The Impact of Market Segmentation on the Value-Relevance of. Accounting Information: Evidence from China

Transcription

1 The Impact of Market Segmentation on the Value-Relevance of Accounting Information: Evidence from China Shwu-hsing Wu Tainan Universy of Technology Stephen Lin* Florida International Universy Shu-hsing Li National Taiwan Universy Meihua Koo California State Polytechnic Universy, Pomona May 2008 *Corresponding author Stephen Wen-jen Lin S.W. 8th Street Ryder Business Building, RB 237B Miami, Florida Tel: (786)

2 The Impact of Market Segmentation on the Value-Relevance of Accounting Information: Evidence from China Abstract This paper examines the extent to which the market segmentation policy in China has affected the value-relevance of accounting information, measured by earnings and book value of shareholders equy. Using Chinese data during the period of , we find that the price difference between A- and B- shares significantly decreased after the change in the market segmentation policy was announced in China in However, the association between A-share price and accounting information significantly increased after Although the price-accounting association becomes much weaker for B-shares during the transion period of , this association also significantly increased after Different from previous studies that examine how the qualy of accounting information affects share valuation, this study contributes to the lerature by providing further evidence on the roles of market condion and information environment in the association between share price and accounting information in an emerging capal market such as China.

3 The Impact of Market Segmentation on the Value-Relevance of Accounting 1. Introduction Information: Evidence from China This study examines the extent to which the market segmentation policy in China has affected the association between share price and accounting information, measured by earnings and book value of shareholders equy. We focus on the A- and B- shares because they are the only two types of public shares traded in the Chinese stock markets. Consistent wh previous studies in the lerature (e.g. Francis and Schipper, 1999; Core, Guay, and Buskirk, 2003; Lin and Chen, 2005; Sami and Zhou, 2004), we measure the value-relevance of accounting information as the explanatory power of earnings and book value of shareholders equy for share price (i.e. the adjusted R-squared value of regressing share price on earnings and book value of equy). Our study compares the value-relevance of accounting information before and after the change in the market segmentation policy in 2001 for both the A- and B- share markets. Both A- and B- shares carry the same dividends, voting, and liquidation rights but were traded separately by domestic and foreign investors respectively and had substantial differences in share prices and trading volumes before the market was fully segmented in China. For example, the average share price (trading volume) of A-shares was more than five (three) times higher than those of B-shares in Many studies find that the price difference between the two-class shares is caused by the completely segmented markets (e.g. Bailey, 1994; Fung, Lee, and Leung, 2000), relatively illiquid B- share market (Chen, Lee, and Rui, 2001), and information asymmetry between the A- and B- share markets 1

4 (Chui and Kwok, 1998; Chakravarty, Sarkar, and Wu, 1998; Yang, 2003). Some studies, however, provide other reasons. For example, Chen, Lee, and Rui (2001) find that B- shares tend to move more closely wh market fundamentals than A-shares. Mei, Scheinkman, and Xiong (2005) find that Chinese domestic investors mainly trade A-shares based on speculation. For the first time in history, the Chinese government allowed domestic investors to trade B-shares from February 19, 2001 to transform the market segmentation policy that had been implemented in China for more than a decade. The inial market reactions to this policy change were enormous. For example, the trading volumes of B-shares had doubled and were even higher than those of A-shares during In contrast, the mean and median prices of A-shares dropped to be only two times higher than those of B-shares in 2001, while the mean and median prices of A-shares were more than five times higher than those of B-shares in The mean and median price differences between the two-class shares, however, dropped to around 55% and 70% respectively in Chinese listed companies follow two sets of accounting standards if they issue both A- and B- shares. Chinese Accounting Standards (CAS) are used to prepare and aud primary financial statements for the A-share market where shares are denominated in local currency (Reminbi). International Financial Reporting Standards (IFRS) are used to prepare and aud financial statements for the B-share market where shares are denominated in US dollars on the Shanghai Stock Exchange (SHSE) or in Hong Kong dollars on the Shenzhen Stock Exchange (SZSE). Moreover, financial statements are auded by local and international accounting firms for the A- and B- share markets, 2

5 respectively. In practice, Chinese companies issuing both classes of shares are allowed to prepare their primary financial statements using CAS first and then reconcile the key financial figures and ratios to those using IFRS. Finally financial statements for the A (B) market are reported in Chinese (English). Many studies have investigated the value-relevance of accounting information in China and provided very mixed results for the A- and B- share markets. Some studies (e.g. Abdel-Ktalik, Wong, and Wu, 1999; Eccher and Healy, 2000; Hu, 2002; Lin and Chen, 2005) find that CAS accounting information is value-relevant and provides incremental price information beyond IFRS accounting information. Other studies (e.g. Bao and Chow, 1999; Sami and Zhou, 2004), however, find that IFRS accounting information is valuerelevant and provides incremental price information beyond CAS accounting information. There are two potential reasons why previous studies fail to provide consistent evidence as to whether CAS or IFRS can provide more useful information for share valuation in the Chinese stock markets. First, previous studies may not provide comparable results because they have used various samples and research designs (Sami and Zhou, 2004). Second, the information environment (Abdel-Khalik, Wong, and Wu, 1999) and market condions and efficiencies (Aboody, Hughes, and Liu, 2004) of the Chinese stock markets have been largely ignored in previous research. Previous studies use data before 2001 and do not directly investigate the potential impacts of the market segmentation policy on the condions and efficiencies of the Chinese stock markets, which may have significant implications for the value-relevance of accounting information. The change of the market segmentation policy in China provides a unique setting to 3

