The labor market is not a reliable guide to policy
By Scott Sumner
A week ago I criticized an article in the Economist, which saw a “puzzle” in the co-existence of low unemployment and sluggish wage gains. I said there was no puzzle, as wage growth was slow because nominal GDP growth was slow. The labor market can eventually adjust to almost any (reasonable) trend rate of growth in NGDP.
Now the Economist has responded, and in doing so has somewhat misinterpreted my position. That’s actually not their fault–my post was very terse and the nuances of this issue are quite subtle, with many different interpretations. For instance, take their concluding paragraph:
Janet Yellen’s critics tend to say she is excessively devoted to the idea that low unemployment portends higher inflation. They characterize this as adherence to the Phillips curve, which is associated in many peoples’ minds with Keynesian thinking. But it was monetarists who first argued that policymakers ultimately cannot control unemployment, because if loose money drives unemployment too low, inflation will accelerate. It is right for the Fed to look for the labour market for signals as to whether monetary policy is tighter or looser than the economy can sustain. Wage growth is the clearest of those signals. And it suggests that we are not in the long run yet.
Many people will read this and infer that it was the monetarists who came up with the idea that the Fed should look to the labor market for signals about the appropriate stance of policy. That’s not the case. When Milton Friedman developed his Natural Rate Hypothesis, he argued that unexpectedly high inflation caused low unemployment, and unexpectedly low inflation caused high unemployment. Many Keynesians took that idea and reversed it—that low unemployment causes high inflation. That was not Friedman’s view (nor mine). In addition, many Keynesians came to believe that the natural rate of unemployment could be estimated, and used as a guide to policy. That was not Friedman’s view. Friedman believed (as do I) that the natural rate of unemployment changed over time, and was difficult to estimate. He favored steady growth in M, whereas I favor steady growth in MV. But neither of us have been supportive of using the natural rate as a guide to policy.
Now I don’t want to take this too far. I concede that if unemployment suddenly soars from 4.5% to 10%, as it did in 2007-09, and NGDP falls, that’s prima facie evidence that unemployment is above the natural rate. What I deny is that we can know whether unemployment is above or below the natural rate today, when unemployment is 4.4%. In between those two extremes we might have varying levels of confidence in estimating the position of the labor market. Both Friedman and I believe that the Great Inflation of 1966-81 was partly caused by people underestimating the natural rate of unemployment, and the central bank responding with an excessively expansionary monetary policy. I don’t want that to happen again.
That’s not to say I’m necessarily opposed to easier money. The Fed is falling short of its 2% inflation target, and you can make a good argument that easier money is needed to hit the target. What I deny is the claim that the labor market provides justification for an easier money policy today.
Mr Sumner’s view implies there is no output gap, the labour market is in equilibrium, and wage growth is low because inflation expectations have been low. I find this unconvincing, for two reasons. First, it means that unless the Federal Reserve brings job growth back in line with working-age population growth, inflation will soon rise. (Central banks cannot for long keep unemployment below its natural rate.) Yet inflation is in fact falling. Second, as Adam Ozimek at Moody’s Analytics has shown, wage growth as measured by the employment cost index is almost exactly where you would expect it to be given the employment-to-population ratio for 25- to 54-year-olds (see chart). If the rate of wage growth consistent with labour market equilibrium had changed to account for a shift in trend nominal GDP growth, you would expect this relationship to have broken down. It has not.
Actually, I would argue that we don’t know whether there is an output gap or not. Economics it too imprecise a science to make these sorts of fine distinctions. We underestimated the natural rate of unemployment in the 1970s and overestimated the natural rate of unemployment in the late 1990s. But I’m willing to play the devil’s advocate, and defend the hypothesis that we are at the natural rate of unemployment (and employment).
What is the Economist’s evidence against this claim? They point to the fact that employment is rising faster than expected, based on growth in the prime age labor force. And I plead guilty to not expecting this to occur this year. (Japan has also had more growth in employment than I expected.) Like many economists, I did not expect unemployment to fall to 4.4%. But contrary to the Economist, this does not show that we still have economic slack in the economy. The strongest argument for economic slack is a falling unemployment rate coinciding with a stable rate of wage inflation. We have that, so I’m happy to admit that it now looks like there was some slack a year ago. But will unemployment keep falling? Maybe, but I’m not convinced. If it’s about 3.5% next year then I was wrong. If it’s 4.2% to 4.5% next year then that would count in my favor.
A second (and more recent) argument for economic slack is that even if unemployment bottoms out, there are people on the sidelines just waiting to enter the labor force. According to this view, an easier money policy may not reduce unemployment much further, but will lead to more employment by expanding the labor force. This is basically a brand new theory, and I just don’t see the data to support this claim. I also find it hard to see the theoretical justification.
I find it much more plausible to assume that the rapid growth in employment reflects secular trends in labor force participation, especially among the elderly. Perhaps various social trends are leading to an extra 50,000 non-prime age workers each month, and these workers are entering the labor force for reasons unrelated to diminishing economic slack. If this seems implausible, let me frame it differently, and tell me which is more plausible:
1. Over a 12-month period, an extra 600,000 older people start working because boomers are retiring, they are still healthy, service jobs don’t require brute strength, and American pensions are often inadequate.
2. Over a 12-month period an extra 600,000 people enter the labor force because they think jobs are easier to find when the unemployment rate is 4.4% as compared to when it is 5.0% (a year earlier)
It’s quite possible that both of these are true to some extent. But I find the second one to be less plausible. I can understand “discouraged workers” when there is 8% or 10% unemployment—but 5%? Would a bit easier money by the Fed actually draw new workers into the labor market? Yes, a few, but how many? My hunch is that employment has been growing faster than working age population because unemployment is falling and because more older people are entering the labor force. I expect unemployment to stop falling soon, and I don’t think the recent trend of the elderly entering the labor force is any sort of “economic slack” phenomenon that can be sped up with easier money. Trying to do so risks destabilizing the economy.
As far as the wage/employment graph, I believe Phillips curves are roughly vertical in the long run. So I don’t see how that evidence refutes any of my arguments.
Again, a bit easier money might be appropriate—to hit the Fed’s 2% inflation target. But don’t try to target the labor market. That didn’t work out well in the 1970s, and there’s no reason to expect it would work any better today.
PS. I picked the assumption of more elderly people working as just one example of why the labor force might be growing faster than the working age population. There are many other possible assumptions. I picked the elderly because the data seems to point in that direction: