Forecasting US GDP: assessing the quantity theory of money as a device

Abstract

We revisit Friedman’s Quantity of Money Theory and analyze the ability to forecast US real GDP using lagged M2 money supply. The Granger Causality Test indicates that M2 money supply Granger causes real GDP. However, we find that the causality is one-way, from M2 to real GDP, as real GDP fails to Granger cause M2. We surveyed a range of ARIMAX models to ascertain whether M2 money supply still forecasts economic activity, both before and during the global COVID pandemic. While economic theory suggests that money is neutral, we found that hybrid models that combine both standard ARIMA parameters plus M2 money supply growth as an exogenous regressor(s) actually improve forecast accuracy. Both before, during, and in the full sample, a hybrid model combining money supply and autoregressive parameters outperform atheoretical Box Jenkins models as well as the naïve model. Our results suggest that, while money supply has fallen off the radar screen of many economists, employing M2 in a hybrid model still improves forecasting ability.