F

MD

a

ARRAA

KRVG

1

(iiisIt

acA1ARRc1tm

1h

The Quarterly Review of Economics and Finance 52 (2012) 219– 226

Contents lists available at SciVerse ScienceDirect

The Quarterly Review of Economics and Finance

jo u rn al hom epage: www.elsev ier .com/ locate /qre f

inancial crisis, REIT short-sell restrictions and event induced volatility

ichael Devaney ∗

epartment of Economics and Finance, Southeast Missouri State University, Cape Girardeau, MO 63701, United States

r t i c l e i n f o

rticle history:eceived 9 June 2010eceived in revised form 24 January 2012ccepted 13 April 2012vailable online 21 April 2012

eywords:EIT

a b s t r a c t

From September 19 through October 8, 2008 the SEC issued a short sale moratorium on approximately800 financial stocks. The emergency order justified the ban based on concerns “that short selling in thesecurities of a wide range of financial institutions may be causing sudden and excessive fluctuationsin the prices of such securities” (see Securities and Exchange Commission, 2008). Although Real EstateInvestment Trusts (REITs) were initially excluded, the management of fourteen REITs requested that theybe added to the restricted list. Diamond and Verrecchia (1987) develop a model in which short saleconstraints decrease trading and increase the time required to adjust to new information resulting in

olatilityARCH

greater price reaction. This research employs a GARCH version of the market model to test the impact ofthe SEC policy on the risk/return of the fourteen restricted REITs and a sample of fifty REITs not on thelist. Rather than mitigate volatility it was determined that fifty of the sixty-four REITS in the combinedsamples exhibited significant event induced risk as a consequence of the ban with a significantly largerincrease occurring among restricted REITs. A cross-sectional test failed to identify significant negative orpositive abnormal returns as a consequence of the short sell ban.

stees

ime“tbA1bscct

fmsir

© 2012 The Board of Tru

. Introduction

Relative to direct ownership, Real Estate Investment TrustsREITs) provide a highly liquid investment vehicle for real estatenvestors. There are essentially three types of REITs; equity REITsnvest in and operate commercial property while mortgage REITsnvest in commercial real estate mortgages and mortgage backedecurities. Hybrid REITs invest in both properties and mortgages.n terms of size, the equity REIT category comprises about 92% ofotal REIT market capitalization.

Equity REIT investment generally concentrates on property typend/or geographic market in an effort to exploit management spe-ialization to the benefit of shareholders. In 2010 the Nationalssociation of Real Estate Investment Trusts (NAREIT) reported53 publicly traded REITs in twelve different property categories.bout 22% of REITs are retail REITs that invest in retail property.etail, office and residential represent over 55% all publicly tradedEITs. REITs became much more attractive in the 1980s. The so-alled “new REIT era” is associated with The Tax Reform Act of986 that eliminated the passive nature of REITs that did not allow

rustees, directors or employees of a REIT to actively engage in the

anagement or operation of REIT property.

∗ Tel.: +1 573 651 2319.E-mail address: mdevaney@semo.edu

tI“iwru

062-9769/$ – see front matter © 2012 The Board of Trustees of the University of Illinoisttp://dx.doi.org/10.1016/j.qref.2012.04.003

of the University of Illinois. Published by Elsevier B.V. All rights reserved.

In 1991 Kimco Realty was the first significant initial public offer-ng of a modern, vertically integrated REIT designed to be internally

anaged and advised. (Brueggeman & Fisher, 2008). This mod-rn and more attractive REIT structure coincided with a real estatecredit crunch” and the creation of the Resolution Trust Corpora-ion that was established to dispose of problem properties acquiredy government as a consequence of the Savings and Loan crisis.ccording to the NAREIT, REIT market capitalization during the990s increased by a factor of 20 and in 2006 exceeded $389illion but had declined to $275 billion in 2009. Despite the impres-ive growth in REIT markets it is estimated that their total marketapitalization represents only 6 percent of the total value of coreommercial real estate in the U.S. suggesting that there is substan-ial room for continued growth.

Growth in the REIT market has fostered growth in REIT mutualunds and exchange traded funds. The expansion of REIT invest-

ent over the last twenty years reflects the new managementtructure, securitized real estate’s increased acceptance by thenvestment community and the success of using REIT IPO’s toepackage problem properties following the real estate crisis ofhe 1980s, a strategy that may be relevant to current markets.n February 2010 The Congressional Oversight Panel report titledCommercial Real Estate Losses and the Risk to Financial Stabil-

ty” predicted that $1.4 trillion in commercial real estate mortgages

ould reset in the years 2010–2014 with almost half of these cur-ently underwater. The REIT structure may once again prove to be aseful vehicle for re-capitalizing many of these problem properties.

. Published by Elsevier B.V. All rights reserved.

2 Economics and Finance 52 (2012) 219– 226

Irkw

cssiaTfttc2rsimRon

srlotaofSsmmnsar

2l

ftLicRciFR

rrlRfctl

0

2

4

6

8

10

12

OctA07

NovB07

JanA08

FebB08

AprA08

MayB08

JulyA08

AugB08

OctA08

NovB08

JanA09

FebB09

AprA09

MayB09

Sh

ort

In

tere

st

Ra

tio

Bi-Weekly

REIT

Market

pc

gooithdfti2

2ita1ribtw

cllsp(coc

tfiRessklt

20 M. Devaney / The Quarterly Review of

nvestors prefer markets that are liquid and conform to the sameules and conventions as other well organized public security mar-ets. An important element of securitized real estate’s integrationith public capital markets is the ability to short sell.

The debate over short selling resurfaced during the financialrisis. Following the demise of Bear Stearns in early 2008 shortelling restrictions were imposed on a small number of financialtocks. After the collapse of Lehman Brothers investment bankn September of 2008, the SEC short sell restricted list includedpproximately 800 financial companies. Real Estate Investmentrusts (REITs) were initially excluded however the managers ofourteen REITs requested that they be added to the list. Regula-ory justification for the restriction was based on the presumptionhat it would mitigate “excessive fluctuation in the prices” of finan-ial institution securities (see Securities and Exchange Commission,008). Diamond and Verrecchia (1987) argue that consistent withational expectations market participants anticipate short sale con-traints when formulating pricing decisions and that they resultn larger price reactions. Their model predicts that rather than

itigate volatility short sell restrictions will cause it to increase.ational expectations would also suggest that the creditable threatf expanded restrictions could even influence the risk of securitiesot currently on the list.

