Toshiaki Watanabe

Professor, Institute of Economic research, Hitotsubashi University

The DSGE-VAR model is used for estimating the parameters in a VAR model in a Bayesian fashion with the prior selected as if the artificial data were generated from a DSGE model. The more data is generated from the DSGE model, the more information from the DSGE model is incorporated. How important is the information from the DSGE model can be evaluated by changing the number of the data generated from the DSGE model and comparing the fit to the actual data. This article explains the DSGE-VAR model and applies it to the macroeconomic data in Japan. The DSGE model used in this article is a New-Keynesian model with several frictions such as stickiness in price and wage, habit formation and adjustment cost in investment, developed by Christiano, Eichenbaum and Evans (CEE, 2005). The comparison of the marginal likelihoods of the DSGE-VAR models with different sample sizes of the artificial data generated from the CEE model provides evidence that there is misspecification in the CEE model but the information from the CEE model is still useful, which is consistent with the finding of Del Negro et al. (2006) for the US economy.