FIRM-SPECIFIC VARIATION AND OPENNESS IN EMERGING …

FIRM-SPECIFIC VARIATION AND OPENNESS IN EMERGING MARKETS

Kan Li Randall Morck Fan Yang and Bernard Yeung

AbstractmdashThis paper compares the comovement of individual stockreturns across emerging markets Campbell et al and Morck et al haveshown that the United States saw rising firm-specific stock return varia-tions and thus declining comovement over the second half of thetwentieth century We detect a similar albeit weaker pattern in most butnot all emerging markets We further find that higher firm-specificvariation is associated with greater capital market openness but not goodsmarket openness Moreover this relationship is magnified by institutionalintegrity (good government) Goods market openness is associated withhigher marketwide variation

The price system is just one of those formations whichman has learned to use (though he is still very far fromhaving learned to make the best use of it) after he hasstumbled upon it without understanding it

Friedrich August von Hayek (1945)

I Introduction

THE extent to which individual stock prices move inde-pendently varies both across countries and over time

Morck Yeung and Yu (2000a) find the ratio of idiosyncratic(firm-specific) variation to total variation in individual stockreturns to be higher in higher-income economies in the mid1990s Campbell et al (2001) and Morck et al (2000a) finda long-term rise in idiosyncratic variation in US stockreturns A series of studies including Wurgler (2000) BrisGoetzmann and Zhu (2002) Bushman Piotrovsky andSmith (2002) Durnev et al (2004 pp 3ab) and othersrelates greater idiosyncratic variation or lower comovementin individual stock returns to a range of measures of theinstitutional development regulatory sophistication andcapital allocation efficacy of the stock market Other worktoo extensive to list documents a relationship betweenfinancial development and economic openness This studydocuments a strong statistical correlation between capitalmarket openness in emerging markets and idiosyncraticstock return variation We focus on emerging markets be-cause developed stock markets are fully open to foreign

investors for the full period over which large panels ofreturns data are available

We first show that the findings of Campbell et al (2001)and Morck et al (2000a) of rising idiosyncratic variation inUS stocks are also evident in the majority of emergingmarkets over the 1990s We then show that higher idiosyn-cratic variation is significantly correlated with greater cap-ital market openness in emerging market economies withsound institutions However capital market openness andpoor institutions may actually increase comovement Tradeopenness generally increases comovement

The remainder of the paper is arranged as follows Sec-tion II describes our conceptual starting points Section IIIdescribes our methodology and section IV presents ourempirical findings Section V presents a case study andsection VI concludes

II Individual Stock Return Comovement and Openness

The total variation in an individual stock return can bedecomposed into idiosyncratic variation which is specificto the stock and systematic variation which is explained bymarket returns A natural measure of comovement is thussystematic variation as a fraction of total variation Sincecomovement can be large either because systematic varia-tion is large or because idiosyncratic variation is small italso makes sense to look at these quantities explicitly Forbrevity we refer to all of these variables as measuringldquocomovementrdquo

Campbell et al (2001) and Morck et al (2000a) docu-ment rising absolute and relative firm-specific variation inUS stocks Morck et al (2000a) also find that firm-specificvariation is a greater part of total variation in more devel-oped countries They are unable to explain these differenceswith differences in macroeconomic stability country ormarket size economy structure or firm-specific variation infundamentals (returns on assets) Rather greater officialcorruption is highly correlated with more comovement andin countries with below-average corruption stronger inves-tor protection laws are associated with higher firm-specificvariation

Comovement and related phenomena matter for two gen-eral classes of reasons These reasons interact with theincreasing global integration of capital markets as docu-mented by Bekaert and Harvey (1995 1997) and others

The first class of reasons relates to portfolio risk Camp-bell et al (2001) note that many investors are not fullydiversified and so are exposed to greater risk when firm-specific variation is greater They further show that greaterfirm-specific variation means investors need larger portfo-lios to diversify fully and argue that greater firm-specificvariation should affect option prices which depend on

Received for publication March 6 2003 Revision accepted for publi-cation September 5 2003

University of Alberta School of Business University of AlbertaSchool of Business and NBER University of Alberta School of Businessand Stern School of Business New York University respectively

We are grateful for helpful comments from Yiorgos Allayannis RobinBrooks John Campbell Kristen Forbes Ben Esty William GoetzmannLarry Harris Andrew Karolyi Wei Li Ashoka Mody Robert Shiller ReneStulz and James Tobin as well as participants at the American EconomicAssociation Washington meeting in 2003 and the International MonetaryFund preconference on Global Linkages in Washington DC We thankHali Edison and Frank Warnock as well as Abdul Abiad and AshokaMody for providing their capital openness measures and for other helpfulcomments We are grateful to Timothy Vogelsang for providing the Gausscode for his algorithm Special thanks are owing to John Campbell andAshoka Mody for serving as discussants at the AEA and IMF respectivelyTimely research assistance by Abigail Hornstein is also greatly appreci-ated Randall Morck gratefully acknowledges partial funding from theSSHRC

The Review of Economics and Statistics August 2004 86(3) 658ndash669copy 2004 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology

firm-specific plus market-related variation in the return ofthe underlying asset

Chen and Knez (1995) argue that as barriers to capitalflows drop cross-border arbitrage affects asset prices acrossmarkets Thus information about industries exposures ordiscount rates that affects prices in one country also affectsprices elsewhere Some of this information is about thewhole market but much is about sectors and should showup as idiosyncratic risk

The second class of reasons has to do with the realeconomy We now review each of these reasons in turn

First La Porta Lopez-de-Salines and Shleifer (1999)find most large publicly traded firms outside the UnitedStates and United Kingdom organized into corporategroups A single controlling shareholder usually a verywealthy family controls all the firms in such a groupmdashdirectly or indirectly Group firms finance each other dobusiness with each other adopt common strategies andotherwise coordinate decisions This could give their fun-damentals a common component Such intercorporate deal-ings might steadily transfer wealth from one group firm toanother and public shareholders of a wealth-losing groupfirm understandably view this as a corporate governanceproblem which Johnson et al (2000) dub tunneling Per-haps comovement gauges the importance of corporategroups and of tunneling

As Coffee (2002) Karolyi and Stulz (2003) MorckStangeland and Yeung (2000b) Rajan and Zingales (2003)Stulz (1999) and others note capital market opennesspressures regulators to adopt international best practices indisclosure governance and regulation It also creates localdemand for information professionals like accountants andanalysts Both changes might render tunneling more diffi-cult The result might be less comovement in both firmfundamentals and returns

Second economic growth arises from technologicalprogress Schumpeter (1912) holds that this occurs as inno-vative firms rise to displace established industry leaders ina process he dubs creative destruction More intense cre-ative destruction thus causes the fundamentals of innovativeand laggard firms to differ more

Caves (1982) argues that openness to outward foreigndirect investment (FDI) raises the rewards to innovators byallowing greater economies of scale and that openness toinward FDI allows technology spillovers from multination-als to local firms Rajan and Zingales (2003) and Morck etal (2000b) argue that foreign portfolio investment (FPI)openness bolsters local technological progress by lettingentrepreneurial startups obtain financing from abroad Thusgeneral capital market openness might let innovative firmsoutpace sedate rivals faster magnifying firm-specific differ-ences in returns

Third neoclassical trade theory links trade openness tospecialization Trade openness should thus reduce an econ-omyrsquos diversification across industries perhaps turning in-

dustry factors into market factors and raising fundamentalscomovement Also if capital openness accompanies tradeopenness a positive correlation between capital opennessand comovement might ensue

However tunneling innovation and specialization allaffect returns comovement only by affecting fundamentalscomovement Morck et al (2000a) cannot explain returnscomovement with fundamentals (return on assets) comove-ment1 Either their fundamentals comovement measure isinadequate (a possibility given the low frequency of fun-damentals data) or some other explanation is paramount

