To the extent that the Chairman of the Fed is responsible for setting monetary policy and interest rates, I find myself wondering whether a computer algorithm could do the same job. Are the inputs knowable and robust enough that such an algorithm could be created, or is there still a role for personal judgment and feel?
Economics at that level is a pretty black art rather than a hard science.
I say that the Chairman of the Federal Reserve could be replaced with a computer as long as they hired him full-time as a programmer to build the code, tweak the code, and interpret the code for policy decisions.
Most of their data comes from computer databases and computer analysis anyway. They just interpret it to make decisions.
I suspect that you’ve extrapolated this question from the writing of Milton Friedman, who did in fact propose that the Fed Board of Governors be replaced by a computer program that would grow the money supply at a constant rate equal to the long-run growth rate of the economy.
Could this be done? Yes, absolutely–the program could even execute the necessary open market operations by program trading. Would this be a good idea? You’ll have to start a GD thread.
Actually not. I just wondered whether there was broad enough consensus about how the Fed should manage things, and about the kind and reliability of economic indicators that went into the decision-making, that the whole thing could be reduced to a computer algorithm.
I don’t recall the name of it, but I saw a TV show (yeah, I know, reliable cite that TV shows are…) about how they were modelling the US economy with computers. They had two different models, and both of them tracked very well with what was going on in the real economy for quite some time. Then, the economy went through a bit of a change, and all of a sudden one model was predicting the complete opposite of the other model, and basically at that point all of the computer simulations fell apart.
I think a similar thing would happen if you tried to replace the chairan of the federal reserve with a computer program. For a while, as long as nothing drastic happened in the economy, the program would probably make very good decisions. But, sooner or later, something odd is going to happen, and the computer model will go to hell in a handbasket real quick.
That’s all very true of predictive modeling, but how much predictive modeling is there in, say, interest rate setting? Wouldn’t you be looking backward at indicators of actual economic activity?
I think the question hinges on what legal authority the computer would have. If we were to truly replace the Chairman of the Fed with a computer, then when the computer started making poor decisions (or politically unpopular ones) based on a faulty economic model, how do we step in and fix it? Currently, the Chairman is appointed for a 4-year term. Is there any way for the appointment to be withdrawn? If the computer program is appointed to a term like that, we’d just have to sit it out.
It seems just as good (and much safer) to let the computer model things but maintain human oversight over actual decisions. Which is pretty much what we do now, isn’t it?
A better question might be: Has he been replaced…already?!
While formulating monetary policy involves reams of data and statistical analysis, at the level of the Federal Reserve Board, decisions aren’t so much about crunching numbers as establishing valuations for outcomes and the acceptable level of risks involved in pursuing them. No computer can tell you, for instance, what the discount rate should be, because that “should” entails an assumption about what we as a society want out of our monetary policy, and what consequences we’re willing to risk to get it.
I suppose you could argue that we could simply decide what outcomes we want from our monetary policy and tell the computer to make it happen. But somebody would still have to make that decision about what we want. And governments have proven to be pretty poor at making those decisions, hence the growth of independent central banks.
I would say that there is not. Today you have arguments about the importance of different variables at a given point in time–“housing starts are down, so we should be easing up on interest rates . . . yes, but factory utilization is up, so the economy might over-heat”. And so on.
Friedman’s point was that these kind of judgments are impossible, and the Fed shouldn’t even attempt them; they should just set the economy on “automatic pilot” at the long-run growth rate. However, his view is not the consensus. The consensus is that the Fed should at least attempt to micro-manage the economy, by stimulating it during recessions, cooling it down during expansions, and attempting to mitigate the impact of short-term shocks like hurricanes and oil price increases. These things cannot be done by computer.
So a computer could replace the Fed, but only if we adopted a view much closer to Friedman’s as to its mission.
It was my general (albeit not necessarily well-informed) sense that we *did *have such a consensus. In other words, I can’t remember there being much controversy, over the last couple of Fed chairmanships, over decisions about interest rates. I don’t doubt that there are extremists on both left and right who have differing views, but the broad middle seems generally supportive of decisions that the Fed makes.
Sure, there’s been a strong consensus on monetary policy over the last couple decades, with a focus on formal or informal inflation targeting. But that very consensus arose out of a values decision: the Fed and other central banks decided that the benefits of taming inflation outweighed the pain of tight credit and short term unemployment. Who knows what changes in the world economy could cause us to reasses what we want out of monetary policy? Heck, once upon a time, the prospect of deflation was the terror of the American economy.
My point is that SOMEONE has to make these big-picture decisions about what direction we want monetary policy to take. Once you figure out where you want to go, computers are great tools to help you figure out how to get there.
I recall the 2000-era rate increases at that time were widely considered completely wacky and made the recession both inevitable and much worse. That a recession was brewing was well known and ramping up rates so steeply was the opposite of what the Fed would normally do. Jeez, The Bubble burst and they still kept the rates too high.
A computer would have done a much better job at that time.
It has been my experience that when you understand a process or system well enough to construct an accurate simulation, and to be able to verify the simulation results with “common sense” then you pretty much don’t gain any significant insight from the pain of producing the simulation in the first place.
Your memory of that is much different than mine. My memory of that was the main critisim was that the rate increases did not come soon enough.