6 investigate how the changes in price difference and trading volume between A- and B- shares have affected the association between share price and accounting information. Following previous studies, we investigate this issue by comparing the explanatory power of earnings and book value of equy for share price in the A- and B- share markets in the pre- and post- market segmentation periods. In doing so, this study integrates and contributes to two strands of lerature. They are (a) the value-relevance of accounting information in the Chinese stock markets and (b) the effects of the market segmentation policy on share valuation in China. Using Chinese data during the period of , we investigate and find that the association between share price and accounting information, measured by earnings and book value of equy, is much weaker for A-shares than B-shares when the Chinese stock markets were fully segmented before The association between A-share price and accounting information, however, increased after In contrast, the price-accounting association significantly decreased for B-shares during the transion period of when the share price and trading volume of B-shares reached their highest levels. The price-accounting association for B-shares, however, significantly increased after The above empirical results suggest that the market segmentation policy in the Chinese stock markets has significantly affected the usefulness of accounting information for share valuation in China. Different from most previous studies that examine how the qualy of accounting information affects share valuation, this study contributes to accounting lerature by providing empirical evidence on the role of market condion and information environment 4

7 in the association between share price and accounting information in an emerging market such as China. The rest of the paper is organized as follows. Section 2 discusses the market and accounting reforms in China. Section 3 summarizes prior research and develops the testable research hypotheses. Section 4 describes the sample selection creria and data collection. Research methods are discussed in Section 5. Section 6 reports empirical results. Section 7 summarizes empirical results from the robustness tests. The final section summarizes and concludes. 2. Market and Accounting Reforms in China Chinese stock markets are unique because of their various classes of shares and types of ownership. Chinese companies have issued A-shares to raise capal from domestic investors since International investors have been allowed to trade B-shares in the Chinese stock markets since February 21, Domestic investors have been permted to trade B-shares since the market segmentation policy was modified on February 19, The average price difference between A- and B- shares was significantly reduced from five times in 1999 to approximately two times in Besides the change in the market segmentation policy in 2001, China has harmonized s domestic accounting standards to be in line wh IFRS since CAS has been revised in 1992, 1998, and 2001, respectively, although significant differences in accounting measurements still exist between IFRS and CAS. They include inventory, longterm equy investments, revenues, leases, the valuation of fixed assets, amortization of 5

8 intangible assets, methods of income tax accounting, measurement and amortization of goodwill and intangible assets, and research and development costs etc. (Delote Touch Tohmatsu, 2002). Generally speaking, CAS is a more cash- and tax- oriented accounting system, whereas IFRS is a more fair value-oriented accounting system. Despe the significant discrepancies between CAS and IFRS, the correlation coefficients between CAS and IFRS earnings appear to be highly correlated. This may be because Chinese managers attempt to manage their accounting numbers to avoid reporting large disparies between the two accounting systems (Eccher and Healy, 2000). However, China has improved the regulatory environment in recent years to increase the usefulness of accounting information provided by Chinese companies, For example, the Chinese government has changed regulations to increase accounting disclosure 1, prevent inside trading, and implement the best practices of corporate governance for listed firms since the early 2000s. 2 (Wu, Koo, and Kao, 2007). Moreover, there have been changes in company ownership over the last few years. Chinese listed companies normally have four different types of ownership. First, state shares are those shares owned by central and local governments, which can only be transferred to domestic instutions and are not allowed to trade on the stock exchanges. Second, legal person shares are those shares owned by domestic instutions and foreign 1 The CSRC requires listed firms management, assets, and finance and treasury functions to be independent from major shareholders. All listed firms must disclose party transactions and intangible asset information. In addion, all listed firms are required to post their annual reports on designated Web ses. 2 If companies are prepared to issue 300,000 or more shares, they must conduct an addional aud in accordance wh IFRS. The Chinese Supreme Court has also established procedures for investors to sue corporate directors and managers for missing or false information. The code requires independent directors on the board, shareholder rights, and related-party transactions be disclosed in an adequate and timely manner. 6

