A Conditional Goodness-of-Fit Test for Time Series

This paper proposes a uni ed approach for consistent testing of linear restrictions on the conditional distribution function of a time series. A wide variety of interesting hypotheses in economics and nance correspond to such restrictions, including hypotheses involving conditional goodness-oft, conditional homogeneity, conditional mixtures, conditional quantiles, conditional symmetry, distributional Granger non-causality, and interval forecasts. The nite-sample properties are investigated in a set of Monte Carlo experiments. The proposed tests are conservative but perform well in samples of the size relevant for empirical nance.