1. Introduction

Information functions as the core of capital market in the sense that all sides of market participants make decisions based on their accessible information sets. As one of the major sources of information in public equity market, company financial reporting lies in the center of company performance relevant information set and is the basis of most investors' analysis and decision-making process. ‘Fundamental analysis' is how people usually refer to for this kind financial performance information analysis, and it is critical especially to long-term value stock investors who have much heavier reliance on analyzing ‘fundamentals' among other factors than short-term investors to implement investment strategy.

At the same time, multiples including price to earnings (P/E) or price to book (P/BV) are two most commonly used metrics now in assisting the ‘fundamental' based investment decision-making. Studies on the value relevance of company financial indicators or multiples become potentially interesting from a value investor perspective.

There had been some studies along the above-mentioned directions and especially together with the fact of segmented Chinese stock markets, in an effort to derive implications on asset pricing or potential trading strategies by looking at price or return premium for dual-listed companies as well as value relevance with accounting information for Chinese listed company. Key research from these various angles, which will be reviewed in more detail in later section, gave some insightful findings although empirical findings may not always be consistent due to different research design or samples difference. There is still considerable space for further research to contribute. By taking up one specific angle, this paper aims to provide more recent evidence on Chinese dual-listing return association with the two multiples (P/E and P/BV, the inverses are used in the model) in the condition of segmented Chinese share market. Most prior studies were conducted based on AB dual-listing samples, and not so much recent similar studies were done on AH dual-listed share samples. This paper therefore looked at Chinese dual-listed companies on A and H share market to supplement this less researched area. Another reason why a AH dual-listing study can be more practically interesting than that on AB dual-listings is due to the fact that B share market was far less liquid as H share market thus less meaningful for most investors.

The paper is organized as follows. Section 2 discusses the background of this study, including introductions of Chinese stock market especially A and H share markets, followed by a sub-section introducing Chinese accounting and auditing system. Section 3 reviews prior research on stock value relevance of company financial information given market segmentation as well as return predictability based on price multiples. Hypotheses and assumptions of this paper are derived accordingly Section 4, elaborated with Section 5 and 6 on data and research methodology respectively. The last section reports the empirical results and provides explanations for the results. References for this research are listed in the appendix.

2. Background

2.1 China A and H stock market

Chinese stock market has been segmented ever since 1991 when the first stock exchange opened in Shanghai, along with its gradual development path was …(brief history)

Currently there are four major types of shares in Chinese corporations, namely government shares, legal entity shares, employee shares and traded shares which further comprise of A-shares, B-shares, H-shares and N-shares. Differences among these shares mainly lie in the ownership restrictions, as further demonstrated in Table XX. The ownership restrictions therefore create another layer of segmentation on top of the investment barrier the central government set between mainland China stock market as a semi-open market and Hong Kong stock market as an international market.

The long term direction of Chinese capital market reform is towards a more integrated and open market, and the Chinese government had carried out a series of important barrier releasing actions including ‘Reform of Non-tradable Shares' in A share market that aims to release the trading restrictions on government shares and legal entity shares. Removal of certain restrictions laid between markets, for example, impoving A and B share market integration by allowing mainland China investors buying B shares directly, had also been good example of Chinese capital market liberalization. However, the not yet integrated market of A and H shares and investment strategy implications of it is the focus of this paper.

Table XX. Comparison of main Chinese stock markets

Exchange

Investor

Currency

Settlement

A share

Shanghai Stock Exchange (SSE)

Domestic individual investor (mainland China citizens), institutional investors including QFII

CNY

T+1

Shenzhen Stock Exchange (SZSE)

B share

Shanghai Stock Exchange (SSE)

Foreign and domestic (opened to domestic individuals since Feb 2002) individual or institutional investors

USD

T+3

Shenzhen Stock Exchange (SZSE)

H share

Hong Kong Stock Exchange (HKSE)

All investors in HKSE, except for China mainland individual investors, who can only indirectly invest through mainland QDII

HKD

T+0

T+2

HKSE

SSE

SZSE

No. of listed companies

Market cap

No. of listed companies

Market cap

No. of listed companies

Market cap

1,426

2,751,672

905

2,905,570

1246

1,336,394

(no. of H shares: 163)

Note: market cap in USD, all numbers are those as of 31 Mar 2011.

B share market became a more integrated market with A share market after the policy change allowing domestic individual investors trading B shares in Feb 2002. H share market used to see a similar policy ease on allowing domestic individual investor buy Hong Kong listed shares (including H shares) directly through a program called ‘Hong Kong Stock Express', however the program was halted after being implemented for a while in the concern of over-release of China's capital account. Cautious but progressive efforts were being made after that to further strengthen the integration of Hong Kong and mainland China capital markets, but so far Chinese mainland investors can only directly invest in H shares through certain institutional investors authorized as qualified domestic institutional investors (QDII).

