been less accurate in predicting economic “turns”) and of inflation, except in
instances of large inflation shocks (as in the 1974 oil shock or the 1982 reces-
sion). Energy forecasts have improved dramatically. But in the last decade, fur-
ther incremental gains in accuracy have been small.
Second, many economic variables still elude accurate forecasting. To be
useful, any prediction also should report a margin of error or confidence inter-
val around its estimate. One way to appreciate this uncertainty is to survey a
great many forecasters and observe the range of forecasts for the same eco-
nomic variable. (But even this range understates the uncertainty. A significant
portion of actual outcomes falls outside the surveyed range; that is, the out-
comes are higher than the highest forecast or lower than the lowest.)
Third, the time period for making forecasts matters. On average, accuracy
falls as the forecasters try to predict farther into the future. The time interval
forecasted also matters. (Forecasts of annual changes tend to be more accu-
rate than forecasts of quarterly changes.) Fourth, no forecaster consistently
outperforms any other. Rather, forecast accuracy depends on the economic
variable being predicted, how it is measured, and the time horizon. But the
differences in accuracy across the major forecasters are quite small. Overall,
macromodels performed better than purely extrapolative models, but, for
many economic variables, the advantage (if any) is small.Final Thoughts
Estimating and forecasting demand are as much art as science. This chapter has
presented some of the most important statistical techniques currently avail-
able. But the analyst (and the manager) must never let these techniques them-
selves be the final arbiter of the quality of demand equations and forecasts.
Judgment plays as important a role as statistics in evaluating demand equa-
tions. Thus, it is important to answer the following questions:- Does the equation (or equations) make economic sense? What is the
underlying economic relationship? Are the “right” explanatory
variables included in the equation? Might other relevant variables be
included? What form of the equation is suggested by economic
principles? - Are the signs and magnitudes of the estimated coefficients
reasonable? Do they make economic sense? - Based on an intelligent interpretation of the statistics, does the model
have explanatory power? How well did it track the past data?
If the equation successfully answers these questions, the manager can be con-
fident that it makes good economic sense.166 Chapter 4 Estimating and Forecasting Demandc04EstimatingandForecastingDemand.qxd 9/5/11 5:49 PM Page 166