You are on page 1of 23

Pacific-Basin Finance Journal 18 (2010) 403–425

Contents lists available at ScienceDirect

Pacific-Basin Finance Journal
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p a c f i n

On the predictability of Chinese stock returns☆
Xuanjuan Chen a,1, Kenneth A. Kim b,⁎, Tong Yao c,2, Tong Yu d,3
a

Kansas State University, College of Business Administration, Manhattan, KS 66506, United States
State University of New York at Buffalo, School of Management, Buffalo, NY 14260, United States
University of Iowa, Tippie College of Business, Iowa City, IA 52242-1994, United States
d
University of Rhode Island, College of Business Administration, Kingston, RI 02881, United States
b
c

a r t i c l e

i n f o

Article history:
Received 8 September 2009
Accepted 14 April 2010
Available online 27 April 2010
JEL classification:
G12
G15
Keywords:
Stock return predictability
Cross-section of stock returns
China

a b s t r a c t
We examine stock return predictability in China. We take 18 firmspecific variables that have been documented to predict crosssectional stock returns in the U.S. and examine their relation with
stock returns in China for the sample period from 1995 to 2007. We
find relatively weak predictability for Chinese stocks. Only five firmspecific variables predict returns in the Chinese market. Tests on U.S.
stock returns find that more predictors can explain cross-sectional
stock return variation. We test two explanations for the cause of weak
returns predictability in China. First, perhaps return predictors in
China are less heterogeneously distributed than they are in the U.S.
Second, stock prices are less informative in China than they are in the
U.S. We find support for both explanations.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction
The stock market in China has grown rapidly in recent years. It plays an increasingly active role in
China's economic growth, and is increasingly accessible to international investors. Naturally, understanding the economic forces and individual firm characteristics driving stock price movements in this market
becomes an important issue. However, to date, academic research in this area is still nascent. The purpose

☆ The authors thank an anonymous referee, Victor Huang, Haiwei Li, Ghon Rhee (the editor), conference participants at the 2009
China International Conference in Finance meetings in Guangzhou, China and the 2009 Financial Management Association meetings
in Reno, and seminar participants at Barclays Global Investors and at AIG-Huatai Asset Management, for constructive comments. We
especially thank Peng Xiong for providing the SinoFin data. Kim gratefully acknowledges an SOM Summer Research Grant for
financial support.
⁎ Corresponding author. Tel.: +1 716 645 3266; fax: +1 716 645 3823.
E-mail addresses: jchen@ksu.edu (X. Chen), kk52@buffalo.edu (K.A. Kim), tong-yao@uiowa.edu (T. Yao), tongyu@uri.edu (T. Yu).
1
Tel.: +1 785 532 6844; fax: +1 785 532 6822.
2
Tel.: +1 319 335 3924; fax: +1 319 335 3690.
3
Tel.: +1 401 874 7415; fax: +1 401 874 4312.
0927-538X/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.pacfin.2010.04.003

404

X. Chen et al. / Pacific-Basin Finance Journal 18 (2010) 403–425

of this study is to provide a systematic analysis on the relationship between firm characteristics and crosssectional stock returns in this emerging yet already quite large stock market.
The fact that stock returns can be predicted by various firm characteristics such as size, book-to-market
ratio, and past stock returns — which is generally termed “cross-sectional return predictability” — has been
well-documented for the U.S. market and many other equity markets around the world. The cause of such
predictability has also been long debated. Unexplained by standard asset pricing models such as CAPM,
cross-sectional return predictability is viewed by many as prima facie evidence of market inefficiency.
Meanwhile, other researchers attribute predictability to unmeasured risk factors or rational dynamics in
conditional risk-return tradeoff. In this study, we take a large set of return-predictive firm characteristic
variables that are documented in the U.S. market, and examine whether these variables can predict stock
returns in the Chinese market. While our approach to document return predictability is systematic, our
goal is modest. We do not intend to provide a definite answer on whether such predictability is a rational
asset pricing effect or due to market inefficiency (answering this question is an ambitious task that we
leave for future studies). However, some of our empirical results are informative in evaluating a subset of
explanations. Moreover, we believe that our findings can serve as valuable basis for further discussions on
the causes and explanations of stock return predictability in emerging markets.
A few studies have shown that patterns of cross-sectional return predictability initially identified in the
U.S. market often exist outside the U.S. (e.g., Fama and French 1998; Rouwenhorst 1998, 1999; Titman et
al., 2004; Pincus et al., 2007). To the extent that return predictability is caused by investor misreaction to
information or irrational preferences, and to the extent that investors in the U.S and Chinese markets
exhibit similar behavioral biases, there are reasons to believe that a return predictor that prevails in the U.S.
market may also predict returns in the Chinese market. For example, similar to U.S. investors, several
studies have documented strong behavioral biases among Chinese investors, such as over-confidence,
disposition effect, representativeness bias, and herding; see, e.g., Chen et al. (2007), Feng and Seasholes
(2005), Kim and Nofsinger (2008), Tan et al. (2008), and Shumway and Wu (2006). In addition, stock
trading is dominated by individual investors, who are potentially subject to stronger biases relative to
sophisticated institutional investors (e.g., Chen et al., in press; Feng and Seasholes, 2008; Lee et al., 2010;
Ng and Wu, 2006). Further, short-sales are not allowed in China, making it difficult for mispricing to be
quickly arbitraged away.4 These latter observations imply that any differences between markets (including
any potential difference in market efficiency which we discuss shortly) may lead to differences in the
magnitude of the predictive power of a firm-specific variable, but not likely lead to opposite predictions
between markets.
There are several studies on the predictability of Chinese stock returns; however, these studies typically
focus on a small set of predictive variables. Wang and Xu (2004) find that firm size explains cross-sectional
differences in returns but that the book-to-market ratio does not. Kang et al. (2002) and Naughton et al.
(2008) document price momentum in the Chinese market while Wang (2004) finds the contrary. In
addition, Wang and Chin (2004) and Wang and Cheng (2004) document the return-predictive power of
trading volume. Eun and Huang (2007) further find that although stock beta is not priced in cross-sectional
stock returns in the Chinese market, a few firm characteristics, including size, book-to-market ratio, firmspecific risk, dividend-yield, and the existence of offshore shares (B and H shares), can be used to predict
returns. They argue that many of these effects are related to market imperfections, and that “[g]iven
imperfections, stocks are priced rather rationally in China, despite the widespread perception to the
contrary.”
We specifically identify 18 firm-specific variables, known to be predictive of cross-sectional stock
returns at the annual horizon in the U.S. market, and test their predictability on Chinese stock returns. The
variables range from the classic size, book-to-market, and momentum variables, to other value indicators,
and also include measures of earnings quality, firms' investments, external financing activities, firms'
growth in size and profit margins, and finally, volatility, trading turnover, and liquidity.

4
To the extent that stock return predictability is driven by rational asset pricing effect, the existence of such predictability is
interesting on its own; it suggests that there are similar risk factors or similar dynamics in risk-return tradeoff in play in both the U.
S. and Chinese markets.

If there is not much cross-sectional dispersion in a return-predictive variable. 2000. our test here is indirect. we compare the cross-sectional differences in the return predictors themselves between the two markets. market. If initial mispricing is large.S.5 A number of studies have concluded that return synchronicity is inversely related to stock price informativeness. earnings management and accounting manipulation have been found to be prevalent (e. market.X. almost all of the proposed predictors predict the direction of subsequent one-year stock returns in a way consistent with their predictive patterns in the U. there is also evidence of rampant market manipulation (e. 2005. but the effect is weaker relative to the U. where stock prices are very noisy and at the same time fundamental information is often not reliable. indicating that stock prices in the Chinese market incorporate firm-specific information to a substantially less degree. . What explains the weaker predictability of Chinese stock returns? To start with. Haw et al. 2004. the market with higher return predictability should be the one with higher degree of market inefficiency. when we divide Chinese stocks into subsamples based on synchronicity. 2003). It is often considered that. that is. market. Further.46 while it is 0.g. / Pacific-Basin Finance Journal 18 (2010) 403–425 405 We find that during the period from 1995 to 2007. To examine this hypothesis. Therefore. however. we find that return predictability is stronger among stocks with low R2. However.S. the return difference between top and bottom decile portfolios sorted by each individual return predictor is mostly positive in the Chinese market. We further confirm such results by using Fama and MacBeth (1973) cross-sectional regressions. suggesting a weaker anomaly effect in the Chinese market than in the U..S. suggesting low pricing efficiency.g.S. but stock price is persistently noisy. However. to the extent that mispricing is seldom corrected. asset growth. To be specific. after adjusting for the sign to make each variable positively related to stock returns in the U. Under the behavioral asset pricing framework. is unlikely to explain our findings as the Chinese market is usually perceived to be much less efficient. For example. 2004). we find significant return spreads between top and bottom decile portfolios for 10 out of 18 variables. The overall message is that. 1988. Four out of the 18 variables produce significant alphas for the top–bottom decile return spread in the Chinese market while nine predictors produce significant alphas in the U. and Durnev et al. market. Chen and Yuan.. weaker predictability in China does not mean that Chinese stocks 5 Admittedly. For robustness we examine the alphas of stock portfolios sorted on firm characteristics. lower return R2 is associated with greater capitalization of firm-specific information (Roll. the cross-sectional dispersion in stock return associated with the variable will also be small. That is. and then examining the joint process of the observed stock price and the intrinsic value. 2004. market.S. Morck et al. we resort to a well-documented measure of valuation efficiency — stock return synchronicity. based on the Fama–French three-factor model. net operating assets. market. It is also well-known that in less developed capital markets. Our empirical evidence is consistent with this hypothesis. this approach is difficult to implement due to lack of reliable fundamental information in the Chinese market. To test this hypothesis. Tingting. and illiquidity. An ideal direct test of the hypothesis would involve identifying an “intrinsic value” for a stock based on fundamental information.. return predictability strength depends on two factors — investors' initial misreaction and subsequent price correction. This is likely for an emerging market such as China. the average R2 in China is 0. Therefore. relative to stocks with high R2. Jian and Wong.S. we first point out an intriguing contradiction of our empirical evidence with conventional perception. because return predictability is an indication of market inefficiency (conditional on behavioral explanations). One possible explanation of weak return predictability in the Chinese market is that Chinese firms are more homogeneous in the firm-specific characteristics examined. or the R2 of regressing individual stock returns onto market returns. Shengzhen Stock Exchange. 2005). The variables with significant return predictability include the book-to-market ratio. then return predictability will be weak. stocks during the same period. Five of the 18 proposed predictors are statistically significant in explaining the cross-sectional variation of subsequent returns. This. the lack of return predictability in China is actually due to long-lasting market inefficiency. During our sample period. We next consider a hypothesis that is opposite to conventional wisdom. When performing the same tests on U.. R2 tends to be particularly high.12 in the U. Chen et al.S. We find that indeed return predictors are more homogeneous in China. we also find that the sensitivity of stock returns to per unit crosssectional difference in many of the variables is weaker in China. R&D spending. lack of cross-sectional dispersion of firm characteristics is not the whole reason for weaker return predictability in China. in cross-country comparisons.. stock return predictability exists in the Chinese market.

