WORKING PAPER SERIES

Transcription

1 College of Business Administration University of Rhode Island William A. Orme WORKING PAPER SERIES encouraging creative research A Tale of Two Markets: Stock Return Predictability in China and US Xuanjuan Chen, Tong Yao and Tong Yu 26/27 No. 13 This working paper series is intended to facilitate discussion and encourage the exchange of ideas. Inclusion here does not preclude publication elsewhere. It is the original work of the author(s) and subject to copyright regulations. Office of the Dean College of Business Administration Ballentine Hall 7 Lippitt Road Kingston, RI

2 A Tale of Two Markets: Stock Return Predictability in China and US Xuanjuan Chen, Tong Yao, and Tong Yu* Preliminary Draft June 26 (Please Do NOT Quote Without Authors Permission) * Chen is at University of North Carolina, Wilmington ( chenx@uncw.edu), Yao is at University of Arizona ( yaot@ .arizona.edu), and Yu is at University of Rhode Island ( tongyu@uri.edu). We thank Peng Xiong at SINOFIN for providing us with the Chinese market data. 1

3 A Tale of Two Markets: Stock Return Predictability in China and US Abstract This study compares the return-predictive performance of a large number of firm-specific variables in the US and Chinese markets. Despite substantial difference in stages of economic development as well as legal and regulatory environment, there exist similar patterns of stock return predictability in these two markets. Similar to the US evidence, size and B/P effects in stock returns exist in China and can be explained by the Fama- French three-factor model. In both markets, accruals, net operating assets, capital expenditures, and asset growth negatively predicts stock returns, whereas research and development expenditure, advertising costs and illiquidity positively predict returns. On the other hand, contrary to the US evidence, momentum, E/P, C/P, sales growth, turnover, external equity and debt financing, idiosyncratic return volatility, and trading volume do not have significant predictive power in China. Finally, results from multivariate regressions suggest that despite fewer return-predictive variables at work, the overall stock return predictability is stronger in the Chinese market. 2

4 A Tale of Two Markets: Stock Return Predictability in China and US I. Introduction A large number of empirical studies have documented that cross-sectional stock returns are predictable in the US equity market. The academic literature has also heatedly debated on two issues that are central to financial economics -- whether the patterns of return predictability are robust, and whether they can be interpreted as evidence of market inefficiency or be reconciled with rational asset pricing theories. To shed light on these issues, some researchers have looked at evidence outside the US market. For example, Fama and French (1998) find a pervasive value effect in international markets and show that it can be explained by a factor-based asset pricing model. Rouwenhourst (1998; 1999), Griffin, Martin, and Ji (23), and Chui, Titman, and Wei (23) find that stock return momentum exists in many developed markets, but is weaker or nonexistent in some Asian and emerging markets. Partly due to its late start, the stock market in China has been so far excluded from most of the studies on stock return predictability. The two official stock exchanges in Shanghai and in Shenzhen were set up in early 199s, and until recently a major part of the Chinese stock market is closed to international investors. Nevertheless, this market has grown substantially. By the end of 25, there are over 13 stocks listed in the two Chinese stock exchanges, with a total market capitalization of above $4 billion, representing the second largest equity market in Asia. Due to its rapid growth and its significance in the international financial arena, institutional characteristics and investor behavior in the Chinese market has recently received increased attention of researchers; 3

5 see, e.g., Sun and Tong (23), Mei, Scheinkman, and Xiong (25), Feng and Seasholes (25), and Seasholes and Wu (26). Besides the significant size and the large cross-section of stocks, several other features of the Chinese market make it an interesting out-of-sample test ground for stock return predictability documented in US and elsewhere. First, in this market, sophisticated institutional investors are still in a nascent form and stock trading is dominated by individual investors. 1 Typically in such a market there is a strong tendency for marginal investors to exhibit behavioral biases and misreact to information. Indeed, several studies have documented strong behavioral biases among investors in this market, such as overconfidence, disposition effect, representativeness bias, and herding; see, e.g., Chen, Kim, Nofsinger, and Rui (24), Feng and Seasholes (25), and Shumway and Wu (26). Second, short-sales are not allowed in China, making it difficult for mispricing to be quickly arbitraged away. While the above two features suggest that return predictability would possibly be stronger in the Chinese market, there are also factors that may work to the opposite effect. For example, earnings management and accounting manipulation have been prevalent (e.g., Chen and Yuan, 24; Jian and Wong, 24; and Haw, Qi, Wu and Wu, 25), making accounting information noisy and unreliable. There is also evidence of rampant market manipulations (e.g., He, 1998, Shenzhen Stock Exchange Report, 25), which may cause the properties of stock returns and especially patterns of predictability to be quite different from the more mature market in US. A further factor that makes the Chinese market an interesting case of study is its two-tier shares system, where foreign investors are limited to a small segment of the market (B shares) and the 1 For example, in analyzing a comprehensive dataset of all trades on the two stock exchanges for the period from April 21 to April 22, Wu (25) report that 99.5% of trading accounts are labeled as individuals. Institutional investors such as mutual funds emerged only recently. 4

6 majority of shares listed at the two stock exchanges (A-shares) are accessible only to domestic investors. 2 Thus, the A-shares market -- which is the focus of our study and is hereafter refereed to as the Chinese market -- is relatively insulated from foreign capital flows and the spillover of international investor sentiment. If there exist patterns of stock return predictability similar to those documented in the US, they must be due to either similar risk-return trade-offs or commonality in local investor behavior across these two markets. On the other hand, different return predictability patterns between US and China would suggest further research to explore the institutional or behavioral differences in the two markets. In this paper, we explore patterns of cross-sectional stock return predictability in the Chinese stock market and compare them with those documented for the US market. To offer a comprehensive comparison between the two markets, we select a list of 18 firm-specific variables, which are referred to as stock return predictors. For the US market, these variables are known to be predictive of cross-sectional stock returns at the annual horizon, and can be constructed using data from CRSP and Compustat. They include: 1) log market capitalization (SIZE), 2) book-to-market ratio (B/P), 3) momentum or past 12-month stock returns (MOM), 4) earnings-to-price ratio (E/P), 5) cash flow-to-price ratio (C/P), 6) sales growth (SG), 7) accounting accruals (ACC), 8) net operating assets (NOA), 9) capital expenditure (CAPEX), 1) R&D expenditure (RD), 11) advertising expenditure (ADV), 12) asset growth (AG), 13) equity financing ( EQ), 14) debt financing ( DT), 15) dividend growth (DG), 16) idiosyncratic return volatility 2 See also Bailey (1994), and Mei, Scheinkman, and Xiong (25). The segmentation of A and B share markets have been reduced in recent years. Starting from February 21, domestic investors with foreign currency holdings can trade on B shares, and a small group of approved foreign investors (qualified foreign institutional investors, or QFII) are allowed to trade A shares with a limited amount starting from December 22. But substantial QFII investments in the A-shares market did not materialize until 25. 5

7 (STDR), 17) trading turnover (TURN), and 18) the illiquidity ratio of Amihud (22) (ILLIQ). Among the 18 variables, size, book-to-market ratio, and momentum are most extensively scrutinized in the US literature. B/P, E/P, C/P, and SG are often known as value indicators. ACC and NOA are inversely related the quality of accounting information. CAPEX, RD, ADV are measures of firms tangible and intangible investments. AG summarizes the growth of firm size, whereas EQ and DT reflect firms external financing activities. DG is regarded as signaling firms private information about future cash flows. STDR, TURN, and ILLIQ are variables related to information uncertainty and/or liquidity of stocks. Related academic studies are reviewed in Section II of the paper. Some of these return-predictive variables are only recently documented in the literature, to the effect that they have not been examined jointly even for the US market. To our knowledge, ours is the first study to systematically characterize stock return predictability in the Chinese market. Related to this paper are three existing studies that report evidence on four individual anomalies. Kang, Liu, and Ni (22) and Wang (24) examine the profitability of momentum strategies. Due to different sample selection criteria, these two studies reach opposite conclusions on the existence of the momentum effect. 3 Wang (24) additionally finds that small stocks subsequently earn higher returns, but the return-predictive power of book-to-market ratio is not significant. Finally, Wang and Chin (24) report that stocks with higher trading volume tend to have lower future returns, but the significance of return difference depends on past returns. Thus, the return-predictive performance of a majority of variables we analyze has not 3 In examining the relation between momentum profits and macroeconomic risks across a large number of markets, Griffin, Martin, and Ji (23) also find that momentum strategies are unprofitable in China. 6

8 been documented in existing studies. Further, the sample periods examined by these three existing studies all end in 2. Using data that covers all A-shares stocks traded in the two Chinese stock exchanges up to the end of 24, we offer updated results on the momentum, size, B/M, and trading volume anomalies. Some of our findings are different from previous studies. For each stock return predictor, we offer side-by-side comparison of its performance in the US and Chinese markets. For the US market, during the sample period from 1974 to 24, our findings are consistent with the existing literature. That is, MOM, B/P, E/P, C/P, RD, ADV, DG, ILLIQ positively predict future one-year returns, whereas SIZE, SG, ACC, NOA, CAPEX, AG, EQ, DT, STDR, and TURN negatively predict future one-year returns. For the Chinese market, the sample period is from 1994 to 24. Start with the two most prominent stock return predictors SIZE and book-to-market ratio (B/P). Similar to the US evidence, size negatively predicts returns and book-to-market ratio positively predicts returns. Using the sorted-portfolio approach, we find that the equalweighted portfolio with stocks in the lowest size decile outperforms the equal-weighted portfolio with stocks in the highest size decile by 2.45% per year (t=1.93). The portfolio with stocks in the highest B/P decile outperforms the portfolio with stocks in the lowest B/P decile by 11.56% per year (t=2.15). Our finding on the existence of the size effect is consistent with Wang (24). However, different from Wang (24), we find a significant B/P effect. The difference is found to be due to the inclusion of four years of recent data in our study. 7

