Richard Clarida presented “Monetary policy in open economies–Practical perspectives for pragmatic central bankers” at John Taylor’s 2014 conference “Frameworks for Central Banking in the Next Century” . The following is a cleaned up version of my comments and question to Prof. Clarida. The original transcript is at the end of this post.
I’m Ken Judd over at the Hoover Institution. I liked Clarida’s paper because it was precise on many points. One point that he was precise on was the assertion that there was a unit root in the [equilibrium system] of your model. It is not clear what that unit root implies. [Clarifying comment for those not familiar with this literature: I presumed that Clarida’s log-linear system was a linearization of a nonlinear system]. There’s a unit root in the linearized system. In a course I took 35 years ago as an economics grad student at the University of Wisconsin, I learned that the presence of a unit root in a linearized system meant that one needs to use Hopf bifurcation methods to ascertain stability. I believe Jess Benhabib, who is just a few subway stops away from you, wrote some of those papers 35 years ago. Your tacit argument that the presence of a unit root in the linearization implies a non-stationary but weakly stable solution does not follow from your analysis. You haven’t done the bifurcation analysis necessary to make that claim. The true local solution of the underlying nonlinear model could be stationary and stable, or there could be a explosive instability there which would make any local approach invalid.
Another example of where people made errrors using linearizations were the designers of the Takoma Narrows bridge. They used linear approximations to conclude that fluctuations in the bridge would be stable in response to wind. The bridge collapsed and is a famous example of an engineering mistake. They realized — that was 75 years ago — that linear approximations were not adequate, and have since used methods far more reliable. We haven’t had those collapses since.
You and Maury made many comments about how dificult it is to solve macroeconomic models. Maury even used the word “analytically” to describe one concept of a solution. I suspect whatever you meant by solution was also very limited. You and Obstfeld have a very limited concept of what can be solved. Consider, for example, the plane on which you flew out here. The people who designed that plane had to deal with computational challenges far beyond those present in economic problems and solved them with computational tools far more powerful than anything used in economics. I could say the same for the engineers who designed the car that Maury drove down here. My question to both of you and any macro professor in the room is, “What are you doing so your students can solve problems that you can’t solve?”
[Note: There was scattered nervous laughter at this point.]
Clarida’s response to my question was surprising:
I will send them to speak with Jess Benhabib if they want to solve difficult models.
More recently, Clarida claimed that he said he would send them to some other NYU macroeconomists, but that is not important. My immediate thought was that perhaps Clarida’s students should just go to NYU instead of Columbia. I wondered what his dean would say if he knew that his professors were telling students to go talk to NYU professors if they wanted to do difficult work. In fact, if he was in the Columbia Business School, I would have contacted the dean!
I am familiar with the Ivy League’s policy of outsourcing student training in numerical methods. From 2005-2012, I helped run ICE, a workshop sponsored by and paid for first by the University of Chicago and then by Jim Heckman. One student from Harvard told me “Professor X says I should use Y methods in my thesis”. I said, “Professor X knows nothing about Y methods”. The student replied “That’s why he told me to come to ICE to talk with you.” I also note that this Harvard professor (who was not Xavier Gabaix) sent his student to a workshop run by Karl Schmedders (Professor, Northwestern University), Che-Lin Su (postdoc, Northwestern University), Todd Munson (PhD from the University of Wisconsin) and me (all degrees from the University of Wisconsin).
The mathematical point I raised is known to anyone who studied economic dynamics in the 1970’s. I had the advantage of having Buz Brock as a professor, but many economists — including Buz’ collaborators Michael Magill, Jose Scheinkman, Dee Dechert, Mukul Majumdar, Lenny Mirman and others — were doing serious work on economic dynamics. Bill Barnett at the University of Kansas has written many papers on bifurcations in macro models, most recently “Shilnikov Chaos, Low Interest Rates, and New Keynesian Macroeconomics” in JEDC.
I was amused by the title of his paper: “Practical perspectives for pragmatic central bankers”. Do you really think that a pragmatic central banker would think it practical to rely on a dynamic analysis that is not robust to even small model misspecifications?
Clarida ignored all of this work. In the published version, he did thank me for “insightful suggestions for future work.” However, the first comment in his conclusion told me that I should not hold my breath waiting for that future:
“The models outlined above are simple — but solvable!”
In 2014, Clarida was just a macro professor and I did not see why I should care more about the quality of education at Columbia than he did. However, in the autumn of 2020, Clarida gave another talk at a Hoover macro seminar. He was a Vice-Chairman of the Federal Reserve. His “Simple — but solvable!” motto might have a direct impact on my economic life. So, I reminded him of my 2014 question and asked him
“What are you doing to improve the ability of Federal Reserve economists to do serious quantitative analysis of monetary policy issues?”
On that occasion, he gave no response. Again, I thought I should contact his boss, but I quickly realized that would be foolish. His boss was President Trump, who appointed him to the Board.
“Simple — but solvable!” may work for some macro models, but I doubt it works for the real world.
=========. 2014 Conference Transcript ==========
[Here is the transcript of my comments as provided to me by Marie Christine Slakey]
“Kenneth Judd: I’m Ken Judd over at the Hoover Institution. I liked Clarida’s paper because it was precise on many points. One point that he was precise on was the assertion that there was a unit root in the system. Now, the fact is that’s not clear. There’s a unit root in the linearized system. Of a non-linear system, however, I learned in a course I took 35 years ago in graduate school, at the University of Wisconsin, that if you face that, you should pull out methods called Huff bifurcation methods. I believe Benhabib, who is just subway stops away from you, wrote some of those papers 35 years ago. So therefore, that economic assertion that there’s a unit root to deal with here, non-stationarity, does not follow from your analysis. You haven’t done the analysis to make that claim. It could be stationary and stable, or there could be a fundamental instability there [Dash] nonlinearity.
Now another example of where people screwed up on this were bridge designers, that also used the same kind of linear methods until 1940, after the collapse of the Takoma Narrows bridge. And they realized [Dash] that was 75 years ago [Dash] that linear approximations were not adequate, and they used methods far more reliable. And we haven’t had those collapses since.
Now there’s many references in your comments and also in Maury’s comments about models that we can solve. But your definition of solution[Ellipsis] Maury even said “analytically.” And I think whatever you meant by solution was also similarly very limited. And in fact, the people who designed the plane you flew here on were not limited to those methods, nor the car that Maury drove down here. They used other methods that are far beyond what is used in economics. My question to both of you and any macro professor in the room is, “What are you doing so your students can solve problems that you can’t solve?”