This is a letter from James Bullard, president of the St. Louis Fed, in response to this post from Tim Duy:
14 February 2012
I appreciate your commentary, and all the commentary, on my Chicago speech from last week. I take the gist of these comments to be “you didn’t show us a model.” That is fair enough, I did not. (Readers may also wish to check Scott Sumner, Noah Smith, Paul Krugman, David Andolfatto, Brad DeLong, David Beckworth, and Steve Williamson at their respective blogs, and possibly others I have not seen yet.)
As you know, I am not too keen on “output gap” ideas as they are knocked around in the business press and in policy circles. I just do not think the output gap rhetoric matches up very well with the state of knowledge in the macroeconomics literature, either conceptually or empirically.
Neil Irwin at the Washington Post does an excellent job of telling the standard story concerning the output gap.
I know you like this story, and you have a lot of company, because it dominates much of the discussion about the U.S. economy. I said this potential output calculation is basically an extrapolation of real GDP from 2007Q4 using growth rates from the years immediately preceding. Of course it is not, it is just … statistically indistinguishable from an extrapolation of real GDP from 2007Q4 using growth rates from the years immediately preceding!
For more detail on approaches to measuring potential output, see the St. Louis Fed’s 2008 conference “Projecting Potential Growth: Issues and Measurements,” published in 2009.
Readers might be especially interested in the paper by Susanto Basu and John Fernald, “What Do We Know (And Not Know) About Potential Output?“
As Basu and Fernald make clear, a lot rides on what is meant by potential output, and one really needs an explicit general equilibrium model to give an appropriate definition.
First, I want to restate my bubble idea in more geeky terms based in part on the basic story presented by Irwin. I know I am an army of one on this issue, but I think it is important to debate the output gap concept because it is having a huge impact on policy choices. And, I think my approach makes more sense given the very damaging housing bubble in the U.S. during the mid-2000s.
Second, as I am under no illusions that I can get you to come to reason on the fallacies behind the Irwin graph any time soon, I want to make a plea to at least use the available literature to define potential output appropriately for monetary policy purposes. As Basu and Fernald stress, we need a full DSGE model to be able to discuss the appropriate measure of the output gap for monetary policy. Fortunately, outstanding work by Mike Woodford at Columbia and co-authors has at least given us a benchmark model. In that work, the key gap concept is the difference between the sticky price and flexible price level of output, not the difference between actual output and a measure of trend output as in the Irwin graph.
The housing bubble in the 2000s
Here is a shorter and geekier version of the Chicago talk: If we look at Irwin’s graph, actual output is essentially at CBO potential during 2005, 2006, and 2007. There is nothing about the CBO potential calculation that allows “bubble” levels of output. That is just not part of the analysis–it is off the radar screen. Potential in this picture is simply a projection based on a production function approach.
At the same time, we often say that these years were characterized by a bubble in housing. One way to interpret this is that fluctuations in real variables were driven by beliefs alone. We certainly have a very good candidate for what this widespread belief was–namely, “house prices never fall.”
If households and businesses had ignored the house price developments as a sort of amusing side show, it would not have been so important. But our rhetoric about the decade suggests otherwise. Households consumed more through cash-out refinancing, developers built more, borrowing increased, Wall Street produced new financially-engineered products to feed the boom, and ancillary industries like transportation thrived. Output went up, and labor supply was higher than it otherwise would have been.
Rhetorically, this is consistent with what most analysts say happened. But we also have a large literature on so-called sunspot equilibria which tells us that fluctuations can be self-fulfilling (driven by beliefs alone) and consistent with rational expectations. According to that literature, the technology for the production of goods would not have to change at all, but the amount of output, consumption, labor supply and other real variables would fluctuate solely in response to the belief. These fluctuations lower welfare for risk-averse households. Potential output via a production function approach would then be sensibly described as that amount of output which would have been produced in the absence of the belief. I think it is plausible that such a line would be lower than the CBO potential line in Irwin’s picture, and thus that the current output gap even by a production function metric would be smaller than the one in the picture.
So, what Irwin’s picture is doing is taking all of the upside of the bubble and saying, in effect, “this is where the economy should be.” But that peak was based on the widespread belief that “house prices never fall.” We will not return to that situation unless the widespread belief returns. I am saying that the belief is not likely to return–house prices have fallen dramatically and people have been badly burned by the crash. So I am interpreting your admonitions on policy as saying, in effect, please reinflate the bubble. First, I am not sure it is possible, and second, that sounds like an awfully volatile future for the U.S., as future bubbles will burst once again.
This is admittedly a qualitative story, but much of the rhetoric about the U.S. economy during this period fits this description. So it is not that the bubble destroyed potential, instead it is that actual output was higher than properly-defined potential during the mid-2000s, and then it crashed back as the bubble burst.
The macroeconomic literature on sunspot equilibria is dense and filled with conditions under which such phenomena could occur. But I will say that one key condition keeps recurring: low real interest rates.
As I noted earlier, the Irwin description is the dominant view of the U.S. economy. But, as you and many others have stressed, we have not seen the bounce back toward potential that would be suggested by that picture. That is giving me pause, and frankly I think it is giving everyone who follows the U.S. economy pause. We owe it to ourselves to at least consider alternative possibilities.
Basu and Fernald
Ok, let’s now forget about self-fulfilling beliefs and simply assume that whatever was going on in housing during the mid-2000s was more benign.
As Basu and Fernald discuss, “… few, if any, modern macroeconomic models would imply that, at business cycle frequencies, potential output is a smooth series.” One possibility would be to use the leading monetary policy literature (e.g., Woodford [2003, Interest and Prices, Princeton University Press]) available to tell us what potential output should be. According to the New Keynesian literature, the relevant output gap is the distance between the actual level of output under sticky prices and the flexible price level of output. It is the flexible price level of output that represents the potential in the economy. The flexible price level of output would fluctuate continuously in response to shocks hitting the economy. This gap has been estimated in the literature, and I think it is fair to say that the concept is quite different from what is in the traditional story as told by Irwin. There are also unemployment versions of this (that is, NK models with search unemployment included)–I might recommend papers by Mark Gertler at NYU and co-authors. But the concept is the same.
So, if you do not believe my sunspot story, then fine, we can assume that housing price appreciation during the 2000s did not importantly affect output and other key macroeconomic variables. But let’s at least use the appropriate definition of the output gap according to the available NK literature.
I know this last point was not in my talk in Chicago, but it is a theme that I often return to because I think is important in the output gap context.
Thanks again for the comments. As always, I find them stimulating and insightful. I think ongoing debate concerning these difficult issues is important.
This post originally appeared at Economist’s View and is posted with permission.