Evidence that short selling “may be causing sudden and exces-ive fluctuations in the prices of [financial] securities” and that aegulatory ban can mitigate volatility as suggested by the SEC isargely anecdotal. This research employs an event study method-logy based on a GARCH market model to examine the impact ofhe SEC short sell restriction on the fourteen restricted REITs as wells a random sample of fifty REITs not on the list. The first objectivef this research is to determine if the short sale ban was success-ul in reducing volatility in either restricted or unrestricted REITs.econd, if short sell constraints result in economically exploitablekewness, it should be reflected in abnormal returns. The GARCHodel allows a cross-sectional event test for the presence of abnor-al negative or positive returns as a consequence of the ban. The

ext section briefly examines the literature on REITs, constrainedhort-selling and event studies. The third section describes thedvantages of the GARCH methodology. Section 5 presents theesults of the model followed by a summary of conclusions.

. REITs, constrained short-selling and the event studyiterature

Much of the REIT literature has focused on the risk/return per-ormance of alternative property types, diversification features andhe behavior of returns relative to securities markets (see Glascock,u, & So, 2000; Gyourko & Nelling, 1996; Young, 2000). The increasen REIT return volatility leading up to the financial crisis raised con-erns among those in the industry. Much of the traditional appeal ofEIT investing derives from the perception that REITs are an assetlass that offers diversification, income and relative price stabilityn REIT shares. Some REIT managers believe that Exchange Tradedunds (ETF) and “naked shorting” created artificial volatility in theEIT market (see The Durable Investor, 2008; Troianovski, 2008).

“Naked shorting” occurs when traders have not actually bor-owed the shares they sell short as required by securities law. It canesult in “failure to deliver” which is the number of stocks in whicharge positions have not been properly delivered to investors. SECegulation SHO was supposed to address the problem of delivery

ailure. Implemented in 2005, SEC Regulation SHO mandated dailyompilations of stocks that had experienced at least five consecu-ive days of delivery failures totaling at least 10,000 shares and ateast a half percent of their outstanding shares. When a stock was

asrt

Fig. 1. Short interest ratio.

laced on the threshold list, traders were presumably required tolose out failed deliveries by the 13th day after the trade.

The daily average number of stocks on the SHO threshold listrew from 300 issues in 2007 to 400 for the first nine monthsf 2008. Prior to the short sell moratorium the list increased tover 500 issues representing in excess of $600 billion in value andncluded many REIT and financial stocks. On September 18, 2008he SEC implemented a new rule that required short sellers whoave sold shares but not delivered them to the buyer within threeays of the trade, to deliver shares by the start of trading on theourth day. Following the new rule and the short sell moratorium,he daily average of stocks on the threshold list declined to 100ssues in December of 2008 and 79 issues in March and April of009.

The time pattern of REIT short selling before and after the fall008 credit crisis is shown in Fig. 1 which plots the bi-weekly short

nterest ratio (total shares shorted divided by total average dailyrading volume) from October 2007 through May 2009 for all stocksppearing on the Short Squeeze (2009) data base (approximately6,000 issues and 120 REITs). Increases in the short interest ratioeflects rising bear sentiment but some technicians interpret it as anndication of pent up demand since shorted shares must ultimatelye repurchased and returned to the lender. In the year precedinghe September 2008 financial crisis, the short interest ratio for REITsas two to three times higher than for the entire market.

The large short interest ratio for REITs preceding the financialrisis is in contrast to previous REIT studies that examine ear-ier periods. Blau, Hill, and Wang (2011) found investors are lessikely to short REITs than other securities. They suggest that REIThort sellers target REITs that are performing well instead of under-erforming REITs. Li and Yung (2004) found that only high levelsthe 90th percentile) of short interest are associated with signifi-ant negative REIT returns. They speculate that the bearish contentf short interest may have been mitigated by the favorable riskharacteristics of real estate securities for the period 1994–2001.

The percent of institutional ownership is also larger for REITshan it is for the market. As mentioned previously, some in thenancial press attributed the increase in market volatility amongEITs and financial stocks to short selling by institutional investors,specially leveraged Exchange Traded Funds. Over the time periodhown in Fig. 1, institutions owned an average of 60% of REIThares while institutional ownership was only 37% for the mar-et. REIT institutional ownership fluctuated from a high of 69% to aow of 57% during this period. The correlation coefficient betweenhe percentage change in institutional ownership and the percent-ge change in the short interest ratio was a positive .05 but was

tatistically insignificant. Although there was not a significant cor-elation between percentage change in institutional ownership andhe change in the short interest ratio, institutional investors may

Econo

hii

stLhtMsvatLrsm

cVotcsaapprlcsoccc

GttwRawsriama

dpMoAtBtti

s(

mDRast(w

rwiSrai

GacqodRosr

mttetam

3

(tmiavova

aaARt

y

V

ε

w

M. Devaney / The Quarterly Review of

ave still contributed to increased REIT volatility via short sell-ng. The ownership statistic tracks the percent of shares owned bynstitutions, not the percent of short positions held by institutions.

Miller (1977) maintains that short sale constraints prevent pes-imistic traders from short selling without restricting optimisticraders from buying, thereby causing stock prices to be over-valued.amont (2004a) summarizes the literature in support of the Millerypothesis. Harrison and Kreps (1978) argue that the overvalua-ion from short sell constraints may be greater than suggested by

iller. They develop a model that demonstrates how restrictinghort sales will result in the price of the security exceeding thealuation of even the most optimistic current investors. Lamontnd Thaler (2003) provide several case studies of overvaluationshat likely occurred as a consequence of difficulty in short selling.amont (2004b) examines a sample of 300 firms that attempted toestrict short selling by various means including publicly attackinghort sellers. He found these stocks tended to under-perform thearket.In contrast to Miller (1977) who analyzed the impact of

onstrained short selling on security valuations, Diamond anderrecchia (1987) examine the relationship between the rising costf short selling and stock returns. They develop a rational expec-ations model in which market participants anticipate short saleonstraints when formulating pricing decisions. Short sale con-traints decrease trading which increases the time required todjust to private information. Because short sale constrained stocksre slow to incorporate new information they have relatively largerrice reactions when private information is made public. Since therice adjustment to negative information is especially slow, theeactions to negative news tend to be relatively larger leading to aeft skewness in the distribution of returns. Negative skewness as aonsequence of constrained short selling should manifest itself asignificant abnormal negative returns. Rational expectations the-ry would suggest that the imposition of a short selling constraintould induce greater volatility not only in stocks that are currentlyonstrained but also securities that confront a creditable threat ofonstrained short selling.