The latter possibility leads them to consider alternativeexplanations Comovement may be symptomatic of marketinefficiencies such as bubbles or herding If more opencapital markets experience fewer bubbles and less herdingthey might exhibit less comovement Alternatively critics ofglobalization argue that openness spreads crises For exam-ple the Bernama News Agency quoted Malaysian PrimeMinister Mahathir Mohamad blaming Malaysiarsquos economiccrisis on international financiers who ldquorobbed the Palestin-ians of everything but in Malaysia they could not do sohence they do this depress the ringgitrdquo2 More sagaciouslyBhagwati (1998) argues that capital market openness canspread financial crises and that only product market open-ness is justified and Forbes and Rigobon (2002) studycontagion as a factor in market fluctuations

However Morck et al (2000a) and Campbell et al(2001) show that changes in comovement are due in part atleast to changes in idiosyncratic variation as opposed tomarketwide variation Roll (1988) argues that idiosyncraticvariation reflects trading by investors with private firm-specific information and Morck et al (2000a) and DurnevMorck and Yeung (2004) speculate that more idiosyncraticvariation reflects periods of especially intense trading bysuch investors and hence more accurate pricing at least inthe short term However Campbell et al (2001) dispute thisnoting correctly that West (1988) links less information tohigher returns variation Their intuition is that as informa-tion is revealed a previously more erroneously priced stockexhibits larger return fluctuations

Regardless a growing body of empirical work linksgreater idiosyncratic variation to variables that on thesurface at least are plausible proxies for the informationcontent of stock prices Some of this work links higherfirm-specific return variation to variables readily interpret-able as gauging informed arbitrage Greater idiosyncraticvariation is evident in countries with stronger insider tradingprohibitions (Beny 2000) more developed financial analy-

1 Morck et al (2000a) run market-model analog regressions on returns onassets defined as earnings plus depreciation plus interest over net assets Thisallows them to estimate firm-specific and systematic fundamentals variationControlling for these variables does not affect their results

2 From an October 10th 1997 speech to Muslim villagers quoted inldquoMalaysia Premier Sees Jews behind Nationrsquos Money Crisisrdquo by SethMydans New York Times October 16 1997

FIRM-SPECIFIC VARIATION AND OPENNESS 659

sis industries and a freer press (Bushman et al 2002) andfewer short sales restrictions (Bris et al 2002) Other worklooks at the information content of stock returns moredirectly Durnev et al (2003b) find significantly higherfirm-specific returns variation following a major historicaltightening in US disclosure law for affected stocks but notfor others Durnev et al (2003a) find returns more accu-rately predicting future earnings changes in industrieswhose stocks move more idiosyncratically Collins Kothariand Rayburn (1987) and others regard such predictivepower as gauging the ldquoinformation contentrdquo of stock pricesYet more work links greater idiosyncratic variation to bettercapital allocation Wurgler (2000) finds capital flows moreresponsive to value added in countries where returns co-move less Durnev et al (2004) show US industries inwhich idiosyncratic variation is higher exhibiting fewersigns of both overinvestment and underinvestment Tobin(1982) defines the market as functionally efficient if pricechanges induce efficient capital allocation and his twostudies speculate that less comovement and greater idiosyn-cratic variation signify more functionally efficient marketsReconciling these findings with the West (1988) frameworkand related literature such as Campbell and Shiller (1987)is an exciting avenue for future research

Our objective here is more limitedmdashto see if the USpattern of rising idiosyncratic variation extends to emergingmarkets and to see what factors correlate with the magni-tudes of this change across countries In doing this we needto consider factors that affect many countries but to differ-ing degrees

One such factor is the increasing globalization of capitalmarkets We make no pretense that globalization is the onlysuch factor However a study of all possible factors isbeyond the scope of this effort We focus on globalizationbecause economic openness especially to capital flows haschanged to different degrees in different countries over thepast decade and in ways that can be measuredmdashalbeit withdifficulty By considering how capital market opennessmight interact with the different explanations of comove-ment advanced above we can explore their validity andimplications

III Methodology

A Estimating the Comovement Variables

Our main comovement measures are based on modifiedmarket model regressions for individual securities Let thereturn on stock j in period t be rjt the domestic marketreturn for country n at t be rnt and the US market returnat t be rmt (converted to local currency) To assess thecomovement of individual stocks in country n during period we run the regression

rjt j0 j1rnjt j2rmt εjt (1)

separately for each stock j n using all Tj observationst The transformed domestic market return rnjt is theequal-weighted average return of all stocks in n except jitself

rnjt in ij rit

Jnt 1 (2)

where Jnt is the number of stocks in country n at time t Wethus use a different domestic market return for each regres-sion This is because we are interested in the comovement ofstock j with other stocks not with itself In economies witha small number of traded stocks this eliminates a potentialupward bias in our comovement measures

A simple variance decomposition expresses the sum ofsquared variation in rjt denoted sj

2 as the sum of thesquared variation explained by equation (1) msj

2 and theresidual variation εsj

2 The systematic variation in stock jduring interval is mj

2 [1(Tj 1)] msj2 the firm-

specific variation is εj2 [1(Tj 1)] εsj

2 and the totalvariation is j

2 [1(Tj 1)] sj2 where Tj is the number

of return observations for firm j in during To estimate country-level analogs we take an average of

the Jn firm-level measures in each country n weighted bythe number of observations on each firm Thus the averageabsolute firm-specific return variation for stocks in countryn during interval is

εn2

jn εsj2

jn Tj Jn (3)

We interpret a larger εn2 as signifying less comovement in

individual returnsAn analogous procedure generates the average absolute

systematic return variation for stocks in country n duringtime interval

mn2

jn msj2

jn Tj Jn(4)

We interpret a greater mn2 as signifying more comovement

in individual returnsScaling the firm-specific by the total variation n

2 ob-tained analogously to equations (3) and (4) we obtain theaverage R2 statistic of the regression (1) for stocks incountry n during time interval

1 εsn

2

sn2 1

εn2

n2 Rn

2 (5)

To gauge the importance of systematic variation as a frac-tion of total variation in country n we can define a country-level analog

Rn2

msn2

sn2

mn2

n2 (6)

THE REVIEW OF ECONOMICS AND STATISTICS660

the average relative systematic variation in the stocks ofcountry n during interval We take a lower Rn

2 as signi-fying less comovement

To construct these measures we download a time series ofWednesday-to-Wednesday returns for every stock inDataStream deleting returns with zero or missing volume ateither endpoint Using weekly returns economizes on down-loading time DataStream contains coding errors especially forLatin America due to misplaced decimal points An algorithmchecks for such errors and drops affected observations

B Regression Framework

We seek to explain comovement with measures of open-ness to the global economy taking into account the differentlevels of institutional development in different countriesWe thus run panel regressions of the form

comovement measure fixed effects 1

openness measure 2 openness measure (7)

institutional development n

We follow Morck et al (2000a) in using as dependentvariables the natural logarithm of country-level averagefirm-specific variation ln(εn

2 ) that of the systematic vari-ation ln(mn

2 ) and the difference between them which wedenote by the Scandinavian letter oslashn3 Note that oslashn is alogistic transformation of the R2 measure for

oslashn lnmn2 lnεn

2 ln Rn2

1 Rn2 (8)

Since mn2 and εn

2 are bounded below by 0 and Rn2 is in

the unit interval these transformations are necessary toprovide approximately normal dependent variables

We use several alternative measures to capture differentaspects of openness

We define trade openness as suggested by Frankel(2000)

trade opennessn Mn

Yn 1

Yn

n Yn (9)

where Mn is total imports and Yn is gross domestic product(GDP) If national borders do not affect buying patternsimports divided by GDP equals 1 minus the nationrsquos shareof world production leaving the value of the opennessmeasure equal to 0 In a completely closed economy thevariablersquos value is 1 plus the countryrsquos GDP as a fractionof world GDP As the country becomes more open themeasure rises towards 0 Trade openness can rise above 1for an entrepot state Frankel (2000) recommends this mea-sure in lieu of the traditional imports plus exports divided byGDP which tends to be larger for smaller economies