Mainland Chinese companies have significant presence in Hong Kong stock market. As of March 2011, there are 163 H shares (companies incorporated in mainland China and listed in Hong Kong), 103 red chip stocks (companies incorporated outside China and listed in Hong Kong), and 334 Non-H share mainland private enterprises, totaling 57.2% of market capitalization and 65% equity turnover in Hong Kong stock exchange. 66 of the 163 H shares have dual-listing in mainland's two A share markets, called A-H companies. These A-H companies are the basis of the empirical study sample of this paper.

Table XX

Mainland Enterprises (Main Board and GEM)

No. of H shares

163

No. of Red chips Stocks

103

No. of NHMPE

334

Market capitalisation (% of market total)

57.20%

Turnover value (% of equity turnover)

65.00%

Note: 1) NHMPE = Non-H Share Mainland Private Enterprises.
2) March 2011, Month-end figures

2.2 China accounting and auditing practices for A and H shares

Companies that issue A- and H-shares follow different financial reporting systems, under which share companies prepare reports under PRC GAAP (ABSE) that are audited by local CPA firms, while H share companies prepare reports under and the International Financial Reporting Standards (IFRS) (IFRS) or Hong Kong GAAP that are audited by international auditors (Big 4). AH dual-listing companies have to file financial reports under both systems until December 2010, when, Hong Kong Stock Exchange announced accepting H share companies reporting under Chinese GAAP as a way of reducing company compliance costs.

Differences between Hong Kong GAAP and IFRS is minimal since 2005 except for a few minor differences (Deloitte 2005), when great efforts were made to integrate Hong Kong GAAP and IFRS. Differences between PRC GAAP and IFRS are bigger than that of Hong Kong GAAP and IFRS, therefore in many cases, key company financials such as earnings or net assets appear differently in the same financial reports. This, together with segmented A and H share markets provide a practical need for research conducted in this paper. The latest approved one system reporting/auditing regulation by HKSE removed the policy barrier financial reporting gap now existing among AH companies, and it is expected that in the long run more AH companies will only adopt PRC GAAP to cut their compliance costs. However in the short term this policy will not change current practice of the majority AH companies mainly due to their concern on whether international investors can adapt well to the PRC GAAP reporting instead of IFRS reporting. As of 18 April 2011, only about 10% of the small and medium-sized AH companies shifted towards PRC GAAP reporting only (Sina.com.hk news), big companies took a wait-and-see strategy and mostly had concern on investors negatively interpret them shifting to one accounting system only immediately.

Table XX. Comparison of accounting and auditing practise in Chinese stock markets

Share

Accounting standard

Auditors

A-shares

PRC GAAP

  • ocal auditing firms

  • B-shares

    IFRS

    International auditing firms

    H-shares

    IFRS/HK GAAP

    International auditing firms

    AB-shares

    PRC GAAP & IFRS dual reporting

  • ocal CPA & international auditing firms

  • AH-shares

    PRC GAAP & IFRS/HK GAAP dual reporting

  • ocal CPA & international auditing firms

  • 3. Previous research

    3.1 Chinese stock market segmentation and A-H share premium disparity

    There had been considerable amount of literature on how market segmentation affected share price premium for multi-listings. Empirical studies on this issue usually found out that countries where ownership restrictions of stocks exist, the foreign shares is trading at a price premium over the domestic counterpart shares. Countries of such example include Finland (Hietala, 1989), Thailand (Bailey and Jagtiani, 1994), Switzerland (Stulz and Wasserfallen, 1995), and Mexico (Domowitz et al., 1997). However the case for China is reversed in the way that Chinese foreign shares had been trading at price discount over domestic counterpart shares (Bailey, 1994; Su, 1997; and Fernald and Rogers, 1998). Reasons for Chinese foreign share price discounts were researched also, including four major hypotheses: Differential risk hypothesis, which assumes that foreign investors require lower risk premium than domestic investors on an unrestricted stock (Hietala 1989); Differential demand hypothesis, which assumes different stock demand elasticity facing different investor groups; liquidity hypothesis which is based on stocks traded with varied liquidity leve, and information asymmetry hypothesis addressing premium caused by information gap.

    Most literature research that test these hypotheses were based on empirical studies on A and B share market rather than A and H share markets but a few recent studies had provided more evidence on A and H share comparative studies which also laid out foundations for this paper.

