Empirical Likelihood Test For Asset Return Predictability

This paper constructs a method for testing whether an asset return can be predicted ina predictive regression model.The predictability of asset return has always been a hot topicno matter in theory or practice,and people often predict the profitability of assets through certain lagging economic variables.However,the observable economic data in reality is of-ten not stable enough and we should make many assumptions when judging the charactersof data,such as whether the sequence is stationary or unit root process.Even though somemethods can be used to determine the characteristic of the data in advance,this will not only make predictive the return of asset returns to be more complex,but also easy to make mistakes.This paper attempts to propose a more convenient method of verifying the predictability of asset return through empirical likelihood methods based onsome weighted score equations.Our approach has a strong test effect for both stationary andnonstationary data,so the complex data can be directly tested without the need to determine the type of data in advance.The innovation of this paper lies in the application of the empirical likelihood method to the standard predictive regression model.The other contribution is to modify the standard model,our method of correction is to add a hysteresis to the predictor in the regression and to convert the independent and identically distributed error term into autoregressive.Finally,the effective test method is also obtained in the correction model.At the end of this paper,the Monte Carlo simulation and numerical simulation of the true financial data show that the empirical likelihood test method proposed in this paper is robust to the cases when predictive variables are stationary processes,unit root processes,near unit roots processes and error anomalies.