9 partners of joint ventures. They are non-tradable and virtually controlled by the government. Third, employee shares are those shares offered to managers or employees by Chinese listed companies; these shares can only be tradable after filing for permission wh the Chinese Securies Regulatory Commtee (CSRC). Finally, tradable shares are those tradable A-shares, B-shares, H-shares, and N-shares. Of these, H-shares are only issued and traded at the Hong Kong Stock Exchange; N-shares are only listed and traded at the NYSE. A breakdown of the ownership structure of Chinese listed companies during the period of is reported in the Panel D of Table 1. Although the CSRC allowed non-tradable shares to be gradually sold to the public investors from June 2001, the CSRC had to stop this activy in June 2002 because both the SHSE and SZSE Indices had significantly dropped due to the impact of the fall of internet share prices. Average state ownership stood at about 70% by the end of On April 29, 2005, the CSRC announced the spl-share structure reform plan to convert all non-tradable shares to be publicly tradable, which has triggered a significant increase in the share price and liquidy of both A-share and B-share markets, though the B-share market is not even directly involved in the reform. In June 2006, 1006 A-share companies completed the transformation and significantly improved the liquidy of Chinese stock markets (CFA, 2007). The market segmentation policy in China has been modified because, since February 19, 2001, B-shares can be traded by both domestic and foreign investors. However, the Chinese stock markets are somewhat segmented due to the following three factors. First, A-shares can only be traded by domestic investors. Second, the requirement 7

10 that US or Hong Kong dollars be used to trade B-shares remains the same. Finally, the trading volumes of B-shares are still much smaller than those of A-shares, although they significantly increased after Prior Research and Hypothesis Development Two strands of lerature are related to this study. They are (a) the effects of the market segmentation policy in China on share valuation, and (b) the value-relevance of accounting information in the Chinese stock markets. We provide separate summaries of previous findings in these two areas and develop two testable hypotheses in the following sections. 3.1 Impacts of Market Segmentation on Share Valuation Bailey and Jagtiani (1994) and Domowz, Glen, and Madhavan (1997) examine how market segmentation induced by ownership restrictions affects stock price in Thailand and Mexico, respectively. Both studies report substantial price premiums for foreign shares (i.e. shares traded by foreign investors only) over domestic shares (shares traded by domestic investors only). The empirical evidence for the Chinese stock markets, however, indicates that foreign investors actually pay much less for B-shares than domestic investors do for A-shares. As mentioned previously, the mean and median prices of A-shares were over five times higher than those of B-shares in Many studies investigate the underlying reasons for the substantial price difference between A- and B- shares. For example, Chakravarty, Sarkar, and Wu (1998) argue that the discounts in B-share prices exist because foreign investors have difficulties in 8

11 acquiring and assessing information about local Chinese companies due to language barriers, different accounting standards, and lack of reliable information about the local economy and companies. Chen, Lee, and Rui (2001), however, argue that the demand elasticy for A- and B- shares could be different because the investment opportunies for domestic investors are limed, while foreign investors can easily diversify their investments in global capal markets. They find that the big discounts of B-share prices are primarily caused by the illiquid B-share market. Under this relatively illiquid market, the expected return is higher and share price is lower to compensate investors for increased trading costs. Zhang and Zhao (2004) also argue that the substantial price difference between A-share and B-share prices can be an indicator for the polical risk in Chinese stock markets perceived by foreign investors due to inadequate protection of shareholders rights under such a transional economy. Using Chinese data for the period of , Chakravarty, Sarkar, and Wu (1998) provide evidence that domestic investors are more informed than foreign investors because foreign investors rely more on financial reporting information, while Chinese domestic investors have access to alternative local information sources in addion to financial reporting information. Some studies (e.g. Chui and Kwok, 1998; Yang, 2003; Froot and Ramadorai, 2007), however, document that foreign investors have a relative information advantage because foreign investors are mainly instutional investors who are more experienced and well equipped wh advanced technology to analyze the accounting information than Chinese domestic investors. Chan, Menkveld, and Yang (2007) find that the signed trading volume and quote revision of the A-share market had strong predictive abily for B-share quote returns, but not vice versa before February 19, In other 9

12 words, domestic investors had a relative information advantage before the change in Chinese market segmentation policy. However, they find reverse causaly from the B- share market to the A-share market after February 19, 2001, indicating that the information transmission between the A- and B- share stock markets increased the information advantage of foreign investors after the change in the market segmentation policy. Since Chinese domestic investors are now allowed to trade B-shares in the Chinese stock markets, there are two important implications of this change. First, the liquidy of the B-share stock market could be significantly improved because of an increase in the demand of B-shares after this policy change. This may result in an increase in B-share price. Second, part of the investment funds that were invested in the A-share stock market might be shifted to invest in the B-share stock market simply because B-share price was much lower than A-share price. This may result in an increase (a decrease) in B- (A-) share prices. The above two factors may have jointly reduced the difference in price and trading volume between A- and B- share stock markets. Furthermore, allowing Chinese domestic investors to trade B-shares may have reduced the information asymmetry or improved the information environment in the segmented Chinese stock markets, which could significantly reduce the price difference between A- and B- shares and therefore increase the association between share price and accounting information. We therefore predict that the change in the market segmentation policy in China may have improved the association between share price and accounting information in both the A-share and B-share stock markets. Hence, we develop the following hypothesis (in alterative terms). 10