We construct the variable using data from World Devel-opment Indicators 2002 produced by the World Bank Forour sample the variable is always negative Note howeverthat we exclude the city-states of Singapore and HongKong which are probably the most important entrepotcountries Hong Kong is a particularly special case becauseof its switch from a UK colony to a Chinese specialadministrative region during our sample period

Measuring capital market openness is more difficult forinvestment stock and flow measures are often highly problem-atic We therefore use a carefully developed capital marketopenness measure provided by Edison and Warnock (2002)This is a direct measure of the openness of each countryrsquos stockmarket to foreign investors Essentially it reflects the value ofstocks that can be purchased by foreign investors as a percent-age of total domestic market capitalization4 It is closer to 1 ifa market is more open and closer to 0 if it is more closed

The index is available for most emerging markets from1990 through 2001 though it is unavailable for some in thevery early 1990s Since developed country stock marketswere essentially fully open to foreign investors throughoutthe 1990s the index has no variation for these marketsConsequently we restrict our attention to emerging markets

The capital and trade openness are not highly correlated(13 0001 p 099) Although many countries with opencapital markets have open goods markets there are notableexceptions Indonesia (capital market openness rises whiletrade openness shows no consistent trend) Malaysia and thePhilippines (capital market openness shows no consistenttrends but trade openness rises) and Pakistan (capital marketopenness rises while the goods market becomes more closed)

To assess institutional development we use the good gov-ernment measure constructed by Morck et al (2000a) Thismeasure sums three variables from La Porta et al (1999) thatgauge the respect a countryrsquos government shows for the rule oflaw the efficiency of its legal system and the freedom of itsgovernment and civil servants from corruption Each individ-ual measure ranges from 0 to 10 so good government liesbetween 0 and 30 with larger numbers connoting better insti-tutions This variable is available only as a cross section

All our regressions control for country year and crisisfixed effects

Including country fixed effects nets out own-countryaverages We do this because Morck et al (2000a) linkcomovement to a variety of variables having to do witheconomy structure economy size and fundamentals co-movement We have no reliable measures of how these

3 Pronounced as a Briton would the ldquourrdquo in ldquomurkrdquo

4 This measure is based on an ldquoinvestablerdquo index reflecting the marketcap available to foreign investors divided by the capitalization of the wholemarket Both are from the International Finance Corporation (IFC) To controlfor ldquoasymmetric shocks to investable and non-investable stocksrdquo the measureis adjusted using price indices computed by the IFC for the two categories ofstocks Since the stocks available to foreigners may trade at different pricesthan the stocks available to locals the value of stocks available to foreignerscan in theory exceed total domestic stock market capitalization The indexused in Edison and Warnock (2002) is actually 1 minus this openness ratioand measures the intensity of capital controls


factors change through time for each country and thereforesubsume them into general fixed effects We recognize thatthis may not capture their full effects If their changes arecorrelated with changing openness our openness variablemight pick up effects that more properly should be ascribedto changes in these other variables If these other effects arethemselves also associated with economic openness this isdefensible If they are not we must interpret our opennessvariable more broadly as perhaps capturing part of abroader range of institutional or other changes Year fixedeffects capture global macroeconomic factors and controlfor any general time trend in our data

A number of emerging economies experienced financialcrises during the 1990s Hence we include three crisis dum-mies to capture transitory changes in comovement associatedwith the unusual conditions prevailing in the affected marketsThe Asian crisis dummy is 1 for East Asian countries in 1997and 1998 and 0 otherwise The Mexican peso crisis dummy is1 for Latin American countries in 1995 and 0 otherwiseFinally the Brazilian real crisis dummy is 1 for Latin Americancountries in 1998 and 0 otherwise

C Sample

Table 1 lists the countries in our final sample The list ofcountries in Table 1 is the intersection of those for whichEdison and Warnockrsquos (2002) capital openness measure isavailable those for which the good-government index isavailable and those for which DataStream stock returns areavailable We go back only to the 1990s because stockreturn data for earlier years are unavailable on DataStreamfor many countries We thus have annual comovementmeasures from 1990 to 2001 for most countries We requirethat five years of comovement data be available to include

TABLE 1mdashSAMPLE DESCRIPTIVE STATISTICS

Market Rn2

εn2

mn2

CapitalOpenness

TradeOpenness

GoodGovernment

Brazil 01757 00080 00018 074 087 2024Chile 01625 00033 00007 081 070 1960Colombia 01448 00040 00007 066 079 1897Greece 02874 00054 00025 081 074 2101India 02516 00078 00027 019 085 1844Indonesia 01837 00093 00025 056 071 1540Korea 03006 00080 00028 036 065 2220Malaysia 04342 00037 00033 075 010 2276Mexico 02463 00036 00013 065 072 1861Pakistan 01987 00072 00018 059 078 1347Peru 01652 00077 00017 101 082 1492Philippines 01963 00085 00023 049 054 1294Portugal 01172 00039 00005 068 062 2485South Africa 00965 00087 00009 100 077 2307Taiwan 03922 00032 00027 023 2513Thailand 02701 00065 00024 038 053 2017Turkey 03753 00087 00056 099 075 1813

Mean 02352 00063 00021 064 069 1941Std dev 00978 00023 00012 025 018 367Minimum 00965 00032 00005 019 087 1294Maximum 04342 00093 00056 101 010 2513

Comovement is measured by the average market model R2 (Rn2 ) the average firm-specific variation (εn

2 ) and the average systematic variation (mn2 ) Capital openness is value-weighted fraction of the market

open to foreign investors Trade openness is importsGDP relative to GDP(world GDP) Good government is a cross-section index taking low values where corruption is worse Data are for 1990 through 2001

FIGURE 1mdashCHANGING COMOVEMENT IN INDIVIDUAL STOCKS

IN EMERGING MARKET

Annual comovement measures are derived from market model regressions of weekly individual stockreturns on domestic and US market returns and include the average regression R2 the systematic(explained) variation in the average stockrsquos returns and the firm-specific (residual) variation in theaverage stockrsquos returns


a country in our panel Our trade openness variable isunavailable for Taiwan (ROC)

The resulting panel contains annual measures for 17countries from 1990 to 2001 with 183 countryndashyear obser-vations We have less than a full panel because data forsome countries are unavailable in the early 1990s Table 1displays univariate statistics

IV Findings

Figure 1 summarizes the pattern across all emergingeconomies Panel A weighting each country equally re-veals falling Rn

2 and mn2 and rising εn

2 though none aremonotonic Panel B weighting each stock equally with noregard to its country reveals a similar picture

Table 2 presents a statistical description of the patterns inFigure 1 Because we have only 12 observations lag esti-mation and unit root tests are problematic Still panel Ashows consistent positive trend point estimates in tests with

and without first-order lags and both assuming and disavow-ing a unit root Unfortunately the statistical significance issporadic For comparison we present the same tests usingUS data for 1990 to 2001 and obtain similarly inconclu-sive findings and even a positive trend in R2 When weextend the US data back to 1963 (not shown) we repro-duce the trends detected by Campbell et al (2001) andMorck et al (2000a)mdasha rising ε

2 and a declining R2Panel B estimates εn

2 mn2 and Rn

2 from firm data forthe first and last halves of our sample periodmdashdropping themiddle two years to mitigate autocorrelation problemsF-tests show an unambiguous rise in firm-specific variationa smaller rise in systematic variation and a resultant declinein Rn

2 mdashall highly significant As a robustness check weconduct a bootstrapping (B) test by recalculating the aver-age first subperiod εn

2 mn2 and Rn

2 in each country 100times using 30 randomly selected stocks each time Thisgenerates a distribution for the first-subperiod average firm-specific variation The p-level for rejecting equal average

TABLE 2mdashCHANGES IN COMOVEMENT MEASURES BETWEEN 1990 AND 2001

Panel A Trends in comovement from 1990 to 2001 are estimated with annual data assuming unit roots and then assuming their absence Becausesmall samples render augmented DickeyndashFuller (DF) unit root tests problematic we report both Figures shown are estimates 102

