback tradingstrategies, regulatory measures intended to displace themcan be a powerul tool to stabilize stock markets. Bans on

selected stocks in six countries during the recent globalfnancial crisis provide a natural experiment to compare

banned stocks to assimilable unbanned stocks with respect

to eedback trading behaviour.

In the United States, the United Kingdom, Germany,France, South Korea and Australia, regulators imposedshort-selling regimes o dierent severities aecting onlyfnancial or even only selected fnancial stocks. Comparingthe group o restricted stocks to careully matched controlgroups o unrestricted stocks allows us to discriminate

between eects o the fnancial crisis and the ban. Foreach test and control group, the eedback trader modelproposed by Sentana and Wadhwani (1992) is estimated,augmented by dummy variables to capture changes in thedegree o eedback trading behaviour under short-selling

constraints. To check or robustness, the model was re-estimated including asymmetric eects in the varianceequation as proposed by Glosten, Jagannathan and Runkle(1993).

The evidence does not support the view that short sellersadhere to positive eedback strategies that may ampliystock market downturns and drive prices away romundamental values. Conversely, in the majority o marketsconsidered in this paper, displacing these investors isassociated with intensifed positive eedback trading.

t t t t

t t t t t t

2 2

7/28/2019 Cigi Paper 15

17/22



Thus, short sale constraints actually play a destabilizingrole and may ampliy market crashes. It is well known inthe literature that short-sale constraints create uncertaintyabout undamental asset values, as negative inormationcan only be exploited with delay. In our view, this lack oreliability o undamental based pricing renders it moreattractive to use positive eedback trading strategies.All things considered, together with plenty o studies

reporting a deterioration in pricing efciency and marketquality under short-sale constraints such as rising bid-askspreads, our fndings suggest that the bans have a negativenet eect.

WorkS cited

Aitken, M. J., A. Frino, M. S. McCorry and P. L. Swan(1998). Short Sales Are Almost Instantaneously BadNews: Evidence rom the Australian Stock Exchange.The Journal o Finance 53, no. 6: 22052223.

Asquith, P., P. A. Pathak and J. R. Ritter (2005). Short

Interest, Institutional Ownership, and Stock Returns.Journal o Financial Economics 78, no. 2: 243276.

Baker, M. and J. Wurgler (2006). Investor Sentiment andthe Cross-Section o Stock Returns. The Journal oFinance 61, no. 4: 16451680.

Battalio, R. and P. Schultz (2011). Regulatory uncertaintyand market liquidity: The 2008 short sale bans impacton equity option markets. The Journal o Finance 66, no.6: 20132053.

Beber, A. and M. Pagano (2013). Short-Selling Bans

around the World: Evidence rom the 2007-09 Crisis.The Journal o Finance 68, no. 1: 343381.

Blau, B. M., B. F. Van Ness, A. R. Van Ness and R. A.Wood (2010). Short Selling during Extreme MarketMovements.Journal o Trading 5, no. 4: 1427.

Boehme, R. D., B. R. Danielsen and S. M. Sorescu (2006).Short-Sale Constraints, Dierences o Opinion andOvervaluation. Journal o Financial and Quantitative

Analysis, 41, no. 2: 455487.

Boehmer, E., C. M. Jones and X. Zhang (2008). WhichShorts Are Inormed?Journal o Finance 63, no. 2: 491

527.

Boehmer, E., Z. R. Huszar and B. D. Jordan (2010). Thegood news in short interest. Journal o FinancialEconomics 96, no. 1: 8097.

Boehmer, E., C. M. Jones and X. Zhang (2011). ShacklingShort Sellers: The 2008 Shorting Ban. Working Paper.

Boehmer, E. and J. Wu (2013). Short Selling and the PriDiscovery Process. Review o Financial Studies 26, no287322.

Bollerslev, T. (1987). A Conditionally HeteroskedasTime Series Model or Speculative Prices and Rato Return. Review o Economics and Statistics 69, no.542547.

Bollerslev, T. and J. M. Wooldridge (1992). Quamaximum likelihood estimation and inerence dynamic models with time-varying covarianceEconometric Reviews 11, no. 2: 143172.

Boudoukh, J., M. P. Richardson and R. F. Whitelaw (199A Tale o Three Schools: Insights on Autocorrelatioo Short-Horizon Stock Returns. Review o FinancStudies 7, no. 3: 539573.

