transparency, price informativeness, and stock return synchronicity theory and evidence. (2010)....

JOURNAL OF FINANCIAL AND QUANTITATIVE ANALYSIS Vol. 45, No. 5, Oct. 2010, pp. 1189–1220COPYRIGHT 2010, MICHAEL G. FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON, SEATTLE, WA 98195doi:10.1017/S0022109010000505

Transparency, Price Informativeness, andStock Return Synchronicity: Theory andEvidence

Sudipto Dasgupta, Jie Gan, and Ning Gao∗

Abstract

This paper argues that, contrary to the conventional wisdom, stock return synchronicity (orR2) can increase when transparency improves. In a simple model, we show that, in moretransparent environments, stock prices should be more informative about future events.Consequently, when the events actually happen in the future, there should be less “sur-prise” (i.e., less new information is impounded into the stock price). Thus a more informa-tive stock price today means higher return synchronicity in the future. We find empiricalsupport for our theoretical predictions in 3 settings: namely, firm age, seasoned equityofferings (SEOs), and listing of American Depositary Receipts (ADRs).

I. Introduction

Financial economists generally agree that in efficient markets, stock priceschange to reflect available information—either firm-specific or marketwide. Re-cent literature has addressed the question of how a firm’s information environment(disclosure policy, analyst following) or its institutional environment (propertyrights protection, quality of government, legal origin) affect the relative impor-tance of firm-specific as opposed to marketwide factors (Jin and Myers (2006),Piotroski and Roulstone (2004), Chan and Hameed (2006), and Morck, Yeung,and Yu (MYY) (2000)). This literature has taken the perspective that if a firm’senvironment causes stock prices to aggregate more firm-specific information,

∗Dasgupta, [email protected], and Gan, [email protected], Department of Finance, Hong KongUniversity of Science and Technology, Clear Water Bay, Kowloon, Hong Kong; and Gao,[email protected], Manchester Accounting and Finance Group, Manchester Business School,University of Manchester, Booth Street West, Manchester, M15 6PB, United Kingdom. We thankUtpal Bhattacharya, Michael Brennan, Kalok Chan, Craig Doidge, Robert Engle, Wei Jiang, Li Jin,Ernst Maug, Anil Makhija, Bill Megginson, Randall Morck, Stewart Myers, Mark Seasholes, JeremyStein, Martin Walker, Bernard Yeung, and participants at the 2006 Financial Intermediation ResearchSociety Conference and the 2006 Western Finance Association (WFA) Conference for helpful com-ments and discussions. We are grateful to Hendrik Bessembinder (the editor) and Joseph Chen andArtyom Durnev (the referees) for their comments and suggestions, which greatly improved the paper.

1189

1190 Journal of Financial and Quantitative Analysis

market factors should explain a smaller proportion of the variation in stock re-turns. In other words, the stock return synchronicity or R2 from a standard marketmodel regression should be lower.

This perspective, while intuitive, is at odds with another equally intuitive im-plication of market efficiency. In efficient markets, stock prices respond only toannouncements that are not already anticipated by the market. When the informa-tion environment surrounding a firm improves and more firm-specific informationis available, market participants are also able to improve their predictions aboutthe occurrence of future firm-specific events. As a result, prevailing stock pricesare likely to already “factor in” the likelihood of the occurrence of these events.When the events actually happen in the future, the market will not react to suchnews, since there is little “surprise.” In other words, more informative stock pricestoday should be associated with less firm-specific variation in stock prices in thefuture. Therefore, the return synchronicity should be higher.

In this paper, we present a simple model to illustrate the point that a moretransparent information environment can lead to higher, rather than lower, stockreturn synchronicity. This is because, for a more transparent firm, more infor-mation is already available to market participants, reducing the “surprise” fromfuture announcements. In our model, we distinguish between 2 types of firm-specific information. One pertains to time-varying firm characteristics, reflectingthe current state of the firm, such as the next quarter’s earnings. The other istime invariant, such as managerial quality.1 Stock return synchronicity can in-crease subsequent to an improvement in transparency through disclosure of bothtypes of information. First, greater transparency can lead to early disclosure oftime-variant information. This can happen around major events such as seasonedequity offerings (SEOs) or cross-listings, during which a big chunk of informationabout future events is revealed. Thus when future events actually happen, there isless “surprise” and hence less additional information to be incorporated into thestock price, resulting in higher return synchronicity.2 While the positive effect ofgreater transparency on return synchronicity is most significant in the case of a1-time “lumpy” disclosure, we show that it also holds in the more general settingwith regular, early disclosure of information. In particular, we show that in a dy-namic setting, if at the beginning of every period, outsiders get to know (1 periodahead of time) some of the information that otherwise would come out at the endof the period, the return synchronicity is actually higher.

The second channel through which greater transparency increases stock re-turn synchronicity is due to learning about time-invariant firm-specific character-istics, such as managerial quality. In particular, better disclosure allows marketparticipants to learn about time-invariant firm fundamentals with greater preci-sion (e.g., in the extreme case where the fundamentals are completely known,

1Strictly speaking, all firm characteristics are time varying in the very long run. Here we refer tothose characteristics that do not change frequently or do not change much over time (so that they donot affect valuation significantly) as “time invariant.”

2Shiller (1981) notes theoretically that if dividend news arrives in a lumpy and infrequent way,stock price volatility becomes lower. If much of the dividend news reflects firm-specific information,one would also expect return synchronicity to become higher.

Dasgupta, Gan, and Gao 1191

there is no new learning). Therefore, with more disclosure the priors about thesefundamentals will be revised less drastically as new information comes in. Asa result, there will be less firm-specific variation in stock prices (i.e., the returnsynchronicity will be higher).

We present 3 pieces of empirical evidence consistent with our model’s pre-dictions. We first provide evidence of learning about time-invariant firm-specificinformation. The idea is that, as a firm becomes older, the market learns moreabout its time-invariant characteristics (e.g., the firm’s intrinsic quality). There-fore, return synchronicity should be higher for older firms, since more of the(time-invariant) firm-specific information is already reflected in the stock price.This prediction is strongly supported by the data.

Second and third, we exploit the fact that the effect of greater transparencyon stock return synchronicity is likely to be especially clear when the disclosureis lumpy, in the sense that the market receives a big chunk of information relevantfor future cash flows. Therefore, we focus on SEOs and cross-listings in the U.S.3

It is well known that both events are associated with significant amounts of in-formation disclosure and market scrutiny (see, e.g., Almazan, Suarez, and Titman(2002) for SEOs, and Lang et al. (2003) for American Depositary Receipt (ADR)listings). Our model suggests a dynamic response of return synchronicity to animprovement in the information environment. At the time when new informationis disclosed and impounded into stock prices, the firm-specific return variationwill increase, as suggested by conventional wisdom. However, since a big chunkof relevant information is already reflected in stock prices, we would expect thefirm-specific return variation of SEO and cross-listed firms to be subsequentlylower. This dynamic response of the firm-specific return variation around SEOsand cross-listing events is the main focus of our empirical exercise, and we findstrong support for it in the data.4

Overall, in this paper we make 2 contributions to the literature. First, weaddress the literature on transparency, informativeness of stock prices, and stockreturn synchronicity by arguing that a more transparent firm can have a higherreturn synchronicity, contrary to the conventional wisdom. Therefore, our paperhighlights that it is important to understand the nature of information disclosurein trying to interpret any particular association (or its absence) between trans-parency and stock return synchronicity. Second, we add to the growing literatureon information disclosure around security issuance events such as SEOs or ADRsby showing that stock price synchronicity changes in a way that is consistent withlumpy information disclosure associated with these events.

The rest of the paper is organized as follows. Section II reviews related lit-erature. Section III presents the model. Section IV reports the empirical findings,and Section V concludes.

3While firms can list their shares in the U.S. exchanges either through ADRs or through directlistings, the literature sometimes uses the terms “cross-listings” and “ADR listings” interchangeably(see, e.g., Lang, Lins, and Miller (2003)). In the rest of this paper, we follow this convention, exceptwhen we discuss our sample.

4A common concern about the empirical identification of the SEO/ADR effects is the potentialself-selection of SEO and ADR listings. We discuss later how our empirical specification addressesthis issue.


II. Related Literature

A. Stock Return Synchronicity (R2)

Recent literature has documented a link between the synchronicity of stockreturns and the informativeness of stock prices at the country level. MYY (2000)first report that, in economies where property rights are not well protected, thesynchronicity of stock returns (measured by a market model R2) is significantlyhigher. The authors argue that weaker property rights discourage informed arbi-trage activity based on private information, and stock prices are driven more bypolitical events and rumors. In a recent paper, Jin and Myers (2006) examine thelink between measures of corporate transparency and return synchronicity. Theyargue that in a more transparent environment, proportionately more firm-specificinformation is revealed to outside investors. As a result, marketwide informationexplains a smaller proportion of the overall return variation, resulting in lowerreturn synchronicity.

Others have investigated whether results at the country level carry over tothe firm level. They find mixed results. On the one hand, Durnev, Morck, Yeung,and Zarowin (2003) find that higher firm-specific stock price variation is as-sociated with higher information content about future earnings. On the otherhand, Piotroski and Roulstone (2004) find that return synchronicity increases withanalyst coverage. They interpret this as evidence that analysts specialize by in-dustry, and, as a result of greater analyst coverage, more industrywide and mar-ketwide information gets impounded in stock prices. Using data from emergingmarkets, Chan and Hameed (2006) report that greater analyst coverage increasesreturn synchronicity. Barberis, Shleifer, and Wurgler (2005) find that inclusion in(deletion from) the Standard and Poor’s (S&P) 500 index, which presumably in-creases (decreases) firm-level transparency, increases (decreases) a stock’s returnsynchronicity.

Given these inconsistencies, it is useful to review the determinants of themarket model return synchronicity. Consider a simple regression of firm returnon market return. In this case, R2 = SSR/SST = β2Sxx/

(β2Sxx + SSE

). Thus, an

increase in return synchronicity can come from 3 sources: i) an increase in mar-ketwide return variation (Sxx), ceteris paribus; ii) a decrease in the “idiosyncraticreturn variation (SSE),” ceteris paribus; and iii) an increase in beta (β), or thestock’s comovement with the market, ceteris paribus. The results in MYY (2000)for country-level R2 could be primarily attributable to higher marketwide returnvolatility associated with weaker property rights protection (which discouragesinformation acquisition and creates more space for noise trading); results in Jinand Myers (2006) are attributable to lower idiosyncratic return variation in coun-tries with poor transparency.

Note that at the country level, as the aggregated β is exactly 1 by definition,the country-level studies have generally associated a lower average R2 with eithera higher firm-specific return variation or lower aggregate market volatility. This,however, is not the case at the firm level. The mixed results on R2 at the firm levelcan be reconciled by this β effect: S&P additions (Barberis et al. (2005)) or moreanalyst coverage (Piotroski and Roulstone (2004), Chan and Hameed (2006)) leadto an increased comovement with the market and thus the β. Barberis et al., for


example, argue that when making portfolio decisions, investors group assets intocategories (such as small-cap stocks or value stocks) and allocate funds at thelevel of these categories. Additions to the S&P 500 may move the stock into acategory that is more popular with investors, with a resultant increase in β andR2. Likewise, as analysts help to impound more marketwide information into thestock price, the stock return exhibits higher comovement with the market, result-ing in higher β and return synchronicity. This highlights a need to control for theβ effect in firm-level studies of R2 when one is primarily interested in how theinformation environment affects the idiosyncratic return variation.

