By Takeshi Yagihashi
The paper conducts a comprehensive survey on the current state of the dynamic stochastic general equilibrium (DSGE) models developed by policy institutions, including central banks, government agencies, and international organizations around the world. Our main sample consists of 84 models developed by 58 institutions, and many of them were developed or updated after the 2008 financial crisis. We first document the evolution of macroeconomic models used for policy purpose, and then provide summary statistics on the models by type of institution, region, and number of authors of the publication. We find that there is a steady increase in the development of DSGE models by policy institutions. While central banks have been the main users of DSGE models, more government agencies in Europe have been actively developing their own DSGE models in the years following the 2008 Global Financial Crisis. We also find that some institutions have multiple DSGE models serving different purposes. Next, we narrow our focus to a subset of 42 models that are owned and actively used by policy institutions, and conduct a model comparison based on five key model features. Although the models share common basic structures, there are large variations in parameter values and modelling strategies, some of which do not necessarily reflect the findings of the empirical literature. Finally, we create a score card for each model depending on whether the model incorporated recent empirical findings on the five model features. Two models have a score of 4 out of 5, and the overall average is 2.21. In conclusion, there is a greater need for future DSGE policy models to adopt more recent findings in the empirical literature.
There is widely held belief that DSGE models are not used in policymaking. This is far from the truth, and this papers gives a nice compilation of some of the uses in ministries, central banks, and international organizations. As I have shown several times on this blog, they are incredibly powerful for understanding scenarios, especially when one faces a historically new situation, when there is a lack of data, or when one needs to disentangle several shocks.