We end up with 18 such variables. Jegadeesh and Titman (1993) show that stocks with better past returns subsequently earn higher returns. (2004) and Beneish et al. Titman et al. However. They point out that intangible investments on R&D and advertising are expensed rather than capitalized in accounting treatment (to the effect of depressing current earnings at the benefit of future earnings) and that investors appear to . Brief discussion of stock return predictors The literature on the relation between firm-specific variables and the cross-section of expected stock returns is vast. Section 5 concludes. Chan et al. (2001) find that ratios of corporate R&D spending and advertising spending to market capitalization are positively correlated with future returns. Smaller market capitalization firms and firms with higher earningsto-price ratios tend to earn higher returns. trading turnover. (2004) show the explanatory power of accruals and NOA do not subsume each other when used jointly to predict returns. (1994) report that future returns are positively correlated with cash flow-to-price ratio and negatively correlated with past sales growth. suggesting that marginal investors do not understand that high NOA implies decreasing returns to assets scale. plus capitalized investments. and group them into the following categories: (i) three return predictors most extensively scrutinized in the U.. (2004) attribute this finding to an overinvestment tendency of corporate managers and investor underreaction to information. Lakonishok et al. (iii) two measures for earnings quality: accounting accruals and net operating assets. 2. Fama and French (1992) show that firms with higher book-to-market ratios subsequently have higher returns. They suggest that investors overvalue firms' past performances. Hirshleifer et al. literature: market capitalization. Reinganum. Sloan (1996) reports an accruals anomaly. it is a symptom of persistent noisy valuation and persistent mispricing. 1981. despite their potential relation. His explanation for the anomaly is investors overvalue firms with high accruals thinking that accruals will persist. 1999) or irrational investor preferences (Grinblatt and Han.S. Banz. Basu. and Amihud's (2002) illiquidity ratio. we offer a brief literature review of these variables. As shown by Fama and French (1996). Daniel et al.. and momentum (i. Chen et al. which summarize the firms' growth in size and profit margins. The remainder of the paper is organized as the following. and Chinese markets. Similarly. (2004) find that firms with higher net operating assets (NOA) have lower future returns.. 1977. and advertising expenditure.e. we attempt to identify the most important firm-specific return predictor variables. the book-to-market ratio. It would be naïve to label a market more efficient simply because there is less return predictability. (v) asset growth and change in gross profit margin. The buying of past winners and the selling of past losers are widely known as momentum strategies. Several behavioral models have been proposed to explain the momentum effect based on investor underreaction or overreaction to information (Barberis et al. is that there exists an intricate relationship between return predictability and market efficiency. In this section. Section 3 discusses data and methodology for constructing stock return predictors for both the U. the momentum effect is not subsumed by the size or value effects. at horizons ranging from 3 to 12 months. research and development expenditure. and sales growth. These two anomalies are somewhat related as net operating assets is the accumulation over time of the difference between net operating income and free cash flows..S. (vi) equity financing and debt financing for firms’ external financing activities. Stocks with higher accounting accruals tend to have lower future returns. (2001) find that firms with high capital investments subsequently have low returns.g. Hirshleifer et al. 1981). The size effect and the value effect are among the earliest proposed stock return predictors (e.406 X. and finally. 2005). / Pacific-Basin Finance Journal 18 (2010) 403–425 are more rationally priced. Hong and Stein. 1998. Section 2 provides a review of the existing literature on stock return predictability. Section 4 provides empirical results. 1998. (ii) three conventional value indicators: earnings-to-price ratio. Titman et al. past 12-month stock returns). In contrast to evidence on tangible capital investments. (vii) variables related to information asymmetry or liquidity of stocks including idiosyncratic return volatility. cash flow-to-price ratio. An interesting implication that one can generalize. Within this literature. at least in the context of comparing international financial markets. (iv) measures of firms' tangible and intangible investments. including capital expenditure. so these variables are often jointly regarded as value or contrarian indicators. rather. These two effects are often referred to as earnings quality anomalies.

We do not attempt to develop unique ex ante hypotheses for each predictor on Chinese stock Table 1 Hypothesized signs of stock return predictors. leads to higher expected operating performance and future stock returns. The effects of capital expenditure.X. seasoned equity offerings (Loughran and Ritter. Datar et al. (2006) show that stocks with high idiosyncratic volatility risk have low subsequent returns. Lakonishok and Vermaelen. and bank borrowings (Billett et al.S. Chen et al. future stock returns are abnormally high following stock repurchases (Ikenberry et al. 1990). We categorize all of these variables and findings as financing-related anomalies. Greater changes in gross margins indicate an improvement in the firm's terms of trade. All of our variables are posited to be positively related to future returns.S.e. Bradshaw et al. We list these variables in Table 1.. we transform the SIZE variable to be: − SIZE. information uncertainty and stock liquidity have also been found to be related to future returns. R&D. They find a strong inverse relation between firm asset growth and future stock returns. investors might assume this relation will persist into the future thus underappreciating that return on assets diminishes as asset size increases. Variable Description − SIZE B/P MOM E/P C/P − SG − ACC − NOA − CPX RD ADV − AG ΔGPM − ΔEQ − ΔDT − STDR − TURN ILLIQ − Firm size (market capitalization) Book-to-price ratio Momentum (past returns) Earnings-to-price ratio Cash flow-to-price ratio − Sales growth from prior year − Accruals-to-total assets ratio − Net operating assets-to-total assets ratio − Capital expenditures to total assets Research and development expenditures-to-market value of equity ratio Advertising expenditures-to-market value of equity ratio − Assets growth from prior year Change in gross profit margin − Net cash flow received from external equity financing − Net cash flow received from debt financing −Idiosyncratic risk −Trading volume turnover Illiquidity . / Pacific-Basin Finance Journal 18 (2010) 403–425 407 misreact to this accounting effect.. Because firms in their early growth stages experience a positive relation between assets growth and subsequent returns. because firm size (SIZE) and subsequent returns are posited to have a negative correlation. In addition. in turn. Related to the corporate investment and financing anomalies is the asset growth anomaly reported by Cooper et al. Ang et al. (2009) link both anomalies to adverse selection in corporate disclosure and investor underreaction. Jiang et al. Finally. (2006) summarize firms' external financing activities into two variables: external equity financing and debt financing. 1999). Abarbanell and Bushee (1998) view changes in gross profit margins as part of a fundamental analysis strategy. They report that these two variables are negatively correlated with future stock returns. For example. and advertising can be referred to as corporate investment anomalies. market. market. A brief description of each variable is provided. (1998) and Lee and Swaminathan (2000) show that stocks with high trading volume earn lower future returns. and they attribute this pattern to investor optimism and firms’ efforts to time the market in raising capital. For variables that are hypothesized to be negatively correlated with subsequent stock returns in the U. which. 1997). variable). 1991).. we transform them by changing their signs and we indicate this transformation by putting a “−” sign in front of variable. debt offerings (Spiess and Affleck-Graves. we adjust them so they can positively predict returns. (2008). A large number of corporate event studies have shown that future stock returns are abnormally low in the years following initial public offerings (Ritter. we identify 18 firm-specific return predictor variables. They associate their finding to investor misreaction to assets growth. 1995. Conversely. For each stock market predictor (i. For those variables that have been posited to be negatively related to subsequent returns in the U. 2006). we test its relation to subsequent stock returns. From the above discussion. Amihud (2002) finds that stock liquidity inversely predicts stock returns.

which account for more than 85% of the tradable market value of Chinese stocks at year-end of 2007. 3. and investment companies. More specifically. a joint product of the PACAP Research Center at the University of Rhode Island and SINOFIN Information Service Ltd. For our Chinese firms. we require that a stock have end-of-June price of no less than ¥1 to be included in our sample for year t. we construct 18 stock return predictor variables. a stock must have available information on stock price.S. which is the firm's market price multiplied by its common shares outstanding at the end of June of year t. Since February 2001.S. Our filters on U. First.7 We eliminate banks. taken as the ratio of the firm's book value to its market capitalization for the year-end at calendar year t − 1. For data on U. data are as follows. As with the Chinese sample period. as well as corporate financial statement information. Approximately 5% of Chinese stocks have a price below ¥1. Third. The Latinized symbol for the yuan is ¥. Firm size (SIZE) SIZE is the natural logarithm of the firm's market value of equity. the yuan. is CNY.S. return. There are different share types in China. / Pacific-Basin Finance Journal 18 (2010) 403–425 returns as our study is exploratory in nature. state owned enterprises. . Tradable A-shares are ordinary shares primarily made available to Chinese citizens and institutions.9 Further. B/P ratio (B/P) B/P is the book value of equity. and banks). information on stock price. More detailed information on all variables can be found in the Appendix. with the restriction that state and legal-entity shares cannot be traded publicly. 8 The official abbreviation for the Chinese currency. the U. Second. Here. sample period also covers 1994–2005 for financial statement data and July 1995 to June 2007 for stock returns data. American Depository Receipts. 3. return. 9 Our results remain similar without this minimum price threshold and when we set the minimum price at ¥5. Information on corporate financial statements is from COMPUTSTAT. which literally translates to “people's currency”). real estate investment trusts.408 X. 6 Stocks of a typical Chinese firm may consist of state shares (those owned by the central or local governments). Construction of stock return predictor variables Based on our discussion in Section 2. The holding period for stock portfolios in the subsequent year is from July of year t to June of year t + 1.S. B-shares are available to domestic investors.and B-share classes. AMEX.6 Our sample covers financial statement data from 1994 to 2005 and stock return data from July 1995 to June 2007. and NASDAQ at the end of June in each year t. we only include A-shares. whereas B-shares are primarily made available to foreign investors. we briefly describe the construction of each of those variables. close-end funds. Data and variables construction 3. 1. markets as the default applicable hypotheses for Chinese markets. Data For data on Chinese firms. information on stock price. firms. is obtained from the PACAP-CCER China database. and foreign companies. Instead. 2.1. we select all common stocks traded in NYSE. and tradable shares.8 To avoid potential market microstructure related issues in measuring returns. the return holding period corresponding to accounting information reported at the end of fiscal year 2005 (December 2005 as December is the fiscal year end for all Chinese firms) is from July of 2006 to June 2007. we accept the hypotheses that have been developed for U. return predictors are constructed using accounting information at the end of fiscal year t − 1. market capitalization. 7 In our tests. but it is commonly abbreviated as RMB (renmindi. In this study. Tradable shares are further classified into tradable A. Chen et al. real estate firms. and trading volume. only the outstanding A-shares are used in the market value computation. we require stocks to have a minimum price of $1 at the end of June. we eliminate primes. close-end funds. and at least one valid stock return predictor in addition to firm size at the end of June in order to be retained in our sample. and trading volume is from CRSP. legal-entity shares (those held by domestic legal entities such as listed companies. A recent regulatory change makes A-shares available to a small group of qualified foreign institutional investors.2.