B. Information Revelation and the Informativeness of Stock Prices

The idea that a more transparent firm has stock prices that are more infor-mative about future events is not new. Fishman and Hagerty (1989), for exam-ple, present a model in which firm disclosure increases the informativeness ofstock prices about future cash flows, which in turn enhances the resource alloca-tion efficiency. Gelb and Zarowin (2002) empirically find that better disclosurepolicies are associated with stock prices that are more informative about futureearnings changes.5 In an interesting paper, Bhattacharya, Daouk, Jorgenson, andKehr (2000) find that shares in the Mexican Stock Exchange react very little tothe announcement of company news. This is not because firms listed in the stockexchange in Mexico are more transparent, but rather because, due to insider trad-ing, the superior information of insiders is already incorporated in stock prices,so there is little surprise on announcement.

Several recent papers have made an association between the informativenessof stock prices as measured by stock return synchronicity and the efficiency ofresource allocation. For example, Durnev, Morck, and Yeung (2004) and Wurgler(2000) find that higher firm-specific return variation enhances investment effi-ciency. Chen, Goldstein, and Jiang (2007) use return synchronicity as a measureof private information incorporated in the stock prices and find that investmentresponds more to stock prices when the stock return synchronicity is lower. Sim-ilar to our view, they note that investment has a weaker response to informationif the manager was already knowledgeable about the information (and hence hadalready taken the relevant action).

III. Disclosure, Transparency, and Stock ReturnSynchronicity: Theory

In this section, we present the arguments about how new disclosure and im-provement in transparency affect return synchronicity. To facilitate comparison,

5Lang and Lundholm (1996) examine the relation between firms’ disclosure policies, analyst fol-lowing, and the accuracy of analysts’ forecasts. They find that within a particular industry, firms thatare more forthcoming in their disclosure policies have a larger analyst following, more accurate ana-lyst earning forecasts, less dispersion about individual analyst forecasts, and less volatility of forecastrevisions. While they do not directly address the issue of informativeness of stock prices, their resultssuggest that future outcomes are easier to predict when firms are more transparent.


we frame the arguments in the context of a model developed in a recent paper byJin and Myers (2006).

As in Jin and Myers (2006), we assume that the firm’s cash-flow generatingprocess is

Ct = K0Xt,(1)

where K0 is initial investment, and Xt is the sum of 3 independent shocks to thefirm’s cash flow:

Xt = ft + θ1,t + θ2,t.(2)

Here, ft captures market factors that are observed by all; θ1,t and θ2,t arefirm-specific shocks. Outsiders only observe θ1,t, whereas insiders observe bothθ1,t and θ2,t. As in Jin and Myers (2006), we assume that ft, θ1,t, and θ2,t are allstationary AR(1) processes with the same AR(1) parameter ϕ, where 1 > ϕ > 0:

ft+1 = f0 + ϕft + εt+1,(3)

θ1,t+1 = θ1,0 + ϕθ1,t + ξ1,t+1, and(4)

θ2,t+1 = θ2,0 + ϕθ2,t + ξ2,t+1.(5)

Let κ=Var(θ1,t + θ2,t)/Var(ft) denote the ratio of firm-specific to market vari-ance in cash flows. Also following Jin and Myers, let η = Var(θ1,t)/(Var(θ1,t) +Var(θ2,t)), the proportion of the variance of the firm-specific component that isdue to the part that is observable to the outsiders. A higher η is associated withbetter firm transparency.

The “intrinsic value” of the firm from the point of view of investors at anypoint of time t is the present value of future cash flows conditional on their infor-mation set It:

Kt(It) = PV{E(Ct+1|It),E(Ct+2|It), . . . , r},(6)

where the discounting is done at the risk-free rate r.Outside shareholders can seize control of the firm through collective action

and manage the firm on their own. The value of the firm under the outsider share-holders’ management is αKt where α < 1. This sets the ex-dividend market valueof the firm (i.e., its value to outside investors) at

Vext (It) = α · Kt(It).(7)

We have Vext (It) = (E(Yt+1|It) + E(Vex

t+1|It))/(1 + r), where Yt+1 is the dividendat t + 1. Jin and Myers ((2006), Prop. 3) show that the equilibrium dividend is aconstant fraction α of the investor’s conditional expectation of cash flow:

Y∗τ = αE(Cτ |It), ∀ τ ≥ t.(8)

We now depart from Jin and Myers (2006) by assuming that there is a changein the firm’s disclosure policy and the firm becomes more transparent. Specifi-cally, we consider 2 different types of changes in disclosure policy: One is relatedto time-variant firm-specific information; the other concerns time-invariant infor-mation about firm characteristics.


A. Disclosure of Time-Variant Information

1. Lumpy (1-Time) Information Disclosure

During SEOs or ADR listings, the firm becomes more transparent in thesense that a big chunk of information comes out that otherwise would have comeout later, or perhaps not at all. To model this type of disclosure, we assume thatthe market learns, at time t0, of δt0+1, where

ξ1,t0+1 = ξ′1,t0+1 + δt0+1,(9a)

E(ξ′1,t0+1|δt0+1) = 0.(9b)

The interpretation is as follows: Equations (9a) and (9b) imply that the mar-ket learns 1 period ahead of time some information that is relevant for the t0 + 1cash-flow innovation. We call this information disclosure lumpy because this is a1-time early disclosure of information that reduces the variance of the cash-flowshock at t0 + 1, so that the quantum of information revealed at t0 exceeds that atany other subsequent point of time. A major event such as the listing of ADRs islikely to be associated with the revelation of information relevant for firm-specificevents that could affect future cash flows. This information, however, should beless relevant for events that occur further in the future. For simplicity of exposi-tion, we make the extreme assumption that the information revealed at disclosureaffects only the cash-flow shock 1 period later (i.e., it is only relevant for eventsthat occur 1 period later).

Denote σ = Var(ξ′1,t0+1)/Var(ξ1,t0+1) < 1. This parameter measures howmuch information is revealed early regarding the cash-flow shock 1 period later:The lower is σ, the less is the residual uncertainty regarding the innovation that isrevealed at t0 +1 (i.e., the greater is the information content of the disclosure at t0).

We are now ready to compare the effect of the change in disclosure policy att0 on stock return synchronicity.

Proposition 1a

i) The proportion of the realized variation in period t0 (i.e., between t0 andt0 + 1) explained by market factors is higher for a firm that experiences animprovement in disclosure at t0 than for a firm that does not.

ii) The proportion of realized variation explained by market factors for periodt0 − 1 is less for a firm that experiences an improvement in disclosure at t0than for a firm that does not.

Proof. See Appendix A.

Lumpy information disclosure consists of a 1-time early disclosure of newinformation that otherwise would have been revealed later. When the informa-tion is revealed and impounded into stock prices, the return synchronicity willdecrease. However, the return synchronicity will increase subsequently—there isless information content to later announcements, since part of the information isalready impounded in the stock price.


2. Regular Early Disclosure of Information

One notion of transparency is simply that news is announced in a timelymanner, so that the surprise component from future events is lower. To formalizethis notion of transparency, we assume that at the beginning of every period, thereis some disclosure that reduces the variance of the cash-flow shock revealed to thepublic at the end of the period. More formally, we assume

ξ1,t+1 = ξ′1,t+1 + δt+1, for all t,(10a)

E(ξ′t+1|δt+1) = 0, and(10b)

Var(ξ′t+1)

Var(ξt+1)= σ < 1.(10c)

We then have the following:

Proposition 1b. Suppose the risk-free rate is strictly positive, and the trans-parency improves in the sense that in every period, some δt+1 is revealed tooutsiders, where δt+1 satisfies equations (10a)–(10c). Then the stock return syn-chronicity in every period is strictly higher than that of an otherwise identical firmthat does not experience an improvement in transparency.


The result that the return synchronicity actually increases in this case maybe somewhat surprising. Each period, some of the information affecting the cash-flow innovation is disclosed early and reduces the subsequent “surprise”; however,a new piece of information relevant for the cash-flow innovation still 1 periodlater is revealed at the end of the period. Why do these two effects not washeach other out completely? The reason is that the information revealed at the endof the period regarding the cash-flow innovation still 1 period later is discountedrelative to the information revealed at the beginning of the period, since the formeris relevant for a more distant cash flow. Thus, the return synchronicity is higher.6

B. Disclosure of Time-Invariant Information about Firm Characteristics

We next show that disclosure that conveys information about time-invariantfirm characteristics such as managerial ability can also raise return synchronicity.The intuition is that if managerial ability has to be inferred (e.g., on the basisof observable cash flows), then the value of the firm will fluctuate more due toobservable cash-flow shocks, compared to a situation where managerial quality isalready known to the market on account of greater transparency and disclosure.Consequently, the proportion of the overall variation in returns that is explainedby market factors will be lower for a less transparent firm.7 Unlike the case of a

6See Peng and Xiong ((2006), p. 577) for a very similar result illustrating the effect of early arrivalof information and discounting.

7West (1988) considers a very general framework that has a similar implication. Suppose that I1and I2 are 2 information sets and I1 is a subset of I2. West shows that the forecast of the presentdiscounted value of dividends will be revised more often if the forecast is made on the basis of I1rather than I2.


1-time early disclosure of information that would have come out later, the effectof this type of disclosure on return synchronicity is likely to be more durable.

To formally demonstrate how the return synchronicity can increase, assumethat θ1,0 in equation (4) represents some firm-specific characteristic (such as man-agerial quality). The true value of θ1,0 is not known to the market, which onlyknows that it is drawn from some distribution. Moreover, define the informationset It to include the entire history of the realizations of (ft, θ1,t). We then have thefollowing:

Proposition 2. Fix a history〈(ft′ , θ1,t′): t′ ≤ t〉 up to time t. The proportion of therealized variation explained by the market factor in period t will be higher if θ1,0

is revealed to the market at any time prior to t than if it is not.


To summarize, the nature of disclosure associated with an improvement intransparency can take different forms. As in Jin and Myers (2006), it can takethe form of more firm-specific information being revealed to outsiders on a regu-lar basis, in which case the return synchronicity will decrease. Alternatively, andas we show in this section, it can be associated with either early disclosure oftime-varying firm-specific information, or disclosure of time-invariant informa-tion about firm characteristics, which may cause return synchronicity to increase.In particular, for lumpy information disclosure, return synchronicity will first de-crease when new information is impounded in stock prices but increase subse-quently. This dynamic behavior of return synchronicity around lumpy disclosureevents is what we attempt to capture in our empirical analysis in the subsequentsection.

IV. Empirical Evidence

This section provides evidence consistent with the theory outlined above, in3 different settings. The first explores the effect of variation in the informationenvironment as proxied by firm age. The other 2 settings correspond to discretechanges in the information environment due to SEOs and cross-listings.

