James Bullard, the President of the St. Louis Federal Reserve Bank, is a very fine economist, having worked his way up the ranks at the St. Louis Fed after joining the research department at the St. Louis Fed in 1990, as newly minted Ph. D. from Indiana University, publishing his research widely in leading journals (and also contributing an entry on “learning” to Business Cycles and Depressions: An Encyclopedia which I edited). Bullard may just be the most centrist member of the FOMC (see here), and his pronouncements on monetary policy are usually measured and understated, eschewing the outspoken style of some of his colleagues (especially the three leading inflation hawks on the FOMC, Charles Plosser, Jeffrey Lacker, and Richard Fisher).
But even though Bullard is a very sensible and knowledgeable guy, whose views I take seriously, I am having a lot of trouble figuring out what he was up to in the op-ed piece he published in today’s Financial Times (“Patience needed for Fed’s dual mandate”) in which he argued that the fact that the Fed has persistently undershot its inflation target while unemployment has been way over any reasonable the rate consistent with full employment, is no reason for the Fed to change its policy toward greater ease. In other words, Bullard sees no reason why the Fed should now seek, or at least tolerate, an inflation rate that temporarily meets or exceeds the Fed’s current 2% target. In a recent interview, Bullard stated that he would not have supported the decision to embark on QE3.
To support his position, Bullard cites a 2007 paper in the American Economic Review by Smets and Wouters “Shocks and Frictions in US Business Cycles.” The paper estimates a DSGE model of the US economy and uses it to generate out-of-sample predictions that are comparable to those of a Bayesian vector autoregression model. Here’s how Bullard characterizes the rationale for QE3 and explains how that rationale is undercut by the results of the Smets and Wouters paper.
The Fed has a directive that calls for it to maintain stable prices as well as maximum employment, along with moderate long-term interest rates. Since unemployment is high by historical standards (8.1 per cent), observers argue the Fed must not be “maximising employment”. Inflation, as measured by the personal consumption expenditures deflator price index, has increased to about 1.3 per cent in the year to July. The Fed’s target is 2 per cent, so critics can say the Fed has not met this part of the mandate. When unemployment is above the natural rate, they say, inflation should be above the inflation target, not below.
I disagree. So does the economic literature. Here is my account of where we are: the US economy was hit by a large shock in 2008 and 2009. This lowered output and employment far below historical trend levels while reducing inflation substantially below 2 per cent. The question is: how do we expect these variables to return to their long-run or targeted values under monetary policy? That is, should the adjustment path be relatively smooth, or should we expect some overshooting?
Evidence, for example a 2007 paper by Frank Smets and Raf Wouters, suggests that it is reasonable to believe that output, employment and inflation will return to their long-run or targeted values slowly and steadily. In the jargon, we refer to this type of convergence as “monotonic”: a shock knocks the variables off their long-run values but they gradually return, without overshooting on the other side. Wild dynamics would be disconcerting.
What is wrong with Bullard’s argument? Well, just because Smets and Wouters estimated a DSGE model in 2007 that they were able to use to generate “good” out-of-sample predictions does not prove that the model would generate good out-of-sample predictions for 2008-2012. Maybe it does, I don’t know. But Bullard is a very smart economist, and he has a bunch of very smart economists economists working for him. Have they used the Smets and Wouters DSGE model to generate out-of-sample predictions for 2008 to 2012? I don’t know. But if they have, why doesn’t Bullard even mention what they found?
Bullard says that the Smets and Wouters paper “suggests that it is reasonable to believe that output, employment and inflation will return to their long-run or targeted values slowly and steadily.” Even if we stipulate to that representation of what the paper shows, that is setting the bar very low. Bullard’s representation calls to mind a famous, but often misunderstood, quote by a dead economist.
The long run is a misleading guide to current affairs. In the long run we are all dead. Economists set themselves too easy, too useless a task if in tempestuous seasons they can only tell us that when the storm is past the ocean is flat again.
Based on a sample that included no shock to output, employment, and inflation of comparable magnitude to the shock experienced in 2008-09, Bullard is prepared to opine confidently that we are now on a glide path headed toward the economy’s potential output, toward full employment, and toward 2% inflation. All we need is patience. But Bullard provides no evidence, not even a simulation based on the model that he says that he is relying on, that would tell us how long it will take to reach the end state whose realization he so confidently promises. Nor does he provide any evidence, not even a simulation based on the Smets-Wouters model — a model that, as far as I know, has not yet achieved anything like canonical status — estimating what the consequences of increasing the Fed’s inflation target would be, much less the consequences of changing the policy rule underlying Smets-Wouters model from inflation targeting to something like a price-level target or a NGDP target. And since the Lucas Critique tells us that a simulation based on a sample in which one policy rule was being implemented cannot be relied upon to predict the consequences of adopting a different policy rule from that used in the original estimation, I have no idea how Bullard can be so confident about what the Smets and Wouters paper can teach us about adopting QE3.
