Sunday, January 14, 2018

Do Changes in Potential Output and Data Revisions Make NGDP Targeting Impractical?

It's Back...
Over the past few months there has been increasing chatter about the need for a new framework for U.S. monetary policy. The Peterson Institute for International Economics (PIIE), for example, recently had its Rethinking Macroeconomic Policy conference where, among other things, Ben Bernanke called for the Fed to adopt a temporary price-level target. PIIE also launched Angel Ubide's new book  on reforming monetary policy. Similarly, at the AEA meetings there was a session titled Monetary Policy in 2018 and Beyond where Christina Romer again made the case for a NGDP level target. Likewise, the Brookings Institute held a recent conference on whether the Fed should abandon its 2 percent inflation target. There, Jeff Frankel shared the arguments for a NGDP level target and Larry Summers endorsed it. Others at the conference, like San Francisco Fed President John Williams called for a price level target.

I am glad this conversation is happening. It is not new--some of us have been having it since 2009--but I get the sense that it is gaining traction. The turnover at the Fed and the opportunity it creates for new thinking makes this conversation about new monetary policy frameworks incredibly important now. 

As this conversation continues to grow, so will the interest in the options available including nominal GDP level targeting (NGDPLT). Obviously, I have much to say here, but for now I want to respond to two critiques often applied to NGDPLT: (1) changes in potential output and (2) data revisions make NGDPLT an impractical rule to implement. I think these concerns are misplaced as explained below.

Critique 1: Changes in Potential Output: A Feature or a Bug for NGDPLT?
Since NGDP growth is approximately the sum of real GDP growth and inflation, many observers are concerned that changes in potential output will cause wide swings in inflation. Here, for example, is Goodhardt et al. (2013):
[A]ny overestimation of the sustainable real rate of growth... could force [a central bank], subject to a level nominal GDP target, to soon have to aim for a significantly higher rate of inflation. Is that really what is now wanted? Bring back the stagflation of the 1970s...?
Put differently, some worry that a NGDP target will not provide a stable nominal anchor. For these folks this is a bug in NGDPLT. This concern is unwarranted for several reasons.

First, a NGDPLT target does provide a nominal anchor. It pins down the long-run growth path of nominal income. Put differently, it approximately stabilizes the growth path of nominal wages. Yes, it will allow more flexibility with inflation but as argued later that may be a good development.

Second, on a practical level, a NGDP target would never have generated 1970s-type swings in inflation given the actual history of potential output growth. The figure below illustrates this point. It assumes a 5% NGDP target and takes the year-on-year growth rate of the CBO's potential real GDP measure (blue line) as given. The trend inflation rate (black line) implied by this counterfactual monetary policy is plotted in the figure.

In this counterfactual, the inflation trend rate averages 1.97 over the entire 1960-2017 period. At most, it temporarily hits just above 4 percent right after the crisis. There is no 1970s inflation. Fears that a NGDPLT will bring back 1970s-type inflation is a red herring.

Interestingly, this counterfactual would have led to lower inflation in periods of high-trend real GDP growth periods like the 1960s and led to higher inflation in low-trend GDP growth periods like the present. To the extent there is hysteresis and potential real GDP is not truly exogenous to monetary policy, the counterfactual higher inflation (as a symptom of higher nominal demand) may have actually reduced the collapse in potential output over the past decade.In any event, a NGDPLT since 1960 would not have led to soul-crushing swings in inflation that some claim.

What many observers miss is that with a NGDPLT target the central bank does not need to worry itself over the latest changes to potential output. As the chart shows above, these changes happen fairly regularly. A NGDPLT lets these potential output changes be absorbed through reasonably-sized changes in the inflation rate. Inflation flexibility provided by a NGDPLT is a virtue here.2

Monetary policy, as it is currently practiced, does not have freedom. Most inflation-targeting central banks loosely follow something like a Taylor rule where they need to know the output gap in real time. That requires knowing both real GDP and potential real GDP in real time--an impossible task.

A NGDPLT acknowledges this ignorance and says to simply focus on stabilizing nominal demand. Potential output changes are therefore a feature rather than a bug for NGDPLT.

Critique 2: Data Revisions Make NGDP an Impractical Framework
There are two replies to this. First, as noted above, inflation targeting as it currently practiced requires real-time knowledge of both real GDP and potential real GDP.  This Taylor-rule framework is therefore is also subject to data revisions. Moreover, it subject to two of them. A NGDPLT target, on the other hand, only faces one variable subject to data revisions. So if data revisions make NGDPLT impractical even more so for the current inflation targeting framework.