Since the theory is about return variation that can be explained by the marketfactors (holding total return variation constant), in our empirical exercises we(inversely) measure stock return synchronicity using log(1− R2). The advantageof this measure is that it is equivalent to firm-specific return variation or the logof the sum of squared errors (LSSE) when log of total return variation (SST)is controlled for.8 Results based on R2 as a measure of return synchronicity arequalitatively the same and are not reported for brevity.

A. Stock Return Synchronicity and Firm Age

We now examine the relation between R2and firm age to provide evidence oflearning about time-invariant firm-specific information. As a firm becomes older,

8This comes from a direct transformation from R2 (a ratio variable) to SSE (a level variable). Inparticular, log(1− R2) = log(SSE)− log(SST), where SST is total variation.


the market learns more about time-invariant firm characteristics (e.g., the firm’sintrinsic quality). Thus stock return synchronicity should be higher for older firms.

We first examine the relation between R2 and firm age by estimating thefollowing basic model:

log(1− R2)i,t = α + βAGEi,t + γ Firm Controlsi,t + ηi + δt + εi,t,(11)

where i indexes firms and t indexes years. The dependent variable is based on R2

estimated from a market model (see Appendix B for details) and, as discussedearlier, is equivalent to firm-specific return variation (LSSE). AGE is the firm agesince initial public offering (IPO). Firm Controls include those commonly used inthe literature, namely, firm size (defined as the natural logarithm of market valueof assets), market-to-book (MB) ratio (defined as the ratio of market value of eq-uity plus the book value of debt over total assets), leverage (defined as the bookvalue of long-term debt over total assets), return on assets (defined as operatingincome before depreciation over total assets), as well as β. Here ηi are firm fixedeffects that control for time-invariant unobserved firm characteristics, and δt areyear fixed effects that control for macroeconomic changes. In all regressions, wecontrol for the log of the total variation of the firm’s stock return. Since informa-tion disclosed during IPO can still affect R2 in the years immediately after theIPO year, we require that firm-years in our sample are at least 3 years after theIPO year.

Table 1 presents the summary statistics of the main variables. Consistent withlearning about time-invariant information, older firms tend to have significantlyhigher R2 (lower LSSE) than do younger firms, both in terms of the mean andthe median (significant at the 1% levels). Older firms tend to be larger, moreleveraged, and more profitable. They have a lower β and a lower MB.

The regression results are reported in column (1) in Panel A of Table 2.Consistent with our univariate analysis, firm age is associated with significantlyhigher R2 (and thus lower LSSE) at the 1% levels, reflecting learning about thetime-invariant information. MB and leverage have negative (positive) effects onR2 (LSSE), whereas higher β, larger size, and higher profitability increases R2 atthe 1% level.

One potential alternative explanation of our results is that the standard mar-ket model is not the correct asset pricing model for firm-level returns. For exam-ple, our measure of R2 does not include industrywide return variation. Thus it ispossible that our age effect is driven by a time-varying industry effect. Therefore,we follow Roll (1988), Piotroski and Roulstone (2004), and Durnev et al. (2003),(2004) by adding industry returns in the standard market model regression. Theresults remain qualitatively unchanged (column (2) in Panel A of Table 2). To fur-ther address the concern that our age effects are simply picking up missing riskfactors, we estimate R2 based on the Fama-French (1993) 3-factor model and a4-factor (including momentum) model and include the firm-specific factor load-ings as independent variables in our regressions (see Appendix B for details onthe construction of these variables). Inclusion of additional risk factors does notchange the age effect on return synchronicity (columns (3) and (4) of Panel A inTable 2).

Dasg

upta,G

an,andG

ao1199

TABLE 1

Descriptive Statistics

Table 1 reports the descriptive statistics for U.S. firms during the sample period of 1976–2004, with IPO age of at least 4 years. The R 2, firm-specific return variation (SSE), and β are estimates from equation (B-1)for each firm-year using weekly data. AGE is the number of years since IPO. Leverage is long-term debt over total assets. Profitability is measured by return on assets. Market-to-book (MB) ratio is the market valueof equity plus the book value of debt over total assets. In Panel B, diversified firms are those that have multiple segments reported in Compustat. In Panel C, the control group, non-SEO firm-years, contains thosefirm-years that do not fall into any 2-year time period before or after an SEO. Significance of the differences between subsamples is based on 2-tailed tests (t-test for mean and rank sum test for median). ***, **,and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Panel A. Descriptive Statistics by Firm Age

All Older Firms Younger Firms DifferenceFirm- (IPO Age (IPO Age (Older –Years ≥ 12 Years) < 12 Years) Younger)

Mean Median Mean Median Mean Median Mean Median

Firm-specific return variation (SSE) 0.274 0.125 0.193 0.092 0.361 0.181 –0.168*** –0.089***R 2 0.130 0.082 0.154 0.105 0.104 0.063 0.050*** 0.042***AGE 16.969 12.000 26.038 21.000 7.036 7.000 19.002*** 14.000***β 0.830 0.755 0.816 0.765 0.845 0.743 –0.029*** 0.023Total assets (mil.) 3,718.697 192.110 5,335.749 376.548 1,947.954 100.380 3,387.795*** 276.169***MB 1.670 1.145 1.556 1.150 1.794 1.138 –0.238*** 0.012***Leverage 0.351 0.342 0.356 0.350 0.346 0.331 0.010*** 0.019***Profitability 0.098 0.111 0.120 0.121 0.074 0.096 0.046*** 0.025***Number of obs. 89,010 46,524 42,486

(continued on next page)

1200JournalofFinancialand

Quantitative

Analysis

TABLE 1 (continued)

Descriptive Statistics

Panel B. Descriptive Statistics by Diversification

DifferenceSingle- (Diversified –

Diversified Segment Single-Firms Firms Segment)

Mean Median Mean Median Mean Median

Firm-specific return variation (SSE) 0.268 0.114 0.397 0.198 –0.129*** –0.084***R 2 0.148 0.100 0.100 0.054 0.048*** 0.046***AGE 21.963 17.000 13.372 10.000 8.591*** 7.000***β 0.839 0.785 0.803 0.722 0.036*** 0.063***Total assets (mil.) 4,192.858 340.881 977.924 66.624 3,214.934*** 274.257***MB 1.483 1.168 1.912 1.249 –0.429*** –0.081***Leverage 0.357 0.354 0.292 0.255 0.065*** 0.099***Profitability 0.105 0.118 0.080 0.108 0.025*** 0.010***Number of obs. 34,039 43,832

Panel C. Descriptive Statistics of the SEO Sample versus the Non-SEO Sample

All Non-SEO 0–2 Years Difference 1–2 Years DifferenceFirm- Firm- Before (Before SEO – After (After SEO –Years Years SEO Non-SEO) SEO Non-SEO)

Mean Median Mean Median Mean Median Mean Median Mean Median Mean Median

Firm-specific 0.274 0.125 0.278 0.123 0.231 0.138 –0.046* 0.0145*** 0.235 0.139 –0.043 0.015***return variation(SSE)

R 2 0.130 0.082 0.127 0.078 0.156 0.119 0.029*** 0.041*** 0.175 0.139 0.048*** 0.061***AGE 16.969 12.000 16.920 12.000 18.038 12.000 1.118*** 0.000*** 16.699 12.000 –0.221 0.000***β 0.830 0.755 0.802 0.731 1.054 0.955 0.251*** 0.224*** 1.146 1.076 0.343*** 0.345***Total assets (mil.) 3,718.697 192.110 3,799.463 177.195 2,780 325.504 –1,019.789** 148.309*** 3,138.474 397.202 –660.989 220.007***MB 1.670 1.145 1.657 1.134 1.876 1.263 0.218*** 0.128*** 1.692 1.254 0.034 0.119***Leverage 0.351 0.342 0.351 0.341 0.368 0.372 0.017*** 0.030*** 0.339 0.333 –0.011** –0.008**Profitability 0.098 0.111 0.098 0.110 0.099 0.120 0.001 0.009*** 0.082 0.110 –0.016*** –0.000***Number of obs. 89,010 80,633 4,604 3,773


Another alternative interpretation of the age effect is that firm fundamentalsare more stable and, therefore, comove more for older companies. Indeed, if thefundamentals of older firms comove more either with market or industry, thenone would observe a higher R2 even without “learning.” We thus follow MYY(2000) and Durnev et al. (2004) to control for return on assets (ROA) comovementwithin the 3-digit Standard Industrial Classification (SIC) code (see AppendixB for details). The coefficient on idiosyncratic ROA movement is significantly

TABLE 2

Age and R2

Table 2 estimates the effect of firm age on R 2 as follows:

log(1− R 2)i,t = α + βAGEi,t + γFirm Controlsi,t + ηi + δt + εi,t ,

where R2 and β are estimated from a market model (equation (B-1)) in columns (1) and (5); from an industry-augmentedmarket model in columns (2) and (6); and from a Fama-French (FF) (1993) 3-factor and a 4-factor model with momentum(equations (B-2) and (B-3)) in columns (3), (4), (7), and (8). AGE is the number of years since the inclusion in CRSP. Sizeis the log of market value of assets. Leverage is long-term debt over total assets. Profitability is measured by return onassets. Market-to-book (MB) ratio is the market value of equity plus book value of debt over total assets. Total volatilityis the standard deviation of weekly returns over 1 year. Diversification indicates whether the firm has multiple Compustatsegments reported. Standard errors are clustered at the firm level and are in parentheses. ***, **, and * indicate significanceat the 1%, 5%, and 10% levels, respectively.

Panel A. Basic Model

Controlling for ROA Comovement

Industry- FF 3 Industry- FF 3Standard Augmented FF 3- Factors Standard Augmented FF 3- FactorsMarket Market Factor Plus Market Market Factor PlusModel Model Model Momentum Model Model Model Momentum

(1) (2) (3) (4) (5) (6) (7) (8)

log(AGE) –0.040*** –0.018** –0.026*** –0.029*** –0.044*** –0.023*** –0.023*** –0.025***(0.005) (0.007) (0.006) (0.006) (0.005) (0.007) (0.007) (0.007)

MB 0.003 0 0 0.002** –0.001 0 –0.001 0.001(0.002) (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.001)

β (Market factor) –0.089*** –0.091*** –0.030*** –0.020*** –0.107*** –0.084*** –0.028*** –0.017**(0.012) (0.003) (0.011) (0.007) (0.005) (0.005) (0.011) (0.007)

β (Industry return) –0.152*** –0.119***(0.007) (0.008)

β (High-minus-low 0.024*** 0.021*** 0.025*** 0.021***factor) (0.003) (0.002) (0.002) (0.002)

β (Small-minus-big 0.001 –0.001 –0.002 –0.002factor) (0.004) (0.003) (0.004) (0.003)

β (Momentum factor) 0.004 0.004(0.003) (0.003)

Log of total volatility 0.932*** 0.990*** 0.872*** 0.861*** 0.947*** 0.923*** 0.873*** 0.861***(0.011) (0.008) (0.012) (0.008) (0.008) (0.008) (0.012) (0.009)

log(assets) –0.030*** –0.036*** –0.048*** –0.053*** –0.021*** –0.037*** –0.044*** –0.050***(0.004) (0.003) (0.003) (0.003) (0.002) (0.003) (0.003) (0.003)