we estimate ACC as the change in noncash current assets less the change in current liabilities (excluding debt in current liabilities and income tax payable) and less depreciation. GPM is the difference between net sales and cost of goods sold divided by net sales. CAPEX for U. 1993). Advertising costs are not separately reported in China but are included in sales and marketing expenses. Net operating assets (NOA) Following Hirshleifer et al. ADV for Chinese firms is. ADV is advertising expenses over market capitalization.S.X. 8. Advertising costs (ADV) Following Chan et al. firms is the capital expenditure of the firm for the fiscal year ending in calendar year t − 1 over the average total assets at the beginning and end of that fiscal year. Accruals (ACC) Accounting accruals is the noncash component of earnings. C/P is the sum of net income and depreciation over its market capitalization at the fiscal year-end of year t − 1. E/P ratio (E/P) E/P is the ratio of earnings for the fiscal year to the market capitalization at the year-end of calendar year t − 1. Chen et al. Following Sloan (1996). 5. Change in gross profit margin (ΔGPM) Consistent with Abarbanell and Bushee (1998). firms is earnings before extraordinary items and depreciation. RD is the ratio of R&D expenditure over market capitalization. scaled by the average total assets at the beginning and end of that fiscal year. (2001). RD for Chinese firms is. during the fiscal year ending in year t − 1. 12. (2001). We skip one month between portfolio formation and holding period to avoid the effects of bid-ask spread. Capital expenditure (CAPEX) Following Jegadeesh et al. 15. C/P ratio (C/P) C/P for U. ΔEQ is the net cash received from the sale (and/or purchase) of common and preferred stock less cash dividends paid for the fiscal year ending in calendar year t − 1. 6. / Pacific-Basin Finance Journal 18 (2010) 403–425 409 3. approximated by the ratio of management expenses for the fiscal year ending in calendar year t − 1 to market capitalization at the end of year t − 1. (2004). Sales growth (SG) SG is the sales revenue for the fiscal year ending in calendar year t − 1 over the sales revenue from the fiscal year-end in year t − 2. the numerator of the CAPEX variable is the change in net fixed assets plus the change in accumulated depreciation from the fiscal year ending in calendar year t − 2 to the fiscal year ending in calendar year t − 1. whereas it is net income for Chinese firms. . (2006). and any lagged reaction (Jegadeesh and Titman. 9. For Chinese firms.S. 7. approximated by sales and marketing expenses for the fiscal year ending in calendar year t − 1 over market capitalization at the end of year t − 1. 11. (2006). 4. AG is measured as the percentage change in total assets from the fiscal year ending in calendar year t − 2 to the fiscal year ending in calendar year t − 1. External equity financing (ΔEQ) Following Bradshaw et al. The earnings measure for U. price pressure. For Chinese firms. 10. ΔDT is the net cash received from the issuance (or reduction) of debt for the fiscal year ending in calendar year t − 1. R&D expenses (RD) Following Chan et al. (2004). R&D is not separately reported in China but is included in management expenses. scaled by the average total assets at the beginning and end of that fiscal year. thus. 13. 14. This definition follows Fama and French (1993). we estimate NOA as the difference between operating assets and operating liabilities for the fiscal year ending in calendar year t − 1. firms is the sum of earnings before extraordinary items and depreciation over the firm's market capitalization at the fiscal year-end of year t − 1. ΔGPM is the percentage change in GPM from the fiscal year ending in calendar year t − 2 to the fiscal year ending in calendar year t − 1. Price momentum (MOM) MOM is the cumulative return of a stock in month-12 through month-1 preceding June of year t. External debt financing (ΔDT) Following Bradshaw et al. similarly. (2008).S. thus. Assets growth (AG) Following Cooper et al.

Trading turnover (TURN) Following Jegadeesh et al.425 0.099 0.681 16.085 0.610 0.555 E/P 0.419 E/P 0.814 − 4.344 1.150 0.548 MOM −0.652 1.232 292.679 13.042 0.287 0. as with the trading turnover measure.105 NOA 0.510 3.320 108.590 0.064 0.430 2.259 0.040 0.048 0.075 5.556 2091.003 0.017 0.121 0.561 0.012 0.020 0.271 993.909 674.871 52. market for our sample period.283 0.733 1.410 X.419 ACC −0.349 0.132 0.140 5.188 0. a percentile ranking is performed separately for NASDAQ and for NYSE/AMEX.129 0.303 1.236 65.957 53.366 0.115 0.130 1.107 0.214 C/P 0.026 −0.301 37.734 69.594 7.443 0.004 0.002 0.371 288.115 ΔGPM −0.011 0.012 296.026 0.267 38.S.754 ADV 0.385 .756 0.009 0.047 0.632 −1.335 1. Equal weight is used to compute cross-sectional averages.096 0. Idiosyncratic risk (STDR) Similar to Ang et al.712 7.081 0.017 0.047 0.370 39.419 4.026 0.799 1617.362 2.358 2.288 4.007 0.140 RD 0.182 ADV 0.806 26.330 − 4.127 STDR 0.592 0. Details of these variables are provided in the Appendix.181 20.015 0.116 ILLIQ(⁎10 ) Panel B: Time-series average of cross-sectional distribution of stock return predictors in United States SIZE(Mil USD) 94.415 16.088 0.261 1105.532 4.473 9.788 14.047 0.674 17.045 0.699 SG 0.952 19.868 RD 0. STDR is the standard deviation of the residuals in the regression of daily stock return on daily value-weighted market return with five lags and five leads for the period from month-12 through month-1 preceding June of year t.398 0.012 0.231 ΔDT −0. Because trading volume is measured differently in NASDAQ than in NYSE and AMEX.161 0.386 0.526 − 0.494 0.008 0.336 1244.057 0.052 9.194 61.881 −7 3.111 6.016 0. Again.006 0.034 0.450 0.852 3.576 11.348 SG − 0.042 0.334 ΔGPM − 0.032 2272.049 0. / Pacific-Basin Finance Journal 18 (2010) 403–425 16.805 40.069 − 0. Table 2 Summary statistics for stock return predictors.097 − 0.070 0.217 0.088 0.369 2014. a percentile ranking is performed separately for NASDAQ and for NYSE/AMEX. across month-12 through month-1 preceding June of year t.846 1126.618 11.026 0.023 0.820 213.453 ILLIQ (⁎10 ) Skewness Kurtosis 8.382 1.022 0.419 MOM − 0. Chen et al. This table presents the time-series averages of cross-sectional averaged statistics of return predictors for China and the U.008 0.178 CPX 0. dev.014 0.196 0.325 0.000 0.036 0.226 0.524 14.370.019 0.365 1.038 0.296 13.296 CPX 0.122 0.088 0.287 104.006 TURN 0.820 3.054 0.048 0.624 1270.012 0.608 B/P 0.687 0.451 − 1.001 0.575 0.015 TURN 0. where daily volume turnover is the ratio of the number of shares traded each day to the number of shares outstanding at the end of the day.705 0.016 0.120 0. (2006).697 ΔEQ 0.063 0.074 0.901 29.368 4.019 0.273 ACC − 0.023 0.291 0.108 0.137 NOA 0.921 C/P 0. Illiquidity (ILLIQ) Following Amihud (2002).693 972.228 −7 0.599 0.546 0.514 6.804 34.028 0.333 106.113 STDR 0.051 − 0.132 0.477 986. Panel A: Time-series average of cross-sectional distribution of stock return predictors in China SIZE(Mil RMB) 442.343 9.125 7. 17.160 0.977 B/P 0.362 − 15.109 AG 0.352 ΔEQ − 0.740 12.923 131. (2004).118 0.022 1.930 34.008 −0.509 0.015 0. 25% Mean Median 75% Std.055 0.235 0.537 0.711 25.082 7.465 0.015 0.793 0.802 0.535 1.025 0.701 1.333 13.142 0.099 96.640 3.098 AG 0.056 35.503 7.022 0.097 ΔDT −0. TURN is the percentile rank of the average daily volume turnover in the 12 months preceding June of year t. 18. ILLIQ is the percentile rank of the average daily ratio of the absolute stock return to its dollar volume.134 0.086 8.025 − 0.307 0.583 1328.024 0.771 13.060 0.346 − 0.441 82.013 −0.000 0.462 2.

ACC. we see firms with high R&D outperform those with low R&D by 0. and in Panel C we report the spread difference between China and U. and kurtosis. we test whether our predictors can explain the cross-sectional variation in subsequent oneyear returns..65). the top B/P-sorted portfolio outperforms the bottom B/Psorted portfolio by 0. perhaps the most important stock characteristic predicting future stock performance in the U. We form equally weighted portfolios in each decile and hold the positions from July of year t to June of t + 1. First. reports the average portfolio returns for each decile for the Chinese market and also the mean return difference (i.e.. interesting points arise. The return difference between D10 and D1 stocks sorted by prior 12-month momentum is 0.05% with a t-statistic of 0.56% per month. reported in column (2). standard deviation. is unprofitable in the Chinese market. and D1 contain the decile of the smallest firms. It shows that only five return predictors significantly predict stock performance in the subsequent year. For example. we see that except for MOM.80). Raw returns of sorted portfolios We first examine the return-predictive performance of our 18 variables using sorted decile portfolios. and so forth. We alternatively measure firm size using firm total market capitalization.16. spread) between D10 and D1 stocks. we rank stocks into deciles based on each predictor using year t − 1 financial statement data. including 25 and 75 percentiles. Recall that among these 18 variables. AG. Further. significant at the 5% level. Second.X. 11 .10 That is. mean. trading on momentum. market. While it is hard to provide a consistent story to justify why some predictors have significant returnpredictive power in the Chinese market while others do not. in column (12). we add a negative sign in front of each of these variables (see Table 1) so that the resulting variables positively predict returns. For ILLIQ rank-sorted portfolios. In column (8). 2004. In the rest of analysis. Chen et al. D10 stocks significantly outperform D1 stocks by 0.S. We first calculate the cross-sectional average return in each portfolio and then compute the time-series averages of the cross-sectional mean for each predictor. More specifically. slightly weaker. and TURN. firm size. for firm size (SIZE). SG. markets. NOA.74% per month at the 10% level (t = 1. 10 are posited to be negatively correlated with future stock returns: SIZE. for both our Chinese data (Panel A) and our U. we see that high −NOA firms outperform low −NOA firms by 0. The statistically significant results for −NOA and R&D is consistent with the interpretation that Chinese investors are largely uninformed and/or un-sophisticated. consistent with a finding in Eun and Huang (2007). consistent with their hypothesized signs (recall that when hypothesized signs are negative we transform those variables so that their hypothesized signs become positive). respectively. median.66% per month at the 10% level (t = 1. data (Panel B).S. ΔDT. indicating that Chinese investors are willing to pay a significant premium for more liquid stocks. Chan et al. they underreact to salient information provided in firms' financial statements regarding firms' net operating assets (accumulated accruals) and R&D expenses (related to firm intangible assets) (Hirshleifer et al. the signs of the mean differences for all other predictors are positive. CPX. For momentum (MOM). In column (10). results when using value-weighted returns. D10 stocks contain the decile of the largest firms. skewness.S. The bottom row in Panel A of Table 3 presents t-statistic for the return difference between D10 and D1 stocks. ΔEQ. we see that firms with high asset growth have significantly higher subsequent returns at the 10% level. In June of each year t. D10 contain firms with the largest past returns. where mean differences between D10 and D1 stocks are reported. Empirical results 4. STDR. and turnover do not significantly predict stock performance in the subsequent year. We obtain similar results. Table 3. This finding is consistent with Wang (2004).S. and in other developed markets. E/P and C/P predictors. return standard deviations. / Pacific-Basin Finance Journal 18 (2010) 403–425 411 Table 2 summarizes the cross-sectional distributions of the 18 stock return predictors.1. Panel A. 4. D10 and D1 deciles contain stocks with the largest and smallest measures of each return predictor. in Panel B we report the return spread between D10 and D1 stocks for the U. In looking at the second-to-last row in Panel A of Table 3.52).87% per month (t = 2.11 10 We obtain consistent. including market capitalization for A and B shares and total book value of assets. 2001).