Second, as Josh Hendrickson and I show in a recent paper, the Fed's forecast of NGDP for the current period is not biased (i.e. the forecast error is stationary). That means even if there are data revisions to the official statistics, the Fed can rely on its forecasts of NGDP to avoid this challenge. Below is chart from our paper that captures this ability of the Fed:

We also show in the paper that the Fed's forecast of the potential GDP is biased (i.e. the forecast error has a unit root). Consequently, the data revision argument against NGDP is not a convincing one, while it is a serious one for flexible inflation targeting.

For me, then, I do not worry about the above two critiques. The real issue with NGDPLT, in my view, is how to credibly implement it. But that is a topic for another post.

As commentator John notes, these problems could also be addressed by Scott Sumner's NGDP futures contracts idea. I would also note that directly targeting a monthly nominal wage index in the manner outline by George Selgin would be another fix.

Related Links
NGDPLT and the Problem of Permanent vs Temporary Monetary Base Injections 
NGDPLT and the Eurozone Crisis

1 A forceful case for hysteresis in the U.S. economy is made by Reifschneider, Wascher, and Willcox (2013)J.W. Mason (2017), and Coibion, Gorodnichenko, and Ulate (2017).
2There are other reasons why the countercyclical inflation created by a NGPDLT would be beneficial. It would lead to better risk-sharing between creditors and debtors as shown in Koenig (2013), Sheedy (2014), and Bullard et al. (2015).


  1. These two issues have been the biggest thing that kept me hesitant on NGDPLT. I think I've resolved them in my own head. On 1) all you have to do is have a RGDP futures market and every 5 years or so adjust the NGDP target growth rate based on the long-run RGDP assumption. On 2) you just need to make sure the NGDP futures are written in a way to be as agnostic to revisions as possible.

    1. Yes, NGDP futures contracts is another solution. Good point.

  2. I'm not an expert but GDP makes me uneasy because even after revision, is it really reliable? You have to record every single transaction in the economy to get a good number. Even ignoring the problems of big brotherism implied in getting good quality data, there could be large variations in the size of the informal or illegal economy caused by say, changes in technology, skewing the data. With inflation you only really need to get a small sample of the economy, a much easier task.

    Couldn't we go towards a PLT that is flexible, maybe telling central banks to use their maneuvering margin to give some weight to stabilizing GDP path.

    PLT also has the additional benefit of making it much easier for people, businesses and banks to negotiate longer term contracts, easier than IT in my opinion. With GDP targets, all contracts are negotiated using a fairly intangible yard stick representing a share of the total future economy. With inflation targets, contracts are negotiated in units tied to a basket of real stuff which is much more tangible, at least to me.

    The only thing that still bothers me about a PLT compared to an NGDP based target is that the NGDP one is more able to absorb huge supply shocks. If a nuclear war suddenly cuts real GDP by 20%, it will be really difficult under an inflation target for most past money denominated contracts to be honored. There simply won't be enough stuff produced out there to honor those contracts. An IT monetary regime will create additional waves of defaults with additional risk of a complete collapse of the financial system. Under an NGDP regime, debtors would owe 20% less real stuff when the world produced 20% less and they might be able to absorb that and not default and the financial system and labor market might survive better.

  3. A 5% NGDPLT might have cooled inflation in the 1970s, but does anyone want to guess what the unemployment rate would have been in that regime? In the 1970s there were times when both the unemployment rate and the NGDP growth rate were above 10%. What happens when NGDP is 5% in such a scenario? 15% unemployment? 20% unemployment?

    "Oh, but we will never find ourselves in that situation again. That was anomalous." Oh really? How can they be so sure? I would rest easier with the market monetarist position if they would just bite the bullet and entertain this sort of scenario and explain how they would react. Maintain iron discipline to the 5% NGDPLT in the face of 10% or higher unemployment? At least that would be a consistent response, and more intellectually honest than claiming that it can't happen in the first place.

    1. Suppose we have stable NGDP growth and also 20% unemployment. You seem to be assuming that in that scenario, some alternate monetary policy that led to even higher NGDP growth would also result in increased employment. Why do you think that? Why would (say) 10% NGDP growth clear the labor markets whereas 5% NGDP growth wouldn't?

      Generally, labor markets fail to clear for two reasons. One is excessive money demand (2008). Another is when the economy has some other binding constraint, such as an energy shortage (1975). If money demand is the bottleneck, then increasing money supply is the solution. If it's not, as in the 1970's, then increasing the money supply brings inflation and tears without increasing employment.

  4. Great post David. NGDPLT would be such a huge improvement over inflation targeting.

  5. Sumner is far better qualified at explaining this, but the NGDP futures market would be complemented by the Fed's commitment to buy or sell bonds if the futures market is above or below target. Professor Farmer would go one step further and commit the Fed to buy or sell equities. Beckworth knows about Farmer's idea because I listened to Beckworth's podcast of Farmer (which, like all of Beckworth's interviews, was terrific). And to finish this comment, Sumner can explain why the Fed would never have to buy or sell bonds even with the commitment (essentially it's the nature of a futures market).