Editor's note: This is part of a three-day series on the changing economy. To see more, click here.
NAIRU and You
JACKSON HOLE, Wyo. -- Maybe some things that seem broken really aren't.
Plot the inflation rate against the unemployment rate from 1951 to present and you will produce a completely patternless scattergram. Why? Because the relationship between inflation and unemployment is not stable over time. Nor would we expect it to be, since money growth, not the unemployment rate, is the principal determinant of inflation in the long run. As
Fed Governor Meyer
so aptly says:
We define full employment as the maximum rate of employment that can be sustained without rising inflation. At full employment -- also defined as the rate of unemployment equal to the NAIRU (nonaccelerating inflation rate of unemployment) -- inflation is constant, but any constant level of inflation is possible at full employment. The rate of inflation in the long run is therefore not determined by the unemployment rate at all. It's determined by the rate of growth of the money supply. This of course gives monetary policy unique responsibility for inflation in the long run.
The relationship between a nominal variable (like the price level or the money stock or an exchange rate) and a real variable (like the unemployment rate) is known as a Phillips curve. It takes its name from British economist
A. W. Phillips
, who published a seminal paper on the subject in 1958. An unemployment-inflation Phillips curve for the period between 1951 and 1998 appears below.
This plot underscores the fact that the relationship between inflation and unemployment is not at all stable over time. Indeed, in the long run, no existing evidence suggests that faster inflation leads to a lower unemployment rate or that policymakers must sacrifice the unemployment rate to win slower inflation.
The short run, however, tells an entirely different story.
In the short run (which, please note, can last several years) a stable relationship does exist between unemployment and inflation. This is easy to see by plotting the two variables over shorter subsets of time in such a way that they match up (roughly) with the business cycle. These curves (save the 1993 to 1998 experience -- leave that one aside for the moment) appear below.
And indeed, they show that there does exist a stable relationship between the two variables in question; they all exhibit a clearly negative relationship between unemployment and inflation. (Incidentally, if the beauty of the 1958-to-1969 curve doesn't bring you to tears, nothing in economics will.)
Note that this short-run Phillip curve shifts over time -- this is best seen by looking at the progression of the maximum values of the x-axis (unemployment) and the y-axis (inflation). How to account for the shifting? By recognizing that the location of the short-run curve depends on the expected inflation rate. History bears out this notion. The expected inflation rate rose steadily to more than 10% in the early 1980s from under 1% in the early 1950s, during which time the short-run Phillips curve continued to move in a northeasterly direction. Then inflation expectations began to decline. They have since dropped steadily to land at roughly 2.2% today, and the short-run Phillips curve has shifted back accordingly.
Which brings us to the present. The Phillips curve for 1993 to 1998 appears below.
What a mess! This is the one plot unlike all the others. If anything, it suggests that policymakers can lower inflation by driving down the unemployment rate!
This plot encapsulates the entire NAIRU/Phillips Curve debate. Indeed, members of the anti-NAIRU/PC camp point to it and claim that (a) even the short-run Phillips curve is now dead and (b) the notion that there exists some threshold level of the unemployment rate below which inflation begins to rise is laughable.
Are they right?
Begin the Begin
A closer look at the evidence does not work in favor of those who would discredit the NAIRU/PC model. This camp's rush to judgement stems from a misinterpretation of the process through which the NAIRU model unfolds.
The process, to paraphrase Meyer, runs like this:
Above-trend growth reduces the unemployment rate; a lower unemployment rate raises wage change relative to productivity growth; an increase in wage change relative to productivity growth raises labor costs; and an increase in labor costs results in higher price inflation.
Members of the anti-NAIRU/PC camp shorten this sentence to three words: Growth
inflation. But that is an unfair reduction. Yet that reduction (as disingenuous as it is) is the one and only thing that allows this group to survive. With it, its members can first cite above-trend growth and a drop in the actual jobless rate below the NAIRU (see table below); then they can say that these factors have failed to produce rising inflation; and then they can conclude that NAIRU is bunk.