Sample Dependent Variable Trend Estimates and p-Levels DF p-Level

Unit root assumed in estimation No No Yes YesLagged dependent variable included No Yes No Yes

Emerging markets stocksfirmndashweek observationsweighted equally

Absolute firm-specificvariation εn

200630 00325 00304 00163 069

(000) (030) (066) (082)

Absolute systematicvariation mn

200126 00183 00053 00118 030

(017) (002) (089) (077)

Relative systematicvariation Rn

200105 00033 00167 00091 000

(011) (031) (043) (061)

Emerging markets stockscountries weightedequally


200210 00080 00020 0343 057

(008) (055) (096) (005)


200002 00046 00173 493 012

(098) (041) (054) (014)


2000577 0205 135 279 000

(011) (028) (034) (000)

US stocks


200013 00003 0002 0001 056

(091) (098) (070) (073)


200016 00018 00008 219 043

(002) (013) (068) (095)


20479 0552 0328 197 061

(001) (007) (045) (060)

Panel B Tests to reject equal firm-specific mean variation across subperiods F-test degrees of freedom correspond to firm observations Bootstrap (B)tests to reject equal average firm-specific variation in the two subperiods recalculate the average first-subperiod firm-specific variations 100 times using30 randomly selected stocks each time This provides a probability distribution for first subperiod average firm-specific variation The p-level is themass in the tail of this distribution beyond the second-subperiod sample average firm-specific variation Subperiods are noncontiguous to mitigateautocorrelation and are of equal length All emerging market stocks are weighted equally regardless of country Variation figures are estimates 102

Comovement Measure

Emerging Markets United States

εn2

mn2 Rn

2εn

2mn

2 Rn2

Mean variation 1990 to 1994 0629 0169 0242 0310 00062 00483Mean variation 1997 to 2001 1236 0308 0234 0295 00173 00723Change in mean variation 0607 0138 0008 0015 00111 00240F-test p-level to reject no change (000) (000) (000) (000) (000) (000)B-test p-level to reject no change (000) (000) (041) (044) (000) (003)


firm-specific variation in the two subperiods is the tail ofthis distribution beyond the second-subperiod sample εn

2 mn

2 or Rn2 These tests show a significant rise in firm-

specific variation in emerging markets as a whole a smallerincrease in systematic variation and an insignificant declinein R2 Analogous US figures for the same period show arise in mn

2 and Rn2 in the 1990s however this is again a

short-term phenomenon Similar techniques applied over alonger time series of US data from 1963 to 2001 confirmthe rising εn

2 and falling Rn2 noted by Campbell et al

(2001) and Morck et al (2000a)The last panel of table 2 reproduces the trend tests and

subsample tests from panels A and B for each individualemerging market Of the 68 trend point estimates 47 arepositive and 8 are statistically significantmdashconsistent withbroadly rising firm-specific variation In contrast only 2 ofthe 21 negative trend estimates are significant F-tests andbootstrapping tests are more definitive indicating signifi-cantly higher firm-specific variation in 12 of our 17 emerg-ing marketsmdashColombia India Indonesia Korea MalaysiaMexico Pakistan the Philippines Portugal South Africa

Taiwan and Thailand F-tests indicate declining firm-specific variation only in Brazil Chile Greece Peru andTurkey and bootstrapping tests are significant only forChile Peru and Turkey

Space constraints prevent the inclusion of detailed de-scriptions of our other comovement variables In 12 of our17 countries we observe a decline in Rn

2 measured analo-gously In 11 countries mn

2 rises During crisis years εn2

and mn2 both spike Because the latter rises more Rn

2 alsospikes These observations are unsurprising at least on thesurface5

As a robustness check we repeat the F and B tests intable 2 using the first and last pairs of years of data for eachcountry rather than the first and last halves of the sample

5 Economic crises by their very nature are systematic They affectbroad swaths of firms and industries simultaneously and so are apparentas elevated systematic variation Firm-specific variation can rise too forthe crisis may affect some firms or sectors more than others and may evenpresent opportunities to some firms If crises also correspond to maniasand panics market swings due to noise trading might also heightencomovement

TABLE 2mdashCONTINUED

Panel C Trends and changes in firm-specific variation by emerging market Trends are estimated as in panel A changes as in panel B

Country Period Trend Estimates and p-LevelsaDF-Testp-Level Subperiod εn

2 a εn2 a

F-Testp-Level

B-Testp-Level

Unit root assumed Yes Yes No NoDependent variable lagincluded

Yes No Yes No

Brazil 95ndash01 0687 0028 0034 0019 065 95ndash97 1538 0136 000 031(018) (072) (030) (046) 99ndash01 1402

Chile 91ndash01 0067 0054 0010 0028 000 91ndash95 0655 0180 000 007(024) (034) (022) (008) 97ndash01 0475

Colombia 93ndash97 0072 0021 0006 0003 015 93ndash94 0489 0583 000 000(060) (059) (029) (068) 96ndash97 1072

Greece 90ndash01 0178 0065 0002 0021 008 90ndash94 0646 0019 000 044(031) (042) (093) (033) 97ndash01 0628

India 90ndash01 0159 0016 0021 0026 018 90ndash94 0924 0540 000 003(055) (081) (025) (010) 97ndash01 1463

Indonesia 90ndash01 0439 0026 0061 0082 035 90ndash94 0714 0990 000 000(012) (090) (027) (008) 97ndash01 1703

Korea 90ndash01 0372 0055 0020 0126 066 90ndash94 0308 1411 000 000(011) (074) (068) (002) 97ndash01 1722

Malaysia 90ndash01 0134 0004 0013 0014 007 90ndash94 0386 0185 000 000(032) (095) (034) (020) 97ndash01 0571

Mexico 90ndash01 0135 0004 0014 0013 012 90ndash94 0455 0260 000 002(041) (091) (012) (005) 97ndash01 0716

Pakistan 93ndash01 0136 0033 0009 0021 017 93ndash96 0903 0795 000 009(070) (063) (060) (029) 98ndash01 1699

Peru 92ndash01 0099 0063 0081 0102 002 92ndash95 1447 0469 000 000(018) (042) (024) (000) 98ndash01 0977

Philippines 90ndash01 0386 0017 0023 0047 066 90ndash94 1015 0551 000 000(011) (082) (034) (001) 97ndash01 1566

Portugal 90ndash01 0117 0004 0002 0004 000 90ndash94 0705 0197 000 009(061) (089) (088) (061) 97ndash01 0902

South Africa 90ndash01 0142 0066 0079 0095 092 90ndash94 0884 0646 000 000(031) (023) (002) (000) 97ndash01 1530

Taiwan 90ndash01 0063 0003 0029 0020 047 90ndash94 0287 0228 000 000(022) (094) (000) (015) 97ndash01 0515

Thailand 90ndash01 0371 0010 0017 0043 047 90ndash94 0517 1064 000 000(016) (095) (053) (020) 97ndash01 1581

Turkey 90ndash01 0470 0046 0023 0029 013 90ndash94 1231 0394 000 001(015) (058) (032) (007) 97ndash01 0837

Comovement is measured by average market model R2 (Rn2 ) average firm-specific variation (εn

2 ) and average systematic variation (mn2 )

aTrend and variation figures are estimates 102


period and generate broadly similar results Unit-root-invariant trend tests as suggested by Vogelsang (1998)indicate uniform insignificance but are barely identified

Our focus is on how openness and institutions correlatewith comovement These same factors may affect the viril-ity and incidence of crises so crises cannot really bedisentangled from them However a thorough analysis ofthe interactions of crises with institutions openness andstock return variation is beyond the scope of this studythough we are pursuing it elsewhere Nonetheless weclearly must control for transitory effects during crises whenwe evaluate the determinants of their more permanentlevels We return to this issue below