Profitability –0.018 –0.015 –0.013 –0.023*** –0.003 –0.003 –0.01 –0.022***(0.013) (0.009) (0.011) (0.008) (0.009) (0.012) (0.013) (0.009)

Leverage 0.051*** 0.055*** 0.067*** 0.074*** 0.051*** 0.070*** 0.073*** 0.079***(0.007) (0.009) (0.007) (0.008) (0.007) (0.009) (0.008) (0.008)

Idiosyncratic ROA 0.003*** 0.003*** 0.003*** 0.003***movement (0.001) (0.001) (0.001) (0.001)

Constant 13.142*** 3.944*** 3.668*** 3.642*** 13.188*** 3.804*** 3.657*** 3.624***(0.035) (0.041) (0.045) (0.036) (0.030) (0.038) (0.047) (0.038)

Observations 89,010 88,133 88,979 88,968 73,156 73,094 73,146 73,137Number of firms 12,015 11,934 12,011 12,011 9,806 9,803 9,804 9,804R 2 0.94 0.89 0.91 0.91 0.94 0.90 0.91 0.91

(continued on next page)


TABLE 2 (continued)

Age and R2

Panel B. The Effect of Diversification

Controlling for ROA Comovement

Industry- FF 3 Industry- FF 3Standard Augmented FF 3- Factors Standard Augmented FF 3- FactorsMarket Market Factor Plus Market Market Factor PlusModel Model Model Momentum Model Model Model Momentum

(1) (2) (3) (4) (5) (6) (7) (8)

log(AGE) –0.040*** –0.026*** –0.018*** –0.024*** –0.043*** –0.028*** –0.020*** –0.025***(0.005) (0.008) (0.007) (0.007) (0.006) (0.008) (0.007) (0.007)

MB 0.001 0.001 0.001 0.001 0 0.002* 0.001 0.001(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

β (Market factor) –0.073*** –0.071*** –0.020** –0.011* –0.074***(0.011) (0.004) (0.009) (0.006) (0.015)

β (Industry return) –0.134*** –0.074*** –0.054*** –0.019** –0.010*(0.007) (0.012) (0.010) (0.009) (0.006)

β (High-minus-low 0.020*** 0.017*** 0.021*** 0.016***factor) (0.003) (0.003) (0.003) (0.003)

β (Small-minus-big –0.003 –0.005 –0.005 –0.004factor) (0.004) (0.003) (0.004) (0.003)

β (Momentum factor) 0.004 0.004(0.003) (0.003)

Log of total volatility 0.832*** 0.974*** 0.790*** 0.781*** 0.822*** 0.810*** 0.778*** 0.767***(0.087) (0.008) (0.082) (0.081) (0.098) (0.095) (0.092) (0.091)

log(assets) –0.028*** –0.032*** –0.045*** –0.049*** –0.027*** –0.041*** –0.044*** –0.047***(0.007) (0.003) (0.006) (0.006) (0.007) (0.007) (0.006) (0.006)

Profitability –0.006 –0.012* –0.01 –0.009 –0.003 –0.006 –0.01 –0.009(0.008) (0.007) (0.008) (0.009) (0.009) (0.010) (0.009) (0.010)

Leverage 0.085** 0.057*** 0.101*** 0.105*** 0.089** 0.102*** 0.105*** 0.109***(0.035) (0.010) (0.033) (0.032) (0.037) (0.036) (0.035) (0.034)

Diversification –0.007* –0.005 –0.008* –0.008* –0.007** –0.006 –0.007* –0.007*(0.004) (0.005) (0.004) (0.004) (0.004) (0.005) (0.004) (0.004)

Idiosyncratic ROA 0.005** 0.005** 0.004** 0.005**movement (0.002) (0.002) (0.002) (0.002)

Constant 12.817*** 3.871*** 3.373*** 3.355*** 12.805*** 3.468*** 3.354*** 3.328***(0.244) (0.044) (0.229) (0.226) (0.265) (0.261) (0.249) (0.245)

Observations 77,871 77,061 77,853 77,845 70,265 70,205 70,253 70,247Number of firms 10,802 10,727 10,801 10,801 9,803 9,801 9,803 9,803R 2 0.91 0.89 0.89 0.88 0.91 0.88 0.89 0.88

positive, consistent with the conjecture that with greater comovement of funda-mentals, stock prices also tends to comove more (columns (5)–(8) in Panel A ofTable 2). However, our age effects remain unchanged.

In Panel B of Table 2, we examine whether some additional firm charac-teristics, other than those commonly used in the literature, might drive the ageeffect. One such firm characteristic is diversification. Older firms tend to belarger and more diversified sectorally. Thus they are more like portfolios, and it iswell known that diversified portfolios are much more correlated than individualstocks with broad market indices. Indeed, as shown in columns (1)–(4) in PanelB of Table 1, diversified firms are older and tend to have lower firm-specificreturn volatility (both differences are significant at the 1% level). To ensure thatwe do not simply pick up a diversification effect, we control for whether or


not the firms have multiple segments as reported in Compustat.9 As shown incolumns (1)–(4) in Panel B of Table 2, diversification is significantly associatedwith higher R2 or low LSSE (at the 10% level).10 However, diversificationdoes not drive out our age effect.11 Finally, since diversified firms tend to bemore mature and stable, we further add idiosyncratic ROA movement in theestimation (columns (5)–(8) of Panel B in Table 2). Our age effects remainqualitatively unchanged. Both diversification and idiosyncratic ROA movementeffects are significant, suggesting that each has independent influence on returnsynchronicity.

In addition to our analysis of the age effect on return synchronicity, thereis evidence that the information content of news announcements is lower forolder firms. Dubinsky and Johannes (2006) develop a numerical method to ex-tract a measure of the “surprise” content from the earnings announcements usingoptions-implied earnings jump volatility. In particular, 2 options expiring rightbefore and after the announcement dates are used. From the implied volatility ofboth options, one can back out the volatility attributable to the jump on earningsannouncement. Based on a sample of firms that have liquid option trading for1998–2004, one can regress option-implied earnings jump volatility on age and aset of controls. As plotted in Figure 1, the option-implied earnings jump volatilityis strongly (negatively) related to firm age, implying that the new informationcontent is larger for younger firms.12

B. Stock Return Synchronicity (R2) and SEOs

As discussed earlier, our point about the dynamic effect of the informationenvironment on return synchronicity is best illustrated in cases where the infor-mation disclosure is lumpy. One such setting is SEOs, which are infrequent eventsthat attract market attention and scrutiny, resulting in disclosure of a substantialchunk of new information. Most U.S. equity issuers choose a traditional mar-ket offering as a method of issuing seasoned equity.13 Typically, the issuer goesthrough a process of book building and road shows much as in an IPO. Duringthe road show, the issuing firm explains to potential investors the changes in the

9The results are robust to some other standard diversification measures in the literature, includingthe number of segments and Hirfindahl indices based on segment sales and assets, both in terms of thesigns of coefficient estimates and their statistical significance (unreported).

10We note that adding the diversification measure results in a reduced sample size. This is becauseour initial sample starts from 1976, whereas Compustat segment information is available only after1979.

11When we include an interaction term between diversification and age, this interaction is notsignificant, suggesting that the age effect does not vary across diversified and single-segment firms. Inthe interest of brevity, this result is not reported but is available from the authors.

12We thank Wei Jiang and Mike Johannes for providing us with the chart based on their project thatanalyzes the information property of the Dubinsky and Johannes (2006) measure.

13In the U.S., especially after 1997, many issuers can also choose to do accelerated offerings ratherthan traditional marketed offerings. These include accelerated book building (where they only do a 1-or 2-day road show or, more often, just a conference call the day before the offering) and block trades,which are similar to sealed bid auctions. However, the traditional method is almost always followedfor large offerings.


FIGURE 1

Age and Option-Implied Earnings Jump Volatility (U.S. sample)

Sample period: 1998–2004. Control for ln(MV) and MB, standard errors adjust for clustering at the firm level.

company (e.g., why it is raising funds now) and thus reveals considerable newfirm-specific information.14 In addition, underwriters are likely to produce infor-mation as part of their “due diligence.” The information may also be generated bynew investors if the process of equity issuance temporarily makes the stock moreliquid.

1. Empirical Specification

To capture the intertemporal response of R2 around SEOs, we pursue a spec-ification that imposes very little structure on the response dynamics. Specifically,we include dummy variables for the year of SEO, for 1 and 2 years after SEO,as well as for the 2 years immediately prior to SEO. These variables should iden-tify the response function of R2 to the passage of time around SEO. In particular,we estimate the following model on a panel of Center for Research in SecurityPrices (CRSP) firms during 1976–2004 (see Appendix B for details on sampleconstruction):

log(1− R2)i,t = α +∑

k

βk(SEO has occurred k periods earlier)i,t(12)

+ γ Firm Controlsi,t + ηi + δt + εi,t.

For the dummy variables indicating “SEO has occurred k periods earlier,” k ∈{−1, 0,+1}, where k = –1 denotes 1–2 years prior to SEO, k = 0 denotes theyear of SEO, and k = +1 denotes 1–2 years after SEO. Firm controls consist of

14At the time of information revelation, it is also possible that the managers may have incen-tives to increase earnings before and around the time of securities issues (Teoh, Welch, and Wong(1998a), (1998b)), which may reduce firm-specific return volatility. Thus earnings smoothing wouldbias against our results by raising R2 prior to ADR or SEO events.


the same set of variables as in Table 2, namely βs, size, leverage, ROA, and MB.Here ηi and δt are firm and year fixed effects, respectively.

The βks are the coefficients of interest, and we test the following hypotheses:Hypothesis 1 derives directly from the first part of Proposition 1a. Hypothesis 2derives from the second part of Proposition 1a.15

Hypothesis 1. To the extent that lumpy and early information is disclosed at orbefore the SEO, R2 should be higher subsequent to the offering. That is, βk < 0for some k > 0.

Hypothesis 2. To the extent that lumpy information is disclosed prior to or at theSEO, the R2 would be lower at the time of disclosure. That is, we expect βk > 0for some k ≤ 0.

One concern about empirical identification of the SEO effects is the potentialself-selection of SEOs. That is, SEOs are not randomly assigned; there might beunobserved firm characteristics that simultaneously affect the SEO decisions andreturn synchronicity. In this paper, we explicitly address this concern in 3 ways.First, we are not relying on a simple regression of R2 on an SEO dummy variable.Rather, we focus on a nonmonotonic dynamic response of return synchronicity tothe SEO. For the self-selection argument to work, it has to be the case that cer-tain SEO-related firm characteristics can influence R2 in both positive and neg-ative directions and that such influences change over time in exactly the sameway as our proposed dynamics in SEO effects. This, however, is by no meansobvious.

Second, we include firm fixed effects in all our estimations. This “within-variation” specification effectively tracks the same firm before and after its SEO.Thus, to the extent that some time-invariant firm characteristics affect the SEOdecisions, these are completely controlled for. Moreover, we include in our re-gressions (time-varying) firm-level control variables that could potentially af-fect return synchronicity and SEO decisions, such as size, profitability, MB, andleverage.

