Noah Smith has a good post on the costs of inflation and recession, but I don't quite agree with this:
Many people seem to think that inflation and recession are equal, symmetric dangers. This is implicit in the idea of nominal GDP (NGDP) targeting, which is promoted by economists like Scott Sumner at George Mason University's Mercatus Institute. Since NGDP growth is just the sum of real GDP growth and inflation, Sumner's policy implies that one percentage point of higher inflation is in some sense just as bad as a one-point reduction in growth. But in reality, a loss of one percentage point of GDP probably is many times worse than a 1 percent rise in inflation.
This claim confuses several unrelated issues:
1. Most of Smith's post focuses on the costs of higher or lower trend inflation. Once people get used to a 4% trend inflation rate, the cost is probably not much higher than an expected 2% inflation rate. But in the long run there is no trade-off between trend inflation and unemployment, with the possible exception of really low inflation rates, where the twin zero bounds on wage increases and nominal interest rates may create problems. Assuming you pick a high enough inflation/NGDP target to avoid these zero bound issues, there should be no further gains to higher inflation rates.
2. Smith's comments make more sense for inflation volatility. But even there I'd suggest a different interpretation. I do not favor NGDP targeting because I view inflation and recession as equally severe problems, rather I believe (rightly or wrongly) that a stable path for NGDP would be most likely to minimize the harmful impact of recessions (unstable employment and unstable financial markets) while continuing to allow real GDP fluctuations that are not harmful, say due to productivity fluctuations with non-monetary causes.
Now in fairness, NGDP is probably not precisely the optimal target for minimizing the welfare costs of employment instability and financial market instability in the US, but I think it's pretty close. In my previous post I suggested that in some cases, such as Ireland, it's not even very close to being optimal, at least for measured NGDP.
PS. Those familiar with New Keynesian models may recognize that I'm making a claim similar to the "divine coincidence" argument for inflation targeting, but applying the concept to NGDP targeting. The world's too messy for divine coincidences to work perfectly, but I believe that NGDP targeting comes pretty close, at least in the US and other major economies.