IN AN EARLIER PAPER presented to the Brookings Panel on Economic
Activity, I attempted to survey the extent to which recent consumer behavior
would have been “understood” by some of the principal U.S. forecasting
models in the absence of concurrent errors emanating from other
sectors of the economy.’ At that time, I concluded that the errors in forecasting
that reflected most seriously on the structure and composition of
the typical forecasting equations concerned consumer durable expenditures,
particularly for automobiles. An attempt to improve the explanatory
power and forecasting ability of a stock-adjustment automobile equation
by incorporating a household wealth variable was notably unsuccessful.