13 H1: The change in the market segmentation policy in China has increased the value-relevance of accounting information in both the A- and B- share markets. Many studies have also examined the value-relevance of accounting information, measured by earnings and book value of equy, for both the A- and B- share markets, but found rather mixed results. For example, using Chinese data for 1994 and 1995 Abdel- Khalik et al. (1999) find a significant association between earnings and abnormal returns for A-shares but not for B-shares. They argue that the high price volatily, dominance of government officials, and illiquid B-share market may have contributed to the usefulness of accounting information for B-share price. Using data during the period , Eccher and Healy (2000) provide evidence that IFRS earnings do not provide incremental price relevant information beyond CAS earnings for companies issuing B-shares. However, CAS earnings can explain the variations of share returns better than IFRS earnings for companies issuing A-shares. More interestingly, they find very high correlation between IFRS and CAS earnings and argue that neher IFRS nor CAS were effectively enforced in China. Using data during the period of , Chen, Chen, and Su (2001) provide evidence indicating that Chinese domestic investors use CAS accounting information when valuing A-shares despe some evidence of inadequate accounting and financial reporting practices in China. Their finding suggests that the price model provides stronger evidence than the return model. A similar study by Chen, Firth, and Kim (2002) finds that both IFRS and CAS accounting information are useful in explaining the variations of A-share and B- share prices. They argue that IFRS accounting information may be used by A-share investors because there are no barriers for domestic investors to have access to the 11

14 financial statements prepared for B-share investors. Chen et al. (2002) also document that accounting information is more correlated wh share price for companies wh higher individual and non-state ownership. Using more recent data (i.e ), Lin and Chen (2005) re-examines the value-relevance of accounting information for companies simultaneously issuing A- and B- shares. They find that CAS earnings and book value of equy jointly can better explain the variations of A- and B-share prices than IFRS earnings and book value of equy. They also find that the CAS earnings are associated wh annual share returns in both stock markets, although the association appears to be stronger in the A-share market. Overall their study provides evidence indicating that CAS accounting numbers are more value-relevant than IFRS accounting numbers in Chinese stock markets at present. At least two studies provide evidence that IFRS accounting information provides incremental price information beyond CAS accounting information. Using data during the period of , Bao and Chow (1999) find that IFRS earnings and book value of equy jointly can better explain the variation of B-share price than CAS earnings and book value of equy. In addion, Sami and Zhou (2004) investigate the relative value-relevance of accounting information prepared and auded by IFRS and CAS respectively, and find that accounting information is reflected in the price process in both the A- and B-share markets and that the accounting information in the B-share market is more value-relevant (i.e. higher explanatory power of earnings and book value of equy for share price) than in the A-share market. They also find that the value-relevance of accounting information in the A-share market was very low in earlier years (i.e and 1995) and decreased in later examined years (i.e and 2000). They do not further examine the reasons 12

15 underlying their findings. In summary, the relative value-relevance of the accounting information prepared and auded by IFRS and CAS appears to vary in previous studies. It is therefore predicted that the change in the market segmentation policy in China may have different impacts on the value-relevance between share price and accounting information in the A- and B- share markets. The second hypothesis is developed as follows (in alternative form): H2: The change in the market segmentation policy in China has different impacts on the value-relevance of accounting information for both the A- and B- share markets We examine the above research question by using the following data and research methods. 4. Research Methods This study primarily uses the accounting-based valuation model suggested by Ohlson (1995) for three reasons. First, this model considers both earnings and book value of shareholders equy, which have been widely used in the lerature to evaluate the value-relevance of accounting information (e.g. Collin at al., 1997; Barth at al., 1998; Francis and Schipper, 1999). Second, previous studies (e.g. Bao and Chow, 1999; Hu, 2002; Chen, Firth, and Kim, 2003; Lin and Chan, 2005; Sami and Zhou, 2004) have used this model to investigate the value-relevance of accounting information in China 3. We use 3 Brown, Lo, and Lys (1999) argue that the between-sample comparisons of R2 from regressions of share price on earnings and book value per share are invalid unless the difference in the scale factor s coefficient of variation is controlled. This study is unable to investigate this issue due to limed observations in our sample (less than 10 years). Sami and Zhou (2005) show that their results using different scales (book value and number of outstanding shares) are qualatively the same. To further 13