Table 3 reports our regression results The leftmost panelshows no relationship between trade openness and absolutefirm-specific variation Capital openness and its cross prod-uct with good government are both significantly related tohigher firm-specific variation When both are included inregression 36 the individual coefficients are insignificantand capital openness per se switches sign Although anF-test shows the two to be jointly significant collinearityproblems make interpreting the point estimates problematicIf we ignore such problems a good-government indexabove 11 induces a positive relationship between capitalopenness and absolute firm-specific variation

The center panel shows openness in trade significantlypositively related to absolute systematic variation and unre-lated to firm-specific variation The cross term betweentrade openness and good government in regression 39 isalso positive and statistically significant Trade openness isassociated with greater marketwide variation

The rightmost panel shows that trade openness ispositively related to the comovement measure oslashn Whenwe include both trade openness and its cross term withgood government in regression 317 trade openness hasa positive coefficient and the cross term has a negativeone both insignificant However the F-statistic indicates

joint significance In contrast capital market openness issignificantly associated with lower systematic variationrelative to the total Again the cross product with insti-tutional development is significantly negative Both re-main significant when included together in regression320 however capital openness takes a positive sign andthe cross term becomes negative Including both tradeand capital openness and both cross terms in regression321 leaves the point estimates of the two capital open-ness terms virtually unchanged from regression 320 Thepoint estimates in regression 321 which are both indi-vidually and jointly significant imply that a good-government index greater than 19 makes the overalleffect of capital openness on comovement negative Themean value for the good-government measure is 19

A Robustness Checks

We repeat all our results using comovement measuresestimated with a simple domestic market model rather thanequation (1) Our results remain qualitatively similar bywhich we mean the signs and statistical significance ofregression coefficients in analogs to the tables shown arepreserved Likewise using DataStreamrsquos global market re-turn rather than the US market return as the second factorin equation (1) produces qualitatively similar results Usingvalue weighting rather than equal weighting in construct-ing the local index also yields qualitatively similar results

As alternative comovement measures we employ theaverage correlation between all possible pairs of thirtystocks randomly selected in each country for each pe-riod and the fraction of stocks moving with the marketRegressions explaining logistic transformations of thesemeasures of comovement closely resemble the regres-sions explaining oslashn

To ascertain that our results are not due to comovementchanges associated with crises we repeat our regressions

TABLE 3mdashPANEL REGRESSIONS

Dependent VariableRegression

Logarithm of Idiosyncratic Variationεn

2Logarithm of Systematic Variation

mn2

Logistic Transformation of SystematicVariation as Fraction of Total Variation Rn

2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)

Trade openness goodgovernment

07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)

Capital openness 78 88 150 35 184 145 43 272 296(00) (51) (34) (29) (32) (46) (08) (05) (04)

Capital openness goodgovernment

04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)

Peso crisis dummy 46 46 47 44 44 45 44 110 111 111 108 109 108 108 64 64 64 64 65 63 64(05) (05) (05) (05) (05) (05) (05) (00) (00) (00) (00) (00) (00) (00) (00) (00) (00) (01) (01) (01) (00)

Asian crisis dummy 47 47 49 37 38 39 46 63 66 66 57 58 56 61 16 18 16 20 20 17 15(01) (01) (01) (03) (02) (02) (01) (00) (00) (00) (01) (01) (02) (01) (30) (24) (31) (25) (24) (31) (35)

Real crisis dummy 15 15 14 11 10 09 13 20 20 20 30 30 29 19 35 35 35 41 41 38 32(58) (58) (60) (68) (70) (72) (63) (55) (53) (54) (40) (40) (42) (56) (15) (15) (15) (13) (13) (15) (18)

F-statistic foropenness terms

19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02

Regression R2 59 60 60 63 63 63 63 72 72 72 67 67 67 73 75 75 75 72 72 73 77Sample 136 136 135 150 150 149 133 136 136 135 150 150 149 133 136 136 135 150 150 149 133

Independent variables include capital openness a value-weighted fraction of the market open to foreign investors trade openness importsGDP relative to GDP(world GDP) and interactions with goodgovernment a cross-section index taking low values where corruption is worse The peso crisis dummy is 1 for Latin American countries in 1995 and 0 otherwise The Asian crisis dummy is 1 for Asian countriesin 1997 and 1998 and 0 otherwise The real crisis dummy is 1 for Latin American countries in 1998 and 0 otherwise Data are for 1990 through 2001 The dependent variables are as indicated All regressionsinclude year and country fixed effects


dropping all observations for which any of the three crisisdummies described above is 1 The results are virtuallyunchanged and the panel regression R2rsquos rise Dropping thecrisis dummies and including all observations also generatesresults similar to those shown as does including onlycountry fixed effects

As a further robustness check we use an alternativecapital openness measure constructed by Abiad and Mody(2002)6 Unfortunately this is available only up to 1996 Ityields a pattern of signs and coefficients similar to thoseshown but with much lower significance levels probablydue to the smaller intersection of that measure with ourcomovement estimates

Substituting the simple trade openness measure of im-ports plus exports divided by GDP for the Frankel (2000)trade measure generates similar patterns of signs and sig-nificance to those shown

Cookrsquos D-statistics indicate that outliers are not drivingour results Tests for heteroskedasticity reject the need formodified t-tests

V A Case Study

A case in point to illustrate the situation is the contrastbetween two eastern European countries Poland and theCzech Republic Both countries experienced a flurry of newlegislation in 1991 and 1992 establishing basic marketeconomy institutions However Glaeser Johnson and Shle-iter (2001) show that the two countries then followed verydifferent trajectories The judicial systems remained ill de-veloped in both countries However strict Polish regulatoryenforcement contrasted starkly with the hands-off regula-tion inspired by the libertarian philosophy of the Czechgovernment Glaeser et al (2001) argue that this stunted thedevelopment of the Czech financial system relative to thatof Poland and stress the need for law enforcement by eitherthe judiciary or regulators to make markets work

Neither country is included in our sample because of theunavailability of complete stock market and institutionaldevelopment data However by downloading daily datafrom DataStream and following precisely the same proce-dure outlined in section III we were able to construct a setof bimonthly comovement measures for these countries7

Figure 2 shows an upward trend in firm-specific variationin Poland and a downward trend in the Czech Republic inthe latter years of the 1990s as both opened their economiesin preparation for accession to the European Union This isconfirmed in Table 4 where Polish data show a highlysignificant positive trend in ε

2 and negative trends in sys-

tematic variation both absolute and relative to total varia-tion Since a unit root cannot be rejected in the last of thesemeasures unit-root-invariant tests are used to confirm asignificant negative trend in R2 In contrast the CzechRepublic shows no trend in ε

2 F-tests and bootstrapping(B) tests akin to those in table 2 show an even greatercontrast indicating an actual decline in ε

2 for Czech stocksPolandrsquos R2 is much higher than its Czech counterpart

early on and falls to levels comparable with the latter asPolish marketwide variation abates However simple con-vergence cannot be a complete explanation because panel Bshows Polish firm-specific variation concordantly risingfrom 0000954 to 0001414 while Czech firm-specific vari-ation falls from 0000828 to 0000573 The contrast illus-trates how opening is associated with reduced comovementand higher firm-specific variation only if the institutionsprotect private property rights

VI Conclusions

Firm-specific variation in individual stock returns risesthough not monotonically during the 1990s in most but notall emerging markets Thus the rising firm-specific varia-tion detected by Morck et al (2000a) and Campbell et al(2001) in US stocks is an international phenomenon

This effect seems related to globalization Greater capitalmarket openness is associated with higher firm-specificvariation and hence lower comovement in countries withinstitutional integrity (good government) In contrast goodsmarket openness is generally associated with higher system-atic variation and hence greater comovement

Unfortunately individual stock returns are not electroni-cally available in most countries before 1990 This meanswe cannot control meaningfully for sectoral shocks and thelike and that the power of time series tests is low Thus ourfinding must be taken as preliminary Moreover the inter-relations among our openness variables and between themand other measures of development are doubtless compli-cated It may be that our openness measures are proxies forother more nuanced aspects of development With thesecaveats in mind we can tentatively consider possible im-plications of our results

