How Effective is the Money Supply in Predicting Nominal Income?: Application of Granger Causuality

The debate as to whether fluctuations in money help predict future fluctuations in in income has stood at the heart of the Monetarist and Keynesian debates as eloquently described by Friedman and Kuttner in their influential article (1992). Undoubtedly, money supply changes have become less useful in predicting the future path of nominal income as evidenced by a test developed by a Nobel Prize winning economist called the Granger Causality test.  This post will develop a simple model for the national income and us the Granger Causality test to statistically quantify the predictive power of money on nominal income with data from the 1990’s and 2000’s.  Unlike research from the 1960’s I find that money is still  statistically insignificant variable in predicting national income in the U.S.

The Model

The equation above states that the logarithmic changes of national income depend on its own past values, and past values of logarithmic changes in the nominal money supply and federal government expenditures.   The question is, in the presence of government expenditures and past values of national income does knowledge of the money supply series provide any further information about future values of national income?

Mathematically this can be described with the hypothesis outlined above.  The null-hypothesis is that all of the coefficients the lags in the logarithmic change in the money supply are equal to zero.  The alternative hypothesis is that at least one of the coefficients in the money supply portion of the income model above is statistically different from zero.


The null-hypothesis is that money does not Granger-cause income.  Friedman and Kutter found that during the period of 1960 and 1979 the value of the Granger’s test statistic was 3.68 so at the 1% significance level it is possible to conclude that money Granger causes income.  However, during teh 1970 to 1990 period one could not reject the null-hypothesis that money did not cause income with an F-statistic of .82.

Here are the results from a Granger-causuality test using data from 1990 to 2010 that I conducted:

The table where the dependent variable is Y represents the model used in this post which is exactly the same as the one used by Milton Friedman except with the most up to date data. One can reject the null-hypothesis that money does not cause income at the 5% significance level, but not at the 10% significance level.  It appears currently that the causal relationship between money and income is stronger than than it was during the late 70’s and 80’s but not as strong as it was during the 60’s and mid-70’s.