16 this model to compare our results wh previous findings. Finally, Chen, Chen, and Su (2001) find that the price model provides stronger results than the return model. We therefore use the return model suggested by Easton and Harris (1991) for robustness purposes. Under this model, annual share returns of A- and B- shares regress on the level of and change in CAS and IFRS earnings. Details of these two models are described as follows. Model 1: Price Model P A( B) = α + α E + α BV + ε 0 1 CAS ( IFRS) 2 CAS ( IFRS) Where A(B) P is the share price per share for A- (B-) shares at the end of the fourth month after the fiscal year-end for firm i for period t; CAS (IFRS) E is earnings per share based on BV CAS (IFRS) CAS (IFRS) for firm i for period t; is the year-end book value of shareholders equy based on CAS (IFRS) for firm i for period t. We report results using cross-sectional (i.e. yearly) and pooled regressions, and also consider the time-series cross-sectional regression suggested by Fama and MacBeth (1974). Model 2: Return Model RET A( B) CAS ( IFRS ) A( B) CAS ( IFRS ) CAS ( IFRS ) A( B) = 0 + [ E / Pi ( t 1) ] + β1[ E Ei( t 1) ) / Pi ( t 1) ] β + ε Where RET A( B) is the cumulative 12-month raw return for A- (B-) shares for firm i, investigate this issue, this study uses the return models for the robustness test purpose and finds very consistent results wh the price model. 14

17 starting from four months after prior fiscal year-end; E / P is CAS (IFRS) CAS ( IFRS ) A( B) i( t 1) earnings per share, deflated by the prior year-end share price of A- (B-) share for firm i; ( E CAS ( IFRS ) E CAS ( IFRS ) i( t 1) ) / P A( B) i( t 1) is change in CAS (IFRS) earnings per share, deflated by the prior year-end share price of A- (B-) share for firm i. In this study, all the slope coefficients of the above two regression models are adjusted for Whe s (1980) unbiased covariance to correct for heteroskedasticy. In addion, we report adjusted R-squared values of the above regression models in the following tables. Following previous studies (e.g. Collin et al., 1997; Francis and Schipper, 1999; Eichenseher, 2000; Sami and Zhou, 2004), the explanatory power of accounting information for share price or share return (i.e. adjusted R-squared value) is used as an indicator of the value-relevance of accounting information. Hence, if H1 is true, then we should observe a higher explanatory power of earnings and book value of shareholders equy (earnings and change in earnings) for share price (share return) in the postsegmentation period versus the pre-segmentation period. On the other hand, if H2 is true, then we should observe that the change in the explanatory power of earnings and book value of shareholders equy for A-share price (return) is different from that for B- share price (return) between the pre- and the post- segmentation periods. Following previous studies, we also consider the statistical significance of slope coefficients of earnings, book value of equy, and change in earnings in the above two regression models. 15

18 5. Sample Selection and Data Collection We include all the industrial firms issuing both A- and B- shares on the SHSE or SZSE wh all the accounting and market data available in the Taiwan Economic Journal (TEJ) for the period of (i.e. four years before and after 2001). We report our empirical results using the whole sample because of two reasons. First, our empirical results after deleting potential outliers (top and bottom 1% observations for each examined variable) are qualatively consistent wh the results reported in this study; second, this allows us to maximize our sample observations. Table 1 describes our sample characteristics. Panel A shows that our sample is comprised of 705 (669) firm-years wh price (return) data available in the TEJ database. Only 622 firm-years have complete ownership data from TEJ. Panel B reports the industrial sectors in which our sample companies operate. Around 42% of our sample companies are in the manufacturing sector, which is the dominant industry in China, followed by the real estate (13%), service and transportation (11%), electronic (9%), and chemicals and metal (9%) industries. Panel C shows that more Chinese companies are listed on the Shanghai Stock Exchange than Shenzhen Stock Exchange. Panel D reports the ownership distributions. More than one third of the shares of our sample companies were owned by the State, although the percentage of state shares was slightly reduced after Tradable share was increased to nearly 50% from 45%, due to the spl-share structure reform plan launched in 16

19 2005, which was both statistically significant and economically material. Both domestic and foreign instutional shareholdings slightly decreased. Public shareholding, i.e. shares owned by individual investors, increased about 2.4%; shares traded on the Hong Kong Exchange remain stable. [Insert Table 1] 6. Empirical Results 6.1. Descriptive Statistics Panel A of Table 2 shows that the mean and median share prices of A-shares are consistently higher than those of B-shares, although their differences become smaller in later years. A- and B- share prices reached their peak in 2000, an unusual year for the B- share market because the average B-share price increased around four times compared to the average price in B-share price appears to be significantly inflated after the announcement of the change in the market segmentation policy. Both A- and B- share prices significantly dropped in 2001 (about 16% and 27%, respectively) and continuously decreased afterwards because of the impact of the fall of Internet share prices. Figure 1 shows that both A- and B- share prices and their price differences continuously decreased after Untabulated results show that the mean price difference between A- and B- shares reached s highest level in 1999 and was then reduced by 33%, 26%, 21%, and 53% in years 2000, 2002, 2003, and 2004, respectively. The mean price difference, however, increased by 1% in 2005, which was caused by increased share prices due to the spl-share plan. The mean A-share price was over six times higher than the mean B-share price in 17

