How forecastable is consumption growth? New evidence on the Hall random walk hypothesis
I provide new evidence on the accuracy of the Hall random walk hypothesis that consumption growth is not predictable. Using 1971-94 US data, I find that consumption growth is substantially more predictable than previously believed. A regression of consumption growth on lagged nominal interest rate changes, lagged consumption growth and forecasts by Data Resources, Inc. yields an adjusted R-squared of 0.32. The empirical tests further suggest that excess sensitivity of consumption to current disposable income, rather than time-varying real interest rates or non-separable utility, is the most likely explanation for the predictability of consumption growth.