    On what makes A and H share markets segmented, Wang and Jiang (2004) and Li et al. (2006) argue that stock ownership as well as listing and trading locations manifests the kind of segmentation between the A- and H-share markets. Hing-Wah Lee (2009) reinvestigated the liquidity hypothesis following a market-microstructure approach, and found out that first, China A-shares on average provide better market liquidity than their Hong Kong H-share counterparts do. Second, after controlling for traditional liquidity measures and variables related to competing hypotheses, the percentage differences in quoted spread and depth between A-shares and H-shares still explain significantly the price premium.

    3.2 Value relevance of company financials

    The effect of different accounting systems along with China's segmented share market had been researched in the past decade. And most of the research made efforts to address the issue of value relevance of accounting information in China as one of the emerging markets. There were mainly two types of models that were used, price model and return model (Ball and Brown, 1968; Collins and Kothari, 1989; Kothari and Zimmerman, 1995; Kothari, 2001) that used share price or return as dependant variables on a series of independent financial variables such as earnings and book value of equity. Evidence was found to support A share value relevance of earnings reported under PRC GAAP from studies including Haw et al. (1999) on their price model built on entire population A-share during 1994 and 1997, and Chen et al. (2001) with both price and return model built on a sample of A share companies during 1990 and 1997, who also found that the Chinese stock market perceives accounting information to be more value-relevant for firms issuing both A- and B-shares than firms issuing only A-shares. Further researches that examine A, B share markets at the same time and found that along with A-share accounting information, the estimated B-share prices from the IAS model are significantly related to the actual B-share prices, indicating that the IAS model has additional explanatory power over that contributed by PRC GAAP model (Bao and Chow, 1999). Hu (2002) used sample companies only listed on Shanghai Stock Exchange to replicate Hu (2002)'s methodology but found opposite evidence that book value and earnings reported under PRC GAAP are more relevant than those based on IFRS. Samia and Zhou (2004) studied on AB dual-listed companies from 1994-2000 and obtained evidence that the accounting information in the B-share market is more value-relevant. Liu and Liu (2007) provided multi-faceted insights on the value relevance issue in the Chinese stock market with more comprehensive analysis on data from 1999 to 2003. They found that accounting information issued by B-share and H-share companies that is prepared and audited under IFRS/H.K. GAAP is more value relevant than that prepared under PRC GAAP of the A-share firms. In addition, by further examining the A-share market, they find that during the bearish market situation (since 2001), the incremental explanatory power of accounting information for equity book value increases, whereas it decreases for earnings. Within the A-share market, no significant difference of the value relevance of accounting information can be found among only-A-share, AB- and AH-share subgroups.

    3.3 Return explanation based on price multiples

    The earliest and most well-known model to explain stock return difference was the Capital Asset Pricing Model (CAPM) built on the Markowitz paradigm, by assuming that expected return of any risky assets linearly depend on its co-movement with the market portfolio. However after the development of CAPM model in 1960s, there were some cross-sectional studies on stock returns showing contradicting results against what CAPM model predicted, supported by large sample tests with the help of a few well-developed databases such as Compustat. Additional factors were found to explain stock return differences such as earnings/price, firm size, long-term return reversals, book-to-market equity, leverage and momentum. Another model developed by Fama and French (1992) synthesized empirical findings of some of these factors (size, leverage, book-to-market equity and beta) in a single one, although the model and the findings received some criticisms later on such as that the empirical finding was a result of data-mining (Black 1993a, 1993b), and/or there was survivorship bias in the sample used as well as beta mis-measurement (Kothari, Shanken and Sloan, 1995). Despite these controversies, more and more research seem to support the effect of other factors initiated in Fama and French (1992) model and the research focus shifted to explain why there are other factors explaining stock return. The Fama and French (1993) three-factor model found results supporting two additional factors besides access return specified in CAPM model, namely SMB - based on investment strategies of long in small-cap stocks and short in large-cap stocks, and HML - based on long in high book-to-market equity stock (value stocks) and short in low book-to-market stocks (growth stocks). And the three-factor regression reported significant coefficients on all three factors and improved R-square than a CAPM model. Research on finding the new factors to be priced by the market had not ceased since then and new models were tested against different markets outside the US.

    4. Hypothesis & Assumptions

    The author believes that given the capital market control and segmentation in China, the existing significant price or return disparity among segmented stock markets indicate potential arbitrage opportunities although other factors like investment barrier and transaction costs etc need to be taken into account when it comes to practical strategies. A wide range of literature had studied on finding out explanations for the disparity; however these are not the focus of this paper. Specifically, the paper tries to take the return disparity as given and examine how financial information (valuation multiples P/E and P/B) of AH dual-listed companies can be used for interpreting A and H share returns separately and jointed as AH return premium/discount.

