Another look at the evidence on money-income causality

Stock and Watson's widely noted finding that money has statistically significant marginal predictive power with respect to real output (as measured by industrial production), even in a sample extending through 1985 and even in the presence of a short-term interest rate, is not robust to two plausible changes. First, extending the sample through 1990 renders money insignificant within Stock and Watson's chosen specification. Second, using the commercial paper rate in place of the Treasury bill rate renders money insignificant even in the sample ending in 1985. A positive finding is that the difference between the commercial paper rate and the Treasury bill rate does have highly significant predictive value for real output, even in the presence of money, regardless of sample. Alternative results, based on forecast error variance decomposition in a vector autoregression setting, confirm these findings by indicating a small and generally insignificant effect of money and a large and highly significant effect of the paper-bill spread on real output.