Before I write anything else, I want to say thanks to Scott Sumner -- his praise is honestly far more than I think I deserve, but I hope I can earn it retroactively over the next few months of blogging.
If NGDP targeting is going to continue gaining intellectual ground in economics -- and I hope it does, see here for a detailed explanation of why -- I think there are a handful of questions about it for which I don't have answers I regard as satisfactory. These are not hostile questions, and many of them relate to the particulars of implementing an NGDP target, but as does Arnold Kling of "EconLog," I think there's something worthwhile about critiquing one's own argument sincerely: it makes you think.
The first two concerns I had about NGDP targeting related to the stability of the two variables which an NGDP target would no longer target -- inflation and interest rates--and the significance of costs to any undesirable changes in those variables.
I've dealt with both of these concerns in previous posts, finding in March that an NGDP target does not risk destabilizing inflation expectations, contrary to the Fed's minutes in November 2011, in which during a discussion of NGDP targeting "a number of participants expressed concern that switching to a new policy framework could heighten uncertainty about future monetary policy [and] risk unmooring longer-term inflation expectations." Then, looking at how interest rates behaved in the United Kingdom, which had a de facto NGDP target until 2008 -- also Australia -- I found that because an NGDP target would make monetary policy could operate more through expectational channels and less through raising and lowering interest rates.
So here are my new questions.
Real output was revised downward three times during the 2008 recession, per the Bureau of Economic Analysis' statistical review process. As you can see in the graph below, the statistical revisions were no small potatoes: a total difference of $400 billion of real GDP in the first quarter of 2009 -- a 3.1 percent gap -- between the estimate announced in June 2009 and the final number, which came out in 2011. (By the way, I got this data from ALFRED, which is a sister project to the FRED economic database. If you haven't heard of ALFRED, check it out -- it keeps the inaccurate legacy data so economists can do this sort of hindsight analysis.)My concern is that an NGDP target ends up relying on rather inaccurate measurement techniques -- such that we may think we are stabilizing aggregate demand, but what we're really stabilizing is just the BEA's proxy measurement at the time. Now, I recognize that economic data will always be imperfect. But the revisions in real GDP, and by extension nominal GDP, were far, far larger in magnitude than the average revisions in price indices such as core and headline CPI and the PCE price index. Is that not a cause for some practical concern?
Also, what this makes me worry about is how an NGDP target would respond to data revisions -- do we just seek to immediately push NGDP back up to, or down to, its path? Or in such a scenario do we practice some sort of hybrid targeting (this is an amazing 1994 paper by Hall and Mankiw on this topic)?
Here's another line of questioning. National accounts data, which includes NGDP, comes out once a quarter. The Fed is used to working with data which comes out with significantly more frequency -- months or weeks, in the cases of data on prices and labor market conditions. I understand that an NGDP target could look at these measures to gauge the path of NGDP and perhaps use a futures market as well. But it's worth remembering that a lot can happen in terms of falling of the NGDP growth rate path in one quarter.How does the Fed deal with any issues coming from the frequency of data, which may make maintaining nominal stability a slower process in practice than in theory? Do we need to get monthly NGDP estimates? Or will a futures market, plus looking at other measures (such as PCE) in between NGDP readings sufficient?
And now, one more totally distinct area. My previous examination of how an NGDP target would respond to real shocks assumes that aggregate demand is price isoelastic/"unit elastic" -- in other words, a real shock has no effect on NGDP, only the components therein. But if AD is not isoelastic, and instead it has an inelastic and elastic section (like microeconomics thinks of most demand curves in markets), then an NGDP target implies considerably more active management of demand then I had first thought, which means that investment and interest rates, given their role in the monetary policy transmission mechanism, could get pushed around a bit more than we might have thought. Moreover, if the demand curve is not entirely isoelastic, that could make maintaining the target level significantly more challenging in the context of real shocks, given any uncertainty as to the price elasticity of aggregate demand.We can see in this graph that the rate of NGDP growth was very erratic in the 1970s, which suggests that dealing with real shocks and the resultant cost-push inflation may be not as easy as isoelastic AD curve might suggest in theory.