ECONOMISTS: ESPECIALLY DISMAL AT CALLING DOWNTURNS
Economists are way off the mark in predicting recessions
By nearly all measures, the economy is cruising through cloudless skies. Inflation is low, interest rates are steady, stock prices are hitting new highs, the Index of Leading Economic Indicators has risen at a solid 5.5% annual rate during the past six months. And only 20% of economists polled by the Blue Chip Economic Indicators see any chance of recession at all before 1998.
So everybody can relax, right? Not so fast. Whether they do their forecasts using large-scale computer models or on the back of an envelope, economists can do a good job predicting growth most of the time. But when it comes to calling recessions, their track record is miserable. Over the past 25 years, economic forecasters have missed four of the past five recessions. That includes the most recent downturn, which began in July, 1990 (table). Economists are "usually late in making the call," says David S. Wyss, director of research at DRI/McGraw-Hill.
Sure, downturns are often triggered by outside shocks, such as the oil-price jumps of the 1970s or the 1990 Iraqi invasion of Kuwait. Such events are inherently unpredictable by even the best forecasters. But after the shock hits, economists often have trouble incorporating the bad news into their models or into their thinking. Moreover, economic forecasting techniques have not kept up with an increasingly global and interconnected world. Academic economists have almost abandoned the field, while the commercial forecasters who use large-scale computer models do not have the resources needed to conduct long-term research.
Even when the U.S. economy was more insulated from global forces, economists had a tough time calling downturns. Go back to the late 1960s, at the end of a long period of sustained growth much like the current one. In December, 1968, leading economists put the probability of recession in the coming year at less than 15%. Unfortunately, they did not foresee that a combination of defense cuts and tight money would push the economy down into recession by the end of 1969.
The batting average of economists got even worse once the U.S. started getting battered by outside surprises such as the OPEC oil embargo of 1973 and the ensuing price increases, which led to an unanticipated sharp recession in America. The onset of the 1981-82 downturn also fooled most economic forecasters. And the recession of 1990 arrived with barely any warning. "No model predicted that recession at all," says Ray C. Fair, a Yale University professor whose forecasting model is now available on the World Wide Web (fairmodel.econ.yale.edu).
Forecasters defend themselves by arguing that recessions are unpredictable for good reason. A potential downturn that develops slowly--say, because consumers are becoming overextended or inventories are reaching excess levels--can be anticipated and defused by a nimble and well-informed Federal Reserve, especially under an inveterate data hawk such as Alan Greenspan. As a result, recessions happen only when the policymakers--and the forecasters--are surprised by events. "We're at a stage in the control of the economy where most recessions are the result of shocks," says Wyss. "That's very hard to forecast."
But much of the blame for the inability of models to forecast recessions must rest on academic economists. In the 1950s and 1960s, economic forecasting models were at the leading edge of academic research. The first major computer macroeconomic model was created by Lawrence R. Klein, who later won the Nobel prize in economics for his work. Top economists believed that macroeconomic models could enable economists to make predictions about the future path of the economy, just as physicists make predictions about the behavior of subatomic particles.
ACADEMIC STEPCHILD. Today, despite the obvious importance of better predictions, macroeconomic forecasting is an academic stepchild. With a few exceptions, "you just don't have new research going on," says Michael R. Donihue of Colby College, who organizes a regular meeting of the leading economic-model makers.
The downfall of the macro models started with their failure to predict the recession of 1973-75 and the stagflation that followed. "There was great disillusionment when it turned out that these models didn't deliver science," notes Christopher A. Sims, an economist at Yale.
Around the same time, academic economists became distrustful of the entire notion of large-scale macro models. The body blow to the models came from the University of Chicago's Robert E. Lucas Jr., the winner of last year's Nobel prize in economics. He argued that the behavior of consumers and businesses depends on their expectations about government monetary and fiscal policy. As a result, major shifts in policy--such as a big tax cut--could bring about changes in behavior that would be missed by the macro models, since they are not able to track expectations. The "Lucas critique," as it came to be known, discouraged a whole generation of economists from having anything to do with forecasting. Says Joel Prakken, chairman of Macroeconomic Advisers, a St. Louis forecasting firm: "There is a segment of the profession who thinks macromodeling is a defunct exercise."
The few attempts by academic economists in recent years to find better forecasting techniques have generally fallen short. In the late 1980s, two high-powered econometricians, James H. Stock of Harvard University and Mark W. Watson of Princeton University, came up with a new "recession index" that they hoped would better predict downturns. The problem? In its first big test, the new index completely missed the downturn of 1990.
Forecasters and academics agree, however, that there are ways to improve recession predictions. At the top of the wish list is a better model of the international economy. Most macro models are focused on the U.S., even though most recessionary shocks in recent years have come from overseas. Could a military flareup between China and Taiwan lead to a recession in the U.S.? What would be the effect of a violent civil war in Russia on the American economy? The existing macro models do not include the equations needed to answer these questions. Moreover, the necessary economic data on investment and trade flows is either faulty or does not exist. "We underrate the importance of what is happening overseas," says Wyss.
Today's forecasting models also have difficulty dealing with changes in consumer and business attitudes. Economists missed the recession of 1990 in part because they did not expect the invasion to depress consumer confidence as sharply as it did. Especially little is known about small-business attitudes, which are closely tied to entrepreneurial and job-creation activity. That's why forecasters cannot predict whether the economy could shrug off another stock market crash, as it did in 1987, or whether a collapse in stock prices would cause businesses and consumers to become more cautious and lead to a recession. "Understanding exactly what's driving expectations is an awfully tough problem," says Frederic S. Mishkin, research director for the Federal Reserve Bank of New York.
Certainly the economy is strong enough that there's no reason to expect a recession in the next few months. But the experience of the past quarter century shows that when the next recession comes, whether next year or next century, it's sure to be a surprise-- especially to economists.By Michael J. Mandel in New YorkReturn to top