20 1999; the price difference, however, was reduced to around 1.5 times in [Insert Figure 1] Figure 2 shows that the mean and median trading volumes of A-shares were over three times more than those of B-shares before 2000, while they were less than two times more in 2000 and onwards. The average trading volumes of B-shares in 2000 and 2001 were extremely high, which were over 1.4 times and 1.9 times higher than those of A- shares respectively. The above findings suggest that the change in the market segmentation policy significantly increased both trading volume and share price of B-shares during the transion period (i.e and 2001). Overall, we find evidence that the price difference between A- and B- shares was significantly reduced around the change in the market segmentation policy in [Insert Figure 2] Panel B of Table 2 shows that the patterns of share returns for both A- and B- shares are very similar except for 1999 and The mean and median share returns of B- shares in 1999 and 2000 were 352% and 336%, respectively, over 14 times higher than those of A-shares. B-share prices were significantly inflated in 1999 and Table 3 reports the descriptive statistics of the variables used for this study. Consistent wh Panel A of Table 2, the average A-share price is over two times higher than the average B-share price during the entire test period. Untabulated results show that the median A-share price is approximately five times (1.8 times) higher than the median B- share price before 2000 (after 1999). Interestingly, both the mean and median earnings per 18

21 share based on IFRS and CAS appear to be identical during the entire test period. As documented by Eccher and Healy (2000), Chinese managers seem intend on reporting very similar IFRS and CAS earnings. On the other hand, both the mean and median book values of shareholders equy per share based on CAS are slightly greater than those based on IFRS, although their differences appear to be insignificant. Untabulated results also show that earnings and book values of shareholders equy based on IFRS and CAS are much closer after 2000 because CAS has been further reformed to be aligned wh IFRS. Panel B of Table 3 reports that A-share and B-shares have posive means but negative median annual share returns, indicating their distributions are significantly skewed. This is consistent wh the finding that the standard deviations of the share returns for both A- and B- shares are approximately 20 times higher than their mean annual share returns. Untabulated results also show that the correlation coefficients between A-share and B-share annual returns increase over time, indicating that both share markets become more connected in recent years. Consistent wh Panel A, the mean and median earnings based on CAS and IFRS are identical when they are deflated by prior year-end price of A- shares. The average earnings based on CAS are greater than those based on IFRS when both earnings are deflated by prior year-end price of B-shares. The deflated mean and median changes in earnings based on CAS are also slightly greater than those based on IFRS. [Insert Table 3] 19

22 6.2. Regression Results Panels A and B of Table 4 report the statistical associations between share price and earnings and book value of equy for A- and B- shares, respectively. Panel A shows that the explanatory power of CAS and IFRS earnings and book value of equy for A- share price is generally low during (i.e ). The slope coefficients of earnings and book value of equy also appear to be very volatile. For example, the slope coefficients of both CAS and IFRS earnings are negative in 1999 but become posive in 2000, although they are never significant in both years. They, however, become posive and significant in The slope coefficients of both CAS and IFRS book value of equy are extreme, but they are posive and significant only in [Insert Table 4] Three factors may have contributed to the above weak results. First, Chinese domestic investors trade A-shares based on speculation (Mei, Scheinkman, and Xiong, 2005) instead of market fundamentals (Chen, Lee, and Rui, 2001), causing low explanatory power of accounting information for share price and volatile slope coefficients. Second, A-share prices were significantly inflated during this period. Previous studies (Bailey and Jagtiani, 1994; Domowz, Glen, and Madhavan, 1997) find that shares that can only be traded by foreign investors have higher market prices than shares that can only be traded by domestic investors in emerging capal markets such as Mexico and Thailand. In contrast, A-shares that can only be traded by domestic investors in China have 20

23 much higher share prices than B-shares that could only be traded by foreign investors before A-share prices were significantly inflated because domestic investors could not invest in the B-share market due to the market segmentation policy. Inflated and biased A-share prices (i.e. dependent variable of the accounting-based valuation model) could cause low explanatory power of accounting information for A-share price. Finally, is also possible that the implementation of new accounting standards, which first took effect in 1998, was not effective because the new accounting system allows managers to exercise more professional judgment to reflect the firm-specific business environment that their companies face. Previous studies argue that the Chinese market infrastructure was still at s early stage and was not ready to support fair value accounting (Eccher and Healy, 2000; Chen et al., 2001). As a result, biased accounting information (i.e. independent variables of the accounting-based valuation model) could drive the slope coefficients of earnings and book value of equy to be downward biased. We observe completely different results during the period of The slope coefficients of CAS and IFRS earnings and book value of equy for A-share price are generally significant and less volatile. The slope coefficients for earnings (book value) are generally greater (less) than 1 except the earnings prepared by IFRS in More importantly, the adjusted R-squared values of the price-accounting regressions significantly increase especially when accounting information is prepared by CAS, indicating that the change in the market segmentation policy indeed reduced the price difference between A- and B- shares, which might have significantly increased the association between share price and accounting information during this period. 21