Consistent with Diamond and Verrecchia (1987), Bris,oetzmann, and Zhu (2007) compare stock market regula-

ion among different countries and conclude that prices do appearo incorporate negative information more slowly in those countrieshere short selling is either not allowed or not generally practiced.eed (2007) found evidence of greater negative price reactionssociated with constrained short selling, which is also consistentith the Diamond and Verrecchia hypothesis. If the rising cost of

hort trading were directly related to greater price reaction thenegulator mandates against short selling would be expected toncrease rather than decrease volatility. Unlike some firms thatsked to be removed from the list, the market may interpret REITanagers request to be added to the short sell restricted list as an

dmission of weakness.One method for examining the impact of regulatory man-

ates on security risk/return is an event study that was firstroposed by Fama, Fisher, Jensen, and Roll (1969). Binder (1998),acKinlay (1997) and Peterson (1989) provide a survey literature

n the extension and modification of the event study method. critical issue in the application of the event methodology is

he increase in stochastic volatility surrounding the event period.rown and Warner (1980) and Brown and Warner (1985) observehat increases in the variance may cause misspecification of theraditional test statistics and that the power of the tests can be

mproved by appropriately modeling the volatility process.

Most applications of the event method are based on con-tant variance versions of the market model. Howe and Shilling1988) suggest that conventional applications of the market model

oTcs

mics and Finance 52 (2012) 219– 226 221

ay not be a good representation of REIT risk/return. Corgel andjoganopoulas (2000) estimate betas for a sample of sixty equityEITs and report many insignificant and negative betas as wells an average R2 of 0.03. They found that individual betas amongix investment advisory firms differed by as much as 25%. Consis-ent with time varying market risk, Chatrath, Liang, and McIntosh2000) observe asymmetric behavior in REIT betas with larger betashen the market is declining than when it is rising.

Given the inconsistent performance of the market model, someesearchers suggest a model specification that replaces beta riskith variance or total risk. Research finds evidence of both a pos-

tive and negative tradeoff between return and total risk. French,chwert, and Stambaugh (1987) find a positive tradeoff betweeneturn and variance, as did Campbell and Hentschel (1992). Bailliend DeGennaro (1990) conclude that the tradeoff parameter isnsignificant for many of the portfolios in their study.

In regard to REIT risk/return, Devaney (2001) estimates aARCH-M model that finds a positive tradeoff between returnsnd own conditional variances for REIT indices. The efficacy ofonventional market model estimates of REIT betas has also beenuestioned by Devaney and Weber (2005) who use a directionalutput distance function to construct a risk/return frontier thatefines the best-practice management technology in which eachEIT produces a desirable output (return) and an undesirableutput (risk). They found that estimates of REIT inefficiency areignificantly lower when based on total variance rather than betaisk.

Brockett, Chen, and Garven (1999) develop an event-studyethod that assumes a market model with GARCH effects and

ime-varying market risk. Savickas (2003) examines a test statis-ic for abnormal returns in the presence of stochastic volatility andvent induced increases in variance. He finds that the traditionalest is mis-specified in the presence of event-induced variancend that the GARCH model rejects the null hypothesis significantlyore frequently.

. Time varying GARCH model

A GARCH model similar to the one developed by Brockett et al.1999) and tested by Savickas (2003) will be used to conducthe event study of REIT reaction to the SEC restricted short sell

andate. Unlike traditional constant variance event models thatmplicitly assume that the event induced variance is the same forll securities in the sample the GARCH model allows event inducedolatility to vary and for each security’s variance to be stochasticutside the event period. Given the observed increase in marketolatility, this is a particularly important advantage of GARCH rel-tive to conventional applications of the event methodology.

Time varying models have become an important tool in financend economics. The ARCH model was first proposed by Engle (1982)nd generalized to GARCH by Bollerslev (1986) and extended toRCH and GARCH in the mean (GARCH-M) by Engle, Lilien, andobins (1986). The generalized GARCH model can be described byhe following system of equations:

t = ˛xt + εt (1)

t = ı0 +∑

ıiε2t−i +

∑ˇivt−i (2)

t |˝t−1∼N(0, vt) (3)

here yt is returns, xt is an exogenous or predetermined vector

f variables, ε is a random error, and ˝t−1 is the information set.he parameter vectors are ˛, ˇ, ı0, ıi and t is a time index. Theonditional variance, v, is linearly dependent on the past behavior ofquared errors and a moving average of past conditional variances.

2 Econo

TwlPg

tcastlwcs

aitirGaB

G

R

h

�

Rmesaf2nrBbioiRhu

eszi“uv

4

isRt

rdpRicoiadoEp

tsonnmivws“aAabmsr2

tvtfifty unrestricted REITs and the S&P 500 index for the sample periodappear in Table 1. Descriptive statistics for each of the individualREITs are available from the authors. Panel A in Table 1 reports

1 On Friday, September 19, 2008, the SEC released an emergency order prohibitingshort sales in 799 securities (http://www.sec.gov/rules/other/2008/34-58592.pdf).On the following Monday the order was amended with important material differ-ences: http://www.sec.gov/news/press/2008/2008-218.htm that included the SEC’sdelegation of authority to each national securities exchange to identify and addlisted companies that qualify for inclusion on the covered list. As the additions andremovals on the list are subject to change on a daily basis, the NYSE has maintainedand provided daily end-of-day list updates to the covered list since the amendedorder went into effect.As of Friday, September 26, 2008 the NYSE and NYSE Arca short sell prohibi-tion list has been enhanced as part of a collaborative effort with NASDAQ andAMEX, to include securities that will be covered for the next trading day from allfour Exchanges in one convenient document. http://www.nyse.com/attachment/CONSOLIDATED-SSPROHIBTION.xls for daily updates or to subscribe for free to NYSESystem Status Trader Updates https://www.nysenet.com/subscription/smLogin toreceive this notice as soon as it becomes available along with other NYSE CashEquities Market System and Trading Information news.

2 F-tests examining the difference in mean percent of institutional ownership andmarket capitalization for the restricted and unrestricted groups were 1.12 and .51;respectively. The hypotheses of equal means for the two groups cannot be rejected.A summary of REIT property types for the restricted and unrestricted sample are asfollows:

22 M. Devaney / The Quarterly Review of

he ıi values determine the weights attached to lagged innovationshile the squared error term implies that if innovations have been

arge in absolute value, they are likely to be large in the future.arameters ˇ, ı0, ıi must be non-negative to ensure that the returnenerating process is well defined.

