One notable change is that movements in inflation now appear to tell us much less about future inflation than was the case, say, thirty years ago. Here I am talking about predictions of inflation using only information on past inflation, without taking into account any other information. The evidence suggests that, at the peak of U.S. inflation in the late 1970s and early 1980s, the best such “univariate” forecast of inflation--into the indefinite future--was a simple average of inflation over the past few quarters (Stock and Watson, 2007; Cecchetti and others, 2007). In that period, sharp increases in inflation were reversed only slowly. By contrast, shocks to inflation since roughly the mid-1980s have tended to be short-lived, so that the best forecast of future inflation would be a very long average of past inflation. Thus, when inflation moves above its recent long-run average, most of the upswing will likely be quickly reversed, although this result is not guaranteed.