Last week, David Andolfatto challenged proponents of NGDP targeting to provide the reasons for their belief that targeting NGDP, or to be more precise the time path of NGDP, as opposed to just a particular rate of growth of NGDP, is superior to any alternative nominal target. I am probably the wrong person to offer an explanation (and anyway Scott Sumner and Nick Rowe have already responded, probably more ably than I can), because I am on record (here and here) advocating targeting the average wage level. Moreover, at this stage of my life, I am skeptical that we know enough about the consequence of any particular rule to commit ourselves irrevocably to it come what may. Following rules is a good thing; we all know that. Ask any five-year old. But no rule is perfect, and even though one of the purposes of a rule is to make life more predictable, sometimes following a rule designed for, or relevant to, very different circumstances from those in which we may eventually find ourselves can produce really bad results, making our lives and our interactions with others less, not more, predictable.
So with that disclaimer, here is my response to Andolfatto’s challenge by way of comparing NGDP level targeting with inflation targeting. My point is that if we want the monetary authority to be committed to a specific nominal target, the level of NGDP seems to be a much better choice than the inflation rate.
As I mentioned, Scott Sumner and Nick Rowe have already provided a bunch of good reasons for preferring targeting the time path of NGDP to targeting either the level (or time path) of the price level or the inflation rate. The point that I want to discuss may have been touched on in their discussions, but I don’t think its implications were fully worked out.
Let me start by noting that there is a curious gap in contemporary discussions of inflation targeting; which is that despite the apparent rigor of contemporary macro models of the RBC or DSGE variety, supposedly derived from deep microfoundations, the models don’t seem to have much to say about what the optimal inflation target ought to be. The inflation target, so far as I can tell – and I admit that I am not really up to date on these models – is generally left up to the free choice of the monetary authority. That strikes me as curious, because there is a literature dating back to the late 1960s on the optimal rate of inflation. That literature, whose most notable contribution was Friedman’s 1969 essay “The Optimal Quantity of Money,” came to the conclusion that the optimal quantity of money corresponded to a rate of inflation equal to the negative of the equilibrium (or natural) real rate of interest in an economy operating at full employment.
So, Friedman’s result implies that the optimal rate of inflation ought to fluctuate as the real equilibrium (natural) rate of interest fluctuates, fluctuations to which Friedman devoted little, if any, attention in his essay. But from our perspective there is an even more serious shortcoming with Friedman’s discussion, namely, his assumption of perpetual full employment, so that the real interest rate could be identified with the equilibrium (or natural) rate of interest. Nevertheless, although Friedman seemed content with a steady-state analysis in which a unique equilibrium (natural) real interest rate defined a unique optimal rate of deflation (given a positive equilibrium real interest rate) over time, thereby allowing Friedman to achieve a partial reconciliation between the optimal-inflation analysis and his x-percent rule for steady growth in the money supply (despite the mismatch between his theoretical analysis of the rate of inflation in terms of the monetary base and his x-percent rule in terms of M1 or M2), Friedman’s analysis provided only a starting point for a discussion of optimal inflation targeting over time. But the discussion, to my knowledge, has never taken place. A Taylor rule takes into account some of these considerations, but only in an ad hoc fashion, certainly not in the spirit of the deep microfoundations on which modern macrotheory is supposedly based.
In my paper “The Fisher Effect under Deflationary Expectations,” I tried to explain and illustrate why the optimal rate of inflation is very sensitive to the real rate of interest, providing empirical evidence that the financial crisis of 2008 was a manifestation of a pathological situation in which the expected rate of deflation was greater than the real rate of interest, a disequilibrium phenomenon triggering a collapse of asset prices. I showed that, even before asset prices collapsed in the last quarter of 2008, there was an unusual positive correlation between changes in expected inflation and changes in the S&P 500, a correlation that has continued ever since as a result of the persistently negative real interest rates very close to, if not exceeding, expected inflation. In such circumstances, expected rates of inflation (consistently less than 2% even since the start of the “recovery”) have clearly been too low.
Targeting nominal GDP, at least in qualitative terms, would adjust the rate of inflation and expected inflation in a manner consistent with the implications of Friedman’s analysis and with my discussion of the Fisher effect. If the monetary authority kept nominal GDP increasing at a 5% annual rate, the rate of inflation would automatically rise in recessions, just when the real interest rate would be falling and the optimal inflation rate rising. And in a recovery, with nominal GDP increasing at a 5% annual rate, the rate of inflation would automatically fall, just when the real rate of interest would be rising and the optimal inflation rate falling. Viewed from this perspective, the presumption now governing contemporary central banking that the rate of inflation should be held forever constant, regardless of underlying economic conditions, seems, well, almost absurd.