In recent years many different information sources have become available that can be combined for improved econometric forecasting and policy analysis. As a consequence Herman K. van Dijk’s research focuses on two topics:
(i) Forecasting and policy analysis using a combination of econometric models where large data sets are available and advanced algorithms can be applied using parallel computing.
(ii) As a technical background, use is made of the ‘computational revolution’ that occurred in Monte Carlo simulation methods in Bayesian econometrics. A novel class of algorithms is developed that approximates very nonstandard stochastic distributions of econometric models. In particular, a new class of simulation methods using mixtures of probability processes is very efficient.
This research allows for predictive and policy analysis of operational models in economics and finance using, for instance, data on expectations and opinion pools. Ultimately this leads to improved measurement of risk, uncertainty and policy effectiveness in the context of important economic issues like the uncertainty of job placement, duration and retirement; time-varying effects of education on earned income; length and duration of shock effects in the macro-economy and the financial world and better risk analysis of extreme outcomes and crises.
Specific research topics are:
- Simulation-Based Bayesian Econometrics (SBBE)
- Long-run macro-econometric modeling: growth and cycles
- Short-run macro-finance modeling: Volatility and Risk
- Mixture processes/Neural Networks
- Income distributions