    Although AH dual-listed companies had financial statement reports under both PRC GAAP and IFRS therefore information under both systems are accessible to all investors, it is assumed that international investors mainly use company financial reports under IFRS and domestic investors use that under PRC GAAP for the purpose of convenience and comparability with international and other A-listing domestic peers. Concerns of most AH companies on shifting to a single-system reporting stopped them to do so indirectly reflected that reporting under IFRS or HKGAAP is heavily relied on by international investors on H share markets. While in A-share markets, domestic investors heavily rely on PRC GAAP reporting. This situation is not likely to change swiftly and the convergence of all accounting stands sees only in the longer term.

    This paper takes the assumption of Chinese stock market being a weak-form-efficient market (Sun, Zhang, & Zhou, 1997; Chen, Chen, & Su, 2001). And it further assumes that information such as financial reporting is useful for investment decisions and is taken advantage of to predict future stock return (Francis & Schipper, 1999). Therefore the key hypothesis of this research is that financial information of A-H dual-listed companies can be taken advantage of to predict return in each market, and that arbitrage opportunities emerged from AH return premium/discount exist partially for the reason of financial reporting disparity within the same dual-listed company who faces different groups of investors on each market it lists in.

    5. Data

    Sample selection

    66 AH dual-listed companies are the focus of this research, therefore are set as the initial sample. Semi-annual stock return and company financial EPS and BPS (every 6-months) are collected for period 2002 first half (2002 S1) to 2010 first half (2010 S1). However some additional criteria are added on top of these sample companies to get the final sample data, for technical purposes. These criteria include:

    1. Financial sector companies (banks and insurance companies) are excluded due to that certain financial attributes may distort financial information of total sample where non-financial companies are majority.
    2. Same length of data points over time and entities are needed for a highly balanced panel dataset and a consistent comparison, however not all 66 companies are included in the sample due to that the AH companies have different listing time in A share and Hong Kong share market. Therefore AH companies that have long periods data missing on any of the variables are excluded.

    Filtering on these two criteria left a sample of 26 AH companies with A and H share prices (returns) respectively over a semi-annum period, 6-month earnings per share (EPS) and book value of equity per share (BPS) over the period of June 2002 to June 2010. The financials disclosed under PRC GAAP and IFRS were collected for A shares and H shares respectively. Variables are further outlined as below in Table X-X:

    Table X-X. Summary of variables

    2002 S1

    2002 S2

    A/H share price/return

    Price: A/H share price as of 30 June 2002

    Return: A/H share 6-month simple return over the period of 31 Dec 2001 to 30 June 2002

    Price: A/H share price as of 31 Dec 2002

    Return: A/H share 6-month simple return over the period of 30 June 2002 to 31 Dec 2002

    A/H share EPS

    A share EPS: Earnings per share (January to June) of the company under PRC GAAP

    H share EPS:Earnings per share (January to June) of the company under IFRS or HKGAAP

    A share EPS: Earnings per share (June to December) of the company under PRC GAAP

    H share EPS: Earnings per share (June to December)of the company under IFRS or HKGAAP

    A/H share BPS

    A share BPS: Book value of equity per share (January to June) of the company under PRC GAAP

    H share BPS: Book value of equity per share (January to June) of the company under IFRS or HKGAAP

    A share BPS: Book value of equity per share (June to December) of the company under PRC GAAP

    H share BPS: Book value of equity per share (June to December)of the company under IFRS or HKGAAP

    In addition, the whole sample is divided into two periods, before and after the global financial crisis (start time taken as June 2008). Post-crisis sample is taken out and only pre-crisis sample is used for a comparative study with the full sample to see whether and how much the empirical results are influenced by the crisis.

    Descriptive statistics of all variables used are summarized in Table XX

    Table XX. Statistical summary of variables

    A share

    Full period

    Variable

    Obs

    Mean

    Std. Dev.

    Min

    Max

    areturn

    442

    0.107

    0.478

    -0.692

    2.922

    aep

    439

    0.013

    0.091

    -1.531

    0.202

    abp

    436

    0.458

    0.305

    -0.652

    1.739

    Pre-crisis period

    Variable

    Obs

    Mean

    Std. Dev.

    Min

    Max

    areturn

    338

    0.104

    0.470

    -0.692

    2.922

    aep

    335

    0.016

    0.092

    -1.531

    0.202

    abp

    332

    0.474

    0.296

    -0.526

    1.476

    H share

    Full period

    Variable

    Obs

    Mean

    Std. Dev.

