Does consumption take a random walk? Evidence from macroeconomic forecasting data

Professional quarterly forecasts of aggregate u.s. consumption series are used to test hall's (1978) random walk hypothesis. results from band spectrum regressions suggest that forecasts predict about 40 percent of the low-frequency variation in growth rates of expenditures on nondurables and services. the forecasts have no explanatory power for high-frequency variations. the rejection of the random walk hypothesis is traced to excess sensitivity of consumption growth to forecasted income growth at the business cycle frequencies.;