Although there is near-uniform agreement among econ-omists that trade openness is welfare-enhancing capitalopenness is subject to debate with many such as Bhagwati(1998) arguing that it creates scope for destabilizing market-wide fluctuationsmdashso-called ldquohot moneyrdquo problems Ourresults suggest that such concerns can be overstated Wefind marketwide fluctuations associated with trade open-ness not capital openness In retrospect this is reasonablefor trade openness is thought to induce greater specializa-tion converting industry effects into marketwide fluctua-tions If this is really happening further study along theselines seems warranted

Why capital openness is associated with higher firm-specific variation and lower comovement is at present un-

6 This measure adds scores for aspects of capital openness directedcreditreserve requirements interest controls entry barriers and procom-petition measures regulation of securities markets privatization andinternational capital flow openness See Abiad and Mody (2002) fordetails

7 We are slowly constructing higher-frequency comovement measuresfor more countries


clear Candidate explanations include reduced tunnelingdue to greater transparency a faster pace of creativedestruction causing greater differences between innova-tors and laggards less investor herding and perhapspresently ill-understood differences in the cost structureof information and hence in the activity of informedarbitrageurs Nevertheless Wurgler (2000) Durnev et al(2004) and others find higher firm-specific variation relatedto greater stock market functional efficiency which Tobin(1982) defines as asset prices inducing an efficient distributionof capital goods

Given these findings and keeping the above-mentionedcaveats in mind the magnitude of firm-specific variation ina countryrsquos stocks presents itself as an interesting variablewith which to examine institutional development as sug-gested by the Czech RepublicndashPoland comparison aboveBetter institutions should cause the market to make a

sharper distinction between firms with good prospects andfirms with poor prospects and thus to allocate capital moreefficiently We believe that these findings suggest a new andpotentially useful measure of the effectiveness of reforms indifferent countries We tentatively propose that increasingfirm-specific variation might be regarded as a gauge of theextent of real institutional reform

The view outlined here is not new In the Pure Theoryof Capital Hayek (1941 p 6) argues that ldquothe stock ofcapital is not an amorphous mass but possesses a definitestructure that it is organized in a definite way and that itscomposition of essentially different items is much moreimportant than its aggregate lsquoquantityrsquordquo In a healthyeconomy Hayek argues different companies undertakedifferent investments because their managers possessdifferent levels of entrepreneurial ability openness toinnovation and foresight Some firms succeed and others

FIGURE 2mdashVARIANCE DECOMPOSITION OF INDIVIDUAL STOCK RETURNS IN POLAND AND THE CZECH REPUBLIC

Bimonthly comovement measures are derived from market model regressions of daily individual stock returns on domestic and US market returns and include the average regression R2 the systematic (explained)variation in the average stockrsquos returns and the firm-specific (residual) variation in the average stockrsquos returns

Systematic and firm-specific variation are plotted on the left axis R2 is on the right axis Table 4 variation tests are based on a single regression for each firm across each subperiod and are therefore not directlycomparable with the annual regression figures displayed here


fail as the economy grows through this ongoing processof creative destruction

We recognize that ours is not the final word and invitealternative explanations of the patterns we detect We wel-come ideas about how to distinguish such possibilities fromthe economic underpinnings we propose

REFERENCES

Abiad Abdul and Ashoka Mody ldquoStatus Quo Bias in Financial ReformrdquoInternational Monetary Fund working paper (2002)

Bekaert Geert and Campbell Harvey ldquoTime-Varying World MarketIntegrationrdquo Journal of Finance 502 (1995) 403ndash445ldquoEmerging Equity Market Volatilityrdquo Journal of Financial Eco-

nomics 431 (1997) 29ndash78Beny Laura ldquoA Comparative Empirical Investigation of Agency and

Market Theories of Insider Tradingrdquo Harvard Law School doctoraldissertation (2000)

Bhagwati Jagdesh ldquoThe Capital Myth The Difference between Trade inWidgets and Trade in Dollarsrdquo Foreign Affairs 773 (1998) 7ndash12

Bris Arturo William Goetzmann and Ning Zhu ldquoEfficiency and theBear Short Sales and Markets around the Worldrdquo Yale School ofManagement working paper (2002)

Bushman R J Piotroski and A Smith ldquoWhat Determines CorporateTransparencyrdquo University of North Carolina working paper(2002)

Campbell John and Robert Shiller ldquoCointegration and Tests of PresentValue Modelsrdquo Journal of Political Economy 955 (1987) 1062ndash1089

Campbell John Martin Lettau Burton G Malkiel and Yexiao Xu ldquoHaveIndividual Stocks Become More Volatile An Empirical Explora-tion of Idiosyncratic Riskrdquo Journal of Finance 561 (2001) 1ndash43

Caves Richard Multinational Enterprise and Economic Analysis (Cam-bridge University Press 1982)

Chen Zhiwu and Peter Knez ldquoMeasurement of Market Integration andArbitragerdquo Review of Financial Studies 82 (1995) 287ndash325

Coffee John ldquoCompetition among Securities Markets A Path DependentPerspectiverdquo Columbia University law and economics workingpaper 192 (2002)

Collins Daniel S P Kothari and Judy Rayburn ldquoFirm Size and theInformation Content of Prices with Respect to Earningsrdquo Journalof Accounting and Economics 9 (1987) 111ndash138

Durnev Artyom Randall Morck Bernard Yeung and Paul ZarowinldquoDoes Greater Firm-Specific Return Variation Mean More or LessInformed Stock Pricingrdquo Journal of Accounting Research 415(2003a) 797ndash836

Durnev Artyom Merritt Fox Randall Morck and Bernard Yeung ldquoLawShare Price Accuracy and Economic Performancerdquo Michigan LawReview 1023 (December 2003b)

Durnev Artyom Randall Morck and Bernard Yeung ldquoValue EnhancingCapital Budgeting and Firm-Specific Stock Returns VariationrdquoJournal of Finance 591 (2004) 65ndash106

Edison Hali and Francis Warnock ldquoA Simple Measure of the Intensity ofCapital Controlsrdquo IMF international finance discussion paper 708(2002)

Forbes Kristin and Roberto Rigobon ldquoNo Contagion Only Interdepen-dence Measuring Stock Market Comovementsrdquo Journal of Fi-nance 575 (2002) 2223ndash2262

Frankel Jeffrey ldquoAssessing the Efficiency Gain from Further Liberaliza-tionrdquo IMF working paper (2000)

TABLE 4mdashTREND TESTS FOR CZECH AND POLISH DATA

Panel A Trends in firm-specific variation from 1990 to 2001 in each country estimated using bimonthly data Estimates assuming no unit root arecoefficients on time index in regressions of levels Estimates assuming a unit root are intercepts in regressions of first differences

Czech Republic Poland

εn2

mn2 Rn

2εn

2mn

2 Rn2

Augmented DickeyndashFuller p-level 000 000 000 003 007 022Trend tests assuming no unit root

Dependent variable lag structurea 1 3 4 5 0 1 1 1Trend estimate 105 387b 228b 000216 651b 416b 00486P-level (014) (000) (000) (000) (000) (000)

Trend tests assuming unit rootDependent variable lag structurea 1 1 6Trend estimate 105 290b 000494P-level (084) (061)

Unit-root-invariant testsc

1 p-level if statistic 265 0447 0135 0722 0149 00123 14325 p-level if statistic 215 0677 0302 113 0288 00688 2245 p-level if statistic 172 0852 0471 144 0414 0178 287

Panel B Tests rejecting equal firm-specific variation across subperiods F-test degrees of freedom correspond to firm observations Bootstrap (B)test recalculates average first-subperiod firm-specific variations 100 times using 30 randomly selected stocks each time This provides aprobability distribution for first-subperiod average firm-specific variation The p-level is the mass in the tail of this distribution beyond the second-subperiod sample average firm-specific variation Subperiods are noncontiguous to mitigate autocorrelation and are of equal length All emergingmarket stocks are weighted equally regardless of country

Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2


aDependent variable lags of one through six 2-month periods are assumed initially A stepwise algorithm eliminates the least significant lag reruns the regression and repeats until all remaining lags are significantat 10 p-levels Zero indicates all were eliminated Czech sample is 44 bimonthly observations (SeptemberndashOctober 1994 to NovemberndashDecember 2001) Polish is 45 observations (JulyndashAugust 1994 toNovemberndashDecember 2001)

bReported value is estimate multiplied by 106cUsing the method of Vogelsang (1998)dReported absolute variations are estimates multiplied by 104


Glaeser Edward Simon Johnson and Andrei Shleifer ldquoCoase versus theCoasiansrdquo Quarterly Journal of Economics 116 (2001) 853ndash899

Hayek Friedrich The Pure Theory of Capital 1941ldquoUse of Knowledge in Societyrdquo American Economic Review 4(1945) 519ndash530

Johnson Simon Rafael La Porta Florencio Lopez-de-Silanes and AndreiShleifer ldquoTunnelingrdquo American Economic Review 902 (May2000) 22ndash27

Karolyi Andrew and Rene Stulz ldquoIssues in International Asset Pricingrdquoin George Constantinides Milton Harris and Rene Stulz (Eds)Handbook of the Economics of Finance (Amsterdam North-Holland 2003)

La Porta Rafael Florencio Lopez-de-Salines and Andrei Shleifer ldquoCor-porate Ownership around the Worldrdquo Journal of Finance 542(1999) 471ndash517

Morck Randall Bernard Yeung and Wayne Yu ldquoThe Information Con-tent of Stock Markets Why Do Emerging Markets Have ComovingStock Price Movementsrdquo Journal of Financial Economics 58(2000a) 215ndash238

Morck Randall David A Stangeland and Bernard Yeung ldquoInheritedWealth Corporate Control and Economic Growth The CanadianDiseaserdquo in R Morck (Ed) Concentrated Corporate Ownership(Chicago University of Chicago Press 2000b)

Rajan Raghuram and Luigi Zingales ldquoThe Great Reversals The Politicsof Financial Development in the Twentieth Centuryrdquo Journal ofFinancial Economics 691 (July 2003) 5ndash50

Roll Richard ldquoR2rdquo Journal of Finance 43 (1988) 541ndash566Schumpeter Joseph Theorie der Wirtschaftlichen Entwicklung (Leipzig

Dunker und Humbolt 1912)Stulz Rene M ldquoGlobalization Corporate Finance and the Cost of Capitalrdquo

Journal of Applied Corporate Finance 123 (1999) 30ndash39Tobin James ldquoOn the Efficiency of the Financial Systemrdquo Lloydrsquos

Banking Review (July 1982)Vogelsang Timothy ldquoTrend Function Hypothesis Testing in the Presence

of Serial Correlationrdquo Econometrica 661 (1998) 123ndash148West Kenneth ldquoDividend Innovations and Stock Price Volatilityrdquo Econo-

metrica 561 (1988) 37ndash61Wurgler Jeffrey ldquoFinancial Markets and the Allocation of Capitalrdquo

Journal of Financial Economics 58 (2000) 187ndash214


firm-specific plus market-related variation in the return ofthe underlying asset

Chen and Knez (1995) argue that as barriers to capitalflows drop cross-border arbitrage affects asset prices acrossmarkets Thus information about industries exposures ordiscount rates that affects prices in one country also affectsprices elsewhere Some of this information is about thewhole market but much is about sectors and should showup as idiosyncratic risk

The second class of reasons has to do with the realeconomy We now review each of these reasons in turn

First La Porta Lopez-de-Salines and Shleifer (1999)find most large publicly traded firms outside the UnitedStates and United Kingdom organized into corporategroups A single controlling shareholder usually a verywealthy family controls all the firms in such a groupmdashdirectly or indirectly Group firms finance each other dobusiness with each other adopt common strategies andotherwise coordinate decisions This could give their fun-damentals a common component Such intercorporate deal-ings might steadily transfer wealth from one group firm toanother and public shareholders of a wealth-losing groupfirm understandably view this as a corporate governanceproblem which Johnson et al (2000) dub tunneling Per-haps comovement gauges the importance of corporategroups and of tunneling

As Coffee (2002) Karolyi and Stulz (2003) MorckStangeland and Yeung (2000b) Rajan and Zingales (2003)Stulz (1999) and others note capital market opennesspressures regulators to adopt international best practices indisclosure governance and regulation It also creates localdemand for information professionals like accountants andanalysts Both changes might render tunneling more diffi-cult The result might be less comovement in both firmfundamentals and returns

Second economic growth arises from technologicalprogress Schumpeter (1912) holds that this occurs as inno-vative firms rise to displace established industry leaders ina process he dubs creative destruction More intense cre-ative destruction thus causes the fundamentals of innovativeand laggard firms to differ more

Caves (1982) argues that openness to outward foreigndirect investment (FDI) raises the rewards to innovators byallowing greater economies of scale and that openness toinward FDI allows technology spillovers from multination-als to local firms Rajan and Zingales (2003) and Morck etal (2000b) argue that foreign portfolio investment (FPI)openness bolsters local technological progress by lettingentrepreneurial startups obtain financing from abroad Thusgeneral capital market openness might let innovative firmsoutpace sedate rivals faster magnifying firm-specific differ-ences in returns

Third neoclassical trade theory links trade openness tospecialization Trade openness should thus reduce an econ-omyrsquos diversification across industries perhaps turning in-

dustry factors into market factors and raising fundamentalscomovement Also if capital openness accompanies tradeopenness a positive correlation between capital opennessand comovement might ensue

However tunneling innovation and specialization allaffect returns comovement only by affecting fundamentalscomovement Morck et al (2000a) cannot explain returnscomovement with fundamentals (return on assets) comove-ment1 Either their fundamentals comovement measure isinadequate (a possibility given the low frequency of fun-damentals data) or some other explanation is paramount

The latter possibility leads them to consider alternativeexplanations Comovement may be symptomatic of marketinefficiencies such as bubbles or herding If more opencapital markets experience fewer bubbles and less herdingthey might exhibit less comovement Alternatively critics ofglobalization argue that openness spreads crises For exam-ple the Bernama News Agency quoted Malaysian PrimeMinister Mahathir Mohamad blaming Malaysiarsquos economiccrisis on international financiers who ldquorobbed the Palestin-ians of everything but in Malaysia they could not do sohence they do this depress the ringgitrdquo2 More sagaciouslyBhagwati (1998) argues that capital market openness canspread financial crises and that only product market open-ness is justified and Forbes and Rigobon (2002) studycontagion as a factor in market fluctuations

However Morck et al (2000a) and Campbell et al(2001) show that changes in comovement are due in part atleast to changes in idiosyncratic variation as opposed tomarketwide variation Roll (1988) argues that idiosyncraticvariation reflects trading by investors with private firm-specific information and Morck et al (2000a) and DurnevMorck and Yeung (2004) speculate that more idiosyncraticvariation reflects periods of especially intense trading bysuch investors and hence more accurate pricing at least inthe short term However Campbell et al (2001) dispute thisnoting correctly that West (1988) links less information tohigher returns variation Their intuition is that as informa-tion is revealed a previously more erroneously priced stockexhibits larger return fluctuations

Regardless a growing body of empirical work linksgreater idiosyncratic variation to variables that on thesurface at least are plausible proxies for the informationcontent of stock prices Some of this work links higherfirm-specific return variation to variables readily interpret-able as gauging informed arbitrage Greater idiosyncraticvariation is evident in countries with stronger insider tradingprohibitions (Beny 2000) more developed financial analy-

1 Morck et al (2000a) run market-model analog regressions on returns onassets defined as earnings plus depreciation plus interest over net assets Thisallows them to estimate firm-specific and systematic fundamentals variationControlling for these variables does not affect their results

