We verify whether parameter-driven and observation-driven classes of dynamic models can outperform each other in predicting time-varying parameters. We consider existing and new dynamic models for counts and durations, but also for volatility, intensity, and dependence parameters. In an extended Monte Carlo study, we present evidence that observation-driven models based on the score of the predictive likelihood function have similar predictive accuracy compared to their correctly specified parameter-driven counterparts. Dynamic observation-driven models based on predictive score updating outperform models based on conditional moments updating. Our main findings are supported by the results from an extensive empirical study in volatility forecasting.

We propose an observation-driven dynamic factor model for mixed-measurement and mixed-frequency panel data. Time series observations may come from a range of families of distributions, be observed at different frequencies, have missing observations, and exhibit common dynamics and cross-sectional dependence due to shared dynamic latent factors. A feature of our model is that the likelihood function is known in closed form. This enables parameter estimation using standard maximum likelihood methods. We adopt the new framework for signal extraction and forecasting of macro, credit, and loss given default risk conditions for U.S. Moody's-rated firms from January 1982 to March 2010.