76 1.00 1.58) 1. Chen et al. To be included in the analysis.23 D10 2.24 2. stocks are sorted into deciles based on each predictor.99 6 2.50 2.30) 0.23 2.97 2.11 1.19 2. sorted by each of the 18 return predictors for China market.14) 1.05 2.11 2.31 (0.97 2.97 2.87 (2.31 (3.24 2.46 2.28 (1.17 2.90 1.61) 0.98) 0.78 1.29 2.01 2.01 1.01 2.06 2. .06 2.49 1. The 18 predictors are defined in the Appendix.10 2.34 1.04 1.27 2.91 (1.24 2.90 1.07 0.33 0. The accounting data is from fiscal year 1994 to 2005 and stock return data is from July 1995 to June 2007.19) X.46 2.27 2.88 2.33 (− 0.92 (4.50 (1.09 (0.42 (1.92 1.23 0.93 1.33 0.38 2.27 2.22 2.36 2.99 1.17 (1.62 (0.17 2. where D10 represents the highest decile of the signed predictor.25 2.34) 1.63 (3.67 1.27 2.37) 0. as well and the return spread between D10 and D1 stocks.74 2.01 2.35 2.06 2.35 (1.18) 1. Panel A shows the decile returns.79 − 0.37 1.97 1.06 0.62) 1.13 1.04) 1.35) 1.66 (1.36) 0.08 2.09 (0. stock price is restricted to be no less than ¥1 for China market and $1 for the U.83 1.18 1.51) 1.25 (0.80) 1.59 (1.11 (1.52) 1.66 (1.59 t-stat (0.35 0. and D1 is the lowest.29 2.74 2.31 2.16) 2.99 2.06 2.99 1.73 (2.39) 0.16 2.10 1.19 2.38 0.26 2.74 (1. In June of each year t.56 2.75) 0.14 2.37 (0.56 2.27 1.39 2.00 2.25 2.71 0.26 0.15 2.08 2.06 1.19 1.01 2.29 2.04 2.27 − 0.97) 1.25) 0.21 2.54) 1.79) 1.03 0.84 2.15 2.43 10−1 0.84 2 2.65 (1.81 1.98 1.15 2.18 2.19 2.02 (2.44) 0.07 2.S.95 2.29 2.87 (2.18 2.98 2.23 8 2.25 2.65) Panel B: US 10−1 0.48 0.22 t-stat (0.84 2.19 2.39 1.21 1.01) 0.67 1.95) 1.74 2.S.02 2.23 2.59) 0.24 (0.83 2.25 (0.412 −SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) − SG (6) −ACC (7) − NOA (8) −CPX (9) RD (10) ADV (11) − AG (12) ΔGPM (13) −ΔEQ (14) −ΔDT (15) − STDR (16) − TURN (17) ILLIQ (18) Panel A: China D1 1.23 2.18 2.99 1.06 2.01) 2.67 1.97 2.94 2.59 1.84 2.99 2.95 1.08 2.95) 0.20 2.07 2. Panel B shows the return spread between D10 and D1 stocks for the U.04 7 2.72) 1. This table reports the average monthly returns of stock portfolio deciles in China and the U.88 1.13 2.97) 1.13 3 1.20 (0.05 (− 0.67 1.57) 2. market in June of each year.14 2.46 2.82) 1.01 (−0.04) 0.93 2.88 2.75 2.91 1.32 2.23 3.02) 1.82 2.98 2.03 2.11 2.07 2.27 1.56 (2.88 2.94 1.94 (1.27 2.12 1.15 2.76 5 1.99 1.96 2.98 2.33 0. Decile portfolios are held unchanged from July of year t to June of year t + 1.20 0.48 1.05 0.22 2.20 9 2.08 (2.28 0. The t-statistics are included in the parentheses.02 1.39 2. / Pacific-Basin Finance Journal 18 (2010) 403–425 Table 3 Returns to decile portfolios sorted on stock return predictors.26 − 0.81 2.91 4 1.S.57 1.15 2.01 1.22 2.33 2. market.42 2.43 2.

13 We follow procedures provided in Fama and French (1993) to compute SMB and HML factors using Chinese stock return data. and analyzing stock returns in subsequent 3 and 6 months. The results are similar to those reported in Table 4. A few studies. Chen et al. B/M.t in each month is the equally weighted stock returns. That is. Similarly. Six size-B/P portfolios are defined as the intersections of the two size and three B/P groups. in each month. R&D. relative to the Chinese stock market. returns.X. Portfolio return Ri. Our analysis is motivated by Fama and French (1998). RMRFt is the market return in excess of the risk-free rate. such as firm size. between the simple average of returns on the three smallstock portfolios (S/L. As discussed previously. In untabulated analysis we examine this conjecture by constructing predictors based on semi-annual data (from 1994 to 2001) and quarterly data (from 2002 to 2006). we see that four return predictors are statistically significant after we control for risk. and high-B/P groups based on the 30% and 70% cutoff points of the book-to-market ratio of all stocks. SMB is the difference. we simply provide mean return differences between D10 stocks and D1 stocks sorted by respective firm characteristics.S. Book value per share is book value divided by totals shares issued by the firm. return standard deviations. SMBt and HMLt are the monthly size and book-to-market factors.t − RFt is the monthly returns of portfolio i in excess of the monthly risk-free rate. S/M.S. Alphas of sorted portfolios An important question that needs to be raised whenever any pattern of stock return predictability is found is whether the pattern is related to risk. we sort all stocks into small and big size groups based on the median market capitalization of all stocks in June of year t. in each month.S. For brevity. . To control for the risk-return relation. and stock turnover are not reliably return predictive. The alpha differences between D10 12 As the average annual turnover ratio is much higher in China than in U. (2002). We calculate the monthly market returns in China as the value-weighted average monthly returns for all A-shares traded in the Shanghai and Shenzhen stock exchanges. 13 We alternatively evaluate the risk-adjusted return using the one-factor CAPM model and obtain qualitatively similar results. We independently sort all stocks into low. however. We construct size and B/P portfolios in June of each year t. HML is the difference. In particular. The risk-adjusted results are reported in Table 4.. Over the same sample period 1994– 2007. return differences for 10 of the 18 predictors are statistically significant at the 10% (or lower) confidence level. although Chinese public companies issue multiple shares. but some firm characteristics. who show that value effects in international markets can be explained by their three-factor model. use the 3-factor model to evaluate the momentum effect and volume-based strategies in the Chinese stock market. Table 4 results are consistent with that in Table 3. Regressions are performed using average monthly returns of each decile-sorted predictor variable during the entire sample period. such as Wang (2004) and Wang and Chin (2004). The market value of stocks is taken by multiplying A-shares' closing prices to their shares outstanding.t .12 4. respectively. ð2Þ where Ri. Following Wang (2004) and Kang et al. / Pacific-Basin Finance Journal 18 (2010) 403–425 413 Panel B of Table 3 reports U. it is likely the included predictors work better in shorter horizons. we use the monthly yield of the three-month household deposit interest rate in China as the risk-free rate. and B/H). we focus on A-shares. results. we use the Fama and French (1993) three-factor model to obtain alphas for all decile portfolios. the return differences for all the return predictors are positive.2. more predictors are statistically significant in explaining U. The three-factor regression model takes the following form: Rit −RFt = αi + bi RMRFt + si SMBt + hi HMLt + εi. The monthly value-weighted average return on each portfolio is then computed. and S/H) and the simple average of returns on the three big-stock portfolios (B/L. between the simple average of returns on the two high-B/P portfolios (S/H and B/H) and the average of returns on the two low-B/P portfolios (S/L and B/L). In our setting. stocks are ranked into deciles in each month by a specific return predictor and i refers to the ith decile portfolio. Overall. As reported in Panel A of Table 4. these results show that. median.

04 − 0.17 − 1.20 − 0.23 − 1.34 (− 1.80 − 1.85 − 1.79 (2.86) 0.02 − 0.06 − 0.99 − 0.00 − 0.95 − 0. / Pacific-Basin Finance Journal 18 (2010) 403–425 Table 4 Alphas to decile portfolios sorted on stock return predictors.67 (5.96 − 1. Panel B shows the alpha spreads between D10 and D1 stocks for the U.72) − 1.87 − 0.70 0.83 − 0.18 − 0.31 − 1.25 (2.00 − 1.22 − 0.35) − 1.85 (3.12 − 0.08 (2.07 −0. This table reports the average monthly three-factor alphas of stock portfolio deciles in China and the U.65 −0.82 − 0.34 − 0. The 18 predictors are defined in the Appendix.38 − 1.14 − 0.92 − 0.51 (3.40 − 0.05 − 1.81 − 1.91 − 0.16 − 1.04 − 0.65) 1.06 − 0. Chen et al. market in June of each year. market.46 −1.19) − 1.86 − 1.85) − 1.58 0.19 (−0.23) −1.32 − 1.26 − 0.43 (1.86 − 0.73 0.16 − 1.27 − 0.20 − 1.09 −1.04 8 − 1.12 (0.07 6 − 1.59) 1.17 − 0. as well and the return spread between D10 and D1 stocks.52 (1.67 0.04 − 1.01 − 0.33 − 1. stock price is restricted to be no less than ¥1 for China market and $1 for the U.75 0.98 −1. .20) 1.02 − 0.45 (2.91 D10 − 0.20 − 1.17 − 1.28 −1.56 (2.46 (2. The accounting data is from fiscal year 1994 to 2005 and stock return data is from July 1995 to June 2007.24 − 1.71) − 0.S.81 − 1. Decile portfolios are held unchanged from July of year t to June of year t + 1.77 0.75 −0.09 − 1.86 (1.86 (2.S.21 − 1.72 − 0.65 0.58 − 0.68 − 0.92 −0.87 −0.31 − 1.30 −1.90 − 1.82 −1.96 − 0.77) 0.59) 0.06 (−0.11 −0. The risk-adjusted return is the intercept from time-series monthly regressions of decile portfolio returns on the three factors: RMRFt is the market return in excess of the risk-free rate.20 (− 0.10 − 1.92 −0.91 −0.20 − 1.04) X.92 − 1.10 4 − 1.59) −1.95 −0.00 − 0.27 −1.89 − 1.23 (1.79 − 0.86 − 1.45 (1.06 −1.62) 1.07 − 0.28 10−1 t-stat (0.95 − 0.93 −0.88 0.99) 0.22 −1.45 (2.87 − 0.80 0. sorted by each of the 18 return predictors for China market.35 (− 1. respectively.32 (1.33) 1.21 −1.30) 0.70 − 0.45) 0.10 − 1.94 − 1.85 − 0.51 − 1.07 −1.00 −0.12) − 1.92 − 1.85 − 0.17 − 1.06 (− 0.15 −1.90) −1.01 − 0.02 − 0.96) 0.16 − 1. and D1 is the lowest.15 −1.77 − 0.85 3 − 1.87 2 − 0.18) −1.97 − 1.57) − 1. S.27) − 1.58) Panel B: US 10−1 0.87 (1.90 − 0.20 − 1.26 −1. stocks are sorted into deciles based on each predictor.414 − SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) − SG (6) − ACC (7) − NOA (8) − CPX (9) RD (10) ADV (11) −AG (12) ΔGPM (13) −ΔEQ (14) −ΔDT (15) − STDR (16) − TURN (17) ILLIQ (18) Panel A: China D1 − 0.45) 0. Panel A shows the decile returns.86 −0.95 −0.26 5 − 1.00 − 1.30 − 1.21) 1.16 − 1.64 − 0. To be included in the analysis.50 − 1.52 (1.16) − 0.18 − 1.15 − 1.90 − 1.32 (1.98 − 1.64 − 1.31 (2.37) 1.32 − 1.94 − 1.97 − 0.95 − 0.12 − 0. where D10 represents the highest decile of the signed predictor.05 −1.30 (2.58 (0. The t-statistics are included in the parentheses.00 − 0.14) − 1.95 − 1.85 (0.98 − 1.05 −1.09 − 0.13 − 1.05) − 0.96 − 1.42 − 0.10 t-stat (0.24 −0.20 − 0.11 − 0.19 −1.06 −1.92 − 0.05 − 0.03 − 0.98 − 0.30) 0. In June of each year t.14 7 − 1.21 (0.10 − 1.75 − 0.31 − 1.26 − 0.96 0.17 −1.48) −0. SMBt and HMLt are the monthly returns on size and book-to-market factors.85 −0.59 0.88 − 0. computed following Fama and French (1993) procedure.08 0.40 (1.88) −0.04 9 − 0.50) − 1.94 − 1.06 − 0.64 0.71 −0.26 − 1.97 − 0.76 − 0.19 (0.14 − 1.10 − 0.66 − 1.92 − 1.90 − 1.