9 Also similar to the US evidence, earnings quality measures negatively predict stock returns in the Chinese market. The annual return spread between the highest and lowest accruals (ACC) deciles is -7.59% (t=-2.5) and that between the highest and lowest net operating assets (NOA) deciles is -5.83% (t=-1.95). In addition, Capital expenditures (CAPEX), asset growth (AG), idiosyncratic risk (STDR) and trading turnover (TURN) negatively predict returns, with annual return spread between the highest and lowest deciles being -6.25% (t=-2.9) for CAPEX, -1.94% (t=-1.69) for AG, -9.7% (t=-1.8) for STDR, and -7.7(-1.72) for TURN. Further, R&D expenditure (RD), advertising costs (ADV), and illiquidity (ILLIQ) positively predict future returns in the Chinese market, also consistent with the US evidence. The annual return spread between the top and bottom deciles is 11.79% (t=4.55) for RD, 4.33% (t=2.33) for ADV, and 12.63% (t=1.7) for ILLIQ. There are several discrepancies between the two markets. First, momentum is nonexistent in the Chinese market. Stocks in the decile of highest prior 1-year returns actually underperforms Stocks in the lowest momentum decile, by a statistically insignificant margin of 11.87% per year (t=.58). The lack of the return persistence is consistent with the finding of Wang (24). Second, potentially due to the less reliable financial information reported by Chinese firms, several conventional value indicators, including earning-price ratio (E/P), cash flow-price ratio (C/P), and sales growth (SG), have no return predictive power. Third, equity and debt financing ( EQ and DT), as well as dividend growth (DG), do not predict stock returns in the Chinese market. To compare the time-series variations in stock return predictability, for each variable we compute the correlation of annual return spreads across the top and bottom 8

10 deciles between the US and the Chinese markets. The correlations are low and insignificant for most variables. That is, local factors are important in explaining the stock return predictability in the two markets. Further, we examine whether the return-predictive performance of these variables can be explained by a risk-based three-factor model of Fama and French (1996, 1998). For the US market, consistent with Fama and French (1996), we find that the three-factor alphas for the top-bottom decile return spreads sorted on SIZE, B/P, E/P, C/P, and SG are statistically insignificant, suggesting that the return-predictive power of these variables can be explained by the risk premium on the market, size and book-to-market factors. The remaining variables exhibit significant return-predictive power under the three-factor model, with somewhat weaker results for CAPEX, ADV, TURN, and ILLIQ. In the Chinese market, we find that a similarly constructed three-factor model can explain the return-predictive power of SIZE, B/P, ADV, STDR, TURN, and ILLIQ. The three-factor alphas for the return spreads based on ACC, NOA, CAPEX, RD, and AG remain significant. We also use the Fama-MacBeth cross-sectional regression approach to evaluate the predictive performance of these 18 variables jointly. The adjusted R-squares from multivariate regressions provide a way to measure the overall predictability of stock returns in a market. As it turns out, despite fewer predictive variables at work, the average of adjusted R-squares from the multivariate regressions in the Chinese market is.12, higher than that for the US market,.7. We obtain similar adjusted R-squares in both markets when using the first 6 principal components of the 18 variables as 9

11 explanatory variables, suggesting that the results are robust to the correlation among these regressors. To sum up, the return predictors share similar characteristics in the two markets despite that the US and Chinese markets are distinct in their institutional backgrounds, the efficacy of corporate governance, and stages of economic development. We find SIZE, B/P, earnings quality measures, capital expenditure, asset growth, and information uncertainty measures (STDR, TURN and ILLIQ) predict subsequently stock returns in China similarly as they do in U.S. The Fama-French three-factor model explains the return-predictive power of SIZE or B/P, consistent with Fama and French (1996) for the US evidence and Fama and French (1998) for international evidence. In addition, the three-factor model explains the return predictability of information uncertainty measures in the Chinese market. Nevertheless, contrary to the US evidence, we do not find evidence on the return predictability of momentum (MOM), several value indicators (E/P, C/P, and SG), or predictors related to external financing and dividend-paying activities ( EQ, DT, and DG). We note that while we attempt to provide systematic analysis of stock return predictability in the Chinese market, the vast body of related literature in the US and other markets makes it impossible for any study to be all-exhaustive in the inclusion of return-predictive variables. For this study, we have left out some known return-predictive variables due to data availability, such as those based on brokerage firm analysts earnings forecasts (e.g., analyst forecast revision, forecast dispersion, and long-term earnings growth forecasts), and those based on transactions or quotes (e.g., trade-based measures of liquidity, and the PIN measure of Easley, Hvidkjaer, and O'Hara (22) on 1

12 private information in trading). We have also left out a few variables that predict returns other than at the annual horizon, such as those related to short-term return reversal (e.g., past daily, weekly, and monthly returns), those related to long-term return reversal (e.g., returns for the past two to five years), and those related to short-term earnings momentum (e.g., quarterly earnings surprise). In addition, carve-outs, spin-offs, and stock repurchase has yet to take place in China and therefore variables related to such corporate events are not included in our study. Finally, we are aware of several firm-specific variables with return-predictive power in the Chinese market but not in the US market. They are worth exploring in future research. We expect this study to be of direct interest to investment practitioners in addition to academic researchers. In well-developed markets such as the US, both fundamental and quantitative approaches have been developed for making stock selection decisions. Stock return predictability documented in academic studies has formed the foundation of the quantitative stock selection approach. 4 The quantitative approach is particularly important for investing in international equity markets, as obtaining quantitative information is of much lower cost relative to global trotting called for by the fundamental approach. In China, in spite of the recent emergence of the asset management industry, it remains an open question whether any stock selection approach is effective. The evidence of our study suggests the possibility of using publicly available information to make stock selection decisions in the Chinese market, especially in light of the high R-squares from multivariate regressions. It should also be pointed out that our study only represents 4 For example, Zhao (26) documents an increasing popularity of the quantitative stock selection approach among US equity fund managers. She reports that the group of pure quantitative equity funds has gained substantial market shares in terms of total assets under management during the period from 1994 to 24. Even fundamental fund managers now often use quantitative stock screening to narrow down their investment universe before conducting in-depth fundamental analysis. 11

13 the first step in exploring the validity of the quantitative stock selection approach in the Chinese market. Issues such as whether accounting information is reliable and can be timely acquired, how various correlated quantitative signals can be optimally combined, and whether the investment strategies are profitable after transaction costs, are important for future research. The remainder of the paper is organized as the following. Section II provides a review of the existing literature on stock return predictability. Section III discusses data and methodology for constructing stock return predictors for both the US and Chinese markets. Section IV provides empirical results. Section IV concludes. II Related Literature Classical asset pricing theories such as CAPM and APT suggest that only exposure to systematic risks affects cross-sectional difference in expected stock returns. Therefore, patterns of stock return predictability not explained by systematic risks are considered market anomalies. In the following, we summarize the literature on market anomalies in the US market and the corresponding behavioral explanations, if any. We then provide a discussion on alternative, rational, explanations of stock return predictability. Finally, we review the Chinese evidence. For the US market, the size effect and the value effect are among the earliest market anomalies documented by academic studies. Banz (1981) and Reinganum (1981), among others, show that small firms tend to earn higher returns than large firms. Basu (1977) and a number of subsequent studies show that firms with higher earnings-to-price ratio tend to earn higher returns. Similarly, Fama and French (1992) show that firms with 12

14 higher book-to-market ratio subsequently have higher returns. Additionally, Lakonishok, Shleifer, and Vishny (1994) report that future returns are positively correlated with the cash flow-to-price ratio and negatively correlated with past sales growth. They propose that the relation of these variables with future returns is due to the extrapolation bias, e.g., investor overvaluation of past performance. Therefore, these variables are often jointly regarded as value or contrarian indicators. Jegadeesh and Titman (1993) show that stocks with higher past returns subsequently earn higher returns, at horizons ranging from 3 to 12 months. They refer to investment strategies of buying past winners and selling past losers as relative strength strategies, which are widely known as momentum strategies in subsequent studies. As shown by Fama and French (1996), the momentum effect is not subsumed by the size or value effects. Several behavioral models have been proposed to explain the momentum effect, based on investor underreaction or overreaction to information (Barberis, Shleifer, and Vishny, 1998; Daniel, Hirshleifer, and Subrahmanyam, 1998; and Hong and Stein, 1999) or irrational investor preferences (Grinblatt and Han 25). Sloan (1996) documents the accruals anomaly that stocks with higher accounting accruals tend to have lower future returns. His explanation for the anomaly is investors misreaction to information -- although the accrual component of earnings is less persistent than the cash flow component, naïve investors fail to recognize such a difference when valuing stocks. Hirshleifer, Hou, Teoh, and Zhang (24) find that firms with higher net operating assets (NOA) have lower future returns, suggesting that naïve investors under-appreciate the decreasing returns to assets scale. These two anomalies are somewhat related as net operating assets is the accumulation over time of the difference 13

15 between net operating income and free cash flows, plus capitalized investments. But Hirshleifer et al. show that the accruals and NOA do not subsume each other when used jointly to predict returns. These two effects are referred to as earnings quality anomalies. Titman, Wei, and Xie (24), and Beneish, Lee, and Tarpley (21) find that firms with high capital investments subsequently have low returns. Titman et al. attribute this to the overinvestment tendency of corporate managers and investor underreaction to information. In contrast to evidence on tangible capital investments, Chan, Lakonishok and Sougiannis (21) find that the ratios of corporate R&D spending and advertising spending to market capitalization are positively correlated with future returns. 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 have mis-reacted to this accounting effect. The effects of capital expenditure, R&D, and advertising are jointly referred to as investment anomalies. There is also a payout anomaly. For example, when firms announce share repurchase, they earn positive long-term abnormal returns (Lonkonishok and Vermaelen 199; Ikenberry, and Vermaelen 1995). In addition, Asquith and Mullins (1983), Healy and Palepu (1988), and Michaely, Thaler and Womack (1995) find that dividend initiation leads to positive market response while stocks with dividend omissions have a negative price reaction. Investor misreaction to corporate signals is a typical behavioral explanation for this anomaly. The payout anomaly can be viewed as part of the more general anomaly related to external financing. A large number of corporate event studies show that future stock 14