    Min

    Max

    hreturn

    442

    0.168

    0.500

    -0.677

    2.557

    hep

    442

    0.014

    0.244

    -3.552

    0.348

    hbp

    442

    1.053

    0.901

    -2.612

    6.588

    Pre-crisis period

    Variable

    Obs

    Mean

    Std. Dev.

    Min

    Max

    hreturn

    338

    0.172

    0.493

    -0.677

    2.557

    hep

    338

    0.026

    0.212

    -3.552

    0.348

    hbp

    338

    1.109

    0.871

    -1.196

    5.071

    A simple correlation summary among all variables used is given in Table XX, for both pre-crisis and full period sample.

    Table XX. Variable correlation table

    A and H variable correlation table

    Full period (obs=433)

    areturn

    hreturn

    hep

    hbp

    aep

    abp

    areturn

    1

    hreturn

    0.739

    1

    hep

    0.051

    0.082

    1

    hbp

    -0.283

    -0.229

    0.171

    1

    aep

    0.041

    0.079

    0.935

    0.133

    1

    abp

    -0.294

    -0.178

    0.245

    0.626

    0.296

    1

    Pre-crisis period (obs=329)

    areturn

    hreturn

    hep

    hbp

    aep

    abp

    areturn

    1

    hreturn

    0.705

    1

    hep

    0.025

    0.050

    1

    hbp

    -0.337

    -0.243

    0.133

    1

    aep

    0.032

    0.069

    0.952

    0.098

    1

    abp

    -0.286

    -0.115

    0.222

    0.580

    0.287

    1

    6. Methodology

    The general fundamental model developed is a series of regressions with stocks returns (next period 6-month return or current period 6-month return) being dependant variable and explained by the inverses of commonly used valuation multiples P/E and P/BV, i.e. E/P and BV/P. Three different models are constructed by taking both time and cross-sectional dimensions into account, to explore relationships between current period return and current price multiples (inversed), future period return and current price multiples (inversed) as well as return premium and price multiples (inverse) gaps. Panel data analysis is applied to each model, elaborated in details as below.

    6.1 Choice of panel data model

    According to Cheng Hsiao (2006), “panel data have several advantages over cross-sectional or time-series data by blending the inter-individual differences and intra-individual dynamics. The advantages include more accurate inference of model parameters (Hsiao, Mountain and Ho-Illman, 1995), greater capacity for capturing complexities, and simplifying computation and statistical inference in certain cases.” There are two major types of panel data regressions, fixed-effects (FE) model and random-effects (RE) model, depending on whether ‘unobserved heterogeneity' in the panel sample is assumed as random variables or fixed parameters. FE and RE specification has its own advantages and limitations, for instance FE specification can allow individual and/or time specific effects to be correlated with explanatory variables but does not allow estimation of time-invariant coefficients while RE specification allow estimation of time-invariant variable's impact by imposing a ‘conditional density assumption' (Hsiao, 2006).

    The choice of FE or RE model in this paper is made with the help of a statistic developed by Hausman (1978) and can be tested under chi-square distribution assumption. Null hypothesis under the Hausman test is that difference in coefficients under FE and RE specification are not systematic, and rejection of the null needs the constructed statistic which follows chi-square distribution is significantly different from zero. STATA command ‘hausman' is used to implement the Hausman test on the sample panel regressions to decide whether FE or RE model should be used in this study. For all panel datasets used, test results identify the suitability of FE specification rather than RE specification therefore FE regressions are used in all three models specified for parameter estimation.

    6.2 Current period return model

    To test if current E/P ratio and BV/P can explain current period (6-month) return in A and H share market respectively. Supposedly companies with higher E/P and BV/P should see higher next period returns.

    H0: current E/P and BV/P ratio cannot explain current return differences among AH companies, i.e. beta coefficients are not significant for both A and H sample companies

    The current period return model is specified as:

    ri,tA=αA+β1AEi, tAPi, tA+β2ABVi,tAPi,tA+ui,t

    ri,tH=αH+β1AEi, tHPi, tH+β2HBVi,tHPi,tH+ui,t

    ri,tA: semi-annual return of A share for company i at time t.

    ri,tH: semi-annual return of H share for company i at time t.

    Ei, tAPi, tA: 6-month earning price ratio for company i A-share at time t.

    BVi,tAPi,tA: 6-month book value price ratio for company i A-share at time t.

    Ei, tHPi, tH: 6-month earning price ratio for company i H-share at time t.