2. Results

Panel C of Table 1 presents the summary statistics of our sample. Comparedto non-SEO firm-years, SEO firm-years differ in almost all firm characteristics,suggesting that firm characteristics need to be controlled for in our later analysis.

Table 3 reports the regression results. Column (1) in Table 3 is a naı̈ve re-gression of 1 – R2 on a dummy variable indicating 1–2 years immediately after anSEO. The coefficient on the post-SEO dummy variable is significantly negative atthe 1% level; that is, contrary to the conventional wisdom, SEO (and presumably

15We note that in the context of SEOs the relative importance of time-invariant information disclo-sure may not be as significant as in some other contexts such as cross-listings (which will be discussedlater) or IPOs. Therefore we do not expect the effect of information disclosure to persist. Indeed, whenwe experiment with alternative specifications with longer horizons, we do not find any significant ef-fects beyond 2 years.


greater transparency) is associated with less firm-specific return variation in theyears immediately after the offering.

TABLE 3

The Dynamic Response of R2 to SEOs

Table 3 reports the dynamics of R 2 in response to SEOs from the model below:

log(1− R 2)i,t = α +

∑

kβk (SEO has occurred k periods earlier)i,t + γFirm Controlsi,t + ηi + δt + εi,t .

Here R 2 and β are estimated from a market model (equation (B-1)) in columns (1) and (2); from an industry-augmentedmarket model in column (3); from the Fama and French (FF) (1993) 3-factor model and a 4-factor model with momentum(equations (B-2) and (B-3)) in columns (4) and (5). AGE is the number of years since the inclusion in CRSP. Size is the logof market assets. Leverage is long-term debt over total assets. Profitability is measured by the return on assets. Market-to-book (MB) ratio is the market value of equity plus book value of debt over total assets. Total volatility is the standarddeviation of weekly return over a year. Standard errors are clustered at the firm level and are in the parentheses. ***, **,and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Industry- FF 3Standard Augmented FF 3- FactorsMarket Market Factor PlusModel Model Model Momentum

(1) (2) (3) (4) (5)

1–2 years prior to SEO 0.006 0.011** 0.012*** 0.009**(0.004) (0.005) (0.004) (0.005)

Year of SEO 0.002 0.003 –0.013* –0.007 –0.005(0.004) (0.005) (0.008) (0.005) (0.006)

1–2 years subsequent to SEO –0.011*** –0.010*** –0.012** –0.012*** –0.016***(0.004) (0.004) (0.005) (0.004) (0.004)

β (Market factor) –0.089*** –0.089*** –0.091*** –0.030*** –0.020***(0.012) (0.012) (0.003) (0.011) (0.007)

β (High-minus-low factor) 0.024*** 0.021***(0.003) (0.002)

β (Small-minus-big factor) 0.001 –0.001(0.004) (0.003)

β (Momentum factor) 0.004(0.003)

β (Industry return) –0.152***(0.007)

Log of total volatility 0.932*** 0.932*** 0.991*** 0.873*** 0.862***(0.010) (0.010) (0.008) (0.012) (0.008)

MB 0.002 0.002 0 0 0.002**(0.002) (0.002) (0.001) (0.001) (0.001)

log(AGE) –0.040*** –0.040*** –0.018*** –0.027*** –0.030***(0.005) (0.005) (0.007) (0.006) (0.006)

log(assets) –0.029*** –0.029*** –0.035*** –0.047*** –0.052***(0.004) (0.004) (0.003) (0.003) (0.003)

Profitability –0.019 –0.019 –0.015 –0.013 –0.023***(0.013) (0.013) (0.009) (0.011) (0.008)

Leverage 0.050*** 0.050*** 0.053*** 0.066*** 0.073***(0.007) (0.007) (0.009) (0.008) (0.008)

Constant 13.141*** 13.141*** 3.942*** 3.666*** 3.641***(0.035) (0.035) (0.042) (0.045) (0.036)

Observations 89,010 89,010 88,133 88,979 88,968Number of firms 12,015 12,015 11,934 12,011 12,011R 2 0.94 0.94 0.89 0.91 0.91

While the above result is consistent with our conjecture that, when lumpyinformation is disclosed, there is less surprise afterward, the specification doesnot consider the possible intertemporal effects of lumpy disclosure. Therefore, incolumns (2)–(5) of Table 3, we introduce the dynamic response of R2 as specified


in equation (12). Consistent with Hypotheses 1 and 2, R2 is lower prior to SEOand increases subsequently (significant at the 10% level or above). The impactsof other firm control variables are similar to those in Table 2.

We plot the R2 dynamics in Figure 2 (Graph A), which reflects point esti-mates in column (3) of Table 3 based on an industry-augmented market model.We start with the R2 during “normal” times (non-SEO firm years), which is 0.17.Coefficients βk translate into R2s that are about 1 percentage point lower be-fore an SEO and 1 percentage point higher during an SEO year and 1–2 yearsafterward.

FIGURE 2

R2 Dynamics around SEO and ADR

Graph A. R2 Dynamics around SEO

Graph B. R2 Dynamics around ADR

C. Firm-Specific Return Variation and Cross-Listings

We now explore the dynamic response to another lumpy information disclo-sure event, namely ADR listings. We use a very similar specification to the one forSEOs. ADR listings are likely to be bigger information events than SEOs, as thelisting firms need to, in addition to the usual disclosure, comply with Securitiesand Exchange Commission (SEC) regulations, which typically require more dis-closure than exchanges in their home countries. Thus, the effects of ADR listingsare likely to happen earlier, starting as soon as the firms begin to prepare disclo-sure and accounts for the listings, and to last longer. This is because, first, the


lumpier disclosure may remove more uncertainty about time-invariant attributessuch as managerial ability, and second, the disclosure environment subsequent toADR listing may change to one that involves continued early regular disclosure.Then according to our Propositions 1b and 2, the return synchronicity may con-tinue to be higher. However, exactly how long the positive ADR effect lasts is anempirical matter.

Thus we estimate the following model:

log(1− R2)i,t = α +∑

k

βk(ADR listing occurred k periods ago)i,t(13)

+ γ Firm Controlsi,t + ηi + δt + εi,t.

For the dummy variables indicating the ADR listing had occurred k periods ear-lier, we consider k = −2,−1, 0,+1, and +2. In particular, k = −2 and k = −1correspond to years 3–4 and 1–2 before listing, k = 0 corresponds to the yearof listing, and k = +1 to +4 correspond to years 1–3, 4–6, 7–9, and more than10 years after listing. Firm controls consist of the same set of variables as inequation (12), namely βs (home β and U.S. β), size, leverage, ROA, and MB. Weaddress the concern of self-selection of ADR listings in a similar manner to theSEOs. In particular, we focus on the nonmonotonic dynamic response of returnsynchronicity to the ADR listing. Here ηi are firm fixed effects that control fortime-invariant firm characteristics that might have affected the ADR decisions,and δt are year fixed effects.

Table 4 provides the descriptive statistics for our sample. Compared to non-ADR firm-years, in ADR firm-years (i.e., a year in which an international firmhas an active ADR), firms tend to have significantly higher R2, larger size, higher

TABLE 4

Descriptive Statistics of the ADR sample

Table 4 reports the descriptive statistics for the ADR sample. The sample contains international firm-years from Datastreamand Worldscope during 1980–2004. An ADR firm-year is a year in which the international firm has an active ADR. Otherwise,a firm-year is a non-ADR firm-year. SSE, Home β, and U.S. β are estimated for each firm-year from an augmented marketmodel (equation (B-4)). Leverage is long-term debt over total assets. Profitability is measured by the return on assets.Market-to-book (MB) ratio is the market value of equity plus book value of debt over total assets. Significance of thedifferences between the ADR firm-years and non-ADR firm-years are based on 2-tailed tests (t-test for mean and rank sumtest for median). ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Non- DifferenceAll ADR ADR (ADR –

Firm- Firm- Firm- Non-Years Years Years ADR)

Mean Median Mean Median Mean Median Mean Median

Firm-specific returnvariation (SSE) 0.192 0.097 0.198 0.094 0.192 0.097 0.006 –0.003

R 2 0.190 0.128 0.260 0.212 0.187 0.126 0.072*** 0.086***Home β 0.946 0.722 1.329 1.084 0.933 0.711 0.395*** 0.372***U.S. β –0.009 0.000 0.058 0.021 –0.012 0.000 0.069*** 0.021***Total assets (mil.) 3,891.104 267.560 20,497.430 1,986.104 3,347 257 17,150.068*** 1,729.061***MB 1.987 1.338 2.663 1.868 1.965 1.320 0.698*** 0.547***Leverage 0.130 0.085 0.178 0.150 0.128 0.083 0.049*** 0.066***Profitability 0.017 0.022 0.007 0.027 0.017 0.022 –0.009*** 0.005***Number of obs. 153,572 4,869 148,703


MB, and higher leverage (at the 1% level).16 Interestingly, the ADR firm-yearstend to have lower profitability measured by ROA in mean but not in median.Since in ADR firm-years, firms on average are more levered, the lower ROA inmean could be due to higher leverage.

Multivariate analysis is presented in Table 5.17 Again, column (1) of Table 5is a naı̈ve regression of 1 − R2 on the ADR dummy variable, indicating whetheror not the firm has an ADR listing. It shows that ADR listing (and presumablygreater transparency) is associated with significantly less firm-specific informa-tion in the stock prices (at the 1% level), contrary to the conventional wisdom.18

Column (2) of Table 5 examines the dynamic responses of R2. Consistent with ourmodel’s predictions, ADRs are associated with a persistent drop in firm-specificinformation in stock prices (i.e., higher R2) in the years after the listings. Thecoefficients on dummy variables indicating years prior to the ADR listings aresignificantly positive (at the 1% level), implying that more firm-specific informa-tion is impounded in the stock prices at the time of disclosure. The coefficients onother control variables are similar to those in Table 3.19

We now examine how the interplay between institutional factors and im-proved information disclosure affect the return synchronicity dynamics. For shareprices to reflect information, arbitrageurs need to expend resources uncoveringproprietary information about the firm (Grossman (1976), Shleifer and Vishny(1997)). Such arbitrage activity, as argued by MYY (2000), may be economicallyunattractive in countries with poor protection of property rights due to the in-fluence of unpredictable political events and uncertainty about the arbitrageurs’ability to keep their trading profits. On the other hand, recent literature on in-ternational corporate governance finds that firms’ incentives to disclose informa-tion and improve transparency are weaker without developed institutions. These

16We note that LSSE does not differ significantly across the 2 groups. This is not surpris-ing, since meaningful comparison of LSSE can only be made when the total return variation iscontrolled for.

17Here we do not use the Fama-French 3-factor model or a 4-factor model, since there is evidencein the asset pricing literature that the size and book-to-market factors do not work very well for at leastsome international stocks (e.g., European or Japanese stocks).