16 returns are unusually low in the years following initial public offerings (Ritter 1991), seasoned equity offerings (Loughran and Ritter 1997), debt offerings (Spiess and Affleck-Graves 1999), and bank borrowings (Billett, Flannery, and Garfinkel 21). Conversely, future stock returns are unusually high following stock repurchases (Lakonishok and Vermaelen 199; Ikenberry, Lakonishok, and Vermaelen 1995). Bradshaw, Sloan and Richardson (26) summarize firms external financing activities into two variables: equity financing and debt financing. They report that these two variables are negatively correlated with future stock returns, and attribute this pattern to investor optimism and firms efforts to time the market in raising financing. Related to the investment and financing anomalies is the asset growth anomaly documented by Cooper, Gulen and Schill (25). They find a strong inverse relation between firm asset growth and future stock returns. Note that accruals, NOA, capital investments, payouts, external financing, and asset growth are obviously correlated. In this paper, we do not take a stance on which variable or combination of variables has the highest return-predictive power. Finally, information uncertainty and stock liquidity are also related to future returns. Ang, Hodrick, Xing and Zhang (25) show that stocks with high idiosyncratic volatility risk have low subsequent returns. Datar, Naik and Radcliffe (1998) and Lee and Swaminathan (2) show that stocks with high trading volume earn lower future returns. Giang, Yao, and Xu (26) link both anomalies to adverse selection in corporate disclosure and investor underreaction. In addition, Amihud (22), among a number of studies, finds that stock liquidity inversely predicts stock returns. 15

17 Just like the existence of behavioral explanations for most anomalies, there is at least a rational asset pricing explanation for most anomalies. For example, Banz (1981) and Reinganum (1981) point out that the size anomaly may be due to misspecification of the CAPM and omitted risk factors. Fama and French (1996) propose a three-factor model to explain the size and value anomalies, and point out that the size and boo-tomarket factors they construct likely capture the economy-wide distress risks. Rational asset pricing models such as Gomes, Kogan, and Zhang (23) and Zhang (26) have also linked the size and value effects to time-varying risks and time-varying expected returns. Models of time-varying risks and expected returns have also been used to explain the momentum effect; see, e.g., Berk, Green, and Naik (1999) and Johnson (22). Momentum is also shown to be associated with dynamic systematic risk factors by Yao (26). As for the accruals anomaly, Khan (26) report that it can be reconciled with a factor model that combines the two factors of Campbell and Vuolteenaho (24) with the Fama-French size and book-to-market factors SMB and HML. Berk, Green, and Naik (1999), and Carlson, Fisher, and Giammarino (24) attribute the investment anomaly to time-varying risks and expected returns. Similar explanations can be applied to the asset growth anomaly. Zhang (26) recently proposes a production-based rational asset pricing model, which can be used to explain a large number of anomalies, such as value, investment, financing, payout, profitability and expected profitability, and post-earnings announcement drift. He shows that under certain conditions return on a firm s investments is equal to the stock return (which can be viewed as a special case of positive correlation between accounting profitability and expected stock return under convex 16

18 capital adjustment costs). This result, plus the assumption of decreasing return to scale in the production function, leads to a negative correlation between capital investments and expected returns. Financing activities, corporate payout activities, as well as asset growth can then be inversely linked to expected returns through their positive correlation with investments. Finally, the relation between illiquidity and future returns has been linked to the illiquidity premium. That is, investors require extra compensation in expected returns for holding illiquid securities (Amihud and Mendelson 1986; Amihud 22). There is also a vast body of literature on stock return predictability in markets outside the US, making it all impossible to provide a complete review with finite space. In contrast, the evidence on the Chinese market is sparse, which to our knowledge consists of the following studies. Using data on 268 Chinese A-share stocks for the period from 1994 to 2, Kang, Liu, and Ni (22) examine the profitability of contrarian and momentum strategies with portfolio formation and holding periods ranging from 1 to 26 weeks. They report significant short-term contrarian profits and intermediate momentum profits for their data sample. A different conclusion is reached by Wang (24), who considers all A-shares traded in Shanghai and Shenzhen stock exchanges for a similar sample period of 1994 to 2. He finds positive but insignificant profits for the 3-month momentum strategy and negative but insignificant profits for the momentum strategies at horizons ranging from 6 to 24 months. In examining the relation between momentum profits and macroeconomic risks across a large number of markets, Griffin, Martin, and Ji (23) also find that momentum strategies are unprofitable in China. In addition, Wang (24) 17

19 finds that small stocks subsequently earn higher returns, but the return-predictive power of book-to-market ratio is not significant during the period from 1994 to 2. Finally, Wang and Chin (24) report that stocks with higher trading volume tend to have lower future returns, but the significance of return difference depends on past returns. III. Data and Methodology A. Data for the US Market For the US market, information on stock price, return, and trading volume is from CRSP. Information on corporate financial statements is from Compustat. The stock sample is selected in the following way. First, from 1974 to 24, at the end of June in each year t, we select all common stocks traded in NYSE, AMEX, and NASDAQ. We eliminate firms that are primes, close-end funds, a real estate investment trust (REIT), an American Depository Receipt (ADR), or foreign companies. To avoid market microstructure related issues in measuring returns, we require stocks to have a minimum price of $5 at the end of June. Further, a stock must have available information on stock price, market capitalization, and at least one valid stock return predictor (detailed in Section III.D) in addition to size at the end of June. Future stock returns are measured for the period from July of year t to June of year t+1. In this way, we have 124,573 firm-year observations in the US sample. B. Data for the Chinese Market Information on stock price, return, trading volume, as well as corporate financial statement information is obtained from SinoFin, a Chinese financial data vendor. Our 18

20 sample covers all A-share stocks traded in Shanghai and Shenzhen stock exchanges for the period between 1994 and 24. In the Chinese market, tradable shares on domestic stock exchanges are classified into A- and B-shares. Tradable A-shares are ordinary shares available exclusively to Chinese citizens and institutions. B-shares were originally designated for overseas investors until February 21, after which they become available to domestic investors as well. The A-shares market has grown substantially in size during our sample period while the growth in B-shares market is limited (see Panel B of Table 1). A recent regulatory change has made A-shares available to a small group of qualified foreign institutional investors. In this study, we only include A-shares for our analysis, which account for more than 9 percent of tradable market value by the end of 24. Similar to the US sample, we eliminate firms that are banks, close-end funds, real estate, and investment companies. Further, to be included in our sample for year t, a stock must have available information on stock price, market capitalization, and at least one valid stock return predictor in addition to size at the end of June of year t. The minimum price grid in the two Chinese stock exchanges is RMB one cent. A large number of Chinese stocks trade at price between RMB 1 to 5, and only a small proportion of stocks have price below 1. To avoid market microstructure related issues in measuring returns, we require that a stock must have end-of-june price of no less than 1 to be included in our sample for year t. In this way, we have 8,546 firm-year observations in our China sample. C. Summary Statistics 19

21 Table 1 presents some summary statistics for both the US and Chinese markets. Focusing on the Chinese market, in 1994, there are altogether 287 stocks traded in the Shanghai Stock Exchange (SHSE) and the Shenzhen Stock Exchange (SZSE). In 24, the number of stocks is increased to By contrast, the number of stocks traded in the US market decreases from 538 in 1973 to 681 in 24. In China, stocks of a typical firm may consist of state shares (owned by the central or local governments), legal-entity shares (held by domestic legal entities such as listed companies, state owned enterprises and banks), and tradable shares, with the restriction that state and legal-entity shares cannot be traded publicly. Tradable shares are further classified into tradable A- and B- share classes. Tradable A-shares are ordinary shares available exclusively to Chinese citizens and institutions. B-shares were designated for overseas investors prior to opening the market to domestic investors in February 21. In 24, the aggregate market capitalization of tradable A shares is RMB 1,48.96 billion, less than a third of the aggregate market capitalization of RMB 3,71.16 billions. China stocks are more frequently traded by investors. In 1994, the average annual turnover ratio, defined as the ratio of trading volume over the firm s market capitalization, is 924% in SHSE, 38 times of the ratio in NYSE and AMEX. The turnover in the Chinese market is stabilized in 24. However, the ratio in SHSE still doubles that for NYSE/AMEX. An abnormally high turnover ratio typically indicates a low level of investor sophistication. Further, Table 1 also suggests that stocks traded in the Chinese market are much smaller than US stocks. The average price for China stocks in 24 is RMB5.81, equivalent to USD.72, relative to USD for an average US stock. 2

22 D. Constructing Stock Return Predictors The 18 stock return predictors are constructed in the following way. More detailed information can be found in the Appendix. 1. Firm Size (SIZE) SIZE is the natural log of the product of a firm s market price and common shares outstanding at the end of June of year t. For China firms only the outstanding A-shares are used in computation. 2. B/P Ratio (B/P) B/P is the book value of equity of a firm for the fiscal year ending in calendar year t-1 over the market capitalization at the end of year t Price Momentum (MOM) MOM is the cumulative return of a stock in month -12 through -1 preceding June of year t. We skip one month between portfolio formation and holding to avoid the effects of bid-ask spread, price pressure, and lagged reaction (Jagadeesh and Titman, 1993). 4. E/P Ratio (E/P) E/P is earnings for the fiscal year ending in calendar year t-1 over the market capitalization at the end of year t-1. The earnings measure for US firms is earnings before extraordinary item and depreciation, whereas it is net income for China firms. 5. C/P Ratio (C/P) C/P for US firms is the sum of earnings before extraordinary item and depreciation in the fiscal year ending in year t-1 over the market capitalization at the end of year t-1. This definition follows Fama and French (1993). For Chinese firms, similarly 21

23 C/P is the sum of net income and depreciation in the fiscal year ending in year t-1 over its market capitalization at the end of year t Sales Growth (SG) SG is the sales revenue for the fiscal year ending in calendar year t-1 over the sales revenue in the fiscal year ending in calendar year t Accruals (ACC) Accounting accruals is the non-cash component of earnings. Following Sloan (1996), we estimate ACC as the change in non-cash current assets less change in current liabilities and less depreciation, during the fiscal year ending in year t-1, scaled by the average total assets at the beginning and end of that fiscal year. 8. Net Operating Assets (NOA) Following Hirshleifer, Hou, Teoh, and Zhang (24), we estimate NOA as the difference between operating assets and operating liabilities for the fiscal year ending in calendar year t-1, scaled by the average total assets at the beginning and end of that fiscal year. 9. Capital Expenditure (CAPEX) Following Jegadeesh et al. (24), CAPEX for US firms is the capital expenditure of a firm for the fiscal year ending in calendar year t-1 over the average total assets at the beginning and ending of that fiscal year. 1. R&D Expenses (RD) Following Chan, Lakonishok and Sougiannis (21), RD for US firms is the ratio of research and development expenses for the fiscal year ending in calendar year t-1 over market capitalization at the end of year t-1. In China, firms do not provide separate 22

