معاملات سفته بازی و سهام بازده: تجزیه و تحلیل تسلط تصادفی از بازار سهام چین

The pricing of A-shares in China has long puzzled financial economists. This paper applies recent tests of stochastic dominance (SD) to examine whether differences in the return distributions of A- and B-shares in China are consistent with market efficiency. As SD is nonparametric, market efficiency can be examined without the joint test problem arising from misspecifications in the asset pricing benchmark. Our results show A-shares have second-order dominated B-shares from 1996 to 2005. This dominance was most significant during the market segmentation period, but has continued, albeit to a lesser extent even after the B-share market was opened to local investors in 2001. Our results are robust to using residual returns from an international asset pricing model instead of raw returns. We conclude that the superior performance of A-shares cannot be attributed to risk. The results are more likely due to a return bias caused by intense speculation among retail individuals under limited arbitrage.

The pricing of A-shares in China has long been a puzzle to financial economists. China is unique among emerging markets because A-shares, which are restricted to Chinese citizens, are priced at huge premiums over B-shares, which are restricted to foreigners. Despite the opening of the B-share market to local investors in 2001, the A-share premium has persisted. A large literature has tried to explain the source of this price differential. Recent contributions to this literature include Mei et al. (2005) and Chan et al. (2008). Several studies argue that China's stock markets are mainly driven by speculation and insider trading. Allen et al. (2006) point out that the desire for quick gains combined with lack of strong legal framework in China fostered a speculative attitude among investors. Mei et al. (2005) point out that share prices probably contain a speculative component due to a limited share float and prohibitions on short sales. Moreover, since the A-share market is dominated by unsophisticated individual investors, the size of the A-share premium is a direct reflection of this speculative component. Using A-share turnover as a proxy for speculation, Mei et al. find that stocks with a higher turnover tend to trade at higher premiums, which is consistent with the speculative demand hypothesis. Chan et al. (2008) provide an alternative explanation of the B-share discount based on the notion of information asymmetry. Specifically, they argue that foreign investors are less informed about the Chinese market than locals. Hence, they demand a price discount for B-shares to compensate for this information disadvantage. Consistent with their hypothesis, they find that B-share discounts are positively related to measures of information asymmetry that are constructed using bid-ask spreads of A- and B-shares. Furthermore, A-share turnover has no significant impact on this discount after accounting for information asymmetry. While these studies shed light on factors that explain the cross-section of A-share premiums, they do not explain why such a premium exists in the first place. Existing debate on the sources of the A-share premium have also missed another intriguing aspect of the Chinese stock market. That is, there exists a higher average return of A-shares relative to the average return of B-shares. For example, between January 1996 and February 2001, A-shares outperformed B-shares with average annual returns of 30% and 12%, respectively. This large mean return differential has led to a dramatic increase in A-share premium over the same period. Even after the B-share market was opened to local investors, A-shares still outperformed B-shares. The average annual returns of A-shares and B-shares in the bear market from June 2001 to December 2005 were −14% and −19%, respectively. Classical finance argues that markets are efficient in promoting price discovery, which leaves no room for noise traders to influence asset prices. This sanguine view has been challenged by mounting evidence from behavioral finance. There is convincing evidence that when stocks are favored by individual investors and difficult to value, they are prone to mispricing. For example, Kumar and Lee (2006) find that firms with high retail concentration and firms that are more difficult to arbitrage have consistently high loadings on measures of investor sentiment. Baker and Wurgler (2006) show that sentiment-driven buying leads to return reversals, which is consistent with ex-ante overvaluations. They show that this effect is more significant among small, young, highly volatile, unprofitable, and non-dividend paying firms as well as among extreme value and growth firms. A common denominator is that these firms have highly subjective valuations. It is natural to think of mispricing as stochastic. Recent work by Brennan and Wang (2007) provide some insights into the relation between stochastic mispricing and the behavior of asset returns. They show that even if a stock trades on average at its fundamental price, its average return may still be biased above its fundamental required return. For example, let View the MathML sourcePt* and Pt denote the fundamental and market price of a stock. Suppose Zt is a stationary stochastic mispricing variable. Brennan and Wang show that even if Zt has a mean of one, so that the stock price is unconditionally rational, Jensen's inequality implies that the expected return of that stock may still be biased above the fundamental expected return View the MathML source(i.e., E(Rt)>E(Rt*)). 1 Empirically, they find that this bias is larger for smaller firms, low-price firms, growth firms, and firms with high turnover. Their results are generally consistent with those of Kumar and Lee (2006). Is the higher average return of A-shares consistent with market efficiency, or does it reflect a return bias? It is plausible that noise trading could be prevalent in the A-share market due to the dominance of uninformed individual investors and short-sale restrictions in the Chinese stock market. On the other hand, advocates of efficient markets may argue that A-share investors face more risk than B-share investors because they are less experienced than B-share investors, who are mainly institutions. Moreover, institutional investors can diversify easily across international markets, while exchange controls in China limit diversification for individual investors. Finally, poor minority investor protection and imperfect regulation of markets may also cause Chinese investors to be more risk averse. As such, the higher average return of A-shares could be a compensation for the higher risk faced by these investors. The goal of this paper is to test whether risk-based explanations can rationalize the high average return of A-shares. The standard approach for testing market efficiency is limiting due to the need for asset pricing model benchmarks. Despite the voluminous literature on asset pricing, this literature has yet to identify equilibrium asset pricing models that can explain a large portion of time-varying expected returns. Ghysels (1998) forcefully argues that in modeling expected returns, conditional asset pricing models fare poorly compared to simpler static models. Harvey (1995) highlights similar difficulties in modeling expected returns of emerging stock markets. Bailey et al. (1999) find little evidence that international asset pricing models can explain stock price differences in segmented markets. This paper relies on a different approach for testing market efficiency. Our approach is based on the idea of stochastic dominance (SD). Because SD is nonparametric, SD tests do not require any specific asset pricing benchmarks and thus avoid the joint test problem inherent in the standard approach. SD rankings also have direct interpretations in terms of expected utility and thus provide an appealing basis to relate investors’ revealed preferences to their risk attitudes. In particular, if the distribution of returns is such that A-shares second-order stochastically dominates B-shares, then all investors with non-decreasing, concave utility functions would unanimously prefer A-shares to B-shares. Put differently, this result implies that no asset pricing model based on risk aversion can rationalize the excess mean returns of A-shares. We investigate the risk-based hypothesis for A-shares using a recent test for stochastic dominance developed by Linton et al. (2005, hereafter LMW). An appealing feature of the LMW test is that it allows for general dependence among assets as well as non-identically and independently distributed observations. Thus, the test accommodates a rich variety of time series properties in returns including weak dependence (α-mixing) and conditional heteroskedasticity. To perform statistical inference, we use a sub-sampling bootstrap method to estimate test critical values. Politis and Romano (1994) show that this technique leads to consistent inference under very weak conditions. The use of bootstrap methods to approximate critical values contrasts most SD tests that assume distributions are known a priori. Simulations by LMW demonstrate that for samples with more than 500 observations, their test has good empirical size and power properties. The rest of this paper is organized as follows. The next section describes the key institutional features of Chinese stock markets. Section 3 describes the data and presents summary statistics of returns for A- and B-shares. As a prelude to stochastic dominance tests, an attempt is made in Section 4 to model the returns of A-shares and B-shares based on a global asset pricing model for emerging markets. Section 5 introduces the stochastic dominance approach for testing market efficiency, and Section 6 reports the results of the LMW test for A- and B-shares. Section 7 concludes the paper with a summary of our main results.