−AG. shows that. we see that 8 firm-specific variables have significant coefficients in the return regression in the U. our results suggest that return predictors have much less power in predicting future stock returns in China than in the U. but return predictors themselves. we further check the robustness of our decile-sorted portfolio results by performing cross-sectional regressions of monthly stock returns on return predictors. the overall explanatory power of firm-specific variables is similar in these two markets as revealed by their similar R2s. / Pacific-Basin Finance Journal 18 (2010) 403–425 415 and D1 stocks are significant for −NOA. reported in Panel B of Table 5. We use the Fama and MacBeth (1973) procedure to compute the time-series averages of the coefficients on each of our predictors. Cross-sectional regressions In this section. To examine this possibility.S. To rule out this explanation. then a multicollinearity problem may inflate the R2s. stock returns.S..3. The dependent variable is the monthly stock return and the explanatory variable is each individual stock return predictor. The differences in the spreads between China and U.S. What explains weak return predictability in China? 4. We find few variables have significantly positive coefficients.S.. (i) the difference between average values of a return predictor for D10 and D1 stocks sorted by the predictor is small in China and (ii) return spreads between D10 and D1 stocks are accordingly smaller in China than in the U. after making risk adjustments.S. Also confirming prior results. As it turns out. Panel B of Table 4. the effect of each variable can be evaluated while holding constant other variables).1. Homogeneity in return predictors So far. It is interesting to see that despite there being fewer statistically significant return predictors in China. and we use these six principal component factors as joint regressors.4. after risk adjustments. but more variables are able to predict returns in the U. The average adjusted R2s from the monthly multivariate regressions. Panel A of Table 5 reports results when we perform univariate regressions. that is. The two main advantages to estimating a multivariate regression is that we can identify the marginal effect of each variable on its ability to predict returns and we can estimate the predictive ability of all of these variables jointly rather than in isolation (i. Note that numbers reported in Panel A are not sign adjusted. From Table 5. provide a way to quantify the overall return predictability in each market. while it is 6% in the U.e. where monthly returns are regressed on either all 18 predictor variables or on the first 6 principal components.S. are . 4. market.X. We also conduct two multivariate regression tests.. and associated t-statistics. −NOA. which reports on U. nine predictors are statistically significant in explaining subsequent returns at conventional confidence levels. RD. we identify the first six principal components of the 18 return predictors. −AG. we compare the spread of D10 and D1 stocks in the predictors themselves in China and U. we replace the missing predictor with the annual cross-sectional median value for that predictor to take advantage of the large cross-sectional data.S. generally validating the sorted-portfolio evidence reported in Table 4. which are also reported in Panels C and D. The results are reported in Panel A of Table 6. we see that the coefficients on B/P. For stocks with missing observations for one or more stock return predictors. One possibility for this finding is that return predictors in China are less heterogeneously distributed than they are in the U. The first row in Panel A reports the spread in return predictors between D10 and D1 stocks in China and the second row reports the spread for the U. market.4. Chen et al. Consistent with our earlier results on raw returns. The time-series average R2 of this principle component-based regression is 5% for both markets. 4. If independent variables in a regression model are correlated.S. are also shown in Panel A.S. and ILLIQ are significantly different from zero.S. RD.S. market. The results of multivariable regressions are reported in Panel C of Table 5. we find that some firm-specific variables are able to predict returns in China. Multivariable regression results using U. −STDR in the Chinese market at the 10% (or lower) confidence level. the average adjusted R2 for the multivariate regression model is 9% for the Chinese market. The differences in the spreads between China and U. data is reported in Panel D of Table 5.S. Panel A.

31) Panel C.62 t-stat (0. markets.26 −0.12) (2.99) (1.61 (0. Adj R2 is the adjusted R2 for multivariate regressions using all 18 predictors and Adj R2* is the adjusted R2 for multivariate regression using the first 6 principal components.48 (0.74 (0.01 −0.61) 21.76 2.416 −SIZE (1) −ACC (7) −NOA − CPX (8) (9) Panel A.22) (1. Univariate regressions for the Chinese market Coeff 0.63) Panel B.61) (2.62 (2. Panels C and D show the result of a joint regression including all the 18 predictors for both markets.18 0.45 0. To be included in the analysis.30 (1.30) B/P (2) MOM (3) E/P (4) C/P (5) −SG (6) 0.00 (1.95) 1.03 0.31 (− 0.17) –– –– –– –– 20. markets.90) 0.00 (0.41) (1.60 (1.78) −0.15) (0. .17 0.07 (1.59 (1.56) 1. This table reports the time-series averages of the coefficients in the regression of monthly stock return on stock return predictors in China and U.05 –– X. Chen et al. market in June of each year.91 (−1.91) 0.13) (−0.62 0.77) 1.04 0.81) 0. market Coeff 0.81 (5.80 (1.86) 1. Panels A and B show the result of univariate regressions for China market and U.61 (1. stock price is restricted to be no less than ¥1 for China market and $1 for the U.20) 0.67) (2.46 (2.S.35 (0.06) (1.68 (1. The accounting data is from fiscal year 1994 to 2005 and stock return data is from July 1995 to June 2007.03 (0.05 t-stat (−0.27 (2. If a stock return predictor is missing.27) −0.00 (0.39) (− 0.75) 0.08) 0.48 (1.10 (1.43) 0.42) 0.01 0.74) (1.67) (− 0.09 –– 0.17 0.77) (0.84) 4.16 − 0.16) 1.32 (1.86 (5.00 (−0.S.01 (0.17) − 0.61 (1.72 (−0.61) 0.25 (− 0.07 (0.83) –– –– –– –– 0.61 (1.20) 0.20) (0.20 (0.28) 0.03) (− 0.04 2.S.S.79) 0.33 − 0.20 (4.22) 1.27 (0.24 − 0.80 (0.06 –– 0.70) (3.09 0.00 (1.84) 0.47) (0.37 (1.44 2.44) Panel D.01 − 0.46) 0.67) (0. Multivariate regressions for the Chinese market Coeff 0.88) 0.34 (2.27 (0.72) 0.33 0.26) (−0. The t-statistics are included in the parentheses.64) 0.44) 1.12) (0.34 0.91) RD (10) ADV (11) − AG (12) ΔGPM (13) −ΔEQ (14) −ΔDT (15) −STDR (16) −TURN (17) ILLIQ (18) Adj R2 Adj R2* 0.30) 1.55) (0.26 0. market Coeff −0.30 0.40) 0.38) 0. / Pacific-Basin Finance Journal 18 (2010) 403–425 Table 5 Cross-sectional regressions. Univariate regressions for the U.26) 0.73 (3. it is replaced with the annual cross-sectional median.88) − 0. Multivariate regressions for the U.67) 0.54) (0.88) 0.90 (−1.12) 0.08 t-stat (0.50) 1.35 0.05 –– −4.64 (0.S.47) 0.10) 0.60 t-stat (1.37) (− 1.14) 0.18) 0.43 (1.28 (0.

79 (1.54) 0.59 0.13 (6. The t-statistics are included in the parentheses.50) (−0.43 1.14) Panel B: Standardized return spreads between D10 and D1 stocks China 0.91) (− 0.48 0.26) 0.18) (−0. To be included in the analysis.77 0.81 0.98 1.58 −0.00) 1.43 − 0.56) (− 2.− SIZE (1) −NOA (8) − CPX (9) RD (10) 0.52) 0. stock price is restricted to be no less than ¥1 for China market and $1 for the U.22 1. markets.74) (−0.76) (1.97) −ΔEQ (14) −ΔDT (15) −STDR (16) 0.36) (1.72) (−1.47) B/P (2) MOM (3) E/P (4) C/P (5) − SG (6) Panel A: Spreads of average return predictors between D10 and D1 stocks China 2.54 1.65) (− 1.60) (2.85) (−1.65 (− 2.29 1.11 t-stat (0.71 0.35) (0.52 0.50 − 0.04 (4.60 0.36 0.77 (1.17) (− 0.21 4.06) (0.68 −0.44) 5.33 1.01 0.41) 9.72 − 2.52) (−0.11) (−0.80 1.05 China–US −3.35 − 0.63) (1.18 t-stat (0.90 0.90 0.46 − 0.52 (0.27) 1.02 0.14) −TURN (17) ILLIQ (18) 0.41 0. The accounting data is from fiscal year 1994 to 2005 and stock return data is from July 1995 to June 2007. Chen et al. 417 .66 (3.19) 1. The difference of the standardized return spread between the China and U.04 −0.36 1.84 − 1.12 (1.23 − 0.50 0.13 −0.99) 0.03 0.67) (−7.27) (− 0.24 0.95) − 0.41 −0.81) (0.25 (−0.82 (0.58 (2.01 (− 0.22) China–US 0.27 0.28) 4.48) (− 0.43 0.67 4.59) 0.S.16 0.04) (0.87) 0. market in June of each year.51 (0.13) (0.21 (2.67 − 1.87) (−1.36 1.30) 0.03) (− 0. / Pacific-Basin Finance Journal 18 (2010) 403–425 Table 6 Standardized return spreads.10 0.20 (0.90 0.51 t-stat (−25.44 −1.84 (0.70 (− 2. Decile portfolios are formed in the end of June and held unchanged from July of year t to June of year t + 1 for each stock decile.30 1.89 − 1.79) 0.47 1.38 (− 1.05 − 0.48 0.65) ΔGPM (13) 0.00 0.63 0.12) (2.32) 3.02) − 1.39 −1.07 t-stat (0.61) (− 0.62) (0.70) (2.38) 0. markets is the difference in the standardized return spreads in these two markets.90 (0.45 − 0.01) X.94) 1.30 0.S.00 (− 0.00 (1.02 − 0.28 0.90 0.82) (1.10 (−0.56 −0.41) −0.06 −1.02 3.60) (0.89) (− 1.45) (3.02 −0.25 − 0.79 (1.08) −4.54 US 6.S.05) (1.81) 0. This table reports the time-series average of standardized stock return spreads between D10 and D1 portfolios sorted by return predictors for the China and U.97) (−0.53 3.17 3.69) (−0.46 (−2.49 0.59) (1. The standardized return spread (D10−D1) of a return predictor is the stock return spreads scaled by the predictor spreads in a corresponding year.52 −0.36 (−6.39 0.03 (− 5.89) −0.07 0.50 (1.14) US 0.99 −0.85) 0.58 1.38) − ACC (7) ADV (11) − AG (12) 1.29) 0.38 (0.88 (−7.98 (0.89) (− 1.