24 information on R&D, but include this item in management expenses. As a result, RD for Chinese firms is 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 Advertising Costs (ADV) Following Chan, Lakonishok and Sougiannis (21), ADV for US firms is the advertising expenses for the fiscal year ending in calendar year t-1 to market capitalization at the end of year t-1. Since advertising costs are not separately reported but included in sales and marketing expenses in China, ADV for Chinese firms is 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 Assets Growth (AG) Following Cooper, Gulen and Schill (25), AG is 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 External Equity Financing ( EQ) Following Bradshaw, Sloan and Richardson (26), EQ 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 External Debt Financing ( DT) Following Bradshaw, Sloan and Richardson (26), DT is the net cash received from the issuance (and/or reduction) of debt for the fiscal year ending in calendar year t Dividend Growth (DG) 23

25 DG is dividend per share from July year t-1 to June year t over dividend payment in the prior 12 months. If a firm pays dividend during July of year t-1 and June of year t but the dividend over prior 12 month is zero, we assign the firm into the top decile portfolio. Alternatively, if dividends are for two consecutive years, we set DG to be missing. 16. Idiosyncratic Risk (STDR) Similar to Ang, Hodrick, Xing, and Zhang (25), STDR is the standard deviation of the residuals in the regression of daily stock return on daily CRSP value-weighted market return with 5 lags and 5 leads, for the period from month -12 through month -1 preceding June of year t. 16. Trading Turnover (TURN) Follow Jegadeesh, Kim, Krische, and Lee (24), TURN is the percentile rank of the average daily volume turnover in the twelve months preceding June of year t, 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. Since the trading volume is counted differently in NASDAQ than in NYSE and AMEX, the percentile ranking is performed separately for NASDAQ and for NYSE-AMEX. 17. Illiquidity (ILLIQ) Following Amihud (22), ILLIQ is the percentile ranking of the average of daily ratio of absolute stock return to its dollar volume, across month -12 through month -1 preceding June of year t. Again, the percentile ranking is performed separately for NASDAQ and for NYSE-AMEX. 24

26 Panel A and B of Table 2 presents summary statistics for each of the above stock return predictors in US and Chinese markets, except for TURN and ILLIQ since they are rank variables. We report cross-sectional characteristics for 25%, mean, median, 75%, standard deviation, skewness and kurtosis. Several patterns are interesting. The mean and median B/P in China (114.35% and 96.49%) are greater than their counterparts in the U.S (5.95% and 45.56%). In most cases, the return predictors are consistent in these two markets. The magnitudes of the other value indicators (E/P, C/P, and SG) are similar. The mean percentage of capital expenditure in total assets in China (6.64%) is slightly higher than the ratio in the US (4.82%). In Panel C and D, we report the time-series means of the cross-sectional correlations among stock return predictors. Similar correlation patterns can be found in both markets. IV. Empirical Results A. Sorted Portfolios We first examine the return-predictive performance of the 18 variables using sorted decile portfolios. In June of each year t, we rank stocks into deciles based on each of the 18 variables. We form equal-weighted portfolios in each decile, and hold the positions from July of year t to June of t+1. We then compute the time-series averages of annual returns for all decile portfolios as well as for the return spreads between the top and bottom decile portfolios. To eliminate the impact of the abnormal high market returns in the Chinese market during 1996 and 1997, we exclude the period from July 1996 to 25

27 June 1997 from our analysis on the Chinese market. 5 The result for the US market is reported in Panel A of Table 3 and the results for the Chinese market is in Panel B of Table 3. In Column (1) to (3), we look at the three most prominent effects in the US literature -- SIZE, B/P and MOM. In the US market, stocks in the top SIZE decile (D1) underperform those in the bottom SIZE decile (D1) by 8.83%, with a t statistic of This result confirms the size effect reported in the prior studies (Banz, 1981; Basu 1983). We find a similar size effect in the Chinese market, where D1 stocks underperform D1 stocks by 2.45% (t= -1.93). These two markets are also consistent in the effect of the B/P ratio, i.e., high B/P ratio is associated with high subsequent stock returns. In the Chinese (US) market, D1 stocks outperform D1 stocks by 11.56% (13.53%), both significant at the 1% level. Interestingly, we do not find a momentum effect in China, while we confirm the momentum effect in US. In the US market, D1 MOM stocks outperform D1 MOM stocks by 9.6% (t=2.52). In sharp contrast, D1 stocks underperform D1 stocks by 11.87% (t=-.86) in the Chinese market. That is, the Chinese market actually has a contrarian effect at the 12-month horizon, albeit statistically insignificant. Our evidence is consistent with Wang (24), who finds that Chinese stock returns exhibit reversal at investment horizons longer than 6 months. The existence of size and value effects and the lack of the momentum effect make the Chinese market quite similar to the Japanese market (e.g., Chan, Hamao, and Lakonishok, 1991; Chui, Titman, and Wei, 23). 5 From July 1996 to June 1997, the averaged annual return weighted by firm market capitalization is 83%, relative to an average of 3.17% in other years. When we include this period, the t statistics for the return spreads between D1 and D1 stocks sorted by most return predictors (except for SIZE and B/P) are no longer significant. 26

28 In the next three columns, we examine the effects of three additional value indicators, E/P, C/P, and SG. High E/P, high C/P, or low SG stocks are considered to be value stocks and have higher future stock returns. Our results in the US market confirm the findings in prior studies: the D1 E/P portfolio significantly outperforms the D1 E/P portfolio by 9.26%, the D1 C/P portfolio significantly outperforms the D1 C/P portfolio by 13.49%, and the D1 SG portfolio significantly underperforms the D1 SG portfolio by 7.71%. Contrary to the US evidence, we do not find the E/P, C/P, or SG measures predict future stock return in the Chinese market. The return spread between D1 stocks and D1 stocks is 2.97% per year (t=.29); the return spread between D1 C/P stocks and D1 stocks is 4.32% (t=.77); the return spread between D1 SG stocks and D1 stocks is 4.99% per year (t= 1.15). As discussed in the introduction, earnings management is prevalent in the Chinese market. The lack of predictive power of these value indicators is potentially attributable to manipulated accounting data on sales, earnings, and cash flow. In unreported tests, we find that operating performance of Chinese firms is less persistent than the US firms. For example, the autocorrelation of ROA in the Chinese market during the sample period is.27, while that of US firms is.57. We then examine if variables related to earnings quality, accruals and NOA, have return predictive power. Consistent with Sloan (1996) and Hirshleifer et al. (24), these two measures are significantly inversely related to stock returns in the US. Likewise, accruals and NOA are negatively associated with stock returns in China. D1 ACC stocks underperform D1 stocks by 7.59% per year (t=-2.5) and D1 NOA stocks underperform D1 stocks by 5.83% (t= -1.85) per year. 27

29 In Columns (9) through (12), we analyze the effect of investment activities on future stock returns. Consistent with prior studies for the US market, we find that CAPEX and AG are inversely related to stock returns while RD and ADV are positively related to stocks returns. For the Chinese market, the return spread for CAPEX is -6.25% (t= -1.91) and -1.94% (t= -1.69) for AG. RD and ADV appear to have significantly positive return predictive power. For RD-sorted portfolios, D1 stocks outperform D1 stocks by 11.79% per year (t= 4.55). For ADV-sorted portfolios, D1 stocks outperform D1 stocks by 4.33% per year (t=2.33). The very similar pattern of return predictability of these investment activity measures between U.S market and the Chinese market is somewhat surprising because the two markets are considered to be strikingly different in many aspects, including investor groups, stages of economic development, political and regulatory environments. As reported as columns (13) and (14) of Panel A, the US results confirm the findings in Bradshaw, Sloan and Richardson (26) that external financing predicts lower future stock returns. The average return for the top EQ decile stocks trails that of the bottom EQ decile stocks by 11.39% (t=-3.25) and the average return for the top EQ decile stocks trails that of the bottom EQ decile stocks by 7.46% (t=-3.25). However, as shown in Panel B, the inverse relation between external equity financing and stock returns no longer holds in for the Chinese market. One notable feature of the Chinese market is that firms external financing activities are still heavily influenced by government policies and sometimes directly regulated by the government. This may reduce the asymmetric information component in firms financing activities. Further, Pilotte (1992), Choe, Masulis and Nanda (1993), Bayless and Chaplinsky (1996) and 28

30 Jung, Kim, and Stulz (1996) suggest that favorable firm growth or economic conditions help alleviate the information asymmetry problems. It is likely that the substantial economic growth in China reduces the information asymmetry of firms conducting external financing, making these activities less likely to contain adverse information about firm value. In Column (15), we report the average portfolio returns across dividend growth (DG) deciles. In US, the return spread between the top and bottom deciles is 7.55% (t=2.55), significant at the one percent level. In China, the return spread is -1.5%, insignificant with a t-statistic of -.9. Therefore, similar to external financing activities, corporate payout activities seem to be not significantly related to asymmetric information in China. One prevailing explanation of the positive return predictive power of DG is that dividend payments reduce firms free cash flows and therefore reduce the agency problems of corporate managers (Easterbrook, 1984; Jensen, 1986). However, the Chinese stock market is characterized by the presence of dominant nontradable state and legal entity ownership, which are entitled to the same claim to dividends as tradable shares. Lee and Xiao (24) suggest that the dividend payment practice in China could be a means for non-tradable shareholders to appropriate wealth from tradable shareholders. The negative appropriation effect may offset the positive effect in alleviating the agency problem of managers, thus we observe an insignificant return predictability of DG in the Chinese market. Finally, our results in the last three columns of Panel A show that in the US, STDR and TURN negatively predict future returns while ILLIQ positively predicts future returns. Our evidence from China shares similar patterns. For STDR-sorted portfolios, 29

31 D1 stocks underperform D1 stocks by 9.7% per year, with a t-statistic of For TURN-sorted portfolio, D1 stocks underperform D1 stocks by 7.7%, with a t-statistic of When sorted by ILLIQ, D1 stocks outperform D1 stocks by 12.63%, with a t- statistic of These results suggest that information uncertainty plays a similar role in both markets. Additional results reported in Table 3 are, in the last two rows, the time-series means of the rank correlations between stock returns for decile portfolios and the fund decile ranks. Albeit weaker, our finding on the rank correlation is consistent with the above analysis on the return differences between extreme portfolios. As a further robustness check to the above results, in Figures 1 and 2 we plot the D1-D1 annual return spreads over the respective sample periods for individual stock return predictors in US and China. The plots are generally consistent with the results documented earlier in this section on average returns. For example, in the US, the momentum strategy of buying high and selling low MOM stocks is profitable in 27 out of 31 years. In contrast, this momentum strategy is profitable in 5 out of 9 years during the sample period for China. Another observation of figure 2 is that E/P, C/P, STDR, and TURN seem to work better after 2, which indicates that the Chinese market better resembles the U.S. market in recent years. B. Correlations of Return Spreads between the US and Chinese Markets In Table 4, we report the correlations for the D1-D1 return spreads between the US and Chinese markets, on each of the 18 return predictors. The sample period is the same as that for the Chinese market, We find positive correlations for SIZE, 3