24 Another interesting finding in the yearly A-share regressions is that the explanatory power of CAS earnings and book value is fairly close to that of IFRS earnings and book value, indicating that IFRS and CAS accounting information are equally considered in the A-share stock market. Although the average explanatory power of accounting information increases from 2% to 5% after 2000, Panel A clearly shows that the average explanatory power for the period of is much higher than that for the period of This indicates that the value-relevance of accounting information for A-shares was significantly improved due to the change in market segmentation policy. Our empirical evidence therefore supports H1. After partioning our sample into the pre- and post- segmentation periods, we find that only earnings (book value of equy) are associated wh share prices in the pre- (post-) market segmentation period. Moreover, only CAS earnings or IFRS book value of equy can explain the variation of A-share price when using the entire test period. The explanatory power of CAS earnings for A-share price is, however, identical to that of IFRS book value of equy (i.e. 3%). Untabulated results using the time-series cross-sectional regressions (Fama and MacBeth, 1974), however, show that eher CAS or IFRS earnings can explain the variation of A-share prices but not any book value of equy, indicating that earnings are the dominant explanatory variable for the variation of A-share price. The above results generally suggest that both CAS and IFRS earnings and book values of equy can explain the variation of A-share prices, which is consistent wh previous studies (e.g. Chen, Firth, and Kim, 2002; Sami and Zhou, 2004). Panel B of Table 4 shows that the adjusted R-squared values are high across the 22

25 entire test period except years 2000 and 2001 when the trading volumes and share prices of B-shares were unusually high (see Figure 2 and Panel A of Table 2). Although the explanatory power of IFRS accounting information is higher than that of CAS accounting information in 5 out of 9 examined years, the overall explanatory power of IFRS and CAS accounting information for B-share price after 2000 are identical (i.e. 19%). The average explanatory power of accounting information for B-share price significantly increases to 19% from 5% or 7% after Overall, the explanatory power of accounting information for B-shares is high during the periods of and This indicates that accounting information is generally value-relevant for B-shares across the entire test period except the transion period of and that the change in market segmentation policy in China temporarily reduced the value-relevance of accounting information for B- shares. The association between accounting information and B-share price returns to normal after Compared to the findings reported in Panels A and B, we find that the increase in the adjusted R-squared values of B-shares between the pre- and post- periods is between 12% and 14% (i.e. 19%-7% and 19%-5% when IFRS and CAS accounting information is used respectively). On the other hand, the increase in the adjusted R-squared values of A- shares is only between 2% and 3% (5%-3% and 5%-2% when CAS and IFRS accounting information is used respectively). Hence, the impact of the change in market segmentation policy on the explanatory power of accounting information for share price varies between A- and B- shares. We therefore find evidence supporting H2. This study finds that both share markets benef from the change in market 23

26 segmentation policy in terms of increased associations between share price and accounting information, although there is some evidence indicating that the B-share market benefs more from the change in market segmentation policy than the A-share market. To further investigate this issue, we compare the adjusted R-squared values of B-share price regressions based on IFRS wh those of A-share price regressions based on CAS. This comparison is based on the argument that financial statements prepared and auded by CAS (IFRS) are designed for the sake of domestic (foreign) investors. Figure 3 demonstrates that the adjusted R-squared values of IFRS earnings and book value for B- share price are consistently higher than those of A-shares across the entire test period, although the differences in their adjusted R-squared values appear to be smaller after However, the explanatory power of accounting information for A-share price is apparently higher in recent years (after 2002) than in early years, indicating that the usefulness of accounting information in the A-share market also significantly benefs from the change in market segmentation policy. [Insert Figure 3] The slope coefficients of CAS and IFRS earnings and book values are generally associated wh B-share prices across the test period except years 2000 and Moreover, the slope coefficients of earnings and book values of equy for B-shares before 2003 are less volatile than those for A-shares. Consistent wh some previous studies, we find that CAS and IFRS accounting information are equally useful for B-share valuation. 24

27 7. Robustness Test Return model We perform three robustness tests in this section. First, we investigate the association between share return and CAS and IFRS earnings and change in earnings in both share markets. Second, we examine whether change in ownership structure in Chinese companies in recent years has any implication for our finding. Finally, we investigate whether further alignment between CAS and IFRS in the Chinese markets in recent years has any implications on our finding. Panels A and B of Table 5 report the regression results using stock returns for A- and B-shares, respectively. Panel A shows that the adjusted R-squared values are generally very small during the period of , especially when using CAS earnings. After 2001, the adjusted R-squared values for CAS and IFRS earnings increase every year except 2005 when A- and B- share prices significantly increase again due to the spl-share plan. Moreover, the adjusted R-squared values based on CAS earnings are generally higher than those based on IFRS earnings, indicating CAS earnings may provide incremental explanatory power for A-share return relative to IFRS earnings. This finding is consistent wh previous studies (e.g. Eccher and Healy, 2000; Chen, Chen, and Su, 2001; Chen, Firth, and Kim, 2002). Overall, the explanatory powers of CAS and IFRS earnings for share return increase 2% (12%-9%) and 3% (10%-8%) respectively after The joint slope coefficients of earnings and change in earnings based on CAS 25