76) (−0.07 t-stat (0.09) (− 1.27 1.96 (p = 0.26) 0.66) (1.27 (1.89 (0.94) (1.67) 0.13) High 0.10 −0.56) 0.t are weekly market returns in each year (July of year t − 1 to June of year t).26) (1.55 (3.12 (0.45) (0. Panels C and D report the results of the U.78 (1.25 −0.37) (1.41) 0.67) 0.07 0.13) (2. The accounting data is from fiscal year 1994 to 2005 and stock return data is from July 1995 to June 2007.03) −0. The last column reports the averages of the 18 predictors.84) (2.51 1.52 1.31 −0.12) (0.05) 0.42) −1.74) (1.29 (0.30) 0.03) 0.24) (0.14) 2.74) (1.91 (1.82 0.78) 1.418 R2 rank −SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) −SG (6) − ACC (7) −NOA (8) −CPX (9) RD (10) ADV (11) Panel A: Return spreads in China Low 0.34 0.12 1.69) 0.67 0.72 −0.10 (0.85) (1.41 0.91 (0.26) 0.71) 0.01) (−0. The Wald statistics on the difference of all return spreads (standardized return spreads) between low and high R2 are reported.77) (−0.49) 0.33 (0.61 (0.00 (0.53 −0.42 (0.03 t-stat (0.25 t-stat (1.68 1.83 (2.27) Low–High 0.27 0.70 (0.05) 0.74) (−0.50 (0. Chen et al.41 (−3.20 (1.49 0.96 (5.06) 1.57 0.54 −0.t + εi.47 0.32 (1.74 (0.21 0.36) 1.90 (2.65 (1.30) 0.40) 0.76 −ΔEQ (14) 1.56 t-stat (0.89 (2.74) (1. / Pacific-Basin Finance Journal 18 (2010) 403–425 Table 7 Return spreads sorted by synchronicity.81) 0.26) −0. stock price is restricted to be no less than ¥1 for China market and $1 for the U.S.45) 0.47 0.91) (2.16) 0.66 1.S.63 (0.91 − 0.22 0.82 (1.25) −0.07 −0.95) − 0.43 (3.18) (0.59 0.42 (0.76) 0.09) (0. markets.54) (0.56) 0.24) 0.S.72) (−1.32 t-stat (1.14 X.69 (0. We break down firms into three groups based on the R2 from the market model regression: ri.29 (− 0.55) 0.46 (0.13) 1. Panel A reports the time-series average of stock return spreads between D10 and D1 portfolios sorted by return predictors for the tercile R2 groups and the difference in low and high R2 groups in the Chinese market.90 (1.88) (3.05 0.42) (0. for the China and U.11) (0.90 − 0.84) (−0.96 0.18 0.26 0.6 (2.44 0.56) (2.38) 0.78) 0.94) (1. market in June of each year.77 (3.11) (1.83) (0.32) (3.50) (2.75 (0.07 1.05 1.63) (−1. The t-statistics are included in the parentheses.02 (2.05 − AG (12) ΔGPM (13) 1.89 1.25 t-stat (0.21 1.23) −0.24) (2.31 (0.69) − 0.18 0.23) 0.05 − 0.17) (0.54) (0.54 (1.83 0.02) (0.46 (1.86) 1.68 (0.43 0. . To be included in the analysis.28) 0.92) 1.19 0.42 0.91) (− 1.09 0.50) − 0.91) 1.34 −0. ri. we further sorted stocks into deciles based on each of the 18 return predictors.64) Mid 0.97) (0.001) Panel B: Standardized return spreads in China Low 0.81) High 0.12 1.79 (1.31 (− 0.62 (0.48) (− 0.56 1.52 0.14) Mid 0.92) 0.87) (0.53 −0. Panel B reports the results based on standardized return spreads in the Chinese market.32) 0.46) 0.31 (2.06 0.89 (1.22) (2.2 0.13 (0.16) 1.29 0.64 0.71 0. In each R2 group.28 0.50) (1.t.09) (−0.43 1.78) (0.22 0.52 0.12 −ΔDT (15) − STDR (16) −TURN (17) ILLIQ (18) Average 0.19 1.75) (0.48) 0.70) 1.02 (0.33 (2.99) − 0.56 (0.29 2. Double-sorted portfolios are formed in the end of June and held unchanged from July of year t to June of year t + 1 for each stock decile.43 (0.53) (0.t are weekly stock returns and rm.20) (0.14 (−0.05 0.89 −0. market.77) − 0.45) −0.87) (0.14 1.11 0.58) −0.65 2.43) (1.80) 0.97) (2.74 (− 1.93) (1.65 (−1.32 0.06) −1.55 0.94) (2.75) Wald statistic for the differences in return spreads between low and high R2 groups being jointly zero: 45.25 0.17 0.03 (1.t = αt + βtrm.

28) 0.38 1.69) −0.27 (1.18 (1.14) (− 0.08 0.99) 0.07) (2.03 (2.46 (−1.62 (3.76) (1.43 (1.72 t-stat (0.90) −0.90) − TURN (17) ILLIQ (18) Average (− 2.57 0.47) (3.55 (1.14) (2.54) 1.27 (2.30) (1.97 (3.75) 0.60) (2.95) −0.99) (1.65 0.16 0.87 (1.51) 0.40) 1.34) −0.92 (3.66 1.43) (3.95) 0.03) 0.07) 0.27) (0.74 (1.69 0.81) (− 0.15 0.85) (1.22) 0.17) (3.46) 0.01) (0.34) (0.48 0.32) (−0.76) 0.44 (0.93 0.42 0.36 (3.24 0.19 0.44) −0.24) 0.88 (0.50) (3.94) (2.98) (2.77) Wald statistic for the differences −AG (12) 419 .29 0.49 0.38 0.44 (1.13) 0.88 0.35) (− 0.67 (2.59) (2.19 0.08 1.73) (2.75) 0.65) (0.42 (0.34) (1.59) (4.92) 1.91) (2.15) (1.64) 0.69) Mid R2 0.86 − 0.19) 1.06 (p = 0.88) 1.06 (0.65 (1.62 t-stat (0.49) (1.94) (1.13) (1.03) (− 0.17) (2.17) (2.73) (1.60 (1.32) (0.39 0.20 0.71 t-stat (0.45 (2.55) 2.08 0.62 (0.84 (3.84 (p = 0.54) (4.27 (−0.72 (0.49) 0.75) − 0.87) (1.45 (0.51) 1.01 0.56 0.93 0.05) 0.31 (2.60 (1.47) (1.20 (0.87) (−0.95) (1.87 −0.28 0.41) (3.17 (1.22 (0.44 (0.04 (2.15 0.05) 0.11) 0.48) (1.39) 0. Chen et al.75) 0.40) 0.43 1.85) (1.57) 0.46) 0.82 (4.03) (1.04) (0.77) Panel D: Standardized return spreads in US Low 0.87 (0.24 (0.62) (1.11 0.35 (1.45) 0.90 (1.64) (−0.86 0.85 0.18) 0.45) 0.12 0. / Pacific-Basin Finance Journal 18 (2010) 403–425 Panel C: Return spreads in US Low R2 0.69 (2.15) 0.91) 1.49 −0.07 0.33) in return spreads between low 1.30 0.38) Mid 0.29) (−0.00 − 0.24) 0.13) (−0.78 1.92 (1.32) (1.18 0.84) (1.88 0.23 (1.10 1.78) 0.07 (− 0.24) 0.66 (1.80 (1.16 0.48 (1.97) (2.56) (1.20 0.93) High 0.78 0.82) 0.42) (−0.91) 2.20 (−0.64 − 0.49) (1.69) 4.73) 0.90) (2.72 (3.34 0.65) (−0.95) 0.91) (0.15) 0.57) 0.27 (0.95 1.87 (1.19 1.08) 0.20 t-stat (1.64) (− 1.7) (−0.97) 0.28 0.11 0.45 (1.76) 0.72) 1.37 t-stat (0.96) 0.45 1.25) (2.02) 0.57 0.20) (1.61) 0.96 t-stat (1.52 1.46 (2.42 1.32) (0.001) 0.15 (1.29) 1.05 (0.95 (2.74) (0.12 1.07) (1.28 (p = 0.65) (0.63) (0.76 (2.Table 7 (continued) R2 rank − SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) − SG (6) −ACC (7) −NOA (8) − CPX (9) RD (10) ADV (11) t-stat (1.50) (1.80) (1.30) 0.30 (−0.88) Low–High 0.08) and high R2 groups being jointly zero: 47.28) (0.40) 0.001) 1.93) and high R2 groups being jointly zero: 41.82 −0.001) ΔGPM (13) −ΔEQ (14) −ΔDT (15) −STDR (16) (1.20 0.86) Wald statistic for the differences in return spreads between low 0.45) Low–High 0.61) 0.57 1.90) (−0.24 (−0.26 0.58 (1.59) (1.39) 0.75) 0.04 0.16) (0.76 (1.58 t-stat (1.65 t-stat (0.27) 0.23 0.33) 0.12 −0.76) (2.27) −0.69 (1.46) (0.19 −0.04 0.62) Wald statistic for the differences in return spreads between low and high R2 groups being jointly zero: 42.78) (− 0.84 (1.55 (1.38) High R2 0.83) X.05 1.09) (0.59 (1.41 (−1.94) (1.51 (1.17) 0.22 (0.80) 6.37 0.29 1.43) (1.12) (− 0.79) −0.86) (1.02 (0.15 (2.32 (2.14 (−0.68 1.64 0.59) 0.87 0.97) −0.97 0.15 1.07) (4.10) (2.27 (0.35 −0.31) (2.56 1.87 (1.49 (0.42 (3.84) (1.34) (2.40 1.50) 0.93 (1.84 (0.02) (1.98 (2.46) (0.42 (1.06) (3.48) (1.94 1.08) (3.16) 0.54 (−1.82 1.90) (1.46) 0.07) Low–High 0.33) 1.90 t-stat (0.08 (0.57) 0.38) (1.99) 0.35) (−1.95) (2.63) (1.45 (1.13 0.28 0.

2000. Taken together. The results. we first average the synchronicity measure across stocks in each year then aggregate the measure over time. −SG. markets by regressing weekly stock returns onto weekly market returns in each year (July of year t − 1 to June of year t) during the period of 1994 to 2005 and obtain the R2 statistics of each stock in each year. significant at the 1% level. In each year. return predictability could still be weak. exhibit patterns similar to those for unstandardized return spreads reported in Table 3. for each predictor. We test the difference in return synchronicity between the two stock markets. Within each market.4. 4.36). Price synchronicity is a measure of informativeness of stock prices. Morck et al. it is unlikely that weaker predictability in the Chinese market is due to higher efficiency. The essential assumption behind the examination of heterogeneity of return predictors across markets is that return spread per unit of predictor spread holds constant.S. E/P.2. −NOA. suggesting return predictors are generally more homogeneously distributed in China than in U. leading to lower quality of Chinese firms. We then use a double-sort procedure to test whether there is a positive link between price informativeness and return predictability. That is. variables.S. We therefore explore another possible reason for low return predictability — stock prices are persistently uninformative in the Chinese market.. the accrual return predictability is only marginally significant in the Chinese market. Under the behavioral asset pricing framework. We find that the average market model R2 in China 0. and correction never takes place. we sort stocks into equal-weighted decile portfolios based on each of the 18 predictors. −AG. Chen et al. (2000).46. However. we resort to a measure well-known in the literature — return synchronicity. we calculate the time-series average of the .. reported in Panel B of Table 6. ADV. the standardized return spreads is also lower in the Chinese market. / Pacific-Basin Finance Journal 18 (2010) 403–425 negative for 15 out of the 18 return predictors. 1988. Given initial mispricing. Overall. Low synchronicity is associated with greater incorporation of firm-specific information on individual returns (Roll. and perhaps just as puzzling. Finally.S. thus higher price informativeness. Then. most Chinese firm-specific variables have lower standard deviations than U. show that standardized return spreads in China and U.S. It is likely that there is large (initial) mispricing in the Chinese stock market. these observations suggest a higher degree of homogeneity in the Chinese market. Differences in standardized return spreads between China and US are negative in 13 predictors. Return synchronicity Despite that return predictability is often considered an indication of market inefficiency. and −STDR spreads in China are significantly lower than that in U. and Durnev et al. though negative.S. as reported in Table 3. Among them. and −ΔDT at the 10% (or lower) confidence level.12 in the U. Following Morck et al. To test this. within each synchronicity group. while it is 0. (0. The much higher R2 in China supports the view that stock prices are generally much less informative in China.S. nine out of the 15 differences in spreads are not significantly different. 2003). China has significantly lower standardized return spreads in MOM. We further assess the impact of the variation of return predictors by examining the standardized stock return spreads between D10 and D1 portfolios sorted by each of the 18 return predictors. We compound daily CRSP value-weighted returns to get weekly market returns. −ΔEQ. Therefore our specific hypothesis here is that the standardized return spreads across D10 and D1 deciles sorted by return predictors are identical in these two markets. it is also likely that stock prices are so noisy that they seldom converge to the fundamental values.49) than in U. we first sort stocks into three groups based on synchronicity.S. C/P. or R2 from regressing individual stock returns on market returns. The weekly stock return is the compounded daily return for a calendar week. −ACC. That is. as reported in Table 2.. stock return predictability depends on both initial mispricing and subsequent price correction. we perform market model regressions for all the stocks in the Chinese and U.420 X. A further interesting point is that the average spread for ACC is larger in China (0. This is consistent with the reality that earnings management is more rampant in China. While SIZE. The standardized return spread of a return predictor is the time-series average of the total stock return spreads between D10 and D1 portfolios scaled by the predictor spreads of that market in corresponding years. if stock prices are persistently noisy. −AG. However. RD. Note that patterns on return predictor spreads are consistent with the cross-sectional standard deviations of these predictors. Table 6 suggests the weaker return predictability in China is not merely due to there being less variation in the predictors themselves.