PS In the comment below Matt Rognlie, with admirable clearness and economy, fleshes out my intuition that the Smets-Wouters paper provides very little empirical support for the proposition that Bullard is arguing for in his FT piece. Many thanks and kudos to Matt for his contribution.
Uneasy Money 
The out-of-sample power of the Smets-Wouters model isn’t very impressive once you dig a little deeper. There is a great article on this by Edge and Gurkaynak.
Edge and Gurkaynak find that over the period examined by Smets and Wouters, no model has much actual forecasting power, especially for inflation; in 1992-2004, inflation pretty much looks like white noise plus a constant. Since the Smets-Wouters DSGE model lacks any nontrivial internal propogation mechanisms (it’s more an “arbitrary decomposition into shocks of dubious interpretation” than a “model”), it does decently in comparisons of mean squared error, because its predictions don’t move around that much over the time period — which is great if the series itself is basically unforecastable! And since the papers were written in the 2000s, the model was constructed with implicit knowledge of the behavior of inflation and the macroeconomy as a whole in the 1990s, which means that it gets the mean for inflation roughly correct.
Basically, if you want to minimize mean squared prediction error of a series that resembles (over the period of interest) unforecastable white noise plus a mean, the optimal thing to do is to write a model that (1) gives nearly constant predictions and (2) gets the mean right. Smets-Wouters does this, since (1) its model is driven almost entirely by (unforecastable) shocks and not by nontrivial propogation mechanisms and (2) there was implicitly some data-snooping when the model was constructed.
LikeLike
David, Welcome back and happy New Year!
What about 1933? There was deflation and a big depression. MP even accompanied by schizophrenic FP did a reasonable job. The Bullards of the time wanted to keep cool and wait for the system to “regenerate itself.
LikeLike
I might be wrong here, but wasn’t the original purpose of the Fed to smooth out economic issues such as the one being experienced today by providing “elastic currency”? If that is not why it still exists, I fail to understand what its new purpose might be. It probably wasn’t the intent, but I think Bullard just made an argument that the institution to which he belongs is obsolete.
LikeLike
Three good comments! You gots sum smart readers, David.
LikeLike
The key assumption behind Bullard’s thinking, as Scott points out, is that the demand shock was somehow “exogenous”. So the Fed’s job is to prevent inflationary/deflationary spirals, and of course not shock the economy itself away from equilibrium, but otherwise sit tight and do nothing rather than risk instability. What he doesn’t realise is that the current malaise is itself the result of instability the Fed itself introduced. If the Fed’s role in this were purely passive it might make sense to wait for things to straighten themselves out. But just as there was no guarantee that business-as-usual would guarantee stability before (the Fed ended up derailing the economy) so too it doesn’t make sense to wait out the effects of this shock, as if business as usual behavior by the Fed could be expected to lead steadily to recovery. As Scott says, there is no such thing as not doing monetary policy. And the lack of clear targets is itself the problem which generates the instability – it is precisely Fed policy which is not always expected to hit its own targets that messes things up. And now Bullard proposes to continue to miss its own targets, as though this were some kind of “neutral” response to an “external” shock? Madness.
LikeLike
So what is the correct interpretation of the Keynes quote, then?
LikeLike
Matt: “….in 1992-2004, inflation pretty much looks like white noise plus a constant.”
An aside: which is exactly what you would expect to find if the Fed were targeting inflation. Inflation should be unforecastable.
LikeLike
There have indeed been attempts to see how Smet-Wouters model forecasts compare to outcomes more recent than 2006. The answer is the model does very poorly. In the linked post the variable being forecast is real GDP growth
http://www.angrybearblog.com/2012/04/more-on-dsge-forecasting.html
see also
http://rjwaldmann.blogspot.it/2012/04/comment-on-macroeconomic-model.html
These are two separate efforts to see how DSGE models deal with the great recession. Both generally conclude that models do OK really. Both conclude this based on ex post ad hoc adjustments to the models.
The fact that someone still assumes that if it is true of the Smets-Wouters model then it it is true of the economy is disturbing (but not really surprising).
LikeLike
David,
I thought the string bounced back. Why is Bullard taking on Bernanke and not Friedman (and Romney, Taylor. Cochrane, et al)
I know a bundler. I am going to give him a call and have him call Bullard for Obama
LikeLike
These discussions on models built on bizarre assumptions seem to me rather pointless as to monetary policy, an attempt to find a figleaf.