28 (IFRS) are statistically significant except 1998 and 1999 (1998, 1999, and 2002). This is generally consistent wh the findings using the price model. The joint slope coefficients of CAS and IFRS earnings and earnings change are all posive and statistically significant in both the pre- and post- segmentation periods. [Insert Table 5] Panel B shows very similar results to Panel A in that the adjusted R-squared values for B-share regressions are high between when using eher CAS or IFRS earnings. More importantly, the adjusted R-squared values of CAS earnings for B-share returns are generally higher than those of IFRS earnings except for years 2001 and This finding is again consistent wh Panel A, but is somewhat different from the results reported in Panel B of Table 4 in that the explanatory power of IFRS earnings and book value of equy is generally higher that that of CAS earnings and book value. Overall, the explanatory powers of CAS and IFRS earnings increase 8% (10%-2%) and 9% (11%-2%) respectively after The joint slope coefficients of earnings and change in earnings based on IFRS (CAS) are statistically significant except for 1998, 1999, and 2001 (1998 and 2001). The joint slope coefficients of CAS and IFRS earnings and change in earnings are all posive and statistically significant in both the pre- and post- segmentation periods. Finally, the average change in the adjusted R-squared values for A-shares (2% to 3%) is lower than that of B-shares (8% to 9%) after 2000, which is consistent wh the findings reported in Table 4. This indicates that B-shares benef more than A-shares from 26

29 the change in the segmentation policy in terms of increased explanatory power of accounting information for share returns. Following our analysis for the price model, we compare the adjusted R-squared values of B-share return regressions based on IFRS wh those of A-share return regressions based on CAS. Figure 4 demonstrates that the adjusted R-squared values of IFRS earnings and change in earnings for B-share returns are generally higher than those for A-shares except years 2001 and 2004, although the differences in their adjusted R- squared values appear to be more volatile than those using the price model. In summary, our empirical results using the return model are generally consistent wh the results using the price model in that the association between share price (and share return) and accounting information, measured by earnings and book value of equy, is improved in both the A- and B- share markets, although the increased association seems be greater for B-shares. We therefore find consistent empirical evidence supporting both H1 and H2. Ownership As shown in the Panel D of Table 1, the tradable and public shares appear to adequately increase in the post-segmentation period. To investigate whether change in ownership structure increases the explanatory power of earnings and book value of equy for share price and return in this period, we partion our sample into two groups of companies based on the median change in tradable and public shares between the pre- and post- segmentation periods. Untabulated results show that there is no significant difference 27

30 in the explanatory power between these two groups of companies in the post-segmentation period. Accounting reform We also examine whether further alignment between CAS and IFRS in the postsegmentation period has any impact on the explanatory power of earnings and book value of equy for share price and return. Untabulated results show that companies who report exactly the same figures for both CAS and IFRS earnings and companies who report very similar figures for both earnings (i.e. the difference between them is whin 10% of IFRS earnings) have higher explanatory power of earnings and book value of equy for share price and return and more significant slope coefficients than the rest of the sample firms. This finding suggests that further alignment between CAS and IFRS may also contribute to the increased association between share price (and share return) and accounting information in the post-segmentation period. Surprisingly, we find that about 34% of Chinese listed companies report identical CAS and IFRS earnings figures in In the same year, 20% of Chinese listed companies report very similar CAS and IFRS earnings figures (i.e. their difference is whin 10% of IFRS earnings). This is unexpected because there are still significant differences between CAS and IFRS. Future research should further investigate the incentives and consequences to report identical and similar CAS and IFRS earnings figures. 8. Conclusions This paper examines whether the association between share price and accounting 28

31 information, measured by earnings and book value of shareholders equy, has increased after Chinese domestic investors were allowed to trade B-shares from February We also predict the impact of the change in the market segmentation policy on A- and B- share markets should be different because previous studies document that the association between share price and accounting information in the two markets appear to vary before the Chinese capal markets were fully segmented. Using Chinese data for the period , we find that the price difference between A- and B- shares significantly reduced after the change in the market segmentation policy was announced by the Chinese government in The association between share price and accounting information, measured by earnings and book value of equy, is much weaker for A-shares than B-shares when the Chinese stock markets were fully segmented before The association between A-share price and accounting information, however, significantly increases after In contrast, the price-accounting association significantly decreased for B-shares during the transion period of when both the share price and trading volume of B-shares were inflated. The priceaccounting association for B-shares, however, significantly increased after The above results generally indicate that the usefulness of accounting information is significantly improved after the change in the market segmentation policy. Further evidence shows that the increased explanatory power of earnings and book value of equy for B-share price (and return) is generally higher than that of A-share price, indicating that the change in the market segmentation policy has a greater impact on the B- share market than the A-share market. We therefore find empirical evidence supporting our 29