Boulton, T. J. and M. V. Braga-Alves (2010). The skinnon the 2008 naked short-sale restrictions. JournalFinancial Markets 13, no. 4: 397421.

Brent, A., D. C. Morse and E. K. Stice (1990). ShoInterest: Explanations and Tests. Journal o Financand Quantitative Analysis 25, no. 2: 273289.

Bris, A. (2008). Short Selling Activity in Financial Stocand the SEC July 15th Emergency Order. WorkiPaper.

Bris, A., W. N. Goetzmann and N. Zhu (2007). Efcienand the Bear: Short Sales and Markets Around tWorld. The Journal o Finance 62, no. 3: 10291079.

Chang, E. C., G. R. McQueen and J. M. Pinegar (199Cross-autocorrelation in Asian stock markets. PacifBasin Finance Journal 7, no. 5: 471493.

Conrad, J. and G. Kaul (1988). Time-Variation in ExpectReturns.Journal o Business 61, no. 4: 409425.

Cutler, D. M., J. M. Poterba and L. H. Summers (199Speculative Dynamics. Review o Economic Studies 5no. 3: 529546.

De Long, J. B., A. Shleier, L. H. Summers and R. Waldmann (1990). Positive Feedback InvestmeStrategies and Destabilizing Rational Speculation. T

Journal o Finance 45, no. 2: 379395.

Dechow, P. M., A. P. Hutton, L. Meulbroek and R. G. Slo(2001). Short-sellers, undamental analysis, and storeturns.Journal o Financial Economics 61, no. 1: 7710

Desai, H., K. Ramesh, S. R. Thiagarajan and B. Balachandran (2002). An Investigation o tInormational Role o Short Interest in the NasdMarket. The Journal o Finance 57, no. 5: 22632287.

7/28/2019 Cigi Paper 15

18/22



Devenow, A. and I. Welch (1996). Rational herding infnancial economics. European Economic Review 40, no.3-5: 603615.

Diamond, D. W. and R. E. Verrecchia (1987). Constraintson Short-Selling and Asset Price Adjustment to PrivateInormation. Journal o Financial Economics 18, no. 2:277311.

Dickinson, A. and J. R. Woolridge (1994). Short Sellingand Common Stock Prices. Financial Analysts Journal50, no. 1: 2028.

Diether, K. B., K. H. Lee and I. M. Werner (2009). Short-Sale Strategies and Return Predictability. Review oFinancial Studies 22, no. 2: 575607.

Figlewski, S. (1981). The Inormational Eects oRestrictions on Short Sales: Some Empirical Evidence.

Journal o Financial and Quantitative Analysis 16, no. 4:463476.

Glosten, L. R., R. Jagannathan and D. E. Runkle (1993).On the Relation between the Expected Value and theVolatility o the Nominal Excess Return on Stocks.


Gonnard, E., E. J. Kim and I. Ynesta (2008). Recent Trendsin Institutional Investors Statistics. OECD Journal:Financial Market Trends no. 2: 122.

Grundy, B. D., B. Lim and P. Verwijmeren (2012). DoOption Markets Undo Restrictions on Short Sales?Evidence rom the 2008 Short-Sale Ban. Journal oFinancial Markets 106, no. 2: 331348.

Harris, L. E., E. Namvar and B. Phillips (2009). PriceInation and Wealth Transer during the 2008 SECShort-Sale Ban. Working Paper.

Helmes, U., J. Henker and T. Henker (2011). How theAustralian Ban on Short Selling During the GFCAected Market Quality and Volatility. Working Paper.

Hurtado-Sanchez, L. (1978). Short Interest: Its Inuenceas a Stabilizer o Stock Returns.Journal o Financial andQuantitative Analysis 13, no. 5: 965985.

Huszr, Z. R. and W. Qian (2011). Short Sale Return

Predictability Revisited: Anomaly or Return Mis-Measurement? Working Paper.

Kaminsky, G. L., R. K. Lyons and S. L. Schmukler (2004).Managers, investors, and crises: mutual und strategiesin emerging markets.Journal o International Economics64, no. 1: 113134.

Kaminsky, G. L. and S. L. Schmukler (1999). What triggersmarket jitters?: A chronicle o the Asian crisis. Journalo International Money and Finance 18, no. 4: 537560.

