股票市场收益率的可预测性:基于中国股票市场的实证研究(英文版)
上QQ阅读APP看书,第一时间看更新

1.4 Conclusion

Forecasting equity market performance has a reputation as a risky activity. Clearly, any forecasting result is closely related to the selection of predictor variables, the construction of forecasting models, and the formulation of forecasting rules. Furthermore, market conditions that lead to time-varying investment opportunities in financial markets may also affect forecasting performance. This narrow view suggests therefore that an investor is better off buying and holding the market rather than risking timing the market.

Market timing is a tough game.[1] However, as Shen (2003) emphasises, asking investors not to forecast market performance is equivalent to asking consumers not to maximise their utility when making consumption decisions. For decades, investors have attempted to upgrade and develop forecasting techniques accounting for various cases, with some success particularly in recent years (Malkiel, 2005). One example is that an improvement has been made in forecasting under model uncertainty, whereby no knowledge is secured about the “true” form of the underlying specification. By contrast, model instability, including the presence, number, location and the size of historical breaks to model parameters as well as the possibility of future breaks, has received far less attention. The fact that model instability has a large effect on long-term asset allocation when compared to the effects of parameter estimation and model uncertainty makes the lack of attention to the subject all the more surprising. [2]

This research examines the predictability of equity market performance in China while allowing for model instability. It attempts to provide insights into the behavior of the mainland equity markets in China by examining the forecasting ability of predictor variables on both equity market downturns and equity market returns.


[1]Richard J. Bauer, and Julie R. Dahlquist, “Market timing and roulette wheels,” Financial Analysts Journal 57 (2001): 37.

[2]See Pettenuzzo and Timmermann (2011) for a further discussion.