1.1 Research Motivation

Predictability of security returns has received a huge amount of attention in the finance literature since the seminal work of Samuelson （1965）. Forecasting techniques have been developed， attempting to help investors explore investment opportunities in financial markets. As mentioned previously in Section 1.0， the focus of these improved techniques has been to generate higher forecasting accuracy/statistical significance （e.g.， Lanne， 2002； Bradley and Jansen， 2004； Hjalmarsson， 2010； Schrimpf， 2010； Rapach， Strauss， and Zhou， 2013）. The question is if model parameters are estimated with error， confronting the investor with what is commonly termed “estimation risk”， to what extent might the evidence influence a rational， risk-averse investor’s investment decision？ Answers to this question provide a metric by which to assess the economic value of return predictability.

On the theoretical grounds， the relationship between statistical and economic value of forecasting performance is far from obvious given that the statistical criteria conventionally focus on the predictability of mean returns whereas the economic value of a return forecast reflects the accuracy of predicted movements in the entire return distribution with weights that depend on the shape of the utility function （Cenesizoglu and Timmermann， 2012）. In other words， return forecasts deemed accurate do not necessarily mean economic value for investors. Empirical studies， such as Cenesizoglu and Timmermann （2012）， find that underperformance along conventional measures of forecasting performance such as root mean squared forecast error and R2 contain little useful information when one examines whether return prediction models help investors generate profits or not. It therefore indicates that testing the economic significance of return predictability warrants further investigation given that it may contain different information of value to investors from the statistical tests.

Another important issue concerns the assumption of constant model parameters， such as constant correlation coefficients， which in fact may vary through time. There are many econometric tests available on examining instability in time series forecasting models （e.g.， Brown， Dubin， and Evans， 1975； Andrews， 1993； Pesaran and Timmermann， 2002； Elliott and Muller， 2003）. However， their applications to equity return forecasts have received limited attention in the extant literature. Instead of formal tests， structural instability is typically addressed by estimating forecasting models for various subsamples， and some of this informal analysis suggests that the pattern of predictability varies over time （e.g.， Goyal and Welch， 2003； Schwert， 2003）. A few more recent papers， such as Paye and Timmermann （2006）， Rapach and Wohar （2006）， and Clark and McCrackern （2009）， have begun to formally investigate this issue for return forecasting models using developed equity market data. Emerging markets， particularly the Chinese equity market， are far less studied. Given the unique investment environment， for example， the uncertainty about government policies and market sentiment， the Chinese equity market provides an interesting arena to test the robustness of the evidence on return predictability when allowing for model instability.

This book， therefore， addresses the apparent gap in the literature with respect to testing the predictability of equity market performance in an emerging economy. Three equity market benchmarks， the SHA， the SHC and the SZC， are used and a comparative assessment of return predictability of different markets is carried out. The analysis is based on various sample periods and large time series observations （relative to the short history of the Chinese equity market） that makes it as broad， if not broader， than any other equity market performance analysis in China. Furthermore， a comprehensive background study on the Chinese equity market have also been undertaken to facilitate an insight into the underlying market characteristics and to outline the current research on this market.