If all coefficients in Eq. (2) are zero (except the intercept) thenhe model will reduce to the traditional constant variance specifi-ation. Both ARCH and GARCH model the conditional variance as

function of past shocks and allow volatility to evolve and per-ist. The ARCH model incorporates a limited number of lags inhe derivation of the conditional variance while GARCH allows allags to exert an influence by including the conditional variance as

ell as lagged values of the squared error. ARCH models are oftenharacterized as short memory models while GARCH models areometimes referred to as long memory models.

An indication of the degree of persistence in shocks to volatilitys measured by the GARCH process is the sum of the coefficientsn the conditional variance equation (ıi + ˇi) which must be lesshan one or unity for stability to hold. The degree of persistence ismportant in determining the relationship between volatility andeturns since only persistent volatility justifies changes in returns.ARCH(1,1) as opposed to higher order models is parsimonious,llows for long memory in the volatility process, and according toollerslev et al. (1992) fits most economic time series.

Based on Savickas (2003) and Brockett et al. (1999) the followingARCH model is estimated for the sample of REIT returns:

i,t = ˛i1 + ˇt · Rm,t + �i · Dt + �i,t (4)

i,t = ˛i2 + bi · hi,t−1 + ci · �2i,t−1 + di · Dt (5)

i,t |˝t∼N(0, hi,t) (6)

i,t are REIT daily returns, Rm,t are the daily returns on the S&P500arket index, ˛i1, ˇi, � i, ˛i2, bi, ci and di are parameters to be

stimated. Rather than Devaney’s (2001) GARCH-M model thatpecified hi,t as the risk measure in the return equation, thispproach follows the more traditional market model but allowsor time varying beta (ˇt). Subsequent to the ban in September of008, there were days in which the T-bill yield was zero and evenegative. Quantitative easing by the Fed drove the 30 day T-billate to an annual average of 10 basis points for late 2008 and 2009.ecause of the unusual behavior of the risk free security for the postan period, daily returns rather than excess returns were used. Dt

s an indicator variable that equals 1 if t is an event day, and 0therwise; and ˝t consists of all information available at time t,ncluding all the current and previous market and security returnsm,u and Ri,u for all u ≤ t, current and previous volatility estimatesi,u for all u ≤ t, and current and previous error estimates �i,u for all

≤ t.The mean of the market model residual Ri,t–˛i–ˇiRm,t during the

vent period will be reflected in the estimate of � i since by con-truction the mean of the disturbance term �i,t must be equal toero. Given that the null hypothesis of zero mean abnormal returns true then the estimate of � i must be close to zero. The meaning ofclose to zero” depends on the volatility of the market model resid-al reflected in hi,t which incorporates the event-induced varianceia the coefficient di.

. Data

The restricted sample is represented by the fourteen REITs

ncluded on the short-sell restricted list along with fifty unre-tricted REITs randomly selected from the approximately 120EITs appearing on the Short-Squeeze data base in the week prioro the implementation of the short sell restriction. Because all

mics and Finance 52 (2012) 219– 226

estricted REITs were equity REITs the unrestricted sample was ran-omly selected from equity REITs with short interest in the weekrior to the ban. Details on the identification and dates restrictedEITs were added to the short sell restricted list.1 Short-Squeeze

s a proprietary database that provides short interest, industryode, institutional ownership and variety of other informationn approximately 16,000 publicly traded stocks with outstand-ng short positions. Data on market and REIT short interest ratioppearing in Fig. 1 was taken from Short Squeeze. REIT returns areaily returns from January 2, 2006 to April 08, 2009 or 822 dailybservations and were provided by the National Association of Realstate Investment Trusts (NAREIT) and Commodity Systems, Incor-orated: Market Data and Trading Software.

The SEC short-sell restriction extended from September 19hrough October 8, 2008. The fourteen restricted REITs were sub-equently added to the list over the four days following the SECrder. The starting date of the sample was limited to the begin-ing of 2006 so that the number of pre-event observations wouldot be excessively large relative to post event days. F-tests deter-ined that there was not a significant difference in the mean

nstitutional ownership or market capitalization of the restrictedersus the unrestricted sample. The results of the F-tests alongith a summary of property type for both the restricted and unre-

tricted samples.2 Because the REITs appearing on the restricted listchose” to be added to the list the event study sample is self selectednd measures market response to a management policy decision.lthough many event studies measure market response to whatre essentially non-choice random events, many other studies areased on management choices such as the decision to merge or noterge, list on an exchange or not list as well as a wide variety of

tudies that examine market response to different choices in theeporting of accounting information (see Bodie, Kane, and Marcus,005).

Diebold (1998) and Bollerslev (1986) observe that return dis-ributions for high frequency daily data exhibit non-stationaryariance along with leptokurtosis inconsistent with the normal dis-ribution. Descriptive statistics for the fourteen restricted REITs, the

Diversified Health Hotel Industrial Office Residential Retail Total

Restricted 4 0 1 2 0 2 5 14Unrestricted 8 6 4 3 8 7 14 50

M. Devaney / The Quarterly Review of Economics and Finance 52 (2012) 219– 226 223

Table 1Descriptive statistic for daily returns.

Restricted REITs Unrestricted REITs S&P 500

Daily returns for the first half of the sample period: A01/01/2006 to 8/20/2007N = 411Mean .000513 .000493 .000375Std. deviation .010247 .0111221 .007351Skewness −.170095 −.101932 −.560737Kurtosis 4.5756 3.9708 5.7888Jarque–Bera 44.575 16.853 154.734

Daily returns for the second half of the sample period:B8/21/2007 to 4/08/2009N = 411Mean −.001671 −.000473 −.001028Std. deviation .053579 .0469394 .023507Skewness .417337 .326202 .178314Kurtosis 6.6725 5.9699 6.9360Jarque–Bera 242.911 158.339 267.481

Daily returns for the full sample period: C01/01/2006 to 4/08/2009N = 822Mean −.000579 .00000985 −.000326Std. deviation .038564 .033734 .017419Skewness .417337 .326202 .178314

tsfipt

osarvtmbsn

5

i11tuwgiiTrri(R

uv

Table 2F-test for equality in the mean between restricted and unrestricted REITs and coef-ficient mean (percent significant at 5%).