32 B/P, E/P, C/P, ACC, NOA, RD, ADV, DT, STDR, TURN, and ILLIQ, and negative correlation for MOM, SG, CAPEX, AG, EQ, and DG. In terms of these numbers, the return spreads of two-thirds of the predictors are positively correlated in the two markets. Nevertheless, as inferred from the one-sided p-values, only the correlations for SIZE, B/P, E/P, ACC, CAPEX, AG, AND EQ are statistically significant. Local, uncorrelated, factors in these two markets drive the respective time-series patterns of stock return predictability. C. Risk-adjusted Return-predictive Performance An important question for any pattern of stock return predictability is whether it is related to risk. For this purpose, we use the Fama-French (1993) three-factor model to obtain the risk-adjusted returns for all decile portfolios and for the D1-D1 return spreads. Our analysis here is motivated by Fama and French (1998), who show that the value effects in international markets can be explained by their three-factor model. For the US market, the risk free rate, market return, and the size (SMB) and value (HML) factors are obtained directly from Ken French s website. For the Chinese market, we follow Fama and French (1993) to calculate the SMB and HML factors using the Sinofin data. Following Wang (24) and Kang et al. (22), we use the monthly yield of 3-month household deposit interest rate in China as the risk-free rate. We calculate the monthly market returns in China as the value-weighted average monthly returns for all A- shares traded in Shanghai and Shenzhen stock exchanges. We construct the size and BM portfolios in June of each year t. In particular, we sort all stocks into small and big size groups based on the median market capitalization of all stocks in June of year t. We 31

33 independently sort all stocks into low, median, and high B/P groups based on the 3% and 7% cutoff points of the book-to-market ratio of all stocks. Six size-bm portfolios are defined as the intersections of the two size and three BM groups. The monthly valueweighted average return on each portfolio is then computed. SMB is the difference, in each month, between the simple average of the returns on the three small-stock portfolios (S/L, S/M, AND S/H) and the simple average of the returns on the three big-stock portfolios (B/L,B/M, and B/H). Similarly, HML is the difference, in each month, between the simple average of the returns on the two high-b/m portfolios (S/H and B/H) and the average of the returns on the two low-bm portfolios (S/L and B/L). 6 In Table 5, we report the results from time-series regressions based on the threefactor model. Specifically, we run the following time-series regression using annual portfolio returns: R it RFt = α i + bi RMRFt + sismbt + hi HMLt + ε i, t, Where R, - RF t is the annual buy-and-hold returns of portfolio i in excess of the annual i t risk free rate. RMRF t is the market turn in excess of the risk free rate; SMB t, and HML t are the annual size and book-to-market factors. The estimated intercept coefficients from these regressions (a i ) are the riskadjusted return of the portfolio relative to the three-factor model. The results for the US market are presented in Panel A of Table 5 while those for the Chinese market are presented in Panel B of Table 5. In the US market, size and value effects (SIZE, B/P, C/P, E/P and SG) disappear after risk-adjustment. The insignificant return differences for 6 Similar to Wang (24), we also construct the size and BM risk factors using cutoff points only based on stocks traded in the Shanghai Stock Exchange and the results are consistent with those reported in panel B of Table 5. 32

34 SIZE, B/P, C/P, EP and SG ranked portfolios confirm Fama and French s (1996) findings. However, for all other predictors but ILLIQ, there exist significant risk-adjusted returns spreads between the top and bottom deciles. For example, the alpha for the top MOM decile is -8.96% while that for the bottom MOM decile is 4.96%, leading to an alpha of 13.92% for the return spread, significant at the 1 percent level. In the Chinese market, we also find the return spreads for SIZE and B/P ranked portfolios become insignificant after the three-factor adjustment. Further, ADV, STDR, TURN, and ILLIQ also become insignificant after risk adjustment. However, the threefactor model fails to explain the return spreads sorted by ACC, NOA, CAPEX, RD and AG in the Chinese market. ACC, NOA, CAPEX, and AG continue to be negatively related to future stock returns after risk adjustment and RD is still positively related to future stock returns. Finally, while the return difference for the top and bottom MOMranked decile portfolios remains insignificant, we notice that risk-adjusted portfolio returns appear to increase as MOM measure increases. For robustness, we have alternatively performed the Gibbon, Ross, Shanken (1989) time-series test of the three-factor asset pricing model. The test is performed on the monthly returns to ten value-weighted, monthly rebalanced, decile portfolios sorted on each of the 18 variables. The monthly factors are constructed following Fama and French (1993). This test is similar to that performed by Fama and French (1996; 1998). The resulting conclusions are consistent with those reported above. D Cross-sectional Regressions 33

35 In this section, we confirm the sorted-portfolio results by performing Fama- MacBeth (1973) regressions. In each year t, we perform regression of annual stock returns from July of year t to June of t+1 on the return predictor(s) constructed for year t. Table 6 reports the time-series averages of the coefficients on the return predictor(s) as well as their time-series t-statistics. We first perform univariate regressions. The dependent variable is the stock return. The explanatory variable is each individual stock return predictor. Results are reported in Panel A and B of Table 6. In the US market, the coefficients of the 18 predictors are all significantly different from zero, consistent with our findings in Table 3 based on sorted portfolios. In the Chinese market, the coefficients on SIZE, B/P, ACC, RD, AG, and STDR are significantly different from zero, also validating the evidence in Table 3. Given the correlations among the 18 variables (Table 2), we expect that their return-predictive power is also correlated. It is therefore interesting to examine stock return predictability by combining these variables. We find that a number of stocks with missing observations for one or more stock return predictors. In order to take advantage of the large cross-sectional data, we replace missing stock return predictor with the annual cross-sectional median value for that predictor. We then perform annual crosssectional regressions of individual stock returns onto the 18 variables jointly. The results of multivariable regressions are reported in Panel C and D of Table 6. For the US market, many stock return predictors remain significant in the multivariate regression, such as SIZE, MOM, C/P, SG, ACC, NOA, RD, EQ, DT, and DG, and STDR. However, the return-predictive power of B/P, E/P, CAPEX, ADV, AG, TURN, 34

36 and ILLIQ is subsumed by other variables. For the Chinese market, SIZE, SG, ACC, RD and AG remain significant in predicting future returns. On the other hand, the returnpredictive power of B/P and STDR is subsumed. The adjusted R-squares from the multivariate regressions provide a way to quantify the overall return predictability in a market. As it turns out, the average of adjusted R-square in the Chinese market is 12%, higher than the 7% in the US market, suggesting stronger return predictability in the former. Of course, one concern we have is that these return-predictive variables are correlated and the multicollinearity problem may inflate the R-squares. To rule this out, in Panel C of Table 6 we perform two additional sets of analysis. First, we only include predictors that are shown to be significant from the univariate regressions in Panel A. The time series averaged R-squared is.11 for the Chinese market. Second, we obtain the first 6 principal components factors of the 18 return predictors, and use the six principal component factors as joint regressors. The time series averaged R-squared of the PCA-based regression is also.11 for the Chinese market. These results suggest that stock return predictability in China is indeed quite high, and robust to any potential multicollinearity among the predictive variables. E. Subsample Results Finally, we look at the return predictability of the 18 predictors in SIZE and B/P sub-samples. In each year, we sort firms based on market capitalization into two groups (i.e., small and big), and independently sort firms based on B/P into two groups (i.e., value and growth). We then form equal-weighted decile portfolios within each of the groups and hold the positions for one year. 35

37 Panel A of Table 7 reports the return spreads between D1 and D1 portfolios ranked by each of the individual return predictors in the US market. By comparing the results for small and big sub-samples, we find that most return predictors work better for small stocks. In addition, the magnitude of return spreads is much larger for growth stocks than for value stocks. For example, the return spread for STDR-sorted portfolios is % for growth stock, while it is only -.23% for value stocks. Consistent with the U.S. evidence, ACC and NOA predictors work better for small stocks. However, other significant return predictors, including B/P, CAPEX, RD, ADV, AG, and STDR, have better return predictability for large stocks the Chinese market. For example, the return spread for RD-sorted portfolios is.23% for small stocks, but 11.22% for big stocks. All these predictors seem to have stronger return predictability in growth stocks than in value stocks in China. We further examine the subsample results based on three-factor adjusted alphas. As reported in Panel C and D of Table 7, the results for both U.S market and Chinese market are consistent with full sample analyses, with stronger effect for small and growth stocks in U.S and for big and growth stocks in China. V. Conclusion In this study, we compare the return predictability of 18 market anomalies on cross-sectional returns in the U.S. and China markets. In terms of stock return predictability there are more differences than similarities between the two markets. While we find similar size and value effects, momentum is nonexistent in China. Further, the Fama-French three-factor model can explain the return-predictive power of SIZE or B/P 36

38 in China, consistent with Fama and French (1996) for the US evidence and Fama and French (1998) for international evidence. Further consistent with the US evidence, we find that future stock returns are positively related to R&D expenditure, but negatively correlated with accruals, net operating assets, capital expenditures, and assets growth. To academic researchers, our findings constitute an out-of-sample test of stock return predictability documented in the US and mature economies in a fast growing emerging market. To investment practitioners, our findings are a useful start to evaluate the quantitative stock selection approach in the Chinese market. For the trading strategies working in the Chinese market, future studies may examine whether the return predictability is reliable across time and whether they are profitable after transaction costs, and how various correlated strategies can be optimally combined. For the strategies failing to work in the China market, it would be interesting to explore how the substantial differences in legal, financial, and corporate governance systems between these two markets contributes to this outcome. 37