18We note that this result is quite different from a contemporaneous paper by Fernandes andFerreira (2008). The differences could be due to methodological differences: Fernandes and Ferreirameasure return synchronicity using log((1 − R2)/R2) (which is log(SSE/(SST − SSE))), and theydo not control for total return variation (SST). Thus even if a variable (X) does not affect SSE, it ispossible to have a significant coefficient for this variable in the regression due to its correlation withSST. This is because

d logSSE

SST− SSEdX

=SST

SSE(SST− SSE)dSSE/dX − 1

SST− SSEdSST/dX,

which is not 0 even if dSSE/dX = 0.19We note that some firms may cross-list in countries other than the U.S. Thus our non-ADR

sample may contain firms which cross-listed outside the U.S. To the extent that some of such cross-listings are from weak law country to countries with better disclosure requirements, our results couldbe weakened. As a robustness check, we drop cross-listings outside the U.S. from the control sample.The results (unreported) remain qualitatively the same and are available from the authors.


TABLE 5

The Dynamics of ADR Effects on R2

Table 5 reports the dynamics of R 2 in response to ADR listings as in the following model:

log(1− R2)i,t = α +

∑

kβk (ADR listing occurred k periods ago)i,t + γFirm Controlsi,t + ηi + δt + εi,t .

Here R 2, Home β, and U.S. β are estimated for each firm-year from an augmented market model (equation (B-4)). Thesample contains international firm-years from Datastream and Worldscope during 1980–2004. Countries with good (weak)institutional environments are classified based on the good-government index from KKM (2004). ADR Listing is a dummyvariable indicating the firm-years with active ADRs. “3–4 years prior to ADR” and “1–2 years prior to ADR” are dummyvariables indicating the number of years prior to the ADR. “Year of ADR” is a dummy variable indicating the year in whichADR is listed. Likewise, “1–3 (4–6, 7–9, and over 10) years subsequent to ADR” are dummy variables indicating the numberof years after the ADR. Firm size is the log of market assets. Leverage is long-term debt over total assets. Profitability ismeasured by the return on assets. Market-to-book (MB) ratio is the market value of equity plus book value of debt overtotal assets. Total volatility is the standard deviation of weekly return over a year. Standard errors are clustered at the firmlevel and are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Overall Good WeakSample Institutions Institutions

(1) (2) (3) (4)

ADR listing –0.069***(0.020)

3–4 years prior to ADR 0.056** 0.063*** –0.079(0.022) (0.022) (0.171)

1–2 years prior to ADR 0.049** 0.055** –0.085(0.024) (0.024) (0.179)

Year of ADR –0.027 –0.015 –0.051(0.024) (0.024) (0.216)

1–3 years subsequent to ADR –0.050** –0.052** 0.017(0.023) (0.023) (0.178)

4–6 years subsequent to ADR –0.058** –0.058** –0.047(0.029) (0.029) (0.207)

7–9 years subsequent to ADR –0.009 –0.010 0.015(0.031) (0.031) (0.209)

10 years subsequent to ADR –0.037 –0.034 –0.015(0.037) (0.037) (0.297)

Log of total volatility 0.033*** 0.033*** 0.023*** 0.134***(0.003) (0.003) (0.003) (0.020)

Home β –0.026*** –0.026*** –0.025*** –0.044(0.002) (0.002) (0.002) (0.057)

U.S. β –0.014*** –0.014*** –0.013*** –0.025(0.004) (0.004) (0.004) (0.018)

log(assets) –0.042*** –0.042*** –0.044*** –0.022(0.003) (0.003) (0.003) (0.014)

Leverage 0.050*** 0.050*** 0.068*** –0.023(0.014) (0.014) (0.015) (0.028)

Profitability –0.047*** –0.046*** –0.054*** –0.012(0.016) (0.016) (0.016) (0.058)

MB –0.001 –0.001 0.000 –0.030***(0.001) (0.001) (0.000) (0.009)

Constant 13.325*** 13.325*** 13.281*** 13.485***(0.026) (0.026) (0.025) (0.176)

Observations 153,572 153,572 140,351 13,221Number of firms 20,544 20,544 17,876 2,668R 2 0.88 0.88 0.89 0.83

considerations suggest that the dynamics of return synchronicity surrounding thelisting of ADRs are likely to be strongest for firms from countries with stronginstitutions.

We divide the sample into firms from countries with better institutional de-velopment and those without, based on the good-government index constructed


by Kaufmann, Kraay, and Mastruzzi (KKM) (2004).20 Specifically, we definecountries with a score above 0, the median of the scores for the good-governmentindex in KKM’s sample, as those with developed institutions, and countries witha score below 0 as without. Among 782 cross-listed firms, 685 are from countrieswith good institutional support. Results in columns (3) and (4) of Table 5 showthat, consistent with our conjecture, the dynamic effects of ADR listings in col-umn (2) are driven by firms in countries with developed institutions. A Chow testindicates that the difference between the 2 groups of countries is significant at the1% level.

Panel B of Figure 2 shows the R2 dynamics based on point estimates in col-umn (2) of Table 5. We start with the R2 during “normal” times (non-ADR firm-years), which is 0.189. The R2 is approximately 4 percentage points lower beforeADR events and 4 percentage points higher afterward. Such an effect is largerthan in the case of SEO events, reflecting the more lumpy nature of informationdisclosure around ADR listings.

So far the findings correspond well with the implications of our model con-cerning changes in firm-specific return variation in response to a change in theinformation environment. We provide 3 pieces of evidence. First, we find that,consistent with learning about time-invariant information, return synchronicity isstrongly positively related to age. Second and third, exploiting settings with lumpyinformation disclosure during SEO and ADR events, we find a dynamic responseof return synchronicity to lumpy information disclosure. In particular, while re-turn synchronicity is lower at the time of information disclosure, reflecting greaterfirm-specific information impounded in the stock prices, return synchronicityafter the disclosure (and thus with greater transparency) is significantly higher.

One remaining concern is that, since SEO or ADR events can be relatedto other significant corporate events, it is possible that information disclosuressurrounding these events, rather than SEO or ADR events themselves, lead toobserved changes in return synchronicity. It is worth noting that while this hy-pothesis changes the interpretation of our results, it does not refute our main pointthat there is a dynamic pattern in return synchronicity surrounding informationdisclosure and that such a dynamic change is inconsistent with the conventionalwisdom. Moreover, the timing of these other events has to be exactly the same asSEO/ADR events; otherwise we would not be able to observe the dynamic patternaround the latter. In fact, as we discussed earlier, this is a strength of our empiri-cal design—it is much less likely for a predicted dynamic pattern (i.e., increasedpreevent SSE and decreased postevent SSE) to arise spuriously. In an effort todistinguish between changes in return synchronicity due to other corporate eventsand changes due to SEO/ADR events, we control for large changes in assets, aswell as their interactions with the SEO-/ADR-related dummy variables, given thatsignificant corporate events are typically associated with major changes in assetsize. It turns out that these interaction terms are generally not significant and that

20KKM (2004) provide 6 indicators on institutional environment. Using the alternative indicatorsdoes not alter our results, which is not surprising, since the correlations between any 2 indicatorsare over 70%. The indicators are available after 1996. Since institutional environment changes veryslowly, for observations before 1996, we use the value in 1996.


our main results remain. In the interest of brevity these results are not reported,but they are available from the authors.

V. Conclusion

Existing literature has taken the perspective that if a firm’s information en-vironment causes stock prices to reflect more firm-specific information, marketfactors should explain a smaller proportion of the variation in stock returns.

This paper broaches, theoretically and empirically, another perspective: thatstock prices respond only to announcements that are not already anticipated bythe market. When the information environment of a firm improves and more firm-specific information is available, market participants are able to improve theirpredictions about the occurrence of future firm-specific events. As a result, thesurprise components of stock returns will be lower when the events are actuallydisclosed, and the return synchronicity will be higher.

Our empirical evidence is drawn from 3 different settings. First, consistentwith learning about time-invariant information, return synchronicity is signifi-cantly higher for older firms. Second and third, exploiting settings with disclo-sure of substantial information about the firm, namely SEOs and ADR listings,we find dynamic responses of return synchronicity that are consistent with lumpyand early disclosure of information relevant for future events, as well as disclosureof information pertinent to time-invariant firm attributes that are relevant for fu-ture cash flows. In particular, return synchronicity decreases prior to these eventsand increases subsequently.

Overall, we make 2 contributions to the literature. First, by showing boththeoretically and empirically that stock return synchronicity can increase withimproved firm transparency, we highlight the importance of understanding thenature of information discovery and the dynamics of the response of stock re-turn synchronicity to changes in information environment. Second, our analysisadds to the growing body of literature on information disclosure around securityissuance events.

Appendix A. Proofs

1. Proof of Proposition 1a

At t0, the investors’ information set is It0 ≡ { ft0 , θ1,t0 , δt0+1}, whereas for any t ≠ t0,It ≡ {ft, θ1,t}.

From equations (2)–(5), we can write

Xt+1 = X0 + ϕXt + λt+1,(A-1)

where X0 = f0 + θ1,0 + θ2,0 and λt+1 = εt+1 + ξ1,t+1 + ξ2,t+1.

Step 1. We can write

Ct+1 = K0(X0 + ϕXt + λt+1) = K0X0 + K0ϕXt + K0λt+1, and for arbitrary k ≥ 1

Ct+k = K0(1 + ϕ + ϕ2 + · · · + ϕk−1) · X0 + K0ϕkXt

+ K0(λt+k + ϕλt+k−1 + · · · + ϕk−1λt+1).


Notice that λt+k = εt+k + ξ1,t+k + ξ2,t+k. For t = t0 and k = 1, we have

λt0+1 = εt0+1 + ξ′1,t0+1 + δt0+1 + ξ2,t0+1.

Thus E(Ct0+k|Ct0 , δt0+1) = K01− ϕk

1− ϕ · X0 + ϕkK0Xt0 + K0ϕk−1δt0+1

and

E (Ct0+k|θ1,t0 , ft0 , δt0+1) = K01− ϕk

1− ϕ · X0 + ϕkK0

(ft0 + θ1,t0 +

θ2,0

1− ϕ)

(A-2)

+ K0ϕk−1δt0+1,

where we use the fact that, for any t,

E(Xt|It) = ft + θ1,t +θ2,0

1− ϕ.(A-3)

For any other t > t0,

E(Ct+k|θ1,t, ft) = K01− ϕk

1− ϕ · X0 + ϕk

(ft + θ1,t +

θ2,0

1− ϕ)

K0.(A-4)

Step 2. The intrinsic value of the firm to the investors at t0 is

Kt0(θ1,t0 , ft0 , δt0+1) =∞∑

k=1

E(Ct0+k|It0)

(1 + r)k(A-5)

=K0X0

1− ϕ ·1r− K0X0

1− ϕ ·ϕ

1 + r − ϕ

+K0ϕ

1 + r − ϕ(

ft0 + θ1,t0 +θ2,0

1− ϕ)

+K0

1 + r − ϕδt0+1

=K0X0(1 + r)r(1 + r − ϕ) +

K0ϕ

1 + r − ϕ(

ft0 + θ1,t0 +θ2,0

1− ϕ)

+K0

1 + r − ϕδt0+1.