Ri,t = ˛i1 + ˇt · Rm,t + � i · Dt + �i,t

hi,t = ˛i2 + bi · hi,t−1 + ci · �2i,t−1

+ di · Dt

Coefficientestimates fromGARCH model

F-test Restricted REIT(percent significantat 5%)

Unrestricted REIT(percent significantat 5%)

˛i,1 .431 .0002164(0%) .000274(0%)ˇi .215 1.27(100%) 1.22(100%)� i 8.49* −.005149(0%) .001836(0%)˛i,2 2.42 .00000108(86%) .0000009(82%)bi 3.70* .082(100%) .026(94%)ci .119 .915(100%) .912(100%)di 9.64** .000321(57%) .000168(84%)

The table lists the F-tests that test for a statistically significant difference in themean coefficients between the restricted (N = 14) and unrestricted groups (N = 50).Restricted REIT and Unrestricted REIT columns show the mean coefficient for theindividual regressions in each group along with the percent that were found to besignificant in parenthesis.

ChsOstisb

aiaopWmciatr

br(psrstaRA

rmAf

Kurtosis 6.6725 5.9699 6.9360Jarque–Bera 3027.35 2026.25 2496.78

he mean, standard deviation, skewness, kurtosis and Jarque-Beratatistics for the first half of the time series which predates thenancial crisis. The mean of daily returns for all three series wereositive, leptokurtic and exhibited Jarque–Bera statistics inconsis-ent with the normal distribution.

Panel B in Table 1 shows the same statistics for the second halff the return series, which includes the financial crisis and the shortell restricted event. The mean return for all three series is negativend the variance increased by a factor of greater than 4 for both theestricted and unrestricted REITs. The large increase in return seriesariance suggests the GARCH method is appropriate for modelinghe REIT short sell event and that the results from traditional event

odels that assume normality and stationary variance would likelye biased. Finally, panel C shows descriptive statistics for the entireample period. Jarque–Bera statistics lead to the rejection of theormality assumption for all three series at 1% significance level.

. Results

The results of the 64 GARCH model estimates are summarizedn Table 2. Details on all 64 GARCH models appear in Appendix. The mean beta coefficient was 1.27 for the restricted REITs and.22 for the unrestricted. F-tests appearing in Table 2 indicate thathere is not a significant difference in the betas of restricted versusnrestricted REITs. All sixty-four REIT time varying beta coefficientsere significant at 5%. Sixty-one of the sixty-four REIT betas were

reater than one. This result is consistent with anecdotal evidencen the financial press that observes an increase in REIT volatil-ty relative to the market index (see The Durable Investor, 2008;roianovski, 2008). It may also be expected given the significantole that real estate and the financial industry played in the recentecession. The strong statistical relationship between the marketndex and REITs is in contrast with Corgel and Djoganopoulas2000), who find statistically weak market model estimates for

EITs based on the constant variance specification.

To further investigate REIT systematic risk, the restricted andnrestricted REITs were combined into two portfolios and con-entional OLS versions of the market model were subjected to a

vrui

* Significant at 5%.** Significant at 1%.

how Breakpoint and Chow Forecast test. Both tests reject the nullypothesis of no structural difference in beta before and after thehort sell restrictions were instituted. CUSUM of squares tests ofLS estimates also provide evidence of beta instability over the

ample period. When the GARCH market model was estimated forhe pre and post ban periods, the beta for the restricted portfolioncreased from 1.12 to 2.23 subsequent to the ban while the unre-tricted portfolio beta increased from 1.17 to 1.83. Coefficients foroth portfolios over both time periods were significant.

As stated previously, during true financial panics the correlationmong risky assets increases. A more robust time varying spec-fication is consistent with Chatrath et al. (2000) who observedsymmetric behavior in REIT betas over market cycles. For muchf 2008 and early 2009 the market index has been driven by theerformance of financial firms and real estate sensitive stocks.hether or not the stronger relationship between REITs and thearket index is a transitory phenomenon attributable to the finan-

ial crisis or represents a permanent shift in REIT systematic risks an empirical question. To the extent that problems in residentialnd commercial real estate continue to impact the banking indus-ry and the real economy it is likely that REIT systematic risk willemain large relative to historical betas.

Except for three of the sixty-four models, the coefficients ˛i,2,i, and ci satisfy the non-negative condition for a well-definedeturn generating process. In addition, the ARCH (bi) and GARCHci) parameters sum to one or less indicating stability. The GARCHarameter was significant in all sixty-four models and ARCH wasignificant in sixty-one of sixty-four. The results suggest that REITeturn data for the time period examined conform to the GARCHpecification. The variance equation intercept term (˛i,2) tests for aime independent volatility component. Although the coefficientsre very small, 86% of restricted REITs and 82% of the unrestrictedEITs exhibited significant time independent volatility when noRCH or GARCH elements are present.

The mean of the coefficient on the dummy variable (� i) in theeturn equation for the restricted sample was negative while theean of the coefficients for the unrestricted sample was positive.n F-test indicates a significant difference in the mean coefficients

or the two groups although none of the coefficients for the indi-

idual models were significant. The coefficient on event-inducedisk was significant in the variance equation (di) for 84% of thenrestricted REITs and 57% of those short sell restricted. An F-test

ndicates that there was a significant difference in the coefficient on

224 M. Devaney / The Quarterly Review of Econo

Table 3Cross-sectional test for abnormal returns. Return with (t-statistic) and date whenREITs were added to the restricted list.

Day t-Statistic short sellrestricted REITs(N = 2–14)

t-StatisticunrestrictedREITs (N = 50)

Event day .884 (0.94) 2 REITs .785 (0.99)Day +1 −.558 (0.56) 3 REITs −.625 (1.10)Day +2 .296 (0.40) 9 REITs −.436 (0.68)Day +3 1.263 (1.05) 12 REITs .758 (1.05)Day +4 .438 (0.22) 14 REITs 1.750 (1.46)Day +5 .826 (0.47) −.339 (0.33)Day +6 −.767 (1.76) −.481 (0.84)Day +7 −.527 (0.93) −.367 (0.40)Day +8 −.509 (0.38) −1.53 (1.09)Day +9 −.046 (0.19) .194 (0.22)Day +10 −.790 (0.43) −.184 (0.12)Day +11 −.752 (0.50) −.156 (015)Day +12 −.362 (0.24) −.691 (0.51)Day +13 2.350 (1.08) 5.28 (2.11)*

Day +14 −.790 (0.59) −2.04 (0.89)Day +15 −.994 (1.39) −1.14 (1.49)Day +16 .161 (0.24) −.150 (0.20)Day +17 −.413 (0.45) .074 (0.08)Day +18 −.008 (0.17) −.008 (0.12)

The daily abnormal returns for the restricted and unrestricted groups, along withtw

ttmsrs

intitbtsraicfSw

∑

w

S

eiactfi

btdRaffito

wrsfssn

6

tqRffifomf

iesdtelDmrraga

mu2apdtbaTofap

-tests shown in parenthesis. Note that the N for the t-test in the restricted groupas only large enough to be statistically meaningful for day +4 and forward.* Significant at 5%.

he dummy variable in the variance equation suggesting that whilehe short sell restriction induced volatility for both samples it had a

uch greater impact on the restricted sample. This finding is con-istent with the hypothesis that a ban on short selling increasesather than mitigates volatility and that it also impacts securitiesubject to a creditable threat of restriction (the unrestricted group).

