NGDP targeting is not a "fragile" policy
By Scott Sumner
This post tries to tie together some diverse observations about the sticky wage/NGDP shock model. The motivation for this (unfortunately long) post was the observation that many non-economists seem to find the sticky wage/NGDP shock model to be appealing. I once met Britmouse, and he told me his background was a technical field, not economics. And yet he’s been able to produce some of the best macro analysis of the British economy in the entire blogosphere. How does that happen? More recently a new blog popped up:
I am not an economist by trade. Instead, I have approached economics in the wake of the 2008 recession with the question of “what in the world explains this?” Paul Samuelson took a strong liberal bent to economics despite having traditional, classical economics teachers because he learned economics during the Great Depression. The equations and graphs on the chalkboard bore little relationship to the suffering outside the classroom. Similarly, at the start of 2008 I had strong libertarian ideals, but my ideals were confronted with many perfectly qualified people I knew out of work due to no fault of their own.
There is a fair question of what do I add when many bloggers hold similar views. Scott Sumner is likely the closest, . . .
His (or her) second post in on wage targeting, so I’m going to assume he/she has at least some similarity to Britmouse.
I think there is a reason for this, but I’m going to have to address it in a roundabout fashion. In the end I’ll argue it has implications for modern DSGE models, as well as Nick Rowe’s conversion to NGDPLT.
A while back I showed the correlation between (NGDP/hourly wages), and the unemployment rate. Arnold Kling dismissed the correlation as a sort of tautology. Let’s think about what would have made Arnold’s claim true. Suppose that instead of NGDP I had used total wages and salaries. And suppose that instead of the unemployment rate I had used total hours worked as my cyclical indicator. Then the graph would have shown (total wages/hourly wages) correlated with hours worked. But those are the exact same thing! That’s why Kling viewed it as a sort of tautology; my model seemed very close to another “model” that was in fact a tautology. Was Kling correct in assuming that my model and this other model were roughly the same thing? I’d say he was. However I don’t view that as a negative, but rather a point in favor of the sticky wage/NGDP shock model.
So let’s say we target total wages and salaries instead of NGDP, and we use a low volatility in hours worked as our policy goal. (We will still allow secular changes in hours worked due to factors such as boomers retiring, just not the sort of sudden plunge in hours worked that we saw in 2008-09.) And let’s say we are successful in stabilizing the path of total wages. Does the “tautological” nature of my correlation mean the policy will work? Unfortunately there is one other potential problem—monetary policy might affect the volatility of wages. The Lucas Critique. If we make total wage income (or NGDP) much more stable, then nominal hourly wages might suddenly become much less stable, making the business cycle just as bad, if not worse.
So you can think of the case for NGDP targeting being dependent on these three factors:
1. Hours worked and the unemployment rate being strongly (and negatively) correlated at high frequencies. That’s pretty much a slam dunk; I doubt any business cycle experts would worry about that issue.
2. NGDP shocks being associated with similar total wage income shocks. That’s also overwhelming likely. Think about what it would take for a stable path of NGDP to produce a recession, solely because total wage income fell as NGDP was stable. For that to happen you’d have to see profits rise sharply during the recession. But in the US profits tend to fall sharply during recessions. So that’s pretty unlikely. And if it is a problem you can target total nominal wages and salaries, instead of NGDP.
3. Another possibility is that if you stabilized the path of NGDP or total wage income, it would suddenly make hourly wage growth much more unstable. Thus in 2008-09, if you kept NGDP growing at 5%, then hourly wage growth might have soared from 3.5% per year to 12%, driving unemployment sharply higher. There are actually new classical types who would claim that’s exactly what would have happened. But I think it’s extremely unlikely, and I believe most other sensible macroeconomists do as well.
What does all this mean? Well the effectiveness of NGDP targeting is not a tautology, but it can be turned into a tautology by making three pretty uncontroversial assumptions about hours/unemployment, about the wage share of national income over the cycle, and about the “inertial” nature of hourly wages. I’d like to emphasize “inertial” wages rather than “sticky” wages, because we don’t really need to make any highly restrictive assumptions about why aggregate wages are so inertial.
