1.3 Structure of the Book
Chapter Two – The Chinese Equity Market： Historical Development and Characteristics reviews the general background on the Chinese equity market， including market history and market characteristics for the period January 1993 （01.1993） to December 2010 （12.2010） which is related to the empirical work conducted later in this book. To facilitate comparisons with developed markets， the chapter001.
The findings are for the most part in line with expectations in that the Chinese equity market is associated with unique characteristics and therefore is considered an interesting arena to test the robustness of both equity market downturn and equity market return predictions. In particular， three specific aspects， namely the risk-return patterns， the extent of market downturn risk， and the level of market integration， are most important in that they distinguish the Chinese equity market from many other world markets.
The underlying dynamics for the market characteristics are also analysed. This sheds light on what precisely constitutes the driving forces of market performance， including limited investment venues， market structures， low dividend yields， and speculative trading behavior.
Overall， the findings in Chapter 2 show that the Chinese equity market is an interesting arena for further study， particularly with respect to the exploitable predictable patterns in both equity market downturns and equity market returns， which is of great importance to investors.
Chapter Three – Academic Literature on the Chinese Equity Market reviews some specific studies on the Chinese equity market， which paves the way for the subsequent empirical research （i.e.， Chapter 6 and Chapter 7） in this book. It begins with the literature on equity pricing factors， including the price determinants of domestic A-shares and the price differentials between A- and B-shares. This provides insights into the behavior of the components of equity indices （i.e.， the SHA， the SHC and the SZC） that will be used to measure equity market performance in both Chapters 6 and 7. As previously shown， the Chinese equity market is unique in some senses. Therefore， the traditional Fama-French’s three-factor model that is useful based on developed market data fails to well explain the expected returns of A-shares. Additions of “systematic” risk factors unique to the market are necessary to enhance the model’s explanatory power. Furthermore， the pricing difference between domestic A-shares and foreign B-shares is in contradiction with both asset pricing theory and empirical evidence from other markets. However， this has been mitigated since the opening of the B-share markets to domestic investors， suggesting that the performance of the A- and the B-share markets is converging as the market develops.
The dynamic linkages of the Chinese equity market with global markets and those between various regional markets are also discussed. This section provides general information on the integration process of Chinese equity markets in advance of Chapter 6 and Chapter 7 which compare the performance of the Shanghai and the Shenzhen markets and test for the predictability of equity market performance considering both local and international influences. Because of the currency shield， restrictions on capital flows， and the separate share ownership structure， Chinese equity markets were not integrated both globally and regionally. However， the lack of integration has been increasingly mitigated particularly in recent decades.
The remainder of Chapter 3 focuses on studies of the Efficient Market Hypothesis （EMH）. This is based on the examination of whether equity prices fully incorporate all publicly available information. One important finding from this part of analysis， which motivates the empirical work conducted in Chapter 6 and Chapter 7， is that the Chinese equity market is not efficient， and it may be possible to forecast equity market performance based on information such as macroeconomic developments.
Chapter Four – Equity Market Returns and Predictor Variables： Theoretical Explanations and Empirical Evidence focuses on extant studies of equity return predictability， with the aim of setting the context for candidate predictor variables used in the subsequent empirical research. Firstly， it reviews the relationships between macroeconomic variables and equity returns， and summarises some of the key results based， in particular， on inflation， money supply， interest rates， industrial production and business cycle indices.
The chapter then moves on to the studies on the relationships between sentiment variables and equity returns. The sentiment variables included are sentiment indices， the EP ratio and spreads between the EP ratio and interest rates. Technical variables， particularly trading volume and market volatility， are considered important in the literature to describe equity returns and therefore their relationships with equity price changes are also discussed in the chapter.
In general， the results from existing studies are mixed when real output variables， such as industrial production， and sentiment variables are considered. The potential reason for the inconsistency in the forecasting ability of these variables is that a vast majority of studies assume constant model parameters， for example， constant correlation coefficients， which may in fact vary over time. The estimated coefficients may therefore fail to identify a candidate variable whose effect switches sign and averages close to zero over time but is occasionally important. The chapter discusses the progress made on model instability to date， both from a theoretical and an empirical point of view， which helps identify this as an area for further research.
The literature on the relationships between equity returns and predictor variables therefore motivates our investigations into the forecasting ability of these variables on equity market performance in China while accounting for model instability as a robustness check. The evidence presented in the chapter allows for a better understanding of how best to design and estimate forecasting models in the presence of structural instability in a Chinese context.
Chapter Five – Data and Methodology describes some of the key data and methodologies used for the empirical research presented in Chapter 6 and Chapter 7. It begins with a detailed description of both the data on financial series and predictor variables （as suggested by both asset pricing theory and empirical evidence in Chapter 4） and how these proxies are measured. The chapter then progresses to look at how the predictability of equity market performance in China is examined.
