I am at John Taylor’s annual macroeconomic conference on monetary policy. The first panel included comments from Richard Clarida and Larry Summers. Time for questions ran out so I put mine here.
Larry Summers said he asked an RA to run simulations of FRB/US, the primary Fed model, and reported that the results said a lot about FRB/US but nothing about the real world. I was shocked and amazed!! Where did Summers find RAs who were willing to run simulations of FRB/US??!!?? The results were not surprising to me, but my RAs would quit if I asked them to run FRB/US.
Summers emphasized that FRB/US was the Fed’s major model. I wanted to ask Richard Clarida, former Vice-Chair of the Fed, why that was still true. The Bank of Canada’s DSGE model for close to 20 years is ToTEM. They have also implemented modern solution techniques, as displayed in the papers by Lilia and Serguei Maliar. ToTEM does not rely on simple linearization methods and also has several sectors and types of consumers.
FRB/US, on the other hand, is written in EVIEWS (excellent time series estimation software but not well-suited for solving macro models) using equation-solving code written by a Fed person instead of professionally written software. It also has far more aggregation than ToTEM. FRB/US goes back to 1996, a time when even Nordhaus was using GAMS to solve DICE. FRB/US has always been far behind many other economists.
I once asked Clarida what he was doing to help his students solve models he could not solve. He basically said nothing. He has declared that he likes models that are “Simple, but solvable.” When he was at the Fed, I asked him the same question about helping train Fed economists; he refused to answer. Now I am asking why the Fed can’t follow the example of the Bank of Canada and build more sophisticated models of the economy?
Many macroeconomists tell me that all their models are bad. Of course, they can only speak of the models that have been written. Of course, most models will be useless, as is the case for most research. However, continuing to use primitive computational tools will help insure more failure. If macroeconomists continue to do modeling, using modern computational tools will surely improve their chances of success. Canada knows this. Why does the Fed refuse to keep up with Canada?