Donald Kohn on The Money Illusion
In my recent book entitled The Money Illusion, I was sharply critical of Fed policy during the Great Recession of 2007-09. Donald Kohn was Vice Chair of the Federal Reserve Board during this period, and has written a very thoughtful review of my book. I encourage people to read the entire piece.
The most controversial aspect of my book is the claim that Fed errors in 2008 caused the recession to be much more severe than otherwise. I based this claim on the fact that, in my view, a plausible alternative monetary policy regime would have prevented the sharp fall in NGDP growth during 2008-09. Kohn argues that during a key period in 2008 the Fed lacked accurate data showing weakening NGDP, due to both data lags and errors in the initial estimates of NGDP growth:
Moreover, the growing economic weakness in the current vintage of data for Q3 2008–which would have set off alarm bells in the Fed–was not evident in the immediate lead up to Lehman’s bankruptcy.Footnote 5 NGDP is now estimated to have grown only 0.9% at a seasonally adjusted annual rate (SAAR) in Q3 and to have collapsed at a 7.6% SAAR in Q4, which Sumner argues should have led the Fed to begin more aggressive easing in Q3. But at the end of July, Board staff was projecting 4.3% SAAR growth in NGDP for Q3 and 3.9% for Q4-2008 to Q3-2009. Private forecasters were in close agreement; the Survey of Professional forecasters in August saw NGDP growing at a 4.3% annual rate in Q3 and at 4.1% for the following four quarters, with rising interest rates. Obviously, none of this suggested an impending collapse that required immediate monetary policy attention in July and August.
In fact, Q3 NGDP growth was first published—one month after the end of the quarter—at a 3.8% SAAR, and it was revised down only slightly through the next two revisions over subsequent months. The current estimate of only 0.9% SAAR growth, which Sumner cites as evidence of too-tight monetary policy, came much later. The difference between first-published and current-estimated growth was similar for Q4—a 3.5 percentage point downward revision (from − 4.1 to − 7.6). This experience underlines several serious weaknesses for NGDP targeting—the difficulty of accurately estimating and forecasting, the availability of only quarterly data with a lag, and the size of revisions; the latter could have a material impact on estimates of the policy necessary to achieve NGDP level targets.
This is roughly the argument I would make if I were asked to defend the Fed’s position. I have three responses to this general argument:
1. Other data clearly showed the economy was sliding into recession in mid-2008. For instance, August unemployment had already risen by 170 basis points from the previous low, and that large a rise in the unemployment rate is a 100% accurate indicator of recession. Indeed even a rise half that large would be a 100% accurate inflation indictor. This fact suggests the government must do a better job of deriving NGDP estimates in real time.
I understand that complaining about our GDP data does not invalidate the core of Kohn’s argument. My next two points that are more essential:
2. Lehman failed in mid-September, and soon after this occurred various market indicators (such as TIPS spreads) clearly suggested that money was too tight, as both inflation and employment forecasts were falling well below the Fed’s implicit policy mandate. Thus even before we had the revised NGDP data, various asset market forecasts clearly suggested that we had a major demand shortfall. The Fed needs to respond to forecasts, not backward looking NGDP data.
Nonetheless, some might argue that by mid-September it was too late to do anything to avert a severe recession. I don’t believe it was too late, but it’s my third point that is the most important:
3. Level targeting. I cannot emphasize enough the need for some sort of level targeting policy regime. The Fed needs to be telling the markets that whatever happens in the short run during a banking crisis, NGDP will be about 8% above current levels two years into the future. They need to emphasize that they will do whatever it takes so that markets expect roughly 4% average NGDP growth over the next two years.
When I mention level targeting, many people assume that I am obsessed with correcting errors, with undoing policy mistakes. That is not the purpose of level targeting. The point is to prevent the initial under or overshoot in NGDP growth (or at least make it milder than otherwise.)
[Here I might use the analogy of the Mutually Assured Destruction doctrine in nuclear war game theory. The point of massively retaliating against a nuclear attack on your country is not to seek revenge, not to “even the score”, the point is to deter the initial attack from occurring in the first place. If you have not credibly committed to that doctrine ahead of time, then it’s pointless (which the theme of the film Dr. Strangelove.]
Now let’s think about how these three pieces relate to late 2008. Despite the flawed NGDP data, markets clearly saw that money was too tight to achieve the Fed’s implicit policy target of roughly 4% NGDP growth. Markets saw high frequency data on everything from ocean shipping rates to payroll employment to US equity prices (and many more data points), and putting all of this data together understood that a recession was developing. Under a level targeting regime, markets would have expected a very expansionary Fed policy to bring NGDP back to the trend line over the next few years.
And leads us to the key point, which is missed in so much discussion of level targeting. Market expectations of NGDP growth over the next few years is basically what Keynes meant by “animal spirits”. Longer run NGDP expectations are the primary factor that drives aggregate demand in the short run. (Michael Woodford has formally modeled the way that future expected demand growth drives current demand in the economy.) This is why the current economy is extremely sensitive to the future expected path of monetary policy.
In a policy regime where the Fed commits to bring NGDP back to the trend line ASAP, the initial deviations from that trend line become much smaller. As an analogy, if a swing oil producer commits to do whatever it takes to bring oil prices back on target in three months, then the effects of a near term oil production disturbance caused by a missile strike on Saudi Arabia become much smaller. Wholesalers sell oil out of inventory, anticipating they will be able to refill in 3 months at a reasonable price.
Of course this is just an analogy, but the same is true for the macroeconomy. Business investment decisions during a temporary period of banking distress will be much different if those making real investment decisions expect a deep and prolonged recession, as compared to the case where they expect the Fed to bring NGDP back to trend within two years. In the latter case, even the initial drop would be much smaller. NGDP soared during 1933, despite much of the banking system being shutdown for months, as dollar depreciation created expectations of higher NGDP in future years.
Many events that look to most people like “exogenous shocks” are actually investment swings driven by a loss of confidence in the future path of monetary policy. There may be some exogenous factors causing that lack of confidence (say financial turmoil or fiscal austerity) but it is the Fed’s job to offset those shocks.
I don’t know if there will be a deep demand side recession in 2023. But I do know that if there is a deep demand side recession in 2023, its cause will be market perceptions that the Fed will not create adequate NGDP growth in 2024 and 2025.
PS. I say “deep” recession, as one can at least plausibly argue that a very mild recession is an acceptable cost of bringing inflation down. A deep demand-side recession would be inexcusable.