    BVi,tHPi,tH: 6-month book value price ratio for company i H-share at time t.

    ui,t: error term

    t: discrete time variable with semi-annual frequency

    i: AH company

    Both pooled regression and Fixed-effect regression are run respectively on A and H shares first for (1) full period 2002 S1 to 2010 S1 (2) pre-crisis period 2002 S1 to 2008 S1.

    6.3 Return prediction model

    To test if current E/P ratio and BV/P ratio can predict next period (6-month return) in A and H share market respectively.

    H0: current E/P and BV/P ratio cannot explain not next period return differences among AH companies, i.e. beta coefficients are not significant for both A and H sample companies

    The return prediction model is specified as:

    ri,tA=αA+β1AEi, t-1APi, t-1A+β2ABVi,t-1APi,t-1A+ui,t

    ri,tH=αH+β1AEi, t-1HPi, t-1H+β2HBVi,t-1HPi,t-1H+ui,t

    ri,tA: semi-annual return of A share for company i at time t.

    ri,tH: semi-annual return of H share for company i at time t.

    Ei, t-1APi, t-1A: 6-month earning price ratio for company i A-share at time t-1.

    BVi,t-1APi,t-1A: 6-month book value price ratio for company i A-share at time t-1.

    Ei, t-1HPi, t-1H: 6-month earning price ratio for company i H-share at time t-1.

    BVi,t-1HPi,t-1H: 6-month book value price ratio for company i H-share at time t-1

    ui,t: error term

    t: discrete time variable with semi-annual frequency

    i: AH company

    6.4 Return premium model

    See if return premium of both current and next period can be partly explained by (1) E/P and B/P in either market (2) A and H gap in E/P and B/P

    H0: current differences on E/P and BV/P ratio between AH companies cannot explain current or next period return differences among AH companies, i.e. beta coefficients are not significant for both AH sample companies

    The return premium models are specified as below:

    (1) Current period return premium:

    ri,tA-ri,tH=αt+β1Ei, tAPi, tA-Ei, tHPi, tH+β2(Bi, tAPi, tA-Bi, tHPi, tH)+ui,t

    (2) Next period return premium:

    ri,tA-ri,tH=αt+β1Ei, t-1APi, t-1A-Ei, t-1HPi, t-1H+β2(Bi, t-1APi, t-1A-Bi, t-1HPi, t-1H)+ui,t

    In all above three models, both pooled regression and Fixed-effect regression are run respectively on A and H shares for result comparison from a technical perspective. FE regression remedies the problem of unobserved heterogeneity in pooled OLS regression and should give more accurate parameter estimation. The same regressions are also done for two samples (1) full period 2002 S1 to 2010 S1 (2) pre-crisis period 2002 S1 to 2008 S1, in order to detect whether estimations change with the inclusion of financial crisis period data.

    7. Empirical results

    Interpretation of variables

    Inverses the two multiples P/E and P/BV are used as independent variables in all the models, therefore it can be useful to first make it clear on how these independent variables, i.e. E/P and BV/P could be interpreted.

    E/P ratio

    Multiple P/E can be interpreted as how much price an investor is willing to pay per dollar of earnings. A high P/E multiple indicates investors' expectation of high earning growth in the future for the concerning company because otherwise the marking is over-paying for the company. Therefore a high E/P ratio reflects investors' expectation of low earning growth because every dollar an investor pays for the stock is backed up by more dollars of earnings. But these stocks may represent value stocks that are ‘good buy' because one can probably pay lower market price for company with good profitability. In conclusion:

    High E/P indicates low earning growth expectations and possibility of value stocks

    BV/P ratio

    Multiple P/BV can be interpreted as how much price an investor is willing to pay per dollar of tangible net asset of a company. A high P/BV multiple indicates high market expectation of cash flow that the company assets can generate in the future in the concerning company (yet again high future growth). Therefore a high BV/P ratio reflects low market expectation and valuation of the company net assets. But similar to high E/P ratio stocks, these high BV/P stocks can also imply ‘good buy' because one can probably pay lower market price for the same value of net assets. In conclusion:

    High BV/P indicates low net asset valuations and possibility of value stocks

    7.1 Current return model

    Fixed-effect model gives improved the fitness and parameter estimation though the results show largely same relationship as in the pooled OLS regression, FE model results should be relied on rather than those of pooled regression.

    Pre-crisis sample (2002 S1 to 2008 S1) results are looked at first since this excludes the possible disturbing factor due to the extreme market situation. In general, the E/P ratios turn to be insignificant in relation to share returns while BV/P ratios hold to be significantly negatively relevant to both A and H share returns. The negative coefficient of BV/P on share return might seem counter-intuitive at first sight. However, note that in the current return model, share return was actually return over the past 6 months, and BV/P are taken as of the current moment ratio with company current book value and price. With book value held the same, companies with higher BV/P have lower current price, which translates to lower past 6-month return.

