How Big is the Random Walk in GNP?
Journal of Political Economy 96 (October 1988) 893-920. Short-order ARMA models suggest that GNP looks a lot like a random walk. But short-order ARMA models are fit to match one-step ahead forecasts, and can do a poor job of capturing long-term forecastability. I used a variance-ratio statistic (variance of long-term differences / variance of one-year differences) to show that there is a lot of mean-reversion in GNP that short-order ARMA models miss. I think the subsequent “permanent and transitory components” answers the substantive question better, but the warning about using long-term implications of short-term models remains worthwhile today.