Research On Return Predictability Based On Predictor Decomposition,Model Switching And Compound Strategies

The research on the predictability of stock returns has undergone a long history,and the attactive force regarding this topic does not even fade a bit.In recent years,the studies on the return predictability mainly center on finding new predictors or new methods to make return forecasts.This article proposes several predictive strategies,including cycle-decomposing approach(CDA)applied to predictors,interest rate-determined model-swithing strategy(IRDMS)applied to predictive models,and compound strategies,to explore their contribution to return forecasts.Main findings based on these strategies are summarized as follows:(1)Founding on existing literature,the decomposition strategy is proposed in chapter 3to decompose each predictor into a long-cycle trend component and a short-term deviation component.In detail,the long-cycle component is the moving average series of each predictor,while the short-term deviations are the difference between the original data series of the economic predictor and the corresponding moving average series.The long-cycle component measures the variation tendency of the corresponding economic variable,and the shor-term component variable reflect the extent of transient economic deviations from the long-term trend.Then according to the application of combination forecasts,I put those component variables into two predictive regressions respectively,and obtain a final forecast by calculating the mean value of the two forecasts from those regressions which contain the whole predictive information of the original economic fundamental.After a general evaluation of those forecasts,I find that the performance of most CDA-based predictive models perform better than the corresponding regressions based on original variables.Besides a better statistical performance of those CDA models,they can also obtain superior economic value considering the portfolio based on new forecasts.In addition,the results from multivariate strategies,forecasts during differenc business cycles,long-horizon forecasts and alternative portfolio prerequisite also support the robust dominance of CDA,which provide further evidence for the predictablity of stock returns.(2)Following the idea of model switching,I propose in chapter 4 a new model swhicting strategy considering the variation of real time interest rate level.The main idea of this strategy is that when the real time interest rate exceeds the mean value of rates over several preceding months,the corresponding rate can be classified as a high level,and the forecasts are from predictive regressions.While the real time interest rate is lower than the mean level,the variation of stock prices is thought to be affected by speculative behavior and cannot rationally reflect economic fundamentals,thus I forecast stock returns through the benchmark of historical average.Empirical results suggest that,for most predictive regression forecasts,the predictive power is enhanced to different extent with the application of IRDMS.Basically,the performance of IRDMS is persistent and effective.First,some IRDMS models perform significantly better than the benchmark considering all 10 judgment criteria.Second,under condition of economic value tests and combination forecasts,the forecasts based on IRDMS still perform better than original regressions.Further more,robust checks regarding compound judgment criteria,long-horizon forecasts and forecasts during different business cycle still support the validity of IRDMS and the predictability of stock returns.(3)The CDA,IRDMS and a forecast restriction is applied together to make compound forecasts in chapter 5.Single strategies are aggregated to construct three double-compound strategies and a triple-compound strategy.Forecasts from these compound strategies are evaluated to analyze the synergstic effect of single strategies.Empirical results suggest that the comprehensive performance of compound strategeis perform robustly better than the single ones,validating the synergistic effect of single strategies considering the two stock markets and the three oil markets.