Here’s what we do know. Wage growth from year to year is pretty stable, unless inflation soars much higher. If inflation soars much higher (as in the 1970s), then nominal hourly wages will follow, with a lag. But the news is even better; it appears that nominal wages track NGDP at least as well as inflation, often better. When NGDP growth slowed during the Great Moderation, so did hourly wages. When NGDP growth became more stable during the great Moderation, so did hourly wages. When prices shot up in 2007-08, and NGDP growth remained slow, wage growth did not shoot upward. And most importantly, when NGDP growth plunged in 2008-09, wage growth did slow, but very slowly. From about 3.5%/year before the recession to 2%/year after the recession. That inertia is a fact, not a theory. Even those who reject sticky wages cannot reject inertial wages. They are a fact about the world, at least when inflation/NGDP growth rates are fairly low.
And since nominal wages tend to track NGDP with a lag, it’s pretty hard to argue that stabilizing NGDP growth would destabilize wages. This makes the logical appeal of NGDP targeting almost overwhelming, and explains its popularity with newbies.
Here’s another way of making the same point, which I hope will cause doubters to reconsider. I’m saying that a serious recession under stable NGDP growth is very unlikely, because at least one stylized fact (of the three above) would have to act very, very strangely, indeed in an almost completely implausible fashion.
In contrast, it’s easy to imagine a serious recession under inflation targeting. The original idea of inflation or price level targeting goes back many decades, to a period where recessions were associated with deflation. That stylized fact was so important that progressive economists like Fisher and Keynes (and many others) were drawn to the idea of using monetary policy to stabilize prices. But that was when commodity prices were:
1. A much bigger share of the economy
2. Highly procyclical
So that made existing price indices in the 1920s and 1930s (like the WPI) very procyclical. More recently inflation got much more inertial, partly due to more comprehensive indices, partly due to us moving from a commodity economy to a service economy. Nonetheless you did still get mild disinflation during most recessions, which the new Keynesians put into their models. But it turns out that inflation targeting was a highly fragile policy. Here’s why:
1. Modern SRAS curves are fairly flat as AD falls sharply. This means a fall of NGDP growth to 9% below trend (as we saw in America during mid 2008 to mid 2009) might be mostly a fall in RGDP, and not much disinflation.
2. The negative demand shock might occur in close proximity to an adverse supply shocks. Whereas AD shocks reduce inflation very gradually (by slowing nominal wage growth gradually), the supply shock causes the CPI to spike suddenly. One can imagine NGDP and RGDP both falling rapidly, even as inflation seems close to target. That’s roughly 2008.
Something like that also happened in Canada during the recent recession, and helped convert Nick Rowe from inflation targeting to NGDPLT. More importantly, whereas a catastrophic failure of NGDP targeting is almost unthinkable (I know, famous last words . . . ) a catastrophic failure of inflation targeting is quite thinkable, and indeed just happened.
I believe this is what makes NGDP targeting so attractive to newbies. They are easily persuaded that nominal wages are sticky because it conforms to both common sense observation about the world, and also a simple glance at the time series data for wages. They also know that mass unemployment is the key business cycle problem. And they know that deep recessions tend to be associated with sharp declines in NGDP growth, but not necessarily sharp falls in inflation. Looking at all those stylized facts, how could someone new not be attracted to NGDP stability?
Modern macro looked for “micro foundations” that could explain business cycles. These models include “P” because prices are a micro variable. NGDP is not, so it doesn’t really appear in the models. Their models are created by people who think in terms of what sort of shocks would give firms an incentive to produce less and lay off workers.
In contrast, the NGDP approach is a very naive model that treats NGDP sort of like a big pot of money, which is shared out among workers with sticky wages. If some day the pot is smaller, then there’s less money to share, and some workers end up disappointed (unemployed.) It’s completely agnostic about the micro foundations, not even yielding predictions about something like real wages (W/P). No behavioral assumptions beyond inertial wages.
But despite all these weaknesses, it’s a very plausible model, because it’s not fragile. It’s very, very hard to imagine plausible scenarios where NGDP growth is stable and we have a lot of employment volatility. (You can imagine RGDP volatility (from say declining North Sea oil output), another mistake by the experts was to focus on RGDP rather than hours worked as the key measure of the business cycle.)
And this means that once the profession starts thinking about NGDP (and it has started) it can never stop thinking about it. NGDP growth will plunge again in the next recession, probably much more dramatically than inflation. This will again show inflation to be (as Nick Rowe puts) it “the dog that didn’t bark.” NGDP always barks. And when it doesn’t bark, you almost never have employment instability. NGDP and sticky wages, they can’t be put back in the bottle. The model is too simple and attractive. It’s up to the math geniuses to catch up with their DSGE models.