BIS research simplifies tests for predictive density evaluation

The paper proposes a simplified framework for evaluating conditional predictive densities, leveraging the probability integral transform (PIT).

This approach accommodates diverse estimation schemes, including expanding and rolling windows, and applies to both stationary and non-stationary processes.

By treating the PIT as a primitive, researchers can apply established tests in settings where their validity was previously uncertain.

This bridges a gap in existing methodologies and allows for more robust forecast evaluation.

Monte Carlo simulations demonstrate favorable size and power properties of the proposed tests.

In an empirical application forecasting US industrial production growth, the research highlights a key finding: incorporating stochastic volatility into an unobserved components model is essential for generating correctly calibrated density forecasts.

This applies to both monthly and quarterly frequencies, underscoring the practical importance of the new framework for accurate economic forecasting.