Evan Soltas
Mar 21, 2012

Shoulders of Giants

Echoes of Robert Hall's mid-80s work in mine

The latest idea I've been exploring, as readers know, has been using different weights on inflation and real output to create "variant" stabilization targets (see here and here for more). This post presumes a basic familiarity with those prior writings.

I had been very excited, because I couldn't find anyone who'd done anything like this before, and I thought it a really cool idea. And then Scott Sumner, by email, directed me to two wonderful papers by Robert Hall which have considerable overlap with mine. They are "Monetary Strategy with an Elastic Price Standard" (1984) and "Optimal Monetary Institutions and Policy" (1986).

So I'm not the first one to break up an NGDP target into price and real output components, based on the same logic that the costs of deviations of the same magnitude in prices and real output are not likely the same. Hall also creates a variable A, similar to my variable H, which describes the central bank's relative tolerance of deviations from trend in either variable.

Hall also does something very interesting which I hadn't thought of. So far in my research, I had only concerned myself with the relative tolerance of deviations -- the problem with this is that if a central bank was equally permissive of deviations (i.e. they let inflation and unemployment highly volatile, but equally so) the bank would get an H-value of 0, which implies the existence of an NGDP target. But Hall also looks at the absolute level of deviations, which leads him to posit the existence of a efficient monetary "policy frontier."(See illustration to right, which depicts how my and Hall's understandings interact.)

Hall also does a few things differently: he chooses to target a constant price level, as opposed to a constant rate of inflation. I think the latter is preferable, given that my model tries to fit the flexible inflation targeting policies of central banks -- and I see targeting a low inflation rate (or I suppose price level targeting, but with a rising level) as in itself better than shooting for complete price stability, given the nominal zero lower bound, downward nominal wage rigidity, etc. He also uses a simulation to project the price level and unemployment rate given various A-values.

I definitely went a number of places where Hall didn't. I had fit my H-values onto a broad array of historical data, including the major flexible inflation targeting central banks and a larger dataset for the Federal Reserve. This tells a lot about the history of monetary policy, and allows us to see more clearly how such a policy could be effected in a practical and realistic fashion: the Fed would keep its "longer-run" inflation target of an annual change of 2 percent in PCE price index, and all it would need to do is explicitly declare an H-value. That way the target is simple but clearly specifies the Fed's reaction function in the event of a real or nominal shock.

In the future, I want to investigate a number of complementary items -- the effect of H-value uncertainty on real variables in the short and long runs, the possibility that my historically fit H-values may be distorted by nonlinear cost functions for deviations in inflation and real output from trends, the relationship of H-values to long-run real variables, and perhaps even an empirical test of Hall's "policy frontier."

There's something exciting in dusting off these studies (they were published almost a decade before I was born) and continuing the search for optimal monetary policy, combining the explicit character of NGDP growth rate targeting and the real-world existence of discretionary flexible inflation targets.