Prof. Bryan Caplan
Week 11: Simple Econometrics and the Phillips Curve
- Economic Theory of the Simple Phillips Curve
- Review of Aggregate Demand and Aggregate Supply
- Classical AS curve is vertical; Keynesian AS curve is upward sloping (or horizontal!)
- The same is true for Classical vs. Keynesian views of the overall labor market (which both assume is closely tied to the overall output market):
- The Classical labor supply curve is vertical.
- The Keynesian labor supply curve is upward sloping or flat.
- It is possible to represent these labor markets from a slightly different perspective: we can show unemployment rather than employment on the X-axis. Then the Classical curve remains vertical, but the Keynesian curve becomes negatively-sloped.
- In the 1950's, a British researcher named Phillips found a negative relationship between unemployment and wage inflation (i.e., a positive relationship between employment and wage inflation). This seemed to be consistent with the Keynesian view of labor markets and the macroeconomy.
- Later researchers (Samuelson and Solow) re-did Phillips' results for the U.S. using unemployment and price inflation data. They called this the "menu of choice between inflation and unemployment."
- Econometric Evidence on the Simple Phillips Curve
- This simple Phillips curve works well for the U.S. for 1961-1969.
- This can be seen with the naked eye, or...
- Look at the regression results. Regressing unemployment on inflation gives a coefficient of -.59 with a t-stat of 4.1.
- This means that every increase in inflation of 1 percentage-point tended to reduce unemployment by .59 percentage points.
- However, if you look at the data for 1961-1995, the results are completely different. Inflation now has a positive sign of .10, and is statistically insignificant.
- Strangely, the simple Phillips curve remains a popular theory among journalists, even though it fails the simplest statistical tests.
- Economic Theory of the Expectations-Augmented Phillips Curve
- Why did the simple Phillips curve quit working? Several theorists (most prominently, Milton Friedman), predicted that it would fail while it was still working.
- The critique: the Keynesian model will only be true if people don't adjust their expectations about inflation.
- If people don't adjust their expectations, then real wages will be eroded by inflation, increasing employment.
- But if people do adjust their expectations, then real wages won't be eroded by inflation, so there will be no tendency to increase employment.
- Lincoln's law: You can't fool all of the people all of the time. In the long-run, trying to increase employment by fooling people about inflation won't work.
- This gives rise to the "expectations-augmented Phillips Curve." It has two Phillips curves on it: the short-run downward sloping curve, and the long-run vertical curve.
- Econometric Evidence on the Expectations-Augmented Phillips Curve
- Try regressing unemployment on inflation and lagged inflation. The expectations-augmented Phillips curve suggests a negative sign on inflation, and a positive sign on lagged inflation. (I say "suggests" because it could take longer than 1 year for expectations to adjust. But this is a good first start).
- The results: it works! The coefficient on inflation is -.35; the coefficient on lagged inflation is +.57. Both are statistically significant.
- General conclusion: the simple Phillips curve worked initially because people had little experience with inflation. After they got some experience with inflation, it became possible to get "stagflation" - high unemployment and high inflation at the same time.
- Currently people have largely forgotten about inflation, so the temptation to take advantage of the situation has re-emerged...
- Estimating the Natural Rate of Unemployment
- An additional part of the critique of the Phillips curve is the idea that there is a "natural rate" of unemployment deriving from job search and demographics. The vertical line in the Phillips curve shows this "natural rate."
- Can we econometrically estimate the "natural rate"?
- Yes. Just plug in 0 for all values of inflation, past and present.
- If you don't include lagged unemployment as a regressor, then the constant gives you an estimate of the natural rate.
- If you included lagged unemployment, set unemployment equal to U for all periods, then solve for U. (Look at some of the specifications that include lagged U; this is also a good check on our results - they still work).
- The natural rate estimates seem to range between 3-7%. (A wide range, but still interesting).
- Caveat: There seems to be some evidence that there is a positive long-run impact of high inflation on employment. (Plug in e.g. 10% inflation in all periods). So if the natural rate view is wrong, then in the long-run inflation is positively harmful rather than merely ineffective.