In an attempt to view macroeconomics through a 1970 lens, I decided to model the Phillips Curve as a simple linear trade-off, using the period 1956-1968. In those years, the inflation rate averaged 2.19 percent and the unemployment rate averaged 5.0 percent. The sum is 7.19 percent. So the obvious linear equation for the inflation rate would be:

inflation = 7.19 – unemployment.

See how this equation works for the periods 1956-1968, 1969-1981, and 1995-2007.

1956-
1968
1969-
1981
1995-
2007
Average Inflation 2.19 7.81 2.67
Standard Error 0.61 7.67 1.04
Largest Underprediction 1.4 12.33 1.68
Largest Overprediction -0.84 none -1.09

My reading of this is that for 1956-1968 and for 1995-2007 the simple linear trade-off works quite well. The standard error is low. There are no major outliers.

For 1969-1981, basically every data point is an outlier, on the upside. Inflation was higher than predicted in every year.

Maybe 1969-1981 still fits a linear model, but with a different line? Since inflation averaged 7.81 percent and unemployment averaged 6.19 percent, we could try

inflation = 14.0 – unemployment.

This new line does not work very well, either. The standard error is 3.54 percent. The largest underprediction is 1980, when inflation was 12.35 percent, or 5.53 percentage points above the line. The largest overprediction was 1972, when inflation was 3.41 percent, or 4.99 percentage points below the line. Note that there were wage and price controls in place in 1972.

The original 1956-1968 line seemed to work again from 1995-2007. What about lately? For 2008, it predicts inflation of 1.38 percent, and we actually got -0.04 percent. For 2009, it predicts inflation of -2.08 percent, and we actually got 2.78 percent, which is an underprediction of 4.86 percent. I have not done the calculation, but my guess is that 2010 will be another outlier on the high side, although probably not as bad.

People who favor monetary and fiscal expansion worry about a 1930’s scenario. People who favor restraint worry about a 1970’s scenario. I do not think we should put all our weight on either one of those scenarios.

[Update: David Andolfatto links to an amusing powerpoint by Mike Bryan on the Phillips Curve breakdown. Thanks to Josh Hendrickson for the email suggesting Bryan’s presentation.]