Discussion of "Putting 'M' Back in Monetary Policy"

1. Introduction and Main Results

The European Central Bank assigns a prominent role for money and monetary analysis in its monetary policy strategy.¹ It will therefore not come as a surprise that I am very sympathetic to the general theme of Eric Leeper and Jennifer Roush's paper. Indeed, let us put "M" back in monetary policy.

In this paper, Leeper and Roush (2003, this issue of JMCB) show quite convincingly that explicitly allowing for a short-term response of the policy-controlled interest rate to money (M2) rather than to contemporaneous output and prices (as in the Taylor rule) helps identify the effects of exogenous monetary policy shocks in a monthly VAR. It reduces the price and liquidity puzzle, which is so often incurred in this type of exercise. It increases the size and the persistence of the output and price effects of a monetary policy shock. It improves the estimation of the short-run money market supply and demand schedules. And, importantly, these findings appear to be robust across various sub-samples (possibly with the exception of the most recent sub-sample). The bottom line is therefore: Let us put M back in the monetary policy reaction.

While I will not disagree with this conclusion and I will provide some additional evidence for the euro area in Section 3 of this discussion, it is worth asking why the estimated effects of policy shocks with a contemporaneous reaction to money are more plausible than those estimated in VARs without such a role for money. First, is the price puzzle really a puzzle? There is no full consensus about this. In particular, one can construct models in which contractionary monetary policy shocks have an initial positive effect on prices due to an interest rate cost channel (e.g., Christiano, Eichenbaum, and Evans 2001). Moreover, there is independent empirical evidence that such interest cost effects may indeed be present in various industries (Barth and Ramey 2001). Second, why is it plausible that a temporary monetary policy shock has large and persistent output effects? Indeed, is it not a puzzle that it has (e.g., Chari, Kehoe, and McGrattan 2000)? To my mind the most convincing feature of the impulse responses in the VARs with money is that monetary policy shocks do eventually lead to a significant and permanent effect on prices in line with the fall in the money stock. This is a feature that all monetary models will agree upon.

In addition, I have two smaller queries. First, the estimates of the short-run money demand function, presented in Appendix B, are not very convincing (M^d © 0.08C - 0.02P - 1.27(R - R^M) + ε ^MD_t ). While the interest rate elasticity is large and significant, the consumption elasticity is very low, and the price elasticity has the wrong sign. Second, there is no attempt at putting Y, P, and M in the short-run policy reaction at the same time. Why not also report those results?

Fig. 1.

Model 1. Impulse responses to a monetary policy shock in the euro area

Fig. 2.

Model 2. Impulse responses to a monetary policy shock in the euro area

2. Interpretation

The empirical result of Leeper and Roush that the Fed appears to respond to current M2 developments (rather than current prices and output) begs for an explanation. After all, as pointed out by the authors, in conventional macro models with recursive money demand, central banks aiming at price stability are well advised not to respond to money demand shocks.

One interpretation is that money plays a structural role in the transmission of monetary policy in addition to its impact on current and future real short-term interest rates. The exercise performed in Section 5.3 of the paper suggests that the authors are sympathetic to this idea. However, the limited attempts at introducing a structural role for money in forward-looking IS and New Keynesian Phillips curves have...