The aim of this project is to develop a formal methodology for assessing the presence of state dependence in the out-of-sample forecast errors generated by predictive regression models.
A vast body of recent empirical research documented the presence of state dependence in the forecast errors produced by models used to forecast a broad range of economic and financial variables such as stock and bond returns, commodity returns, rates of inflation, currency returns among many others. State dependence in this context takes the form of forecast errors having different quality characteristics such as lower variances in periods of economic recessions versus expansions. This state dependence in the behaviour of forecast errors has been typically documented informally through a descriptive comparison of prediction errors (e.g. lower MSEs or during recessions) or the use of recession dummies within the underlying forecasting models.
New test statistics are developed that can accommodate the presence of multiple predictors with potentially heterogeneous dynamics and that are also robust to the choice of the training sample used to initiate forecasts.
This project has been funded by the British Academy.