We are grateful for suggestions from Robert Barsky, Alan Blinder, John Campbell,

ALTHOUGH MOST ECONOMISTS agree that inflation is costly, there is no consensus about why. Many traditionally cited costs, such as deadweight loss from the inflation "tax, " seem too small to justify concern about moderate inflation. One approach is to argue that inflation of 10 percent or 15 percent would not be particularly costly if it were constant and fully anticipated, but that a rise in the level of inflation raises uncertainty about future inflation. In the absence of perfect indexation, such uncer-tainty has significant costs, including arbitrary redistributions, relative price variation, and fewer long-term contracts, such as loans to finance investment. I This view implies that understanding the costs of inflation requires that we understand the connection between the level of inflation and uncertainty. The idea that high inflation leads to greater uncertainty is suggested in Arthur Okun's "The Mirage of Steady Inflation " and in Milton Friedman's Nobel lecture, and many economists treat it as a stylized fact.2 But empirical studies of the inflation-uncertainty relation report conflicting results, and the issue appears unsettled