He writes (can anyone find a free, ungated copy?),

The implied short-run unemployment NAIRU is highly stable, staying in the range of 3.9 to 4.4 percent between 1996 and 2013. However, the rise in the extent and duration of long-run employment causes the NAIRU for the total unemployment rate (short-run plus long-run) to increase from 4.8 percent in 2006 to 6.5 percent in 2013:Q1. As a result, this paper supports the recent research that argues that there has been an increase in structural unemployment taking the form of long-run unemployed.

Gordon talks up what he calls the “triangle” approach to the Phillips Curve, as opposed to the New Keynesian approach (NKPC).

The triangle approach differs from the NKPC approach by including long lags on the dependent variable, additional lags on the unemployment gap, and explicit variables to represent the supply shocks…the change in the relative price of non-food non-oil imports, the eight-quarter change in the trend rate of productivity growth, and dummy variables for the effect of the 1971-74 Nixon-era price controls.

This methodology is so reminiscent of the 1970s that it makes me want to wear bell-bottom jeans and listen to disco.

No inflation model has any credibility unless it deals explicitly with the supply shocks that created the positive inflation-unemployment correlation in the 1970s and early 1980s and the lead of inflation ahead of unemployment. But this was not the only episode. Between 1996 and 2000 the unemployment rate descended far below its 5.84 average to less than 4.0 percent, and yet the inflation rate did not speed up as it had done in the late 1960s and late 1980s. The triangle model explains why inflation was so tame, pointing to beneficial supply shocks during the late 1990s – oil prices were low, the dollar was appreciating, and productivity growth was reviving.

He points out that a simple plot of the Phillips relation between 1970 and 2006 looks simply awful (from the point of view of anyone who thought that the 1960s Phillips Curve would persist).

The overall correlation appears to be positive but very weak, close to zero, and the 1970-80 observations in the scatter plot once led Arthur Okun to describe the PC as “an unidentified flying object.”

Gordon’s productivity trend acceleration variable is really important.

Its deceleration into negative territory during 1964-1980 might be as important a cause of accelerating inflation in that period as its post-1995 acceleration was a cause of low inflation in the late 1990s. Note also that the productivity growth trend revival of 1980-85 may have contributed to the success of the “Volcker disinflation,” a link that has been missed in most of the past PC literature. There has been a sharp deceleration of trend productivity growth since 2004, helping to explain the absence of deflation in the past few years.

Note that there are other ways to arrive at a negative correlation between productivity growth and inflation. You could have measurement error in the rate of inflation (if you overstate price increases, you understate productivity growth, and conversely). You could have the Fed following a policy that approximately targets nominal GDP, so that when productivity goes up inflation goes down, and conversely.

From a modern perspective (meaning any graduate macro syllabus after 1975), the “triangle” model is hard to swallow. In Gordon’s model, the only way for money to affect inflation is through the “output gap.” My guess is that this makes hyperinflation a mathematical impossibility in the Gordon model–to get to 100 percent inflation rates would probably require decades of a negative output gap, which is not what preceded any of the many actual hyperinflations in the world.

For me, the important point is to keep in mind is that there is no unique interpretation of the path of nominal GDP, employment, productivity, and inflation. Take any three of those as given, and you arrive at the fourth. Arithmetically,

[1] growth in nominal GDP minus growth in inflation = growth in employment plus growth in productivity

Scott Sumner says “tell me the path of nominal GDP, and I will tell you the path of employment.” He is bound to be right, as long as you do not have to worry about how inflation and productivity move in the short run. And they will move in opposite directions to the extent that inflation is measured with error.

Robert Gordon says “tell me productivity growth and the rate of short-term unemployment, and I will tell you the rate of inflation.” He is bound to be right about productivity growth, since if we hold nominal GDP growth and employment constant, productivity growth and inflation vary inversely. His problem is that, arithmetically, employment and inflation should be negatively related. But if you eliminate all instances of positive relationships between unemployment and inflation by calling them “supply shocks” (which differ from time-varying productivity), then by golly, the Phillips Curve is indeed “alive and well.”

I say that in order to claim to have a macroeconomic model, you have to treat nominal GDP, employment, inflation, and variations in trend productivity all as endogenous. Otherwise, you are just choosing one interpretation from among many.