Karolyi, G. A. (2002). Did the Asian fnancial crisis scareoreign investors out o Japan? Pacifc-Basin Finance

Journal 10, no. 4: 411442.

Kolasinski, A. C., A. V. Reed and J. R. Thornock (2013).Can Short Restrictions Actually Increase InormedShort Selling? Financial Management (orthcoming).

Koutmos, G. (1997). Feedback trading and theautocorrelation pattern o stock returns: urtherempirical evidence. Journal o International Money andFinance 16, no. 4: 625636.

LeBaron, B. (1992). Some Relations Between Volatility andSerial Correlations in Stock Market Returns.Journal oBusiness 65, no. 2: 199219.

Lee, C. M., A. Shleier and R. H. Thaler (1991). InvestorSentiment and the Closed- End Fund Puzzle. The


Lee, W. Y., C. X. Jiang and D. C. Indro (2002). Stock market

volatility, excess returns, and the role o investorsentiment. Journal o Banking & Finance 26, no. 12:22772299.

Lo, A. W. and A. C. MacKinlay (1988). Stock Market PricesDo Not Follow Random Walks: Evidence rom a SimpleSpecifcation Test. Review o Financial Studies 1, no. 1:4166.

(1990). An econometric analysis o nonsynchronoustrading.Journal o Econometrics 45, no. 1-2: 181212.

Marsh, I. W. and R. Payne (2012). Banning Short Sales

and Market Quality: The U.K.s Experience.Journal oBanking & Finance 36, no. 7: 19751986.

McQueen, G. R., M. Pinegar and S. Thorley (1996). DelayedReaction to Good News and the Cross-Autocorelationo Portolio Returns. The Journal o Finance 51, no. 3:889919.

Mech, T. (1993). Portolio return autocorrelation.Journalo Financial Economics 34, no. 3: 307344.

Merton, R. (1973). An Intertemporal Capital Asset PricingModel. Econometrica, 41, no. 5: 867888.

Miller, E. M. (1977). Risk, Uncertainty, and Divergence oOpinion. The Journal o Finance 32, no. 4: 11511168.

Ogden, J. P. (1997). Empirical Analyses o ThreeExplanations or the Positive Auto- correlation o Short-Horizon Stock Index Returns. Review o QuantitativeFinance and Accounting 9, no. 2: 203217.

7/28/2019 Cigi Paper 15

19/22



Salm, C. A. and M. S. Schuppli (2010). Positive FeedbackTrading in Stock Index Futures: International Evidence.International Review o Financial Analysis 19, no. 5: 313322.

SEC (2008). SEC Release No. 58166 / July 15, 2008.

Senchack, A. J. and L. T. Starks (1993). Short-Sale

Restrictions and Market Reaction to Short-InterestAnnouncements. Journal o Financial and QuantitativeAnalysis 28, no. 2: 177194.

Seneca, J. J. (1967). Short Interest: Bearish or Bullish? TheJournal o Finance 22, no. 1: 6770.

Shiller, R. J. (1984). Stock Prices and Social Dynamics.Brookings Papers on Economic Activity 2: 457510.

7/28/2019 Cigi Paper 15

20/22

About ciGi

The Centre or International Governance Innovation is an independent, non-partisan think tank on internationalgovernance. Led by experienced practitioners and distinguished academics, CIGI supports research, orms networks,advances policy debate and generates ideas or multilateral governance improvements. Conducting an active agendao research, events and publications, CIGIs interdisciplinary work includes collaboration with policy, business andacademic communities around the world.

CIGIs current research programs ocus on our themes: the global economy; global security; the environment and energy;and global development.

CIGI was ounded in 2001 by Jim Balsillie, then co-CEO o Research In Motion (BlackBerry), and collaborates with andgrateully acknowledges support rom a number o strategic partners, in particular the Government o Canada and theGovernment o Ontario.

Le CIGI a t ond en 2001 par Jim Balsillie, qui tait alors co-che de la direction de Research In Motion (BlackBerry). Ilcollabore avec de nombreux partenaires stratgiques et exprime sa reconnaissance du soutien reu de ceux-ci, notammentde lappui reu du gouvernement du Canada et de celui du gouvernement de lOntario.

