[See my previous post for context.]
I started out to research and write a post on why I disagreed with Scott Sumner about NGDP targeting, and discovered an important point of agreement: targeting nominal wages forecasts would probably be better than targeting either NGDP or CPI forecasts.
One drawback to targeting something other than CPI forecasts is that we’ve got good market forecasts of the CPI. It’s certainly possible to create markets to forecast other quantities that the Fed might target, but we don’t have a good way of predicting how much time and money those will require.
Problems with NGDP targets
The main long-term drawback to targeting NGDP (or other measures that incorporate the quantity of economic activity) rather than an inflation-like measure is that it’s quite plausible to have large changes in the trend of increasing economic activity.
We could have a large increase in our growth rate due to a technology change such as uploaded minds (ems). NGDP targeting would create unpleasant deflation in that scenario until the Fed figured out how to adjust to new NGDP targets.
I can also imagine a technology-induced slowdown in economic growth, for example: a switch to open-source hardware for things like food and clothing (3-d printers using open-source designs) could replace lots of transactions with free equivalents. That would mean a decline in NGDP without a decline in living standards. NGDP targeting would respond by creating high inflation. (This scenario seems less likely and less dangerous than the prior scenario).
Basil Halperin has some historical examples where NGDP targeting would have produced similar problems.
Problems with inflation forecasts?
Critics of inflation targeting point to problems associated with oil shocks or with strange ways of calculating housing costs. Those cause many inflation measures to temporarily diverge from what I want the Fed to focus on, which is the problem of sticky wages interacting with weak nominal wages to create unnecessary unemployment.
Those problems with measuring inflation are serious if the Fed uses inflation that has already happened or uses forecasts of inflation that extend only a few months into the future.
Instead, I recommend using multi-year CPI forecasts based on several different time periods (e.g. in the 2 to 10 year range), and possibly forecasts for time periods that start a year or so in the future (this series shows how to infer such forecasts from existing markets). In the rare case where forecasts for different time periods say conflicting things about whether the Fed is too tight or loose, I’d encourage the Fed to use its judgment about which to follow.
The multi-year forecasts have historically shown only small reactions to phenomena such as the large spike in oil prices in mid 2008. I expect that pattern to continue: commodity price spikes happen when markets get evidence of their causes/symptoms (due to market efficiency), not at predictable future times. The multi-year forecasts typically tell us mainly whether the Fed will persistently miss its target.
Won’t using those long-term forecasts enable the Fed to make mistakes that it corrects (or over-corrects) for shorter time periods? Technically yes, but that doesn’t mean the Fed has a practical way to do that. It’s much easier for the Fed to hit its target if demand for money is predictable. Demand for money is more predictable if the value of money is more predictable. That’s one reason why long-term stability of inflation (or of wages or NGDP) implies short-term stability.
It would be a bit safer to target nominal wage rate forecasts rather than CPI forecasts if we had equally good markets forecasting both. But I expect it to be easier to convince the public to trust markets that are heavily traded for other reasons, than it is to get them to trust a brand new market of uncertain liquidity.
The argument that we have good market forecasts of CPI and don’t have good market forecasts of NGDP so its going to be much easier to target CPI forecasts is a good one, though it is also an argument for trying to develop market NGDP forecasts.
The argument that NGDP will lead to deflation when there’s high productivity growth isn’t good. The basic idea is that deflation due to productivity growth is especially easy to respond to because the producers experiencing productivity growth have all the information they need to lower their prices (in contrast to general deflation, where everyone needs to reduce their prices a little bit). Indeed, they’re already lowering their prices (that’s how we’re getting deflation). Actually, calling this deflation, is misleading. Another way of describing this argument is that reductions in the price level due to productivity changes doesn’t lead to monetary disequilibrium.
This paper (http://www.cato.org/pubs/journal/cj10n1/cj10n1-14.pdf) on the topic by George Selgin is pretty good.
John, I agree that productivity growth is not sufficient to generate monetary disequilibrium.
The scenario that worries me involves rapid growth in the amount of economic activity. Let’s say that activity starts growing at a rate that would be 100% per year, using a hypothetical measure of NGDP that counts all goods and services as if the were priced the same as they were last year. That means the Fed would need something like 47.5% per year deflation in order to get actual NGDP growing at 5% per year.
I’m assuming that productivity growth would not be dramatic enough to produce that kind of deflation while also keeping nominal wages stable. So I expect an NGDP-targeting Fed to initially tighten enough to cause declining nominal wages, then make crude guesses about how to raise NGDP targets.
Pingback: Equationizing Market Monetarism | Bayesian Investor Blog