RDP 2010-05: Direct Effects of Money on Aggregate Demand: Another Look at the Evidence 5. Conclusion

The question of whether money has direct effects has become more important as policy rates have approached the zero lower bound in the euro area, Japan, the United Kingdom and the United States. Monetary authorities in these economies have turned to unconventional monetary policies which involve balance sheet expansions of one form or another.

Our contribution in this paper is to reinterpret some of the econometric evidence which on the surface often suggests that, after controlling for the short-term real interest rate, real base money growth can be a significant determinant of total output for a number of countries and sample periods. Our interpretation of these types of results is that they are likely to be biased.

We reach this conclusion by using a model that has no direct effects of money. Yet the model is capable of producing data which leads to positive and statistically significant coefficients on real money growth in a real output regression similar to those that are often found when using actual data. From the perspective of the structural model, it then becomes clear that the reduced-form regressions suffer from an omitted variable bias. In particular, the bias on real money growth is large enough to undermine the validity of any inference about the existence of direct effects of money.

We have also repeated the analysis with a model in which money has sizeable direct effects and have also found a bias, but this time operating in the opposite direction – that is, the estimate of the reduced-form coefficient on real money is not significantly different from zero. The empirical regressions also exclude key variables from this model, in which case the estimated coefficients on real money growth go to zero.

In short, the reduced-form regressions – even when they fit the data well – are misleading. They simply fail to uncover the true structural relationship between money and the rest of the economy. They lead to incorrect inferences on the existence of direct effects and they are an unreliable guide to calibrate monetary policies, in general, including at the zero lower bound.