The Wald tests and the analysis on the average (standardized) return spreads across 18 predictors yield similar conclusions. To summarize. In addition. We reject the null hypothesis in both Panels.S.X. stock returns. we report the difference in the average (standardized) return spreads of the 18 return predictors within each R2 group. in the last column of each panel. albeit a bit weaker — four predictors generate significant positive future stock return spreads in the low-R2 group and only one predictor in the high-R2 group. at least in the context of cross-country comparison. 11 work significantly better at predicting returns in low synchronicity groups than in high synchronicity groups (see last row of Panel B). Another possible explanation for low return predictability in China is that there is high price inefficiency in China. the predictors work much better in the low R2 group — out of 18 predictors. Chen et al. This evidence suggests that weak return predictability is related to low price informativeness. Our results also question the validity of the conventional perception of the positive relationship between return predictability and market efficiency. 5. The results are reported in Table 7. combined with high average R2 in the Chinese market. However. The patterns are similar to those in the Chinese market: return predictability is much stronger in the low synchronicity group than in the high synchronicity group.S. the Wald statistics follows a Chi-squared distribution with 18 degrees of freedom. we find that more predictors work better for the bottom R2 stocks than for top R2 stocks. we also find the stock return sensitivity to return predictors is lower in the Chinese market than in the U. While we find evidence of greater homogeneity in the Chinese market.S. / Pacific-Basin Finance Journal 18 (2010) 403–425 421 return difference between D10 and D1 portfolios (return spread) and the return difference scaled by spread in the predictor itself (standardized return spread) within each synchronicity group. markets. . we find that almost all firm-specific variables predict subsequent one-year returns in their hypothesized direction. Regardless. the return spread and standardized return spread should be higher among stocks with lower synchronicity. 11 predictors generate significant positive return spreads in the low-R2 group while only four predictors have significant positive return spreads in the high-R2 group. in both the Chinese and U. stock returns from the same sample period and find that more predictors are statistically significant in explaining subsequent cross-sectional stock return variation. For the Chinese market. Under our hypothesis. We believe this novel finding. markets.S.S.S. For comparison purpose. only 5 firm-specific variables are statistically significant in their ability to predict raw returns. nine return spreads in the low-R2 group are significantly larger than spreads in the high-R2 group. when comparing the return spread for high and low R2 stocks. and 4 firm-specific variables are statistically significant in their ability to predict risk-adjusted returns. Finally. We conduct the same tests on U.S. provides an interesting explanation for the weaker return predictability in China. Using portfolio-sorted returns and regression analyses. we find much stronger return predictability in the low-R2 group than in the high-R2 group. Under the null hypothesis. We find that the market model R2. Conclusion Using Chinese stock returns data. market. One possible explanation is that return predictors are more homogeneously distributed in the Chinese market than they are in the U. We perform this sort and calculate these spreads for both the Chinese and U. market in Panels C and D.S. We test two explanations for the cause of weak return predictability in China. is much higher for China than for the U. market. For the results based on return spread (Panel A). a measure of stock price uninformativeness. we examine the predictive power of 18 firm-specific variables that have been hypothesized to predict U. we perform similar analysis for the U. We also perform Wald tests in Panels A and B on the hypothesis that all the (standardized) return spreads among low-R2 firms equal to those among high-R2 firms.S. The results based on standardized return spreads (Panel B) are similar. The number is significantly higher among low-R2 stocks than among high-R2 stocks. indicating that the stock return predictability is weaker in China than it is in the U. across Chinese stocks. Further. return predictability is stronger in groups where pricing informativeness is higher.

D10. ΔLTDC is the change in long-term debt included in current liabilities. E/P 5. SG Sales growth Salest−1 7. LTD = long-term debt [D9]. where ΔCA is the change in current assets from previous fiscal year. where CapExt − 1 is the change of net fixed assets in fiscal year t − 1 plus the change in accumulated depreciations in year t − 1. ACC Accruals [(ΔCA − ΔCASH)t − 1 − (ΔCL − ΔSTD − ΔTP)t − 1 − DEPt − 1] / ATAt − 1. where m is June of year t Net Incomet−1 . computation details for the eighteen return predictors used in this paper. [Operating Assets − Operating Liabilities]t − 1 / ATAt − 1. PS = preferred stocks [D130]. NOA Net operating assets [(ΔCA − ΔCASH − ΔSI)t − 1 − (ΔCL − ΔSTD − ΔLTDC − ΔTP)t − 1 − DEPt − 1] / ATAt − 1. X. long-term debt. SIZE Natural logarithm of market capitalization Book to price ln (closing price at the end of June multiplied by A-shares outstanding at the end of June) BVt−1 . operating liabilities = total asset − STD − LTDC − LTD − MI − CE. and minority interest are set as zero if they are missing. DEP is depreciation and amortization expense [D14]. 8. DEP is change in accumulative depreciation. Constructing stock return predictors We provide the variable name. [Operating Assets − Operating Liabilities]t − 1 / ATAt − 1. CAPEX Capital expenditure to total assets CapExt−1 CapExt−1 Salest−2 ½D12 Salest−2 ATAt−1 . For the U.422 Appendix A. and ATAt − 1 is the average of the beginning and ending total assets [D6] of the reporting year t − 1. where ΔCA is the change in current assets [D4] from previous fiscal year. MVt−1 inDecember A-Shares in December / Total Shares Outstanding). taxes payable. where BV of A-Shares = Book Value of Equity ⁎ (Number of outstanding MVt−1 ln (closing price at the end of June [CRSP] multiplied by common shares outstanding at the end of June [CRSP]) BVt−1 . ΔCASH is the change in cash/cash equivalents [D1]. and ATAt − 1 is the average of the beginning and ending total assets (TA) of the reporting year t − 1. 9. where MV t − 1 is market capitalization of A-shares at the end MVt−1 of year t − 1 Net Incomet−1 + Depreciationt−1 . where MVt − 1 is market capitalization of MVt−1 A-shares at the end of year t − 1 Earnings before Extraordinary Itemt−1 ½D18 Salest−1 ½D12 MVt−1 ½D24 D25 Earnings before Extraordinary Itemt−1 ½D18 + Depreciationt−1 ½D14 MVt−1 ½D24 D25 6. and ATA = the average of the beginning and ending total assets [D6] of the reporting year. in that order]. MI = minority interests. ΔCL is the change in current liabilities. and preferred stock are set as zero if they are missing. Predictors are constructed in June of year t. The values of short-term debt. and ATA is the average of the beginning and ending total assets (TA) of the reporting year. ATAt−1 . and ATA = the average of the beginning and ending total assets of the reporting year. MI = minority interests [D38]. where operating assets = TA [D6] − cash and short-term investment [D1]. ΔCL is the change in current liabilities [D5]. operating liabilities = total asset − STD − LTD − MI − PS − CE. where m is June of year t m−1 ∏m−1 i = m−12 ð1 + monthly returnt Þ−∏ i = m−12 ð1 + market monthly returnt Þ. The values of short-term debt. ΔCASH is the change in cash. C/P Cumulative marketadjusted return for the preceding 12 months Earnings to price Cash flow to price m−1 ∏m−1 i = m−12 ð1 + monthly returnt Þ−∏ i = m−12 ð1 + market monthly returnt Þ. ΔSTD is the change in debt included in current liabilities [D34]. MOM 4. CE = common equity [D60]. and ATA is the average of the beginning and ending total assets (TA. we provide data item numbers in the CRSP and COMPUSTAT datasets. long-term debt. [D6]) of the reporting year t − 1. D130. taxes payable. Description Computation detail: China Computation detail: United States 1. ΔSI is the change in net short-term investment. LTDC = long-term debt included in current liabilities. CE = common equity. MVt − 1 is the market value of A-shares at the end of year t − 1. where BV is the book value of stockholders' equity [D216]. minus the book value of preferred tax [D56. For all the financial statement items. if available]. where CapExt − 1 is capital expenditure [D128] in year t − 1. / Pacific-Basin Finance Journal 18 (2010) 403–425 Variable .S. plus balance sheet deferred taxes and investment tax credit [D35. STD = debt included in current liabilities [D34]. B/P 3. year t refers to the fiscal year ending in the calendar year t. ΔSTD is the change in short-term debt. minority interest. ΔTP is the change in income taxes payable [D71]. STD = short-term debt. where operating assets = TA − cash − short-term investment. Chen et al. 2. LTD = total long-term debt. data. description. ΔTP is the change in taxes payable.