For more inormation, please visit www.cigionline.org.

ciGi mAStheAd

Managing Editor, Publications Carol Bonnett

Publications Editor Jennier Goyder

Publications Editor Sonya Zikic

Assistant Publications Editor Vivian Moser

Media Designer Steve Cross

eXeCUTIVe

President Rohinton Medhora

Vice President of Programs David Dewitt

Vice President of Public Affairs Fred Kuntz

CoMMUnICATIonS

Communications Specialist Declan Kelly [email protected] (1 519 885 2444 x 7356)

Public Affairs Coordinator Kelly Lorimer [email protected] (1 519 885 2444 x 7265)

7/28/2019 Cigi Paper 15

21/22

7/28/2019 Cigi Paper 15

22/22

Recent cIGI PublIcatIons

Five Years aFterthe FallThGvrnncLgcsfthGlblFnnclCrss

SpeCiaLRepoRT

FIve YeaRs aFteR the Fall: the

GoveRnance leGacIes oF theGlobal FInancIal cRIsIsJ hiw CIGI expM fiv y f f gb fici cii,

t cts ctiu t t css spctu issus. fi

pp by CIGI xp fm c f i pci p

pvi iig cmmi f buiig gvc

gm qui wi uig gci, wi

vviw b cx i wic pp w

p icu CIGI u cfc.

CIGIPaPersno.12January2013

Another Fine Mess:repAiring the governAnceoF internAtionAlFinAnciAl regulAtionPIerresIklos

anotheR FIne Mess: RePaIRInG the

GoveRnance oF InteRnatIonalFInancIal ReGulatIonPi sikfi ys t t st t g ici cisis, picy s

migy ciu biv viy f y cii-

wu c b ivc fm i uc. Cibiiy i w

ii guy fmwk mu bgi m wi

mii f my pici wk g.

ConferenCe report

False DIchotoMIes: econoMIcsanD the challenGes oF ouR tIMeKvi egiti p fwig CIGI-Inet cfc i nvmb

2012 fcu w cmi u wk wi

f uy, uc iy, w, pycgy piic cmy,

ic c pvi m w-u w

qui fcig cmic cmmuiy y.

leaDeRshIP In a tuRbulent aGef os hps Pu hicsu cmic pici, wic i uppy i m ky

i f w, fum i cuiy

ii ip. t Ui s mu wk gi

ccp wi mim wk g

wiu i wi wi g f pi mgig

gb pbm iu.

CIGIPaPersno.14MarCh2013

The ShorT View:The GlobalConjunCTureand The needfor CooperaTionJaMesa.haley

THe SHoRT VIew: THegLoBAL ConjUnCTURe AndTHe need FoR CooPeRATIonJm a. hysuccfuy ig b - mium-m picy

cg qui picy iz muc g mypic

ii p by my, i pp gu, i wi k

gb cmic ip cu cpi i

ik juiciu bcig f jum bu.

CIGIPaPersno.10January2013

SuStainableDevelopment anDFinancing criticalglobal public gooDSBarryCarIn

sustaInable DeveloPMentanD FInancInG cRItIcalGlobal PublIc GooDsBy CiT i G Sup fu c s psiti su

gm, wi wi-wi-wi ci wu g wip

gb upp uimy b ccp by mj py.

CIGIPaPersno.13February2013

StrengtheninginternationalFinancialinStitutionSto PromoteeFFectiveinternationalcooPerationThomasa.bernes

stRenGthenInG InteRnatIonal

FInancIal InstItutIons toPRoMote eFFectIve InteRnatIonal

cooPeRatIontm a. Bt cu gb fici cii u fm fiu f mj

cmi gb iiui mgig fu i i

gb fici mk gb iiui. n ig cuy

biiy uc fix ig i w -

upc g f cciv ci i qui.

CIGIPaPersno.9november2012

A Policy MisMAtch:cAnAdA And theUnited stAtes inthe AsiA-PAcificRegionJamesmanICom

a PolIcY MIsMatch: canaDa anD

the unIteD states In the asIa-PacIFIc ReGIonJm Micmt Ui s C, w f w c i, v

ivig i ipmic miiy pu w ai-

Pcific gi. o bc, wv, C my b i

Pcific p f Ui s.

a CIGI pubici vib f i www.cigii.g.

Documents

Cigi Paper 15