Files

Abstract

In this final article of our three-part series, we demonstrate why stochastic coefficients models are well suited to predict future variables We analyze the forecasting problem and consider various criteria of prediction If a forecaster must choose one from among several coherent predictors, then the choice should be the one with the best track record Decomposing the forecast error shows that stochastic coefficients models can cover more possible sources of prediction error and correct for them The empirical record shows that stichastic coefficients models can substantially reduce out-of-sample forecast errors more than fixed coefficients models Our assessment of coefficient stability tests is they are contradictory , misleading, and without empirical value

Details

PDF

Statistics

from
to
Export
Download Full History