market MVt−1 value is market capitalization at the end of year t − 1 Change in total assets Sales Expensest−1 MVt−1 end of year t − 1.t Percentile rank τ−12 in SHSE and SZSE.τ t−2 − St−2S−CG . / Pacific-Basin Finance Journal 18 (2010) 403–425 Change in gross profit margin 16. where |Ri.τ is daily stock return in the 12 months preceding June of each year. t−1 17. Rm.τ = αi + ∑ βk. where n is number of days available for 12 months preceding the end of June in each year. ADV 12.τ + k is the valueweighted average of the Shanghai Stock Exchange (SHSE) and Shenzhen Stock Exchanges (SZSE) value-weighted daily index return from 5 lags and 5 leads. respectively. where ΔLTD is the change in long-term debt. where St − 1 and CGt − 1 are net sales [D12] and cost of t−2 goods sold [D41] in year t − 1. where n is number of days available for 12 months preceding the end of June in each year. 423 . and ATA is the average of the beginning and ending total assets of the reporting year.t . where TA is total assets [D6].i 4Rm. TAt−2 TAt−1 −TAt−2 . and ATA is the average of the beginning and ending total assets [D6] of the reporting year. STDR ADVt−1 . t−1 ATAt−2 [D111]. and ATA is the average of the beginning and ending total assets [D6] of the reporting year. . respectively. and n is number of days available for 12 months preceding the end of June in each year.τ is the respective daily volume in dollars.τ is the respective daily volume in RMB. ΔGPM Description Computation detail: China Computation detail: United States Research and Development expenses to market value of equity Advertising expenses to sales SGAEt−1 . where ΔCT is the change in common stock. STDR is the standard deviation of the error term in daily data market model k=5 k=5 regression: Ri. CREt−1 −CPEt−2 −DIVt−1 . ILLIQ Amihud illiquidity measure ∑ Daily Volume = Shares Outstanding Percentile rank t−12 in SHSE and SZSE. ΔCD is change in current debt [D301]. where Ri. Percentile rank τ−12 n where |Ri. and ATA is the average of the beginning and ending total assets of the reporting year. where Ri. VOLDi. TURN Average daily volume turnover 18. where R&D is research and development expenses [D46]. t−1 ∑ Daily Volume = Shares Outstanding Percentile rank t−12 in NASDAQ and n NYSE/AMEX. Rm. STDR is the standard deviation of the error term in daily data market model Idiosyncratic risk where St − 1 and CG are net sales and the cost of goods sold k = −5 + k + εi.i 4Rm.t . respectively. VOLDi. CPE is cash paid for purchase of common and preferred stock [D115]. ΔDT Net cash flow received from external debt financing ΔLTDt−1 + ΔLTNt−2 −ΔSTDt−1 . 15. ΔSTD is the change in total short-term debt. TAt−2 St−1 −CGt−1 St−1 St−1 −CGt−1 St−1 t−2 − St−2S−CG . regression: Ri. DIV is cash dividends paid [D127]. where MVt − 1 is market capitalization of A-shares at the TAt−1 −TAt−2 .t = VOLDi. AG 13.τ + k = −5 k + εi. where CRD is cash received from the issuance debt ∑ Ri. Chen et al. where ADV t − 1 advertising expenses [D45] in year t − 1. t−1 ∑ Ri. general and administrative expenses MVt−1 at the end of year t − 1. where TA is total assets.(continued) Variable 10. where SGAE t − 1 is the selling. RD 11. CRDt−1 −CPDt−2 −CDt−1 .τ| is the return on stock i on day τ within 12 months preceding June of each year. ΔEQ Net cash flow received from external equity financing ΔCTt−1 + ΔCSt−1 . ATAt−2 ΔLTN is the change in long-term note. n respectively. 14. t−2 in year t − 1.t in NASDAQ and NYSE/AMEX.t = VOLDi. CPD is cash paid for reduction of debt [D114]. X. MVt − 1 is market capitalization of A-shares at the end of year t − 1 R&Dt−1 .τ = αi + ∑ βk .τ + k is the valueweighted CRSP market daily returns from 5 lags and 5 leads. and n is number of days available for 12 months preceding the end of June in each year.τ is daily stock return in the 12 months preceding June of each year.τ| is n the return on stock i on day τ within 12 months preceding June of each year. where CRE is cash received from the sale of ATAt−2 common and preferred stock [D108]. MVt − 1 is MVt−1 market capitalization at the end of year t − 1. ΔCS is the ATAt−2 change in capital surplus.

momentum trading and overreaction in asset markets. Journal of Finance 53. MacBeth.. Trading performance. Working paper. 1992.. C. Price momentum and trading volume. Yu.F. 1994. J. overconfidence. C..M.. M.W. W. Do investors overvalue firms with bloated balance sheets? Journal of Accounting and Economics 38. R... Ni... R. Banz.. Fama. K.J... X. 1823–1850.. D.F..A. 2006.A. Lakonishok. 2007.. 3–56. Basu. A unified theory of underreaction. Radcliffe. T. R.. Hou. G.. Lakonishok.. 2007. Hirshleifer... M. Barberis..S. J. Accounting Review 82 (1). Ritter. Nofsinger. Venkatachalam... Bradshaw. H. Fama. A. E. 1–7. M. J. Journal of Finance 47... R.. Beneish... Multifactor explanations of asset pricing anomalies.C. Nofsinger. R. French. Behavioral finance in Asia. Hong. Journal of Finance 52. Market underreaction to open market share repurchases. E. W. 1993. and risk. Journal of Finance 32. Common risk factors in the returns on stocks and bonds. 307–343.. 455–477. N. D. R. B. K. Vermaelen. 2008. 215–260. Billett. Naik. Wang.. Yuan.D.F.. F. Lakonishok. E. Liu. French.. 55–84. 175–192.D. Review of Accounting Studies 6.. The cross-section of expected stock returns. M. Lee. Xing. D. C. Contrarian and momentum strategies in the China stock market: 1993–2000. C. Han. 2004. Y... M. Liquidity and stock returns: an alternative test. Pincus... 1839–1886.. 116–137.. N. G. 305–351..G. Contrarian investment.. Eun.. Journal of Financial Economics 49. F. Yao. S. Does greater firm-specific return variation mean more or less informed stock pricing? Journal of Accounting Research 41. 2000. Investor psychology and security market under. Datar. Chen. N. R.. 2431–2456. 169–203.. J. and momentum.. / Pacific-Basin Finance Journal 18 (2010) 403–425 References Abarbanell. Journal of Financial Markets 1.. A. extrapolation. A. K. Fama. 2009. J. 427–465. Durnev. French. Prospect theory. Pacific-Basin Finance Journal 18.. Y. 1990. Bushee. B.. Journal of Finance 54. Journal of Finance 45. S. J. Feng. J.R.. Vishny. K. M.F. 1998. A.F... analysts' forecasts and stock returns. Accounting Review 73. The stock market valuation of research and development expenditures. 2003.. The accrual anomaly: international evidence. 2006. 1997. I. Lee.. Truong. 1998. Pacific-Basin Finance Journal 14. 2002.. Journal of Financial and Quantitative Analysis 44..424 X. 311–339. K. Shleifer. Market consequences of earnings management in response to security regulations in China. S.... E. Journal of Finance 61. Xu. Swaminathan.. Chen.. Journal of Financial and Qualitative Analysis 41. Rajgopal.L. The relation between corporate financing activities. J. B.M. 2007. 1609–1651.T. S. Jegadeesh.R. 2005.W. Chinese University of Hong Kong Jiang. Kim. Wong. Asset pricing in China's domestic stock markets: is there a logic? Pacific-Basin Finance Journal 15. Investment performance of common stocks in relation to their price-earnings ratios: a test of the efficient market hypothesis. Chan. 1217–1269. Yeung. J.. Individual investors and gender similarities in an emerging stock market. C. 53–58..A. The operating performance of firms conducting seasoned equity offerings. Earnings management and tunneling through related party transactions: Evidence from Chinese corporate groups... R.. Ng. Y. 243–265. Ikenberry. J. Sloan. Journal of Finance 48. 1996. Lee. Stein. Garfinkel. Amihud.. T. The information content of stock markets: why do emerging markets have synchronous stock price movements? Journal of Financial Economics 59. Morck. 1998. Journal of Financial Economics 78. Journal of Financial Economics 9. Daniel. 2008. Anomalous price behavior around repurchase tender offers.T. Gulen.. M.. Loughran. Journal of Accounting and Economics 42. T. Pacific-Basin Finance Journal 16.. 2005. Pacific-Basin Finance Journal 16. Journal of Finance 53. 19–45. B.. and experience of emerging market investors. Contemporary Accounting Research 22.W. Kang... 3–18. 2005. K. 1981. Journal of Political Economy 81. and equilibrium: empirical tests. 1–28. Zhang. Journal of Behavioral Decision Making 20. 2006... Feng. Journal of Finance 55. 1977. W. 2006. Seasholes. 1995.J. 1083–1124. 797–836. 2004. M.and over-reaction. T. Titman... 607–636.Wu. Do investor sophistication and trading experience eliminate behavioral biases in financial markets? Review of Finance 9. disposition effect. Rui.. Zarowin. S. Morck. H. The information content of idiosyncratic volatility.. Qi. Jian. Momentum strategies and stock returns: Chinese evidence.. Asset growth and the cross-section of stock returns. Kim. B. P. representativeness bias. M. Are bank loans special? Evidence on the post-announcement performance of bank borrowers.R.. Sougiannis. V. 203–219. 1975–1999. Hodrick. 2001. 452–480.. M. B.. D. Vermaelen. Pacific-Basin Finance Journal 10. J. Journal of Finance 49. 31–56. 1973. M.. 65–91. W.D. 2000. 2008.. The cross-section of volatility and expected returns. 259–299. S. Lakonishok. Chen et al. Journal of Finance 63 (4). 2001. 733–751. 95–140. Schill. . 1998.. Wu.. Shleifer. French.T. 1541–1578.J. Journal of Finance 56. 297–331. Analyzing the analysts: when do recommendations add value? Journal of Finance 59. T. and T.. 2004. Journal of Financial Markets 5. D... Returns to buying winners and selling losers: implications for stock market efficiency. Naughton. Contextual fundamental analysis through the prediction of extreme returns. K. Value versus growth: the international evidence.. Abnormal stock returns to a fundamental analysis strategy. E. 645–665. 663–682... Zhang. 425–451.J.. Lee.. 2010. Vishny. 1999. J. Hirshleifer. F.... Teoh. 2008. H. R. return. Risk. Richardson.R. Revealed stock preferences of individual investors: evidence from Chinese equity markets. Tarpley.. Wu. K... S. Journal of Finance 51. 1993. S. Li. O. M. M. Subrahamanym. Haw. Accounting Review 79.Y. Journal of Financial Economics 33. Grinblatt. mental accounting. Cooper. Kim.A. L. 2004. Seasholes.. Fama.. Yeung. T. N.J. 2002. 2143–2184. Fama. Journal of Financial Economics 39. K. Jegadeesh. 44–60... M.. J. Illiquidity and stock returns: cross-section and time-series effects.. The dynamics of individual and institutional trading on the Shanghai Stock Exchange.. The relationship between return and market value of common stocks. Krische. Huang. A. L. 476–492.. A model of investor sentiment. 181–208. 65–189. Ang.R. Pacific-Basin Finance Journal 16. Flannery. Veeraraghavan.. 1998... Earnings management and capital resource allocation: evidence from China's accounting-based regulation of rights issues.

Pacific-Basin Finance Journal 12. Wei. Pacific-Basin Finance Journal 12. The long-run performance of stock returns following debt offerings. S. Rouwenhorst. 2008..K. Do stock prices fully reflect information in accruals and cash flows about future earnings? Accounting Review 71. / Pacific-Basin Finance Journal 18 (2010) 403–425 425 Reinganum. 1991. 2004. The long-run performance of initial public offerings. Titman. Affleck-Graves. PacificBasin Finance Journal 16. J. Chin. 1998.. Shengzhen Shumway. Relative strength strategies in China's stock market: 1994–2000. Pacific-Basin Finance Journal 12. Profitability of return and volume-based investment strategies in China's stock market. 19–46. 65–77. Chiang. Shengzhen Stock Exchange. D.. Journal of Financial and Quantitative Analysis 39. Wang. Chen et al. 2004.X. Xu. 2005. Extreme volumes and expected stock returns: evidence from China's stock market. The regulation of stock trading and its enforcement in China. Journal of Financial Economics 54.. Roll. Capital investments and stock returns. R.. 289–315.. Tan. What determines Chinese stock returns? Financial Analysts Journal 60. 1996.. The major illegal behaviors in the Chinese stock market: case study on insider trading and price manipulation. 1999.. 2004. Does disposition drive momentum? Working Paper. R2. 2004. Wang. 577–597. Y. 1981. University of Michigan and University of Houston.. L.. F. Journal of Finance 43.G. Local return factors and turnover in emerging stock markets. Spiess. J. Wang. 267–284. C. University of Konstanz. C. 541–566. R. Wang. M. 2004. 2004. Misspecification of capital asset pricing: empirical anomalies based on earnings' yields and market values.. E. International momentum strategies... 61–77. Mason. 3–27.. S. 1999. Wu. K. Tingting. 45–73.. 1439–1464. J. Cheng.. and G. Nelling. Journal of Finance 54. Masters Thesis in Economics. 541–564. S.G. 677–700. Journal of Financial Economics 9. K. 159–177. T. N. F. Journal of Finance 46. C. . 1988. Germany. Journal of Finance 53.. Xie. J.. Sloan. Ritter.. Rouwenhorst.... 2006.. T. Herding behavior in Chinese stock markets: an examination of A and B shares.