Because of reduced flexibility in asset management somenvestors may prefer to hold securities that provide the opportu-ity to short sell and avoid those that are short sell restricted. Ifhis view is more pervasive among more sophisticated institutionalnvestors, then asset classes with a large proportion of institu-ional ownership such as REITs may be more adversely impactedy short sell restrictions than other sectors. As previously men-ioned, the market may interpret the request to be added to thehort sell restricted list as a signal of financial weakness and thisesult may be reflected in a test for abnormal returns. Diamondnd Verrecchio predict that constrained short selling will resultn negative skewness in returns. If negative skewness is economi-ally significant it should be reflected in abnormal returns. To testor abnormal returns in the presence of event-induced volatility,avickas (2003) suggests the following cross-sectional test statistic,hich is Student’s-t distributed with N-1 degrees of freedom:

N

i=1

Si,t/N√1/N (N − 1) ·

∑Ni=1(Si,t −

∑Nj=1Sj,t/N)

2(7)

here

i,t = �i√hi,t

(8)

The test normalizes abnormal event induced return ( �i) by thestimated standard deviation (hi). The cross-sectional test outlinedn Eqs. (7) and (8) is implemented for the short sell restricted REITS

nd the fifty REITs in the un-restricted control sample. Details of theross-sectional tests are summarized in Table 3. The “event day” ishe first day that REITs were included on the SEC restricted list. Therst two REITs were added to the list on the event day followed

Tofd

mics and Finance 52 (2012) 219– 226

y one REIT on Day +1, six REITs in Day +2 and so on such thathere were 14 short sell restricted REITs by Day +4. Note that theummy variable estimates in the GARCH model for the restrictedEITs were based on the exact day that the particular REIT wasdded to the list. However, the estimates of the dummy variableor REITs in the unrestricted group were based on the day that therst two REITs were added or Day 0. Consequently, t-statistics forhe restricted sample only have statistical validity for Day +4 andnward.

None of the restricted REITs during the time the short sell banas in place exhibited significant negative or positive abnormal

eturns. Among the unrestricted REITs, there was only one day ofignificant positive abnormal returns and this occurred on day 13ollowing the ban. Returns for none of the remaining days wereignificant. Tests for normality of Si,t in Eq. (8) generate Jarque–Beratatistics that suggest the test statistic is approximatelyormal.

. Conclusions

This research utilizes a GARCH version of the market model toest for event induced volatility and abnormal returns as a conse-uence of the SEC mandate that restricted short sales on fourteenEIT stocks. The sample included GARCH market model estimates

or each of the fourteen restricted REITs and a random sample offty REITs not included on the restricted list. The time varying betas

or all sixty-four REITs were found to be significant with sixty-onef the sixty-four betas greater than one. The time varying GARCHarket model was determined to be the appropriate specification

or the period examined.The model supports the contention that the ban induced volatil-

ty in both restricted REITs and the unrestricted sample. Rationalxpectations would attribute induced volatility in the unrestrictedample to the creditable threat of subsequent restriction. Theummy variable coefficient was significant in the variance equa-ion (di) for forty-two of fifty REITs in the unrestricted sample andight of fourteen short sell restricted REITs and was significantlyarger for the restricted REITs. This result is consistent with theiamond and Verrecchia’s (1987) model, which predicts that thearket anticipates the rising costs of constrained short selling and

esults in greater price reaction. Despite evidence of event-inducedisk, there was only one day of significant abnormal positive returnsnd this occurred on the +13 day of the ban for the unrestrictedroup. None of the other trading days for either group exhibitedbnormal returns as a consequence of the ban.

The short sell moratorium along with the four-day delivery ruleay have enabled regulators to sharply reduce the number of fail-

res to deliver in the financial and real estate sectors. On February4, 2010 the SEC adopted the “alternative up-tick rule,” including

circuit breaker feature. The circuit breaker rule applies when therice of a security has decreased by 10% or more from the prioray’s closing price. Short selling in that security will be subjecto the alternative up-tick rule. In particular, when the 10% circuitreaker is triggered, short selling in that security will be prohibitedt a price that is less than or equal to the current national best bid.he price test restriction will remain in place for the rest of the dayn which the circuit breaker is triggered, as well as for the entireollowing day. (Securities and Exchange Commission, 2010) Thelternative up-tick rule replaces the old general up-tick rule thatrohibited short selling except on an up-tick or a zero-plus tick.

he rule went into effect in 1938 and was removed in the summerf 2007. Ulibarri (2009) concludes that had an uptick rule been inorce during the crisis it might have helped stabilize price behaviorepending on the nature of the stochastic process.

Econo

rTwtmwwsh

“asawvmarket.

C

M. Devaney / The Quarterly Review of

Rather than mitigate volatility, the results suggest that REITeturns became more volatile as a consequence of the SEC mandate.he lifting of the SEC short sell restriction after a scant three weeksould appear to be an admission that the policy failed to achieve

he desired result. In his assessment of the ban, former SEC Chair-an Christopher Cox said on December 31, 2008: “Knowing whate know now, [we] would not do it again. The costs appear to out-

eigh the benefits.” (Younglai, 2008). The week preceding the short

ell moratorium was a very tumultuous period in financial marketistory. It included the bankruptcy filing of Lehman Brothers, the

A

oefficient estimates.

