# 13-191/III (2013-11-03)

Nalan Basturk, Erasmus University Rotterdam; Cem Cakmakli, University of Amsterdam; S. Pinar Ceyhan, Erasmus University Rotterdam; Herman K. van Dijk, Erasmus University Rotterdam
History, Bayesian Econometrics
JEL codes:
B4, C01

After a brief description of the first Bayesian steps into econometrics in the 1960s and early 70s, publication and citation patterns are analyzed in ten major econometric journals until 2012. The results indicate that journals which contain both theoretical and applied papers, such as Journal of Econometrics, Journal of Business and Economic Statistics and Journal of Applied Econometrics, publish the large majority of high quality Bayesian econometric papers in contrast to theoretical journals like Econometrica and the Review of Economic Studies. These latter journals published, however, a few papers that had a substantial impact on Bayesian research. The journals Econometric Reviews and Econometric Theory published key invited papers and special issues that received wide attention, while Marketing Science shows an ever increasing number of papers since the middle n ineties. The International Economic Review and the Review of Economics and Statistics show a moderate time varying increase. The early nineties indicate an upward movement in publication patterns in most journals probably due to the effect of the ‘Computational Revolution'. Next, a visualization technique is used to connect papers and authors around important theoretical and empirical themes such as forecasting, macro models, marketing models, model uncertainty and sampling algorithms. The information distilled from this analysis shows the names of authors who contribute substantially to particular themes. This is followed by a discussion of those topics that pose interesting challenges for discussion amongst Bayesian econometricians, namely the computational revolution, unobserved component and flexible model structures, choice models, IV models, dynamic models and forecasting. Three issues are summarized where Bayesian and frequentist econometricians differ: Identification, the value of prior information and model evaluation; dynamic inference and nonstationarity; and vector autoregressive versus structural modeling. A major topic of debate amongst Bayesian econometricians is listed as objective versus subjective econometrics and communication problems and bridges between statistics and econometrics are summarized. The paper ends with a list of four important themes that will be a challenge for twenty-first century Bayesian econometrics: Sampling methods which are suitable for parallelization and GPU calculations, complex economic models which can account for nonlinearities, analysis of implied model features such as risk and instability and incorporating model incompleteness in econometric analysis.