Similarly, for any t ≠ t0 − 1,

Kt+1(It+1) =K0X0(1 + r)r(1 + r − ϕ) +

K0ϕ

1 + r − ϕ(

ft+1 + θ1,t+1 +θ2,0

1− ϕ).(A-6)

Thus, for t ≠ t0 − 1, using equation (A-3), we have

Kt+1(It+1) + E(Ct+1|It+1) =K0X0(1 + r)r(1 + r − ϕ) +

K0(1 + r)1 + r − ϕ

(ft+1 + θ1,t+1 +

θ2,0

1− ϕ).(A-7)

Step 3. Denote by rt the realized return in period t. From equations (7) and (8),

rt =αKt+1(It+1) + αE(Ct+1|It+1)

αKt(It)− 1.

Substituting from equations (A-6) and (A-7), for t ≠ t0 and t ≠ t0 − 1,

rt =

(1 + r)K0

1 + r − ϕ(

ft+1 + θ1,t+1 +θ2,0

1− ϕ)− K0ϕ

1 + r − ϕ(

ft + θ1,t +θ2,0

1− ϕ)

K0X0(1 + r)r(1 + r − ϕ) +

K0ϕ

1 + r − ϕ(

ft + θ1,t +θ2,0

1− ϕ) ,(A-8)


whereas for t = t0, using equations (A-5) and (A-7), we have

rt0 =

[(1 + r)K0

1 + r − ϕ(

ft0+1 + θ1,t0+1 +θ2,0

1− ϕ)

(A-9)

− K0ϕ

1 + r − ϕ(

ft0 + θ1,t0 +θ2,0

1− ϕ)− K0

1 + r − ϕδt0+1

]/

[K0X0(1 + r)r(1 + r − ϕ) +

K0ϕ

1 + r − ϕ(

ft0 + θ1,t0 +θ2,0

1− ϕ)

+K0

1 + r − ϕ · δt0+1

],

and for t = t0 − 1, using equations (A-5) and (A-7), we have

rt0−1 =

[K0(1 + r)1 + r − ϕ

(ft0 + θ1,t0 +

θ2,0

1− ϕ)

+K0

1 + r − ϕ · δt0+1(A-10)

− ϕK0

1 + r − ϕ(

ft0−1 + θ1,t0−1 +θ2,0

1− ϕ)]/

[K0X0(1 + r)r(1 + r − ϕ) +

ϕK0

1 + r − ϕ(

ft0−1 + θ1,t0−1 +θ2,0

1− ϕ)]

.

Consider equation (A-9) first. We can write the right-hand side as

r +α0

X0(1 + r)r

+ ϕ

(ft0 + θ1,t0 +

θ2,0

1− ϕ)

+ δt0+1

,

where

α0 = −X0(1 + r)− rϕ

(ft0 + θ1,t0 +

θ2,0

1− ϕ)− rδt0+1

+ (1 + r)

(ft0+1 + θ1,t0+1 +

θ2,0

1− ϕ)− ϕ(

ft0 + θ1,t0 +θ2,0

1− ϕ)− δt0+1

= −X0(1 + r) + (1 + r)

(ft0+1 − ϕft0 + θ1,t0+1 − ϕθ1,t0 +

θ2,0

1− ϕ −ϕθ2,0

1− ϕ)

− (1 + r)δt0+1

= −X0(1 + r) + (1 + r) (f0 + εt0+1 + θ1,0 + ξ1,t0+1 + θ2,0)− (1 + r)δt0+1

= (1 + r)(εt0+1 + ξ′1,t0+1

).

Thus,

rt0 = r +(1 + r)

(εt0+1 + ξ′1,t0+1

)X0(1 + r)

r+ ϕ

(ft0 + θ1,t0 +

θ2,0

1− ϕ)

+ δt0+1

.

Hence, the proportion of the return variation explained by market factor is

R2t0 =

Var(εt0+1)

Var(εt0+1) + Var(ξ′1.t0+1)=

1

1 +Var(ξ′1,t0+1)

Var(εt0+1)

(A-11)

=1

1 +Var(ξ′1,t0+1)

Var(ξ1,t0+1)· Var(ξ1,t0+1)

Var(ξ1,t0+1) + Var(ξ2,t0+1)· Var(ξ1,t0+1) + Var(ξ2,t0+1)

Var(εt0+1)

=1

1 + σ · η · κ >1

1 + η · κ .


Next, consider equation (A-10). Proceeding similarly,

rt0−1 = r +α1

X0(1 + r)r

+ ϕ

(ft0−1 + θ1,t0−1 +

θ2,0

1− ϕ) ,

where

α1 = −X0(1 + r)− ϕr

(ft0−1 + θ1,t0−1 +

θ2,0

1− ϕ)

+ (1 + r)

(ft0 + θ1,t0 +

θ2,0

1− ϕ)

+ δt0+1 − ϕ(

ft0−1 + θ1,t0−1 +θ2,0

1− ϕ)

= −X0(1 + r) + (1 + r)( ft0 − ϕft0−1 + θ1,t0 − ϕθ1,t0−1 + θ2,0) + δt0+1

= (1 + r)(εt0 + ξ1,t0) + δt0+1.

Thus

rt0−1 = r +(1 + r)(εt0 + ξ1,t0) + δt0+1

X0(1 + r)r

+ ϕ

(ft0−1 + θ1,t0−1 +

θ2,0

1− ϕ) .

Hence,

R2t0−1 =

(1 + r)2Var(εt0)

(1 + r)2Var(εt0) + (1 + r)2Var(ξ1,t0) + Var(δt0+1)(A-12)

=1

1 +Var(ξ1,t0)Var(εt0)

+1

(1 + r)2· Var(δt0+1)

Var(εt0)

<1

1 + η · κ .

Finally, proceeding as above, for any t ≠ t0 or t ≠ t0− 1, or equivalently, for any t for a firmthat does not experience a disclosure event, we have

rt = r +(1 + r)(εt+1 + ξ1,t+1)

X0(1 + r)r

+ ϕ

(ft + θ1,t +

θ2,0

1− ϕ)(A-13)

so that

R2t =

11 + η · κ .(A-14)

Comparing equations (A-11), (A-12), and (A-14), the results follow.

2. Proof of Proposition 1b

Here, the information set of the outsiders at the beginning of every period t is It ={ft, θ1,t, δt+1}. Following steps similar to Step 1 in the proof of Proposition 1a, we get

E(Ct+k|θ1,t, ft, δt+1) = K01− ϕk

1− ϕ · X0 + ϕkK0

(ft + θ1,t +

θ2,0

1− ϕ)

+ K0ϕk−1δt+1.

Hence,

Kt(It) =K0X0(1 + r)r(1 + r − ϕ) +

K0ϕ

1 + r − ϕ(

ft + θ1,t +θ2,0

1− ϕ)

+K0δt+1

1 + r − ϕ


and

Kt(It) + E(Ct|It) =K0X0(1 + r)r(1 + r − ϕ) +

K0(1 + r)1 + r − ϕ

(ft + ϕ1,t +

θ2,0

1− ϕ)

+K0δt+1

1 + r − ϕ.

Proceeding exactly as before, we get

rt = r +(1 + r)(εt+1 + ξ′t+1) + δt+2

X0(1 + r)r

+ ϕ

(ft + θ1,t +

θ2,0

1− ϕ)

+ δt+1

.

Hence,

R2t =

(1 + r)2Var(εt+1)

(1 + r)2 (Var(εt+1) + Var(ξ′t+1)) + Var(δt+2)

=(1 + r)2Var(εt+1)

(1 + r)2Var(εt+1) + (1 + r)2Var(ξt+1)− (1 + r)2Var(δt+1) + Var(δt+2).

Since Var(δt+1) = Var(δt+2), it follows that R2t > Var(εt+1)/(Var(εt+1) + Var(ξt+1)), which

is the return synchronicity for a firm with no early information disclosure.

3. Proof of Proposition 2

When the true value of θ1,0 is already revealed, the analysis of Proposition 1a applies,and the realized return is given by equation (A-13).

Suppose θ1,0 is not revealed, but the market at each t updates its expectation aboutθ1,0 from the realized values of θ1,t for t′ ≤ t. Let the posterior mean estimate of θ1,0 at tbe E(θ1,0|It) = θ

t1,0.

Conditional on the information set It (which now includes the entire history of therealizations of (ft, θ1,t)), the expected value of X0 = f0 + θ1,0 + θ2,0, computed with respectto the posterior distribution of θ1,0, is X0(t) = f0 + θt

1,0 + θ2,0. Proceeding exactly as in theproof of Proposition 1a, we get

rt = r +

(1 + r)1 + r − ϕ

(θt+1

1,0 − θt1,0

)+ (1 + r)(εt+1 + ξ1,t+1)

X0(t)(1 + r)r

+ ϕ

(ft + θ1,t +

θ2,0

1− ϕ) .(A-15)

The result therefore follows immediately from a comparison of equations (A-13) and(A-15), because conditional on It, θt+1

1,0 is a random variable whose value will depend on therealization of θ1,t+1; hence, it has a positive variance.

Appendix B. Data Construction

1. Computation of R2 (or LSSE)

R2 is first computed based on the method proposed in MYY (2000). For the empiricalanalysis of the U.S. CRSP firms in Sections IV.A and IV.B, we run the following modelusing weekly returns for each firm in each year:

rit = ai + birm,t + εit,(B-1)

where i, t index firms and weeks, respectively. Here rm,t is the U.S. market index return de-fined as the value-weighted returns of all CRSP firms.To mitigate the thin-trading problem,we follow Jin and Myers (2006) and estimate the model using weekly returns (Wednesday


close to Wednesday close). LSSE is the log of the sum of squared errors from the regres-sions. Since the regressions are run for each firm in each year, our estimates of market βand R2 (or LSSE) are annual variables for each firm.

For each firm in each year, we also estimate its R2or LSSE from the Fama-French(1993) 3-factor model and a 4-factor model with momentum. In particular,

rit − rft = ai + b1,i(rm,t − rft) + b2,iHMLt + b3,iSMBt + εit and(B-2)

rit − rft = ai + b1,i(rm,t − rft) + b2,iHMLt + b3,iSMBt + b4,iUMDt + εit,(B-3)

where i, t index firms and weeks, respectively. Here rm,t is the U.S. market index return de-fined as the value-weighted returns of all CRSP firms, and rft is the 1-month T-bill rate. Weget the daily returns on HML, SMB, and UMD (momentum) from Kenneth French’s Website (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/) and convert them to weeklyreturns (Wednesday close to Wednesday close). Again the regressions are run for eachfirm in each year; our estimates of factor loadings, as well as R2 (and LSSE), are annualvariables for each firm.

For the empirical analysis in Section IV.C involving international firms, we esti-mate the following model based on weekly returns (again, Wednesday close to Wednesdayclose):

rit = ai + b1,irm, jt + b1,i[rUS,t + ej,t] + εit,(B-4)

where i, j, t index firms, countries, and weeks, respectively. Here rm,jt is the local marketindex return defined as the value-weighted returns of all Datastream companies availablefor that country;21 rUS,t is the U.S. market return, which is computed from CRSP; and ej,t

is the rate of change in the exchange rate per U.S. dollar, which is obtained from Datas-tream and, in cases where it is not in Datastream, from Reuters. The expression rUS,t + ej,t

translates U.S. stock market returns into local currency returns.