Ri,t = ˛i1 + ˇt · Rm,t + �i · Dt + �i,t hi,t =

REIT ˛i,1 ˇi �i

ABR −8.15E−05 1.17* −.01096

ACC .000065 1.01* .00811ADC .000349 1.25* .008935

AEC .00119 .94* −.00535

AIV .000576 1.46* .000839

AMB .000302 1.19* −.00062

ARE .000367 1.26* .000367AVB .000701 1.46* .006812

BDN −.00055 1.16* .001478

BEE −.000668 1.55* −.00767

BRE .000503 1.24* .004539

BXP .000717 1.40* .010381

CBL −.00048 1.12* −.01048

CLI 1.31E−05 1.20* .000305

CLP −6.90E−05 1.16* .005714

CPT 3.61E−06 1.22* .001891

CUZ .000317 1.59* .009557

DDR −2.73E−05 1.27 −.008896

DRE −.000121 1.17* −.000243

EGP .000148 1.12* .006534

ESS .000556 1.13* .004750

FCH 1.01E−05 1.61* −.005346

FR 4.28E−05 1.21* −.015071

FRT .000742 1.22* .000944

FUR 2.75E−06 1.07* −.004103

GGP −.000184 1.48* −.032860

GRT −.0000279 1.23* −.006949

GTY 3.91E−05 1.00* .007444

HCN .000812 .94* .007437

HPT −6.22E−05 1.03* −.009526

HR .000215 1.29* .008764

HRP −8.64E−05 1.03* −.006797

HST −.000131 1.36* −.000931

HT .000866 1.05* −.017836

IRC .000309 1.20* .005713

KIM .000548 1.51* −.002883

KRC .000320 1.26* .008119

KRG .000307 1.07* .002682

LHO 1.65E−06 1.41* −.007862

LRY 3.84E−05 1.22* .002575

LXP −.000525 1.15* −.005548

MAA .000252 1.27* .013780

MAC 1.48E−05 1.18* −.009821

MPG −.000559 1.36* −.005374

NHP .000993 1.22* .011181

NRP .000879 1.37* .004696

PCL .000305 1.20* .001160

PKY .000177 1.09* −.011726

PLD .000500 1.32* 7.49E−05

PPS −.000108 1.17* .005302

PSA .000600 1.37* .010196

PSB .000463 1.08* .003057

REG .000313 1.22* −.001110

SKT .000805 1.15* .001185

SLG .000647 1.47* .000653

mics and Finance 52 (2012) 219– 226 225

breaking of the buck” by the Reserve Primary Money Market Fundnd a new stricter SEC rule addressing failures to deliver. Notwith-tanding the potential abuse associated with naked shorting and

large backlog of failures to deliver, these results are consistentith a finance literature, which suggests that short selling provides

aluable information to investors and results in a more efficient

ppendix A.

˛i2 + bi · hi,t−1 + ci · �2i,t−1

, +di · Dt

˛i,2 bi ci di

−.2.53E−07 .036* .973* .0001653.12E−06* .049* .939* .000149*

4.22E−06* .107* .897* .0001153.09E−06* .033* .960* .000268*

1.78E−07* .028* .966* .000502*

1.55E−06* .051* .947* .000381*

2.18E−06* .056* .935* .000159*

3.38E−06* .052* .937* .000158*

2.75E−06 .027* .973* .000334*

2.31E−06* .036* .959* .001233*

1.81E−06* .037* .957* .000132*

4.52E−06* .058* .929* .000189*

6.92E−07 .042* .958* .000633*

2.53E−06* .029* .961* .000351*

1.65E−06* .060* .938* .000459*

3.62E−06* .069* .920* .000386*

2.11E−06* .052* .943* .000585*

3.66E−07 .036* .965* .000877*

2.76E−06* .076* .918 .000261*

1.98E−06* .041* .948* .000143*

5.48E−07 .024* .976* .000005*

3.67E−06* .024* .970* .001210*

2.10E−06* .087* .914* .001174*

2.50E−06* .041* .943* .000242*

2.50E−06* .169* .826* .0001631.26E−06* .128* .875* .004653*

2.55E−06* .094* .908* .000441*

5.31E−06* .000 .993* .000009*

3.37E−07 .016* .981* .000008*

4.51E−06* .211* .798* .0000014.85E−07 .005 .994* .000106*

7.29E−06* .152* .839* .000244−1.98E−07* .003 .999* .000134*

−1.98E−07* .077* .926* .000321*

1.99E−06* .017* .972* .000102*

1.85E−06* .084* .916* .000398*

8.99E−06* .081* .889* .000409*

2.11E−06* .062* .932* .0004311.62E−06* .039* .960* .000230*

1.51E−06* .037* .959* .000166*

2.70E−06* .104* .896* .000715*

2.01E−06* .072* .923* .000241*

2.56E−06* .107* .892* .0002927.72E−07* .021* .982* .000832*

2.67E−06* .079* .912* .000134*

5.55E−06* .102* .896* .000647*

1.64E−06 .057* .941* 6.77E−053.55E−06* .104* .891* .0001595.54E−06* .133* .861* .0005411.88E−06* .081* .922* .000260*

1.44E−06* .024* .971* .000118*

1.89E−07 .014* .987* 3.87E−05*

1.55E−06* .049* .947* .000166*

2.30E−06* .039* .950* .000109*

9.09E−06* .121* .866* .000487

226 M. Devaney / The Quarterly Review of Economics and Finance 52 (2012) 219– 226

Appendix A (Continued )

Ri,t = ˛i1 + ˇt · Rm,t + �i · Dt + �i,t hi,t = ˛i2 + bi · hi,t−1 + ci · �2i,t−1

, +di · Dt

REIT ˛i,1 ˇi �i ˛i,2 bi ci di

SNH .000966 1.53* .012949 1.73E−06* .045* .950* .000212*

SPG .000371 1.56* .005625 1.01E−05* .142* .838* .000435*

SUI −3.78E−05 1.02* −.004976 8.81E−07* .031* .966* .000106*

TCO .000913 1.28* −.005280 6.69E−07 .021* .976* .000163*

UBA .000400 1.35* .008699 3.28E−06* .042* .949* .000203*

UDR .000436 1.19* .007461 1.80E−06 .075* .923* .000140*

UHT .000829 1.30* .008713 1.78E−05* .132* .827* .000299VNO 9.91E−05 1.44* .0020260 4.32E−06* .098* .895* .000232*

VTR .000464 1.24* .007269 1.24E−06 .046* .951* .000108*

T atesr

R

B

B

B

B

B

B

B

B

B

B

B

C

C

C

C

D

D

D

DE

E

F

F

G

G

H

H

L

L

L

L

M

M

P

R

S

S

S

S

T

T

U

he REIT stock symbol followed by the individual GARCH model coefficients estimestricted REITs in bold.