39 Reference Amihud, Yakov, 22, Illiquidity and stock returns: cross-section and time-series effects, Journal of Financial Markets 5, Amihud, Y., Mendelson, H., Asset pricing and the bid ask spread. Journal of Financial Economics 17, Ang, A., R. J. Hodrick, Y. Xing and X. Zhang, 25, The Cross-Section of Volatility and Expected Returns. Journal of Finance, Forthcoming. Asquith, Paul, and David W. Mullins, 1983, The impact of initiating dividend payments on shareholders wealth, Journal of Business 83, Banz, Rolf W., 1981, The relationship between return and market value of common stocks, Journal of Financial Economics 9, Basu, S Investment Performance of Common Stocks in Relation to their Price- Earnings Ratios: A Test of the Efficient Market Hypothesis. The Journal of Finance 32, Bailey, W, Risk and return on China s new stock markets: some preliminary evidence, Pacific-Basin Finance Journal, 2, Barberis, Nicholas, Andrei Shleifer and Robert Vishny, 1998, A Model of Investor Sentiment, Journal of Financial Economics, 49, Bayless, M., and S. Chaplinsky, 1996, Is There a Window of Opportunity for Seasoned Equity Issuance, Journal of Finance 51, Beneish, M.D., C.M. Lee and R.L.Tarpley, 21, Contextual Fundamental Analysis through the Prediction of Extreme Returns, Review of Accounting Studies 6, Berk, Jonathan, Richard Green, and Vasant Naik, 1999, ptimal Investment, Growth Options, and Security Returns, Journal of Finance 54, Billett, Matthew T., Mark J. Flannery and Jon A. Garfinkel, 21. The long-run performance of firms following loan announcements, Working paper, University of Iowa. Bradshaw, Sloan, and Richardson, 26, The Relation between Corporate Financing Activities, Analysts Forecasts and Stock Returns, Journal of Accounting and Economics, Forthcoming. 38

40 Carlson, Murray, Adlai Fisher, and Ron Giammarino, 24, Corporate Investment and Asset Price Dynamics: Implications for the Cross-section of Returns, Journal of Finance 59, Chan, Louis K.C., Yasushi Hamao, and Josef Lakonishok, 1991, Fundamentals and Stock Returns in Japan, Journal of Finance, 47, Chan, Louis, Narasimhan Jegadeesh, Josef Lakonishok, 1996, Momentum Strategies, Journal of Finance, 51(5), Chan, Louis, Josef Lakonishok, and Theodore Sougiannis, 21, The stock market valuation of research and development expenditures, Journal of Finance 56, Chen, Gong-Meng, Kenneth A. Kim, John R. Nofsinger, and Oliver M. Rui, 24, Behavior and Performance of Emerging Market Investors: Evidence from China, Working paper, Washington State University. Chen, K.C.W. and Yuan, H. 24. Earnings Management and Capital Resource Allocation: Evidence from China s Accounting-Based Regulation of Rights Issues. Accounting Review, 79, Campbell, J., T. Vuolteenaho, 24, Bad Beta, Good Beta, American Economic Review, 94, Cooper, Gulen and Schill, 25, What best explains the cross-section of stock returns? Exploring the asset growth effect, Working paper. Choe, H., R. Masulis, and V. Nanda, 1993, Common Stock Offerings across the Business Cycle: Theory and Evidence, Journal of Empirical Finance, Chui, A., S. Titman, and J. Wei, 23, Momentum, legal systems and ownership structure: An analysis of Asian stock markets, Working Paper, University of Texas at Austin. Daniel, K.D., D. Hirshleifer, and A. Subrahamanym, 1998, Investor psychology and security market under- and over-reaction, Journal of Finance 53, Datar, V. T., N. Y. Naik, and R. Radcliffe, 1998, Liquidity and stock returns: An alternative test, Journal of Financial Markets 1, Easley, D., S. Hvidkjaer, and M. O'Hara, 22, Is information risk a determinant of asset returns? Journal of Finance 57, Easterbrook, Frank, Two Agency Cost Explanations of Dividends. American Economic Review 74(4):

41 Fama, Eugene F. and Kenneth R. French, 1992, The cross-section of expected stock returns, Journal of Finance 47, Fama, Eugene F., and Kenneth R. French, 1993, Common risk factors in the returns on stocks and bonds, Journal of Financial Economics 33, Fama, E. F., and K. R. Frech, 1996, Multifactor Explanations of Asset Pricing Anomalies, Journal of Finance 51, Fama, E. F., and K. R. Frech, 1998, Value versus growth: The international evidence, Journal of Finance 53, Feng, Lei and Mark S. Seasholes, 25, Do investor sophistication and trading experience eliminate behavioral biases in financial markets? Review of Finance 9, Gomes, J., L. Kogan, and L. Zhang, 23, Equilibrium Cross-Section of Returns, Journal of Political Economy 111, Griffin, John, Susan Ji, and Spencer Martin, 23, Momentum Investing and Business Cycle Risk: Evidence from Pole to Pole, Journal of Finance 58, Grinblatt, Mark and Bing Han, 25, Prospect theory, mental accounting, and momentum, Journal of Financial Economics, forthcoming. Haw, I. Qi, D., Wu, D. and Wu, W. 25. Market Consequences of Earnings Management in Reponse to Security Regulations in China. Contemporary Accounting Research, 22, He, Xiannan, 1998, Stock market manipulation behavior and regulatory suggestions, Soft Science in China, Healy, P. M. and K. G. Palepu, 1988, Earnings Information Conveyed by Dividend Initiations and Omissions, Journal of Financial Economics 21, Hirshleifer, D., K. Hou, S. Teoh, and Y. Zhang, 24, Do Investors Overvalue Firms with Bloated Balance Sheets? Journal of Accounting and Economics 38: Hong, H., and J. Stein, 1999, A unified theory of underreaction, momentum trading and overreaction in asset markets, Journal of Finance 54, Ikenberry, D., J. Lakonishok and T. Vermaelen, 1995, Market Underreaction to Open Market Share Repurchases, Journal of Financial Economics 39, Jegadeesh, Narasimhan, J. Kim, S. Krische, and C. lee, 24, Analyzing the Analysts: When Do Recommendations Add Value? Journal of Finance 59,

42 Jegadeesh, Narasimhan and Sheridan Titman, 1993, Returns to buying winners and selling losers: Implications for stock market efficiency, Journal of Finance 48, Jensen, M., 1986, Agency costs of free cash flow, corporate finance, and takeovers, American Economic Review 76, Jian, M. and Wong, T.J. 24. Earnings Management and Tunneling through Related Party Transactions: Evidence from Chinese Corporate Groups. Working Paper. Chinese University of Hong Kong. Jiang, G., T. Yao, and D. Xu, 26, The Informtoin Content of Idiosyncratic Volatility, Working Paper, University of Arizona Johnson, Timothy C., 22, Rational momentum effects, Journal of Finance 57, Jung, K., Y. Kim, and R. Stulz, 1996, Timing, Investment Opportunities, Managerial Discretion, and the Security Issue Decision, Journal of Financial Economics 42, Kang, J., Liu, M. and S. Ni, 22, Contrarian and momentum strategies in the China stock market: , Pacific-Basin Finance Journal, 1, Khan, Mozaffar, 26, Are Acruals Really Mispriced? Evidence from Tests of an Intertemporal Capital Asset Pricing Model, working paper. Lakonishok, Josef, Andrei Shleifer, and Robert W. Vishny, 1994, Contrarian investment, extrapolation, and risk, Journal of Finance 49, Lakonishok, J. and T. Vermaelen, 199, Anomalous Price Behavior Around Repurchase Tender Offers, Journal of Finance 45, Lee C., and B. Swaminathan, 2, Price momentum and trading volume, Journal of Finance 55, Lee, Chi-wen Jevons and Xing Xiao, 24, Tunneling Dividends, Working Paper, Tulane University and Tsinghua University. Loughran, T., and J. Ritter, 1995, The new issue puzzle, Journal of Finance 5, Michaely, R., R.H. Thaler, and K. Womack Price Reactions to Dividend Initiations and Omissions: Overreaction or Drift? Journal of Finance, vol. 5, no. 2 (June): Mei, J., J. Scheinkman, and W. Xiong, 25, Speculative Trading and Stock Prices: Evidence from Chinese A-B Share Premia, Working Paper, Princeton University. 41

43 Pilotte, E., 1992, Growth Opportunities and the Stock Price Response to New Financing, Journal of Business 65, Reinganum, Marc 1981, Misspecification of capital asset pricing: Empirical anomalies based on earnings yields and market values, Journal of Financial Economics 9, Ritter, J., 1991, The Long-Run Performance of Initial Public Offerings, Journal of Finance 46, Rouwenhorst, K. G., 1998, International Momentum Strategies, Journal of Finance, 53, Rouwenhorst, G. 1999, Local return factors and turnover in emerging stock martkets, Journal of Finance 54, Seasholes, M., and G. Wu, 26, Predictable Behavior, Profits, and Attention, Working Paper, University of Houston. Shengzhen Stock Exchange, 25, The major illegal behaviors in the Chinese stock market: case study on insider trading and price manipulation. Shumway, T., and G. Wu, 26, Does Disposition Effect Drive Momentum? Working Paper, University of Houston. Sloan, R Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review 71, Spiess, D. Katherine and John Affleck-Graves, The long-run performance of stock returns following debt offerings, Journal of Financial Economics 54, Sun, Q. and W.H.S. Tong, 23. China share issue privatization: the extent of its success, Journal of Financial Economics, 7, Titman S., J. Wei, and F. Xie, 24, Capital Investments and stock returns, Journal of Financial and Quantitative Analysis 39, Wang, C and S. Chin, 24, Profitability of return and volume-based investment strategies in China s stock market, Pacific-Basin Finance Journal, 12, Wang, C, 24, Relative strength strategies in China s stock market: , Pacific- Basin Finance Journal, 12, Wu, F., 25, Individual Investors' Trading Preferences for Fundamental Stock Characteristics, Working Paper, University College Dublin. Yao, T., 26, Systematic Momentum, Working Paper, University of Arizona 42

44 Zhang, L., 26, Anomalies, Working Paper, University of Rochester. Zhao, J., 26, Quant Jocks and Tire Kickers: Does the Stock Selection Process Matter? Working Paper, University of Arizona. 43