Specifically， Chapter 6 examines equity market downturn predictions and the presence and the importance of structural instability associated with financial time series forecasting models for the period February 2002 （02.2002） to May 2010 （05.2010）. The data used are monthly data on the SHA， the SHC， the SZC， the ChinaBond Composite Bond index （CCB） and the predictor variables （under macroeconomic， sentiment and the technical categories） as previously described. Equity market downturns are defined as when the excess market return， namely the difference between the returns of the equity market and the bond market， is negative. To select predictor variables from the large pool of candidates， two correlation tests - Pearson and Spearman correlation tests are conducted to test for significant relationships between excess market returns and predictor variables. The predictor variables are included for further analysis when both tests return p-values less than 0.050. Using the methodology proposed by Shen （2003）， signals of market downturns are constructed. The statistical significance of these signals is examined by both Henriksson and Merton’s （1981） non-parametric test and Monte Carlo simulation. To take into account investors’ utility function， market timing strategies are formulated based on the signals and their performance is compared with that of three buy-and-hold strategies. A market timing strategy is designated as profitable when （1） all of these statistics return the same indication in support of the strategy’s superiority to the benchmarks and （2） the strategy itself works for trading purposes. A series of robustness tests are also conducted to ensure strategies’ profitability against the model specification and/or parameters employed. These include （1） the length of time invested in the equity market， （2） the selection of percentile thresholds for the forecasting rules， （3） the selection of the equity market index， and （4） a small sample bias. Finally， the chapter investigates whether forecasting models are associated with structural instability and how important it is to the profitability of market timing activity by comparing the results using a fixed rolling method versus an expanding window method.
By contrast， the work presented in Chapter 7 examines equity market return predictions and tests for the presence， number， location and the importance of structural instability associated with financial time series forecasting models for the period January 1993 （01.1993） to May 2010 （05.2010）. The data used is monthly data on the SHA， the SHC， the SZC， and the predictor variables whose data can be dated back as long as January 1993 （01.1993）. To build linear predictive regression models， the stepwise regression methodology first proposed by Efroymson （1960） is employed and a series of sensitivity analyses to general data problems， including outliers， heteroscedasticity， and serial correlation， are conducted. Given these forecasting models， the chapter moves on to testing structural instability in model parameters and how it would affect the results of return predictability using the testing procedure proposed by Bai and Perron （1998， 2003）.
Chapter Six – Equity Market Downturn Predictability and Model Instability begins by looking at the summary statistics on both financial market series and predictor variables （as described in Chapter 5） for the period February 2002 （02.2002） to May 2010 （05.2010）. It is followed by selecting candidate variables for further analysis based， in particular， on two correlation tests – Pearson and Spearman.
The chapter then moves on to examining the ability of the selected predictor variables to predict how vulnerable the excess market return is in the near future， both from a statistical and an economic point of view. Specifically， the statistical approach is based on both the estimation of whether the number of accurate market downturn forecasts that a forecaster makes is statistically large and whether the downturn moments that a forecaster captures are statistically important. By contrast， the economic approach focuses on examining whether market timing strategies based on the signals from these predictor variables are profitable. Robustness tests are also conducted on these results. A number of interesting findings emerge from this part of analysis. Firstly， the useful information contained in inflation variables to consistently predict imminent market downturns provides evidence consistent with prior studies that fundamentals help explain equity returns in China. Furthermore， the superior performance of inflation-based trading strategies to buy-and-hold strategies reinforces previous studies that the Chinese equity market does not immediately fully incorporate important macroeconomic announcements and therefore it tends not to be efficient.
Finally， an investigation of structural instability in forecasting models， which is based on comparing results between the fixed rolling and the expanding window methods， significantly challenges the traditional approach that assumes constant model parameters. It suggests that model instability is a very important source of investment risk for investors when making asset allocation decisions. However， one important issue with the analysis presented in this chapter is that it has not identified the occurrence and the size of these structural breaks as well as their impact on parameter estimates. All these questions， which are of interest to investors， are addressed in the following empirical chapter.
Chapter Seven – Equity Market Return Predictability and Model Instability begins with presenting the descriptive statistics for both the equity market series and the predictor variables （as specified in Chapter 5） for the period January 1993 （01.1993） to May 2010 （05.2010）. It continues by building linear predictive regression models for equity market returns and examining the robustness of the resulting models.
The remainder of the chapter focuses on examining potential issues of model instability， including the estimation of break dates， determination of the presence and the number of breaks， and an investigation of the resulting parameter estimates. Several interesting observations arise from this part of analysis. Firstly， the different mainland China equity markets are driven by different underlying dynamics， indicating that investment in these markets involves different considerations and strategies. Secondly， there is some evidence of structural breaks in forecasting models using the Chinese data even though the precise timing of their occurrence is difficult. Thirdly， the break occurrence tends to have close relationships with market events， consistent with previous studies that market sentiment and large macroeconomic shocks are potential sources for model instability. Finally， conditioning on market conditions results in an improvement in the performance of predictive regression models in predicting aggregate equity returns in China. The apparent changes in return predictability are indeed of considerable economic interest.
Overall， both empirical chapters importantly show the predictability of equity market performance in China while accounting for model instability， highlighting the benefit to consider the possibility of structural breaks in the relationships between predictor variables and aggregate equity returns.
Chapter Eight – Summary， Discussion and Conclusion involves a summary discussion of the research. Firstly， the main research findings are presented along a number of themes； correlation analysis， statistical and economic significance results， robustness checks， and model instability analysis （for Chapter 6）， model setup， parameter sensitivity investigation， estimation of model instability （for Chapter 7）. The limitations of the research are also flagged， as are suggestions for further research into forecasting equity market performance.