Is The Fed Fighting A Phantom Menace?

A new study says no indicator is much good at predicting the inflation rate

With unemployment at a 30-year low, it seemed only natural for the Federal Reserve to raise interest rates on May 16. Conventional wisdom says a tight labor market forces companies to pay more for scarce labor, and this leads to higher prices. Unemployment is "the classic and most watched indicator" of inflation, says Jonas D.M. Fisher, senior economist at the Federal Reserve Bank of Chicago.

But a new study by the New York Fed throws cold water on this widely accepted theory. It finds that inflation cannot be predicted from current unemployment rates. Some 85% of the time, according to the study, taking unemployment into account produces worse forecasts than simply assuming inflation will continue on its current path. And that's not all. Other widely followed indicators of inflation do little better (table). This suggests the Fed's forecasting ability is far weaker than many people believe--and there's at least a chance the Fed is waging war against a phantom inflation menace.

The study, The Unreliability of Inflation Indicators, cannot be dismissed lightly. Its three authors are Stephen G. Cecchetti, an Ohio State University professor and longtime inflation hawk who is a former research director of the New York Fed; Rita S. Chu, a New York University graduate student who used to work at the Fed; and Charles Steindel, a senior vice-president for research at the New York Fed.

The authors sized up the performance of 19 supposed indicators of inflation from 1985 to 1998. These included measures of the real economy, such as the unemployment rate; financial indicators, such as interest rates; and commodity prices, such as the prices of gold and oil. They found that none of these reliably improved on forecasts that simply projected past inflation rates into the future.

GUESSWORK. Each measure had its own problems. For instance, average hourly wages did a decent job of predicting inflation. But that's because the researchers assumed unrealistically that a forecaster using wages to predict inflation would have perfect foreknowledge of the wage level for two years ahead. In reality, forecasters don't know what wages will be any more than they know what inflation will be. In fact, inflation affects wages, so the researchers' shortcut amounts to knowing inflation in order to predict it. Write the authors: "Forecasters need an indicator whose future values can be predicted independently of inflation."

Some other widely followed indicators don't have that interdependence problem because they aren't influenced as much by inflation--but they have their own drawbacks. Financial indicators, for example, tended to be "absolutely rotten," says David A. Wyss, chief economist at Standard & Poor's DRI, a unit of Business Week'S publisher, The McGraw-Hill Companies, who analyzed the Fed study. The poor performance of M1, which is currency plus checking and savings accounts, is particularly notable, Wyss adds, since monetarists believe inflation comes from overly rapid growth in money-supply measures such as M1.

As a group, commodity prices proved to be the best inflation predictors--but in a way that defies all rationality. When the prices of industrial materials, gold, and oil rose, inflation tended to fall. One possibility the authors consider is that the Fed noticed the price increases and reacted to them so strongly with higher interest rates that its moves extinguished the inflation the commodity prices were signaling. But they say the possibility that policymakers regularly overreact to commodity price signals is "far-fetched."

Cecchetti, Chu, and Steindel note that their study concerns indicators used individually, and they suggest it might be easier to forecast inflation using combinations of indicators. James H. Stock of Harvard University and Mark W. Watson of Princeton University, for instance, have used historical inflation data to generate a predictor incorporating more than 160 indicators. But it's still not clear that the composite is consistently much better than single indicators. "We're still in the research and development and testing phases," Stock says.

Another alternative to the single-indicator approach is to build a complex, multi-equation model of the economy that spits out forecasts for all kinds of measures, from inflation to gross domestic product. Examples include the models of the Fed, DRI, and Macroeconomic Advisers LLC in St. Louis.

Cecchetti and his colleagues find the DRI model, the only one they examined, outperformed single-indicator forecasts nearly every time. But that's not much use to the Fed, because the predictions of macro models incorporate expectations of what the Fed itself will do. Before intervening, the Fed needs to know how the economy would fare without its actions. For the Fed to base decisions on this sort of model would be like looking into a hall of mirrors.

If inflation is so hard to predict, what should the Fed do? Inflation hawks argue that it's prudent to raise rates now so inflation doesn't have a chance to take root. An alternative argument is that preemptive interest-rate moves are unwarranted. There's a cost to inflation, but there's also a cost to squelching a healthy economy.

    Before it's here, it's on the Bloomberg Terminal. LEARN MORE