China's stock market presents an interesting setting to examine the effects of market frictions on pricing efficiency. The market is relatively young, dominated by retail investors and lacks a strong legal and institutional framework to protect investors’ rights. Moreover, market frictions such as exchange controls and prohibitions against short-selling also limit arbitrage. Theorists such as Miller (1977) argue that under this environment, speculative bubbles tend to form when investors have diverse beliefs since only optimistic investors will choose to be in the market. Mei et al. (2005) argue that market frictions and speculation may explain why prices of A-shares tend to exceed those of B-shares. To date, however, reasons for the existence of the A-share premium have not been resolved. Although the literature has identified liquidity and information asymmetry as important factors in explaining why some firms have higher A-share premiums than others, the basic question of whether A-shares are rationally priced is still open. This paper shows that stochastic dominance tests can be used to test the efficiency of the A-share market. We find strong evidence that A-shares second-order dominate B-shares in the full segmentation period in China. Since this result implies that all risk averse investors prefer A-shares to B-shares, the superior average return of A-shares in this period cannot be explained by risk, and is more likely a reflection of market inefficiency. The policy change of February 2001 provides us with a natural experiment to assess the impact of market reform on market efficiency. Chan et al. (2008) find that A-shares continued to dominate price discovery at the daily level after the policy change. Likewise, our results show that A-shares continued to SSD-dominate B-shares in the post-reform period. The higher turnover rate of A-shares may be the main reason why investors remain attracted to this market compared to the B-share market. Our results indicate that there is some way to go before the A-share market can be considered fully efficient.