2. Estimating Fundamental Comovement

Following Durnev et al. (2004), we estimate fundamental (ROA) comovement usingthe following model:

ROAi,j,t = ai, j + b1,i, jROAm,t + b2,i, jROAj,t + ei, j,t,(B-5)

where i, j, m, t index firm, industry, market, and year, respectively. We define ROA asnet income plus interest expense and depreciation over total assets. Industries are definedbased on 3-digit SIC code (2-digit SIC code gives very similar results). Both market andindustry ROAs are value-weighted averages excluding the firm in question. We estimatethe regression for each firm in each year using the previous 6 years of data (including thecurrent year). The log of SSE from regression (B-5) is the idiosyncratic ROA movement,which we use as an additional control in Table 2.

3. The SEO Sample

Our initial SEO sample is retrieved from the Securities Data Company (SDC) GlobalNew Issues database. To ensure significant information disclosure, we require the issuesize to exceed $10 million USD and to be at least 5% of the issuer’s market value of equity.We exclude right issues because they are issued to existing shareholders, and the disclosureof information would not be as intense as a public offering. We also exclude units, sharesof beneficial interest, primes and scores, closed-end funds, and real estate investment trusts(REITs). This procedure gives us 7,523 SEOs in the first instance.

21Jin and Myers (2006) require a minimum of 25 stocks. All the countries in our study have over25 stocks except Zimbabwe, which has 16 stocks.


We classify firm-years into SEO firm-years and non-SEO firm-years. To ensure thatwe do not pick up the informational effects of other confounding events, we further dropfrom our entire sample firm-years that are within 3 years before and after another SEO bythe same firm, within 3 years after its IPO, or within 3 years after it changes the listingstock exchange. The SEO firm-years are defined as those firm-years that are within 2 yearsbefore or after an SEO event. The remaining are non-SEO firm-years. Thus our final sampleis a panel of 12,015 firms and 89,010 firm-years, of which 2,354 firms have SEO eventsand 8,377 are SEO firm-years.

4. The ADR Sample

We start with all firms covered by the Worldscope database for the period 1980–2004.For a firm-year to be included in our sample, we require valid information to estimatethe market model in equation (B-4). We also require the firm to have relevant accountinginformation, the shareholders’ equity above 0, an asset size of more than $10 million USD(to make firms across countries comparable in size, see, e.g., Doidge, Karolyi, and Stulz(2004)), and at least 30 weeks of return data in Datastream for a given year (to ensurereliable estimate of return synchronicity, see MYY (2000), Jin and Myers (2006)). Wecompute firm age according to the base date provided by Datastream. To avoid potentiallycontaminating effects of information disclosure at the time of IPO, we require the firmage to be at least 4 years. Consequently, we have a sample of 20,544 firms with 153,572firm-years.

To identify cross-listed firm in the U.S., it is useful to recognize that there are 2 waysfor a non-U.S. firm to be listed in the U.S. One is through an ADR program; the other is todirectly list shares in the U.S. stock market. There are no readily available databases thatprovide systematic information on the identity of the cross-listed firms or the starting andending dates of the listings. Researchers have explored different data sources to identifycross-listings (e.g., Reese and Weisbach (2002), Lang et al. (2003)). In this paper, wefollow the approach in these previous studies.

In particular, we identify directly listed non-U.S. firms by first checking, for firmswith return data in CRSP, their “countries of incorporation” in Compustat (Compustat vari-able FIC). If the firm is not incorporated in the U.S. and if the company name is not markedwith “ADR,” it is a direct listing. We then determine the effective listing dates and the ter-mination dates based on the beginning and ending dates of return data in CRSP. We identify583 directly listed firms, of which 378 have active listings at the end of 2004.

To identify ADR listings, we start with the information provided by major ADRsponsor institutions, namely, Bank of New York, Citibank, and JP Morgan, that provide adatabase of ADR listings. Firms may list different types of ADRs that are subject to dif-ferent levels of disclosure requirement. Level II and Level III ADRs are listed in the stockexchanges (NYSE, NASDAQ, and AMEX), have the most strict disclosure requirement,and are subject to the closest public scrutiny (Lang et al. (2003), Doidge et al. (2004)).Therefore, all else being equal, the improvement in firm transparency should be more sig-nificant for Level II and Level III ADRs than for lower-level ADRs, namely Level I and144A/Regulation S ADRs. To preserve the power of our tests, we follow Lang et al. (2003)and include only Level II and Level III ADRs in our sample. Moreover, as pointed out byReese and Weisbach (2002) and Lang et al. (2003), the legal and informational implicationsof these ADRs and direct listings are essentially the same.

Firms may list multiple ADRs at the same time and terminate their listings. If afirm has multiple ADR programs, we consider the highest level of ADR. When an ADRprogram is terminated, the ADR sponsor institutions remove these inactive programs fromtheir database. Therefore, there is a survivorship bias in reported ADRs in the sponsoringbanks’ databases. To correct this bias, we check, for each firm with return data in CRSP, thecountry of incorporation in Compustat. ADR firms have countries of incorporation outsidethe U.S. and are marked with “ADR” at the end of the Compustat company name. Thebeginning and ending dates of the return time series in CRSP are then taken as the effectivelisting and termination dates, respectively. This search yields 102 additional inactive Level


II and III ADRs. Thus we have in total 507 firms with active ADRs at the end of year 2004and 102 inactive ADR firms.

As pointed out by Reese and Weisbach (2002), not every cross-listing firm can bematched in Worldscope.22 Out of our 1,192 cross-listed firms, 782 have information inWorldscope. Thus we have 782 cross-listed firms from 41 countries/regions, of which 442are ADR firms and 340 are directly listing firms.

ReferencesAlmazan, A.; J. Suarez; and S. Titman. “Capital Structure and Transparency.” Working Paper,

University of Texas at Austin (2002).Barberis, N.; A. Shleifer; and J. Wurgler. “Comovement.” Journal of Financial Economics, 75 (2005),

283–317.Bhattacharya, U.; H. Daouk; B. Jorgenson; and C.-H. Kehr. “When an Event Is Not an Event: The

Curious Case of an Emerging Market.” Journal of Financial Economics, 55 (2000), 69–101.Chan, K., and A. Hameed. “Stock Price Synchronicity and Analyst Coverage in Emerging Markets.”

Journal of Financial Economics, 80 (2006), 115–147.Chen, Q.; I. Goldstein; and W. Jiang. “Price Informativeness and Investment Sensitivity to Stock

Prices.” Review of Financial Studies, 20 (2007), 619–650.Doidge, C.; G. A. Karolyi; and R. M., Stulz. “Why Are Foreign Firms Listed in the U.S. Worth More?”

Journal of Financial Economics, 71 (2004), 205–238.Dubinsky, A., and M. Johannes. “Earnings Announcements and Equity Options.” Working Paper,

Columbia University (2006).Durnev, A.; R. Morck; and B. Yeung. “Value-Enhancing Capital Budgeting and Firm-Specific Stock

Return Variation.” Journal of Finance, 59 (2004), 65–105.Durnev, A.; R. Morck; B. Yeung; and P. Zarowin. “Does Greater Firm-Specific Return Variation Mean

More or Less Informed Stock Pricing?” Journal of Accounting Research, 41 (2003), 797–836.Fama, E. F., and K. R. French. “Common Risk Factors in the Returns on Stocks and Bonds.” Journal

of Financial Economics, 33 (1993), 3–56.Fernandes, N., and M. A. Ferreira. “Does International Cross-Listing Really Improve the Information

Environment?” Journal of Financial Economics, 88 (2008), 216–244.Fishman, M. J., and K. M. Hagerty. “Disclosure Decisions by Firms and the Competition for Price

Efficiency.” Journal of Finance, 44 (1989), 633–646.Gelb, D. S., and P. Zarowin. “Corporate Disclosure Policy and the Informativeness of Stock Prices.”

Review of Accounting Studies, 7 (2002), 33–52.Grossman, S. “On the Efficiency of Competitive Stock Markets Where Trades Have Diverse Informa-

tion.” Journal of Finance, 31 (1976), 573–85.Jin, L., and S. C. Myers. “R2 Around the World: New Theory and New Tests.” Journal of Financial

Economics, 79 (2006), 257–292.Kaufmann, D.; A. Kraay; and M. Mastruzzi. “Governance Matters III: Governance Indicators for

1996, 1998, 2000, and 2002.” World Bank Economic Review, 18 (2004), 253–287.Lang, M. H.; K. V. Lins; and D. P. Miller. “ADRs, Analysts, and Accuracy: Does Cross Listing

in the U.S. Improve a Firm’s Information Environment and Increase Market Value?” Journal ofAccounting Research, 41 (2003), 317–345.

Lang, M. H., and R. J. Lundholm. “Corporate Disclosure Policy and Analyst Behavior.” The Account-ing Review, 71 (1996), 467–492.

Morck, R.; B. Yeung; and W. Yu. “The Information Content of Stock Markets: Why Do EmergingMarkets Have Synchronous Stock Price Movements?” Journal of Financial Economics, 58 (2000),215–260.

Peng, L., and W. Xiong. “Investor Attention, Overconfidence and Category Learning.” Journal ofFinancial Economics, 80 (2006), 563–602.

Piotroski, J. D., and B. T. Roulstone. “The Influence of Analysts, Institutional Investors and Insiderson the Incorporation of Market, Industry and Firm-Specific Information into Stock Prices.” TheAccounting Review, 79 (2004), 1119–1151.

22Indeed, since there is not any reliable (i.e., consistently nonmissing) common identifier betweenCRSP (or Compustat) and Worldscope, we need to manually merge many of our cross-listing firmswith Worldscope, based on company name, country of incorporation, and, if necessary, assets or sales.


Reese Jr., W. A., and M. S. Weisbach. “Protection of Minority Shareholder Interests, Cross-Listings inthe United States, and Subsequent Equity Offerings.” Journal of Financial Economics, 66 (2002),65–104.

Roll, R. “R2.” Journal of Finance, 43 (1988), 541–566.Shiller, R. J. “Do Stock Prices Move Too Much to Be Justified by Subsequent Changes in Dividends?”

American Economic Review, 71 (1981), 421–436.Shleifer, A., and R. W. Vishny. “A Survey of Corporate Governance.” Journal of Finance, 52 (1997),

737–783.Teoh, S. H.; I. Welch; and T. J. Wong. “Earnings Management and the Long-Run Market Performance

of Initial Public Offerings.” Journal of Finance, 53 (1998a), 1935–1974.Teoh, S. H.; I. Welch; and T. J. Wong. “Earnings Management and the Underperformance of Seasoned

Equity Offerings.” Journal of Financial Economics, 50 (1998b), 63–99.West, K. D. “Dividend Innovations and Stock Price Volatility.” Econometrica, 56 (1988), 37–61.Wurgler, J. “Financial Markets and the Allocation of Capital.” Journal of Financial Economics, 58

(2000), 187–214.

Copyright of Journal of Financial & Quantitative Analysis is the property of Cambridge University Press and its

content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's

express written permission. However, users may print, download, or email articles for individual use.

transparency, price informativeness, and stock return synchronicity theory and evidence. (2010)....

Documents

stock priceschange

firmspecific information

stock returns

informative stock price

firm age

university of manchester

andstock return synchronicity

higher return synchronicity