The two dominant aspects of monetary policy in the USA are:
* Political constraints block fiscal policy, so some people turn to monetary policy then, which is nowhere like a substitute.
* Monetary policy for the past 30 years has been very expansionary,
with a colossal private (and public) debt boom fueled by the Fed and
Treasury, and this expansionary policy has had two effects:
– A colossal long term asset price bubble in the USA and elsewhere.
– A colossal long term jobs and wages boom in large parts of China
and some parts of India.
It would be difficult to persuade me that:
* The outcome of even more expansionary monetary policy would be different from that of the past (booms in USA asset prices and in jobs and wages in China and India).
* The current Fed Board policy is not already wildly expansionary, as USA asset prices have been through a series of boomlets and jobs and wages in China and India have been still booming, all this despite a recession in USA jobs and wages.
Perhaps a fraction of bubbles created by monetary policy in domestic asset prices and in foreign jobs and wages turns into support for domestic jobs and wages, but that to me seems an extraordinarily inefficient way of doing it.
Unless of course you are a domestic asset investor, in which case it is marvellous.
LikeLike
Blissex:
The two dominant aspects of monetary policy in the USA are:
“* Political constraints block fiscal policy, so some people turn to monetary policy then, which is nowhere like a substitute.”
Political constraints do not block fiscal policy. Political constraints block non-discretionary fiscal policy. Political gains / losses get in the way of long term economic thinking. This is something that one time Keynesian Milton Friedman realized.
“* Monetary policy for the past 30 years has been very expansionary,
with a colossal private (and public) debt boom fueled by the Fed and
Treasury, and this expansionary policy has had two effects:
– A colossal long term asset price bubble in the USA and elsewhere.
– A colossal long term jobs and wages boom in large parts of China
and some parts of India.”
And if you compare asset to liability ratios in the household sector over say the last 50 years, you will find that households far from becoming “wealthier” and inching towards insolvency.
Household asset to liability ratio (1959): $10.1 in assets to $1 in liabilities
Household asset to liability ratio (today): $5.8 in assets to $1 in liabilities
You do not build wealth by consuming more than you produce and borrowing to make up the difference.
LikeLike
All, Please excuse the tardy response to your comments.
Matt, Again, thanks for your very enlightening comment and the link to the Edge and Gurkaynak paper.
Marcus, Thanks. You are right, of course, about the role of monetary policy in the Great Depression. I was thinking about mentioning it, but was in a hurry and getting tired. But that episode can’t be cited too often.
dajeeps, You are right about the idea of an elastic currency. Unfortunately that idea is subject to various interpretations (or misinterpretations), so it was actually used by some people to say that the money supply should contract as the real economy contracts, or expand as the real economy expands. That view, of course, neglects the possibility that there could be an independent shift in the demand for money, or velocity. But the elastic currency doctrine can be a bit of morass.
Steve, I’m impressed.
Saturos, Nicely put.
About Keynes, my interpretation, whether right or wrong, is that it is not enough to identify an equilibrium toward which the economy will eventually gravitate, and wait for the equilibrium to be achieved spontaneously. There may be obstacles blocking the path to equilibrium, and if monetary policy can help overcome those obstacles, it should be used to do so. That quote by the way comes from Keynes’s Tract on Monetary Reform, when Keynes was a pretty orthodox Marshallian economist, not form the General Theory when he had adopted a more (and, in my view, unnecessarily) activist policy stance.
Nick, Good point.
Robert, Thanks for the links. The thing about DSGE models is that you can shoot first and then draw the bullseye.
John, Let us know how it turns out.
Blissex, We obviously are not on the same page when it comes to relative effectiveness of fiscal and monetary policy.
Frank, The increase in asset values has probably led to increased borrowing against those assets, accounting for some of the decline in the asset to liability ratio.
LikeLike
David,
“Frank, The increase in asset values has probably led to increased borrowing against those assets, accounting for some of the decline in the asset to liability ratio.”
Um no. It is more like the cost of debt service is fixed while the value of assets are market determined on a last transaction basis. Hence, flights to liquidity provided by the Fed can actually make solvency issues worse. To alleviate solvency issues you either need the cost of debt service to float with the market value of the underlying asset (which means the possibility of negative interest rates) or you need an asset whose future value always exceeds the cost of debt service (which means the federal government sells equity).
LikeLike
“a model that, as far as I know, has not yet achieved anything like canonical status”
This post aged poorly
LikeLike