This is from a recent paper on NGDP targeting by Bennett McCallum:
I find it plausible that movements in nominal spending growth would be more closely and reliably related to central bank policy actions—primarily open market sales and purchases—than would movements in inflation and output separately. If so, then the central bank that targets nominal GDP would not have to rely upon its models of the way in which nominal and real variables are related, that is, its model of the “Phillips curve” relationship. That is a significant advantage, because the Phillips curve relationship is the component of quantitative (econometric) macroeconomic models for which professional understanding and agreement is, by far, the weakest. Thus, if the central bank can manage nominal spending growth in a manner that does not involve conceptually the Phillips curve, it can conduct policy without use of that elusive relationship. By contrast, if it focuses on inflation and real GDP separately, or on inflation alone, it cannot possibly avoid its use.
The point of view expressed in the preceding discussion is somewhat reminiscent of Milton Friedman’s approach to price-level determination, in which he famously depicts the central bank as choosing the supply of money in nominal terms while the private sector is choosing the quantity of money demanded in real terms. The interaction of these choices then determines the price level—see Friedman (1987, pp. 3-4). In the present application, the central bank determines the amount of spending in nominal terms, with the private sector’s behavior determining how much of any change in spending will be in terms of (real) output changes and how much will be in (nominal) price level changes.
I like the analogy with nominal money supply and real money demand, which I’ve always seen as the core of monetary economics. Indeed my biggest beef with Keynesians is that they don’t see Friedman’s example as the core of monetary economics. I also like the pragmatism in McCallum’s approach. We really don’t have good Phillips Curve models. I think this is partly because our data is worse than many assume, and is less closely related to the theoretically appropriate concepts than many macroeconomists assume. For instance, one reason why deflation is bad is because it’s less profitable to produce output when prices fall (if wages are sticky.) In that context it’s interesting that official CPI figures show housing prices up 7.5% over the past 5 years, and the Case-Shiller index shows housing prices down 32% over the past 5 years. Which provides a better estimate of the incentive to construct new houses? Which is the inflation number used in actual economic studies by real world macroeconomists? It’s also much easier to measure the nominal output of the health care industry, the consulting industry, or the PC industry, than to separately measure the prices and outputs of those two industries. And even if we had accurate data, the Phillips Curve would be highly unstable due to things like the recent extension of unemployment insurance from 26 weeks to 99 weeks. But the data problems make it even worse.
Here McCallum expresses skepticism about level targeting:
From the foregoing it can be seen that one issue that arises in discussions of nominal GDP targeting is whether the targets should be expressed in terms of “level” or “growth-rate” measures. For an example of the distinction, suppose that the chosen rate of growth of nominal GDP is 4.5% per year. Suppose that in some year, however, the central bank misses that target by a full percentage point on the high side, yielding 5.5% growth consisting of (for example) 3.0 percent inflation and 2.5% real growth. Should the central bank strive for the usual 4.5% growth in nominal GDP again in the following year? Or should it decrease its growth target to 4.0%, aiming thereby to be back at the original path for the nominal GDP level at the end of the next year? In other words, should the nominal GDP targets be set in terms of growth rates or growing levels? In the latter case, the disadvantage will be that policy that decreases nominal growth below its usual target value may be excessively restrictive, whereas the former case leaves open the possibility of cumulative misses in the same direction for a number of periods, i.e., it permits “base drift” away from the intended path. My position on this issue has been that keeping with the target growth rates will, if they are on average equal to the correct value over time, be unlikely to permit much departure from the planned path and so should probably be preferred. This is not at all a universal point of view, however, among nominal GDP supporters.
In this recent post I explained one advantage of level targeting; the fact that it leads market participants to assist monetary policymakers. Perhaps I’ve been overly influenced by the 2008 period, when the advantages of level targeting seem relatively large. I would also point to Michael Woodford’s work on liquidity traps. Woodford argues that level targeting is especially important when a central bank hits the zero bound (as he is even more skeptical about QE than I am.) McCallum may be right that when the central bank is doing its job, growth rate targeting is as good as or even better than level targeting. By “doing its job,” I mean targeting the forecast. But given the spotty track record of real world central banks, it seems to me that level targeting has a great advantage over growth rate targeting, the ability to prevent very large and costly errors. It seems inconceivable to me that NGDP would have fallen 9% below trend between mid-2008 and mid-2009, if markets had known that the Fed was engaging in level targeting of NGDP and would soon return the economy to the trend line.
PS. In my previous post I probably created the impression that I entirely agree with Romer’s post. Those who view me as an inflationist may be surprised to learn that I actually think her recommendation is a bit too expansionary. In 2009 I favored going back to the old (pre-2008) trend line. I still think that policy would be better than the status quo. But any dating of a trend line is arbitrary unless the Fed has set an explicit target. As time goes by more and more debt and wage contracts have been set based on post-2008 expectations. So at this late date it might be more prudent to only go part way back.
And in a sense this discussion is purely academic. My fear is that the Fed will do too little, not too much. My recent discussion of aiming for 6% or 7% growth for a couple years, and 5% thereafter, is actually far more conservative than the Romer proposal, but also represents the outer limits of what the Fed would be willing to do. More likely they’d just shoot for 5%, with no catchup at all. The same discussion occurred during the Great Depression, when some advocated going half way back to pre-Depression prices, and others advocated going all the way back. They only went half way back, but even that would have been enough for a quick recovery if the NIRA hadn’t raised nominal wages by over 20% in the late summer of 1933.
PPS. David Beckworth also comments on McCallum’s paper.
PPPS. Lars Christensen also comments on McCallum.