2 From an October 10th 1997 speech to Muslim villagers quoted inldquoMalaysia Premier Sees Jews behind Nationrsquos Money Crisisrdquo by SethMydans New York Times October 16 1997





III Methodology





rnjt in ij rit

Jnt 1 (2)












εn2

jn εsj2

jn Tj Jn (3)




mn2

jn msj2

jn Tj Jn(4)




1 εsn

2

sn2 1

εn2

n2 Rn

2 (5)


Rn2

msn2

sn2

mn2

n2 (6)













oslashn lnmn2 lnεn

2 ln Rn2

1 Rn2 (8)

Since mn2 and εn





trade opennessn Mn

Yn 1

Yn

n Yn (9)














C Sample



Market Rn2

εn2

mn2

CapitalOpenness

TradeOpenness

GoodGovernment







IN EMERGING MARKET





IV Findings







2 mn2 and Rn



2 mn2 and Rn








200630 00325 00304 00163 069

(000) (030) (066) (082)


200126 00183 00053 00118 030

(017) (002) (089) (077)


200105 00033 00167 00091 000

(011) (031) (043) (061)



200210 00080 00020 0343 057

(008) (055) (096) (005)


200002 00046 00173 493 012

(098) (041) (054) (014)


2000577 0205 135 279 000

(011) (028) (034) (000)

US stocks


200013 00003 0002 0001 056

(091) (098) (070) (073)


200016 00018 00008 219 043

(002) (013) (068) (095)


20479 0552 0328 197 061

(001) (007) (045) (060)


Comovement Measure


εn2

mn2 Rn

2εn

2mn

2 Rn2




2 mn



















2 a εn2 a

F-Testp-Level

B-Testp-Level


Yes No Yes No




























A Robustness Checks








mn2


2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)


07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)



04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)





19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02








V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2
























III Methodology





rnjt in ij rit

Jnt 1 (2)












εn2

jn εsj2

jn Tj Jn (3)




mn2

jn msj2

jn Tj Jn(4)




1 εsn

2

sn2 1

εn2

n2 Rn

2 (5)


Rn2

msn2

sn2

mn2

n2 (6)













oslashn lnmn2 lnεn

2 ln Rn2

1 Rn2 (8)

Since mn2 and εn





trade opennessn Mn

Yn 1

Yn

n Yn (9)














C Sample



Market Rn2

εn2

mn2

CapitalOpenness

TradeOpenness

GoodGovernment







IN EMERGING MARKET





IV Findings







2 mn2 and Rn



2 mn2 and Rn








200630 00325 00304 00163 069

(000) (030) (066) (082)


200126 00183 00053 00118 030

(017) (002) (089) (077)


200105 00033 00167 00091 000

(011) (031) (043) (061)



200210 00080 00020 0343 057

(008) (055) (096) (005)


200002 00046 00173 493 012

(098) (041) (054) (014)


2000577 0205 135 279 000

(011) (028) (034) (000)

US stocks


200013 00003 0002 0001 056

(091) (098) (070) (073)


200016 00018 00008 219 043

(002) (013) (068) (095)


20479 0552 0328 197 061

(001) (007) (045) (060)


Comovement Measure


εn2

mn2 Rn

2εn

2mn

2 Rn2




2 mn



















2 a εn2 a

F-Testp-Level

B-Testp-Level


Yes No Yes No




























A Robustness Checks








mn2


2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)


07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)



04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)





19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02








V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2
































oslashn lnmn2 lnεn

2 ln Rn2

1 Rn2 (8)

Since mn2 and εn





trade opennessn Mn

Yn 1

Yn

n Yn (9)














C Sample



Market Rn2

εn2

mn2

CapitalOpenness

TradeOpenness

GoodGovernment







IN EMERGING MARKET





IV Findings







2 mn2 and Rn



2 mn2 and Rn








200630 00325 00304 00163 069

(000) (030) (066) (082)


200126 00183 00053 00118 030

(017) (002) (089) (077)


200105 00033 00167 00091 000

(011) (031) (043) (061)



200210 00080 00020 0343 057

(008) (055) (096) (005)


200002 00046 00173 493 012

(098) (041) (054) (014)


2000577 0205 135 279 000

(011) (028) (034) (000)

US stocks


200013 00003 0002 0001 056

(091) (098) (070) (073)


200016 00018 00008 219 043

(002) (013) (068) (095)


20479 0552 0328 197 061

(001) (007) (045) (060)


Comovement Measure


εn2

mn2 Rn

2εn

2mn

2 Rn2




2 mn



















2 a εn2 a

F-Testp-Level

B-Testp-Level


Yes No Yes No




























A Robustness Checks








mn2


2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)


07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)



04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)





19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02








V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2























C Sample



Market Rn2

εn2

mn2

CapitalOpenness

TradeOpenness

GoodGovernment







IN EMERGING MARKET





IV Findings







2 mn2 and Rn



2 mn2 and Rn








200630 00325 00304 00163 069

(000) (030) (066) (082)


200126 00183 00053 00118 030

(017) (002) (089) (077)


200105 00033 00167 00091 000

(011) (031) (043) (061)



200210 00080 00020 0343 057

(008) (055) (096) (005)


200002 00046 00173 493 012

(098) (041) (054) (014)


2000577 0205 135 279 000

(011) (028) (034) (000)

US stocks


200013 00003 0002 0001 056

(091) (098) (070) (073)


200016 00018 00008 219 043

(002) (013) (068) (095)


20479 0552 0328 197 061

(001) (007) (045) (060)


Comovement Measure


εn2

mn2 Rn

2εn

2mn

2 Rn2




2 mn



















2 a εn2 a

F-Testp-Level

B-Testp-Level


Yes No Yes No




























A Robustness Checks








mn2


2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)


07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)



04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)





19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02








V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2























IV Findings







2 mn2 and Rn



2 mn2 and Rn








200630 00325 00304 00163 069

(000) (030) (066) (082)


200126 00183 00053 00118 030

(017) (002) (089) (077)


200105 00033 00167 00091 000

(011) (031) (043) (061)



200210 00080 00020 0343 057

(008) (055) (096) (005)


200002 00046 00173 493 012

(098) (041) (054) (014)


2000577 0205 135 279 000

(011) (028) (034) (000)

US stocks


200013 00003 0002 0001 056

(091) (098) (070) (073)


200016 00018 00008 219 043

(002) (013) (068) (095)


20479 0552 0328 197 061

(001) (007) (045) (060)


Comovement Measure


εn2

mn2 Rn

2εn

2mn

2 Rn2




2 mn



















2 a εn2 a

F-Testp-Level

B-Testp-Level


Yes No Yes No




























A Robustness Checks








mn2


2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)


07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)



04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)





19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02








V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2






















2 mn



















2 a εn2 a

F-Testp-Level

B-Testp-Level


Yes No Yes No




























A Robustness Checks








mn2


2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)


07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)



04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)





19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02








V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2



























A Robustness Checks








mn2


2

31 32 33 34 35 36 37 38 39 310 311 312 313 314 315 316 317 318 319 320 321

Trade openness 118 204 28 315 003 38 197 202 11(19) (52) (93) (00) (99) (92) (01) (48) (97)


07 17 05 17 17 16 10 0026 11(12) (29) (77) (00) (39) (44) (02) (99) (46)



04 08 11 01 07 04 02 15 15(00) (21) (15) (37) (41) (64) (03) (02) (03)





19 12 24 00 00 00 00 00 00 01 29 37 41 02 01 02 05 08 03 01 02








V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2

























V A Case Study









VI Conclusions



















REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2































REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2























REFERENCES























εn2

mn2 Rn

2εn

2mn

2 Rn2







Comovement Measure


εn2 d

mn2 d Rn

2εn

2 dmn

2 d Rn2





































Documents

FIRM-SPECIFIC VARIATION AND OPENNESS IN EMERGING …