    Table XX: Current model regression results with pre-crisis sample

    Pre-crisis sample

    A share

    (Pooled)

    A share

    FE

    R-squared = 0.0958

    R-sq: within = 0.1668

    coef

    p

    coef

    p

    aep

    0.635

    0.003

    aep

    0.444

    0.132

    abp

    -0.514

    -

    abp

    -0.863

    -

    _con

    0.342

    -

    _con

    0.512

    -

    H share

    (Pooled)

    H share

    FE

    R-squared = 0.0666

    R-sq: within = 0.0918

    coef

    p

    coef

    p

    aep

    0.193

    0.001

    aep

    0.095

    0.466

    abp

    -0.145

    -

    abp

    -0.202

    -

    _con

    0.327

    -

    _con

    0.393

    -

    Comparing the coefficients for A and H share FE regressions, one can see that A share return has higher relevance (higher absolute value of coefficients) with BV/P under PRC GAAP rule than that of H share return with BV/P under IFRS. This partially reflects a more efficient H market than A share market, in that H share investors cannot use AH company valuation ratios as much as A share investors can do in screening potential well performing AH companies.

    When applying the model on full period (2002 S1 to 2010 S1) samples, there is still significant relationship between current period A or H share return and BV/P ratios under respective financial reporting rule. However E/P ratios turn to be significant as well while implying a positive relationship between E/P and current period returns, reason for this could possibly be significantly disturbed price information during crisis period.

    Table XX: Current model regression results with full sample

    Full sample

    A share

    (Pooled)

    A share

    FE

    R-squared = 0.1047

    R-sq: within = 0.1721

    coef

    p

    coef

    p

    aep

    0.742

    0.004

    aep

    0.666

    0.010

    abp

    -0.529

    -

    abp

    -0.881

    -

    _con

    0.344

    -

    _con

    0.507

    -

    H share

    (Pooled)

    H share

    FE

    R-squared = 0.0680

    R-sq: within = 0.0861

    coef

    p

    coef

    p

    aep

    0.254

    0.002

    aep

    0.235

    0.020

    abp

    -0.140

    -

    abp

    -0.191

    -

    _con

    0.311

    -

    _con

    0.366

    -

    The coefficients estimated with A share sample are still bigger in absolute values than those of H share samples, also supporting a more efficient H share market conjecture as in pre-crisis sample.

    7.2 Return prediction model

    Results based on pre-crisis sample show that current company valuation ratio BV/P can be used to predict next period (6-month) return of AH companies, given the significantly positive coefficients on BV/P for both A and H share regressions. Higher BV/P ratio indicates higher next period return and this positive relationship is higher for A shares than for H shares. However E/P ratios turn to be insignificant in predicting both share-returns.

    Table XX: Return prediction model regression results with pre-crisis sample

    Pre-crisis sample

    A share

    (Pooled)

    A share

    FE

    R-squared = 0.0253

    R-sq: within = 0.0378

    coef

    p

    coef

    p

    aep

    -0.019

    0.932

    aep

    -0.158

    0.629

    abp

    0.264

    0.012

    abp

    0.425

    0.001

    _con

    -0.044

    0.460

    _con

    -0.118

    0.066

    H share

    (Pooled)

    H share

    FE

    R-squared = 0.0351

    R-sq: within = 0.0564

    coef

    p

    coef

    p

    aep

    0.117

    0.362

    aep

    0.029

    0.833

    abp

    0.103

    0.002

    abp

    0.164

    -

    _con

    0.028

    0.509

    _con

    -0.037

    0.461

    The full sample analysis results are in line with those with pre-crisis sample analysis, except for that BV/P coefficients for A share return rise much more significantly than those for H share return from pre-crisis sample to full sample. This indicates that with period that exceptional market condition occurs, investors would be able to rely more on fundamental financial ratios in stock picking than during normal market conditions, and this is more true in A share market than in H share market.