45 Appendix: Constructing Stock Return Predictors The following list of stock return predictors are constructed in June of year t. For all the financial statement items (Compustat items), year t refers to fiscal year ending in the calendar year t. Variable Description Computation Detail: China Computation Detail: US 1. SIZE Natural logarithm of market capitalization ln(closing price at the end of June multiplied by A shares outstanding at the end of June) ln(closing price at the end of June [CRSP] multiplied by common shares outstanding at the end of June[CRSP]) 2. B/P Book to price Book Value of Equityt -1 Book Value of Equityt -1[D216] Market Cap of A Sharest-1 in December Market Cap of Equityt-1 in December 3. MOM Cumulative market-adjusted 1 m 1 return for the preceding 12 i = m 12 (1+monthly return t ) - i = m 12 (1+market monthly Same month return t ), where m is June of year t 4. E/P Earnings to price Net Income Earnings before extraordinary items t-1 t-1 [D18] MarketCap of A sharest-1 MarketCap t-1 [D24*D25] 5. C/P Cash flow to price Net Income t-1 + Depreciationt Earnings before extraordinary items 1 t-1 [D18] + Depreciation r-1[d14] MarketCap of A shares MarketCap t-1 [D24*D25] t-1 6. SG Sales growth Salest-1 Salest-2 7. ACC Accruals [( CA CASH SI) t-1 ( CL STD LTDC TP) t-1 DEP t-1 ]/ATA t1, where CA is the change in current assets from previous fiscal year; Cash is the change in cash, SI is the change in short term investment; CL is the change in current liabilities; STD is the change in short debt; LTDC is the change in long term debt included in current liabilities; TP is the change in taxes payable [D71]; DEP is change in accumulative depreciation; and ATA is the average of the beginning and ending total assets (TA)of the reporting year. 8. NOA Net operating assets NOA = [Operating Assets Operating Liabilities] t-1 /ATA t-1, where operating assets = TA cash short term investment, operating liabilities = total asset STD LTDC LTD MI CE, STD = Debt included in current liabilities, LTDC is long term debt included in current liabilities, LTD = total Sales t-1 [D12] Sales [D12] t-2 [( CA CASH) t-1 ( CL STD TP) t-1 DEP t- 1]/ATA t-1, where CA is the change in current assets [D4] from previous fiscal year; Cash is the change in cash/cash equivalents [D1]; CL is the change in current liabilities [D5]; STD is the change in debt included in current liabilities [D34]; TP is the change in income taxes payable [D71]; DEP is depreciation and amortization expense [D14]; and ATA is the average of the beginning and ending total assets [D6] of the reporting year. [Operating Assets Operating Liabilities] t-1 /ATA t-1, where operating assets = TA [D6] cash and short term investment [D1], operating liabilities = total asset STD LTD MI PS CE, STD = Debt included in current 44

46 9. CAPEX Capital expenditure to total assets Long Term Debt, MI = Minority Interests, CE = Common Equity, and ATA is the average of the beginning and ending total assets of the reporting year. The values of short-term debt, taxes payable, long-term debt, and minority interest are set as zero if they are missing. CapEx ATA t 1 t 1, where Capex t-1 is the change of net fixed assets in fiscal year t-1 plus the change in accumulated depreciations in year t-1, and ATA is the average of the beginning and ending total assets (TA) of the reporting year. liabilities [D34], LTD = Long Term Debt [D9], MI = Minority Interests [D38], PS = Preferred Stocks [D13], CE = Common Equity [D6], and ATA is the average of the beginning and ending total assets [D6] of the reporting year. The values of short-term debt, taxes payable, long-term debt, minority interest, and preferred stock are set as zero if they are missing. CapEx t 1, where Capex is capital expenditure [D128], ATAt 1 and ATA is the average of the beginning and ending total assets (TA, [D6]) of the reporting year. 1. RD Research and Development Administrative Expenset 1, where market value is market R & Dt 1, where R&D is research and development expenses to market value of MVt 1 MV equity t 1 capitalization at the end of year t-1 expenses [D46]; market value is market capitalization at the end of year t ADV Advertising expenses to sales Sales Expensest -1+ Marketing Expensest -1, where ADV t 1, where Advt-1 advertising expenses [D45] in MV t 1 MVt 1 market value is market capitalization at the end of year t-1 year t-1; market value is market capitalization at the end of year t AG Change in total assets TAt-1 TA t 2, where TA is total assets TAt-1 TA t 2, where TA is total assets [D6] TAt-2 TAt EQ Net cash flow received from CTt 1+ CSt 1, where CT is the change in common CREt 1 CPEt 1 DIVt 1, where CRE is cash received external equity financing ATAt-1 ATAt-1 stock, CS is the change in capital surplus, and ATA is the average of the beginning and ending total assets of the reporting year. from the sale of common and preferred stock [D18], CPE is cash paid for purchase of common and preferred stock [D115], DIV is cash dividends paid [D127], and ATA is the average of the beginning and ending total assets [D6] of the reporting year. 14. DT Net cash flow received from external debt financing LTDt 1+ LTNt 1+ STDt 1, where LTD is the change ATAt-1 in long term debt, LTN is the change in long term note, STD is the change in total short term debt, and ATA is the average of the beginning and ending total assets of the CRDt 1 CPDt 1 CDt 1, where CRD is cash received ATAt-1 from the issuance debt [D111], CPD is cash paid for reduction of debt [D114], is CD is change in current 45

47 reporting year. 15. DG Growth rate of dividend DIVt 1 DIVt 2, where Div t-1 is the cash dividend paid in DIVt-2 year t STDR Idiosyncratic risk STDR is the standard deviation of the error term in daily data market model regression: 17. TURN Average daily volume turnover 18. ILLIQ Amihud illiquidity measure R k = 5 i, τ = α + βk * Rm, τ+ k + εi, τ k = 5, where Ri, τ is daily stock return in the 12 months preceding June of each year, Rm, τ + k is the value-weighted average of the SHSE and SZSE valueweighted daily index return from 5 lags and 5 leads. i= m 1 / tan Percentile rank daily Volume Shares outs ding i= m 12, n where n is number of days available for 12 months preceding the end of June in each year. n,, Percentile rank R / 1 i VOLD τ = τ i τ, where Ri, τ is the n return on stock i on dayτ within 12 months preceding June of each year, VOLDi, τ is the respective daily volume in dollars, and n is number of days available for 12 months preceding the end of June in each year. debt [D31], and ATA is the average of the beginning and ending total assets [D6] of the reporting year. DIVt 1 DIVt 2, where Div t-1 is the cash dividend paid DIVt-2 in year t-1[d21]. STDR is the standard deviation of the error term in daily data market model regression: k = 5 Ri, τ = α + βk * Rm, τ+ k + ε, where R i, τ i, τ is daily stock k = 5 return in the 12 months preceding June of each year, Rm, τ + is the value-weighted CRSP market daily returns k from 5 lags and 5 leads. i= m 1 / tan Percentile rank daily Volume Shares outs ding i= m 12 n by exchange, where n is number of days available for 12 months preceding the end of June in each year. Same 46

48 Table 1 Market Overview This table reports year-end summary characteristics of the US and Chinese stock markets. Num is the number of listed firms. Turn is annual trading volume scaled by year-end shares outstanding. Price is the year end stock price. MV, market value, is year-end close price multiplied by year end shares outstanding. Equal weighting is used when we compute averages. Panel A is for the US market and Panel B is for the Chinese market. In Panel A, price is for A shares only. Total market value aggregates the market value of A, B, and H shares plus that of non-tradable shares using year-end price of A shares. Panel A: The US Market Year Num Num Num Turnover Turnover in Price MV NYSE/ AMEX NASDAQ NYSE/ AMEX NASDAQ (US$) (US$ billion) Panel B: The Chinese market Num Num Turn Turn Price MV--A MV--B MV--Total Year Num SHSE SZSE SHSE SZSE (RMB) (RMB billion) (RMB billion) (RMB billion)

49 Table 2 Summary Statistics for Stock Return Predictors This table presents summary statistics of return predictors and correlations among these predictors in the US and China. In Panels A and B, we report cross-sectional characteristics, including 25%, mean, median, 75%, standard deviation, skewness and kurtosis, for each stock return predictor at the end of 24. In Panels C and D, we report the time-series averages of the cross-sectional averages of the correlations between return predictors. SIZE is the year-end closing price multiplied by the number of shares outstanding. A shares market value is used to compute SIZE in China. B/P is the book value of equity divided by market value of equity. MOM is the cumulative market-adjusted return for a stock in the 12 months preceding June in year t. E/P is annual earnings per share scaled by year-end closing price. C/P is the sum of earnings before extraordinary item and depreciation scaled by market value of equity. SG is the growth rate of sales. ACC is accounting accruals defined in Sloan (1996). NOA is net operating assets defined in Hirshleifer, Hou, Teoh, and Zhang (24). CAPEX is Capital expenditure scaled by total assets. RD is research and development scaled by market value of equity. In the analysis of Chinese stocks, RD is management expenses scaled by market value of equity. ADV is advertising expenses scaled by market value of equity. In the analysis of Chinese stocks, ADV is sales and marketing expenses scaled by market value of equity. AG is the growth rate of total assets. EQ and DT are the net cash flow from external equity and debt financing defined in Bradshaw, Richardson, and Sloan (26). DG is the percentage change of annual dividend. STDR is the standard deviation of the error term in the regression of daily stock excess return on daily CRSP value-weighted market return in the 12 month formation period. TURN is the average daily trading volume over shares outstanding for a stock for 12 months preceding June of year t. ILLIQ is the illiquidity measure defined in Amihud (22). Details of these variables are provided in the Appendix. Equal weight is used when we compute cross-sectional averages. Panel A: Cross-sectional Distribution of Stock Return Predictors in US, 24 25% Mean Median 75% STDEV SKEW KURT SIZE (Mil US$) B/P (%) MOM (%) E/P (%) C/P (%) SG (%) ACC (%) NOA (%) CAPEX (%) RD(%) ADV(%) AG (%) EQ (%) DT (%) DG (%) STDR (%)

50 Panel B: Cross-sectional Distribution of Stock Return Predictors in China, 24 25% Mean Median 75% STDEV SKEW KURT SIZE (Mil RMB) B/P (%) MOM (%) E/P (%) C/P (%) SG (%) ACC (%) NOA (%) CAPEX (%) RD(%) ADV(%) AG (%) EQ (%) DT (%) DG (%) STDR (%)

51 Panel C: Correlation among Stock Return Predictors in the US Market SIZE B/P MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ DT DG STDR TURN B/P -.8 MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ DT DG STDR TURN ILLIQ

52 Panel D: Correlation among Return Predictors in the Chinese market SIZE B/P MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ DT DG STDR TURN B/P.1 MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ DT DG STDR TURN ILLIQ

