But as one can see from the red circles in the graph above, the expectations-adjusted Phillips Curve again seems to be missing over the last 5 years, with the observed inflation rate higher than predicted. Coibion and Gorodnichenko (2013) explore a number of possible explanations for this, including structural instability and changes in the labor market. They suggest that the best explanation is a divergence of different measures of the “expected inflation” that serves as a shift factor for the Phillips Curve. Using either the last-year’s average adjustment used in the above figures, or looking at expectations of inflation implied by the yields on Treasury Inflation Protected Securities, or expectations from the Survey of Professional Forecasters, one always finds recent inflation to have been higher than predicted by the historical Phillips Curve. But Coibion and Gorodnichenko note that these measures of expected inflation have recently diverged from the answers given by those households who are sampled in the University of Michigan’s survey of consumers. Those respondents have been consistently saying that they expect a higher inflation rate than the value implied by TIPS or professional inflation forecasters.
Read the whole thing. The charts tell a lot of the story.
In the current draft of the introduction to my macro book (and I am–once again–starting it over), I write,
If we look at the relationship between inflation and unemployment within time periods, the story is mixed. During the Forgotten Moderation, as unemployment came down, inflation increased, showing strong negative correlation. During the Great Stagflation, there is no apparent correlation, positive or negative, between inflation and unemployment. The same is true of the Great Moderation. The Financial Crisis Aftermath has only five years of data, which makes it difficult to establish correlation.
As you probably know, many macroeconomists have employed an equation (often on a diagram) that traces out a negative relationship between inflation and unemployment, and that this Phillips Curve has been at the center of controversy. I will have plenty to say about it in later chapters.
For the moment, I am just pointing out that the Phillips Curve is an example of macroeconomists using an equation that is not necessarily data driven…
The Phillips Curve began as an interesting empirical regularity, with no theoretical foundation. Milton Friedman famously said in 1967 (published 1968) that the attempt by policy makers to use it would cause it to break down. A few laters, it broke down. However, by this point, many economists had become so attached to it that they searched for new variables to add to the equation to make it work again. One of these was, in effect, the lagged dependent variable, supposedly representing “expectations.” In fact, in most economic time series, adding the lagged dependent variable improves the fit dramatically. Then you can play around with specifications to get the relationship you want between the variables you care about (in this case, inflation and unemployment).
The Phillips Curve is the archetype of Tinkerbell relationships in macroeconometrics. It is alive, but only if you believe in it.
I was going to look up “Tinkerbell relationships”, but you explained it. I wonder how many of these relationships are like seeing figures in the clouds which disappear quickly, or patterns in the stars, which disappear more slowly.