    Table XX: Return prediction model regression results with full sample

    Full sample

    A share

    (Pooled)

    A share

    (FE)

    R-squared = 0.0295

    R-sq: within = 0.0508

    coef

    p

    coef

    p

    aep

    -0.333

    0.077

    aep

    -0.400

    0.150

    abp

    0.283

    0.002

    abp

    0.477

    -

    _con

    -0.017

    0.722

    _con

    -0.106

    0.043

    H share

    (Pooled)

    H share

    (FE)

    R-squared = 0.0357

    R-sq: within = 0.0378

    coef

    p

    coef

    p

    aep

    -0.085

    0.285

    aep

    -0.134

    0.189

    abp

    0.106

    -

    abp

    0.165

    -

    _con

    0.057

    0.115

    _con

    -0.004

    0.913

    7.3 Return premium model

    There are some research explaining dual-listed company price/return premium with hypothesis of liquidity, demand and speculative investors etc. However, the author believes it also makes sense to test whether the return disparity can be partly attributed to the gap of financial reporting figures for the same dual-listed company due to segmented financial reporting rules and separation of investor group.

    Results based on the specified model show that the gap of financial ratios E/P and BV/P cannot explain current A-H return disparity but BV/P can explain and be used to predict next period A-H return disparity. And the results hold true for both full sample and pre-crisis sample only as well, and the prediction relevance is significantly higher in pre-crisis period only.

    Table XX: Return premium prediction model regression results

    Full sample

    Pre-crisis sample

    (pooled)

    (FE)

    (pooled)

    (FE)

    A-H return

    A-H return

    A-H return

    A-H return

    R-squared = 0.0800

    R-sq: within = 0.1239

    R-squared = 0.1292

    R-sq: within = 0.1877

    coef

    p

    coef

    p

    coef

    p

    coef

    p

    a_hep

    0.111

    0.914

    a_hep

    0.105

    0.602

    a_hep

    0.083

    0.494

    a_hep

    0.020

    0.899

    a_hbp

    0.026

    -

    a_hbp

    0.026

    -

    a_hbp

    0.178

    -

    a_hbp

    0.251

    -

    _cons

    0.023

    0.412

    _cons

    0.022

    0.015

    _cons

    0.049

    0.064

    _cons

    0.095

    0.001

    Table XX: Current return premium model regression results

    Full sample

    Pre-crisis sample

    (pooled)

    (FE)

    (pooled)

    (FE)

    A-H return

    A-H return

    A-H return

    A-H return

    R-squared = 0.0016

    R-sq: within = 0.0014

    R-squared = 0.0020

    R-sq: within = 0.0027

    coef

    p

    coef

    p

    coef

    p

    coef

    p

    a_hep

    0.051

    0.095

    a_hep

    0.058

    0.604

    a_hep

    0.054

    0.483

    a_hep

    -0.016

    0.928

    a_hbp

    0.019

    0.890

    a_hbp

    0.013

    0.644

    a_hbp

    0.019

    0.428

    a_hbp

    0.030

    0.365

    _cons

    0.025

    0.018

    _cons

    -0.053

    0.029

    _cons

    -0.055

    0.077

    _cons

    -0.048

    0.108

    Conclusion

    This paper take up a specific angel looking into value relevance of two price ratios among Chinese AH dual-listed companies, given current stock market segmentation. Firstly an introduction of Chinese stock market and differently applied financial reporting rules in segmented markets were given, followed by a review section of prior research on relevant issues, including market segmentation and dual-listed stock price/return disparity, value relevance of accounting information and stock return predictability across individual shares. Three fixed-effects models were constructed to explore the relationship between price ratios and stock returns on A and H market respectively, and a comparison of results is done on both pre-crisis samples and full period samples to shed light on the possible impact of recent financial crisis. It was found out that:

    • Current E/P and BV/P explains current period return of both A and H shares, with E/P positively and BV/P negatively relates to current period return among AH companies.
    • Current BV/P can predict next period return for both A and H shares, and it is a positive relationship.
    • Current E/P and BV/P gap under two financial reporting rules cannot explain the current A-H return disparity, but current BV/P gap under two financial reporting rules can predict next period A-H return disparity
    • A share returns show higher relevance with E/P and BV/P ratios than H share returns do in general, reflecting a less efficient A share market.
    • Crisis period twist the informational efficiency of markets and impact the relevance of E/P to current return and B/P to next period return.

    Limitation and possible future research

    This paper found out some interesting empirical results that support for example explanation and prediction of A-H return premium based on BV/P ratio, which practically could indicate an investment strategy. However, the strategy is only valid to the extent that no frictions such as investment barriers, transaction costs etc. Implementation of the strategy therefore is subject to restrictions and the deregulations in reality. All models are constructed based on fixed-effect specification which is an improvement over pooled OLS regressions and has certain advantages over time-series or cross-sectional regression, however, the data size available is another limitation.

    Further research can be explored on improving current limitations discussed in previous paragraph. Also the specified models which include only two price ratios as independent variable could potentially be extended to include more variables that may explain A or H share returns (or disparity), which can be other financial information such as expected growth of earnings per share, or stock liquidity measurement.