53 Table 3 Returns to Decile Portfolios Sorted on Stock Return Predictors This table reports returns across decile portfolios sorted by each of the 18 return predictors in US and China. In June of each year t, we rank stocks into deciles based on each of the 18 predictors defined in Table 2 and the Appendix. D1 portfolios contain stocks in the highest decile of the predictor and D1 portfolios contain stocks in the lowest decile of the predictor. We compute the average buy-and-hold stock performance from July of year t to June of year t+1 for each stock decile. Panel A reports the time series average of raw stock returns in each deciles for the 18 return predictors in the US market. The return difference between D1 and D1 stocks, the correlation between average decile return and decile rank (RCorr) and the associated t-statistics are also reported. To be included in the analysis, stock price is restricted to be no less than $5. Panel B reports the raw return performance of decile portfolios for Chinese stocks. To be included in the analysis, stock price is restricted to be no less than RMB1. The US sample is from 1974 to 24 and the China sample is from Panel A: The US Market SIZE B/P MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ DT DG STDR TURN ILLIQ (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) (11) (12) (13) (14) (15) (16) (17) (18) D D t-stat RCorr t-stat

54 Panel B: The Chinese Market SIZE B/P MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ DT DG STDR TURN ILLIQ (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) (11) (12) (13) (14) (15) (16) (17) (18) D D t-stat RCorr t-stat

55 Table 4 Correlations of Return Spreads between US and Chinese Markets This table reports the correlation of return spreads between US and Chinese markets from 1994 to 24. We calculate the annual return spreads of the D1 stocks and D1 stocks sorted by a return predictor in the US and Chinese market, respectively. We then calculate the correlation of the return spreads between US and Chinese market over 1994 to 24 periods. The one-tailed p-values of the correlation coefficients are reported underneath. To be included in the analysis, prices of US stocks are no less than $5 and prices of Chinese stocks are no less than RMB1. SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) CAPEX (9) CORR p-value RD (1) ADV (11) AG (12) EQ (13) DT (14) DG (15) STDR (16) TURN (17) ILLIQ (18) 54

56 Table 5 Three-Factor Alphas of Decile Portfolios Sorted on Stock Return Predictors This table reports risk-adjusted stock performance across stock deciles sorted by each of the 18 US and Chinese stock return predictors. In June of year t, we rank stocks into deciles based on one of the predictors defined in Table 1 and the Appendix. D1 is the highest decile of the predictor and D1 is the lowest decile of the predictor. We compute the average buy-and-hold stock return from July of year t to June of year t+1 for each stock decile. The risk-adjusted return is the intercept from time-series annual regressions of decile portfolio returns on the three factors: R it RFt = α i + bi RMRFt + sismbt + hi HMLt + ε. R i, t i, t - RF t is the annual buy-and-hold return of portfolio i in excess of the risk free rate (the yield on Treasury bills with one-month maturity, from CRSP). RMRF is the t market turn in excess of the risk free rate; SMB t, and HML t are the annual returns on size and book-to-market factors computed following Fama-French (1993) procedure. Panel A is for US results. To be included in the analysis, stock price is restricted to be no less than %5. Panel B is for China results. To be included in the analysis, stock price is restricted to be no less than RMB1. The t-statistics for the difference in three-factor adjusted returns and the one-sided p-value for rank correlations are provided. The sample period for the US market is from 1974 to 24. The sample period for the Chinese market is from 1994 to 24. Panel A: The US Market SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) CAPEX (9) D D RD (1) ADV (11) AG (12) EQ (13) DT (14) DG (15) STDR (16) TURN (17) ILLIQ (18) t-stat RCorr p-value

57 Panel B: The Chinese Market SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) CAPEX (9) RD (1) ADV (11) AG (12) EQ (13) DT (14) DG (15) STDR (16) TURN (17) ILLIQ (18) D D t-stat RCorr p-value

58 Table 6 Cross-Sectional Regressions This table reports the time-series averaged coefficients on stock return predictors in the US and Chinese markets. Panel A reports the result of univariate regressions (Panel A) and the result of a joint regression including all the 18 predictors (Panel B). The dependent variable is the annual buy-and-hold stock return from July in year t to June of year t+1. The cross sectional median is set as the value of return predictor when it is missing. Panel C reports the adjusted R 2 when we regress stock returns on (1) the significant predictors from univariate regressions (Panel A) and (2) the first 6 principal components of the 18 return predictors. To be included in the analysis, prices of US stocks is restricted to be no less than $5 and prices of Chinese stocks are restricted to be no less than RMB1. Panel A: Univariate Regressions SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) CAPEX (9) The US Market Predictor t-stat The Chinese Market Predictor t-stat Panel B: Multivariate Regressions SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) CAPEX (9) RD (1) RD (1) ADV (11) ADV (11) AG (12) AG (12) EQ (13) EQ (13) DT (14) DT (14) DG (15) DG (15) STDR (16) STDR (16) TURN (17) TURN (17) ILLIQ (18) ILLIQ (18) Adj. R 2 The US Market Coeff t-stat The Chinese Market Coeff t-stat Panel C: Additional Multivariate Regressions Adjusted R 2 using significant predictors Adjusted R 2 based on principal components (2) (1) The US Market.7.6 The Chinese Market

59 Table 7 Subsample Results on Sorted Portfolios This table reports average annual stock returns between for bottom and top deciles sorted on each return predictor for sub-sample groups in the US and China. In June of each year t, we first rank stocks into equal-sized based on market capitalization (small and large stocks) or on the B/P ratios (growth and value stocks). We then rank stocks into deciles within each size group based on one return predictor. We compute the difference between the average buy-and-hold stock performance from July of year t to June of year t+1 for bottom and top stock deciles. Panel A reports the raw return for small and large firms. Q5 stocks are those with the highest value of the predictor and Q1 stocks are those with the lowest predictor. Panel B reports the raw return for growth and value firms. Panels A and B report the raw return results for the US and China. Panels C and D report the three-factor adjusted returns for the US and Chinese markets. The t-statistics for the difference in the average return between top and bottom deciles are provided. To be included in the analysis, prices of US stocks are no less than $5 and prices of Chinese stocks are no less than RMB1. The coefficients on each predictor and t-statistics are provided. Panel A: Net Return Spread in the US Market SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) CAPEX (9) Small Stocks t-stat Big Stocks t-stat Growth Stocks t-stat Value Stocks t-stat RD (1) ADV (11) AG (12) EQ (13) DT (14) DG (15) STDR (16) TURN (17) ILLIQ (18) 58

60 Panel B: Net Return Spread in the Chinese Market SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) CAPEX (9) Small Stocks t-stat Big Stocks t-stat Growth Stocks t-stat Value Stocks t-stat Panel C: Three-factor Alpha in the US Market SIZE (1) B/P (2) MOM (3) E/P (4) C/P (5) SG (6) ACC (7) NOA (8) RD (1) CAPEX RD (9) (1) Small Stocks t-stat Big Stocks t-stat Growth Stocks t-stat Value Stocks t-stat ADV (11) ADV (11) AG (12) AG (12) EQ (13) EQ (13) DT (14) DT (14) DG (15) DG (15) STDR (16) STDR (16) TURN (17) TURN (17) ILLIQ (18) ILLIQ (18) 59

61 Panel D: Three-factor Alpha in the Chinese Market SIZE B/P MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ DT DG STDR TURN ILLIQ Small Stocks t-stat Big Stocks t-stat Growth Stocks t-stat Value Stocks t-stat

62 Figure 1 Annual Return Spreads of Decile Portfolios: the US Market Figure 1 plots by year return spreads for decile portfolios sorted by stock return predictors in the US Decile portfolios for individual return predictors are formed in June of year t. The annual buy-and hold return for each stock from July of year t to June of year t+1 is computed. Then the return spread of a predictor is the difference in average annual returns between D1 and D1 portfolios ranked by that return predictor. 5 SIZE 5 B/P 5 MOM E/P C/P SG ACC NOA CAPEX RD ADV AG EQ STDR DT TURN DG ILLIQ

63 Figure 2 Annual Return Spreads of Decile Portfolios: the Chinese Market Figure 2 plots by year return spreads for decile portfolios sorted by stock return predictors in China. Decile portfolios for individual return predictors are formed in June of year t. The annual buy-and hold return for each stock from July of year t to June of year t+1 is computed. Then the return spread of a predictor is the difference in average annual returns between D1 and D1 portfolios ranked by that return predictor. 5 SIZE 5 B/P 5 MOM E/P ACC RD EQ STDR C/P NOA ADV DT TURN SG CAPEX AG DG ILLIQ

64 Founded in 1892, the University of Rhode Island is one of eight land, urban, and sea grant universities in the United States. The 1,2-acre rural campus is less than ten miles from Narragansett Bay and highlights its traditions of natural resource, marine and urban related research. There are over 14, undergraduate and graduate students enrolled in seven degreegranting colleges representing 48 states and the District of Columbia. More than 5 international students represent 59 different countries. Eighteen percent of the freshman class graduated in the top ten percent of their high school classes. The teaching and research faculty numbers over 6 and the University offers 11 undergraduate programs and 86 advanced degree programs. URI students have received Rhodes, Fulbright, Truman, Goldwater, and Udall scholarships. There are over 8, active alumnae. The University of Rhode Island started to offer undergraduate business administration courses in In 1962, the MBA program was introduced and the PhD program began in the mid 198s. The College of Business Administration is accredited by The AACSB International - The Association to Advance Collegiate Schools of Business in The College of Business enrolls over 14 undergraduate students and more than 3 graduate students. Mission Our responsibility is to provide strong academic programs that instill excellence, confidence and strong leadership skills in our graduates. Our aim is to (1) promote critical and independent thinking, (2) foster personal responsibility and (3) develop students whose performance and commitment mark them as leaders contributing to the business community and society. The College will serve as a center for business scholarship, creative research and outreach activities to the citizens and institutions of the State of Rhode Island as well as the regional, national and international communities. The creation of this working paper series has been funded by an endowment established by William A. Orme, URI College of Business Administration, Class of 1949 and former head of the General Electric Foundation. This working paper series is intended to permit faculty members to obtain feedback on research activities before the research is submitted to academic and professional journals and professional associations for presentations. An award is presented annually for the most outstanding paper submitted. Ballentine Hall Quadrangle Univ. of Rhode Island Kingston, Rhode Island