Without this reductiveness, they are dead. They must face the fact that that the path from growth to inflation is not as clear-cut as they make it out to be, that the causal structure involved is actually fraught with factors that simply cannot be ignored.
In other words, they have to ignore some very convincing answers to this question: Why does the plot from 1993 to 1998 look like it does? Or, why has inflation not accelerated in the face of an unemployment rate that has fallen to its lowest level in 30 years?
- Relatively slow wage growth -- one of the very first links in the growth-inflation chain (see Meyer paraphrase above) -- is one of the biggest reasons. Paul Kasriel of
Northern Trust points out that real (inflation-adjusted) hourly compensation has grown only an average 0.8% per annum during the current expansion against an average 2.6% during the 1961 to 1969 expansion and an average 1.6% since 1947. Why so slow?
Earlier in the expansion, the sluggishness owed much to worker insecurity (employees thought the recovery too tenuous and their skills too dated to ask for raises); now it owes much to excess capacity worldwide (the inability of a firm to raise prices means an inability for it to pay higher wages). And one other important factor was at play: A remarkable deceleration in medical-care and benefits costs. The CPI for medical care was rising at a 9.7% year-on-year rate when the expansion began; it recently posted increases as low as 2.5%. The benefit costs portion of the employment cost index was rising at a 7.4% year-on-year rate when the expansion began; it recently posted increases as low as 1.8%.
All four of these factors kept overall labor costs subdued; relatively slow wage growth, in turn, ended up severely compromising the predictiveness of the NAIRU model in the 1990s.
So, too, did a confluence of exceptionally favorable but temporary supply shocks that impacted the inflation rate directly. (Meyer calls these "developments that have recently lowered the prices or slowed the rate of increase in the prices of specific goods, unrelated to the overall balance between supply and demand" in American labor and product markets.) They include an unambiguously strong dollar. And sharply lower commodity prices (which were rising at a 22.2% year-on-year rate when the expansion began; they're falling at a 17.6% rate now). And sharply lower computer prices (which were falling at a 7.0% year-on-year rate when the expansion began; they're falling at a 31.3% rate now). And sharply lower energy prices (which were rising at a 19.0% year-on-year rate when the expansion began; they're falling at an 8.8% rate now). And less rapid increases in food prices (which were rising at a 5.6% rate when the expansion began; they're rising at a 2.3% rate now). And a reversal in import prices (which were rising at a 3.1% year-on-year rate when the expansion began; they're falling at a 3.1% rate now).
There is no question that all of these favorable but temporary supply shocks have helped to produce an inflation rate more docile than the NAIRU model would have predicted.
This is not to say that other forces are not at work. Surely they are. Most importantly, an increase in trend productivity growth -- and hence an increase in trend gross domestic product growth -- means that the economy can grow at a faster rate than it has in the past without producing undue strain. And it is here that the anti-NAIRU/PC camp pounces: It chants that trend productivity growth has increased so much that there no longer exists a level of the jobless rate -- the NAIRU -- below which inflationary pressure can rise.
Yet it is extremely risky to be banking on an increase that big for two reasons. One is that, from the empirical side, it's terribly difficult (if not impossible) to prove that the bigger productivity numbers we saw between 1996 and 1998 owe to anything save cyclicality. The other is that, from the policy analysis/forecasting side,
doesn't believe in one. To lift from recent testimony:
Through the end of 1998, the economy continued to grow more rapidly than can be currently accommodated on an ongoing basis, even with higher, technology-driven productivity growth.
Note the "even with" and, at the same time, keep in mind the forces that have kept both labor costs and inflation in check. No, not one of them is guaranteed to reverse quickly or forcefully; it might be some time before they all even point in the same "troublesome" direction. But the point is that the huge, once-in-a-lifetime chunk of help that they provided for years is the key to understanding why the short-run Phillips curve and the NAIRU concept are far from dead -- and why it's unwise to assume they won't prove useful in future.
Thinking that it's totaled, people keep wanting to send the NAIRU model to the junkyard; but it's senseless to scrap a car that's only stalled.