* Significant at 5%.

eferences

aillie, R., & DeGennaro, R. (1990). Stock returns and volatility. Journal of Financialand Quantitative Analysis, 25, 203–214.

inder, J. (1998). The event study methodology since 1969. Review of QuantitativeFinance and Accounting, 11, 111–137.

lau, B. M., Hill, M. D., & Wang, H. (2011). REIT short sales and return predictability.Journal of Real Estate Finance and Economics, 42, 481–503.

odie, Z., Kane, A., & Marcus, A. (2005). Investments. Boston, MA: McGraw-Hill Irwin.,pp. 381–384

ollerslev, T. (1986). Generalized autoregressive conditional heteroscedasticity.Journal of Econometrics, 31, 307–327.

ollerslev, T., Chou, Y., & Kroner, K. (1992). ARCH modeling in finance: A review ofthe theory and empirical evidence. Journal of Econometrics, 52, 5–59.

ris, A., Goetzmann, W., & Zhu, N. (2007). Efficiency and the bear: Short sales andmarkets around the world. Journal of Finance, 62(3), 1029–1079.

rockett, R., Chen, H., & Garven, J. (1999). A new stochastically flexible eventmethodology with application to proposition 103. Insurance, Mathematics andEconomics, 25, 197–217.

rown, S., & Warner, J. (1980). Measuring security price performance. Journal ofFinancial Economics, 8, 205–258.

rown, S., & Warner, J. (1985). Using daily stock returns: The case of event studies.Journal of Financial Economics, 14, 3–31.

rueggeman, W., & Fisher, J. (2008). Real estate finance and investments (13th ed.).McGraw-Hill., p. 623.

ampbell, J., & Hentschel, L. (1992). No news is good news: An asymmetric model ofchanging volatility in stock returns. Journal of Financial Economics, 31, 281–318.

hatrath, A., Liang, Y., & McIntosh, W. (2000). The asymmetric REIT beta puzzle. TheJournal of Real Estate Portfolio Management, April–June, 16–29.

ongressional Oversight Panel. Commercial real estate losses and the riskto financial stability. (2010). <http://cop.senate.gov/reports/library/report-021110-cop.cfm>.

orgel, J., & Djoganopoulas, C. (2000). Equity REIT beta estimation. Financial AnalystsJournal, 56(1), 70–79.

evaney, M. (2001). Time varying risk premia for real estate investment trusts: AGARCH-M model. Quarterly Review of Economics and Finance, 41, 335–346.

evaney, M., & Weber, W. (2005). Efficiency, scale economies, and the risk/returnperformance of real estate investment trusts. The Journal of Real Estate Financeand Economics, 31(3), 301–317.

iamond, D., & Verrecchia, R. (1987). Constraints on short selling and asset priceadjustments to private information. Journal of Financial Economics, 18, 277–311.

iebold, F.X. (1998). Elements of Forecasting, Southwestern.ngle, R. (1982). Autoregressive conditional heteroscedasticity with estimates of the

variance of U.K. inflation. Econometrica, 50, 987–1007.ngle, R., Lilien, D., & Robins, R. (1986). Estimating time varying term premium in

the term structure: The ARCH-M model. Econometrica, 55, 391–407.ama, E., Fisher, L., Jensen, M., & Roll, R. (1969). The adjustment of stock prices to

new information. International Economic Review, 10, 1–21.

Y

Y

for the restricted REITs (bold type) and unrestricted REITs are shown. Short sell

rench, K. G., Schwert, W., & Stambaugh, R. (1987). Expected stock returns andvolatility. Journal of Financial Economics, 39, 1141–1153.

lascock, J., Lu, C., & So, R. (2000). Further evidence on the integration of REIT, bondand stock returns. Journal of Real Estate Finance and Economics, 2, 177–194.

yourko, J., & Nelling, E. (1996). Systematic risk and diversification in the equityREIT market. Real Estate Economics, 24(4), 493–515.

arrison, J. M., & Kreps, D. M. (1978). Speculative Investor Behavior in a Stock Marketwith Heterogeneous Expectations. Quarterly Journal of Economics, 92, 323–336.

owe, J., & Shilling, J. (1988). Capital structure theory and REIT security offerings.Journal of Finance, XLIII(4), 983–993.

amont, O. (2004). “Short sale constraints and overpricing,” The theory and practice ofshort selling. New Hope, PA: Frank J. Fabozzi Associates.

amont, O. (2004). Go down fighting: Short sellers vs. firms. NBER Working Paper10659, August.

amont, O., & Thaler, R. (2003). Can the market add and subtract: Mispricing in Techstock carve-outs. Journal of Political Economy, 111(2), 227–268.

i, D., & Yung, K. (2004). Short interest in real estate investment trusts. InternationalReal Estate Review, 7(1), 56–70.

acKinlay, A. (1997). Event studies in economics and finance. Journal of EconomicLiterature, 35, 13–39.

iller, E. (1977). Risk, uncertainty and divergence of opinion. Journal of Finance, 32,1151–1168.

eterson, P. (1989). Event studies: A review of issues and methodology. QuarterlyJournal of Business and Economics, 28, 36–37.

eed, A. (2007). Costly short selling and stock price adjustments to earnings announce-ments. Chapel Hill, NC: University of North Carolina.

avickas, R. (2003). Event induced volatility and tests for abnormal performance.The Journal of Financial Research, XXVI(2), 165–178.

ecurities and Exchange Commission. Emergency order pursuant to Section 12(k)(2)of the Securities and Exchange Act of 1934 Taking Temporary Action to Respondto Market Developments. (2008b). Release: 34-58592/September 18, 2008.

ecurities and Exchange Commission. Alternative up-tick rule. Amendment toregulation SHO. (2010). <http://www.sec.gov/news/press/2010/2010-26.htm>Accessed 24.02.10.

hort Squeeze. (2009). <http://shortsqueeze.com/?gclid=CLv74Nm0gpsCFdSM5wodtB7Xew>.

he Durable Investor. REIT volatility and leveraged ETFs. (2008).<http://durableinvestor.wordpress.com/2008/12/26/reit-volatility-and-leveraged-etfs/> Accessed 26.12.08.

roianovski, A. (2008). REIT moves rub executives the wrong way. The Wall StreetJournal, Real Estate, December 26, 2008.

libarri, C. (2009). Regulating noisy short-selling of troubled firms. Journal of Finan-cial Policy, 1(3), 227–245.

oung, M. (2000). REIT property-type sector integration. Journal of Real EstateResearch, 19(1), 3–21.

ounglai, R. SEC chief has regrets over short selling ban. (2008). Reuters U.S. Edi-tion. <http://www.reuters.com/article/idUSTRE4BU3GG20081231> Accessed31.12.08.