In my previous post I attempted to shoot down the idea of the impotence of monetary policy at the zero bound. Given the issues raised in that post, there are two topics that need to be addressed. One issue is whether the interest rate is useful for monetary policy analysis, especially given its limitation at the zero bound. The second issue is whether there is a preferable alternative to the interest rate that should be used. I will answer these questions in reverse order.
David Beckworth asked which monetary aggregate should be used to examine the current crisis. Following the suggestion of Gary Gorton, who suggests using M3, he plots the year-to-year percentage change in M1, M2, and M3. The plot yields different predictions for M3 than the other aggregates. So, which aggregate should we use? The answer is none of them — or at least as they are traditionally measured.
Traditional aggregates are computed using a simple sum method. In other words, one simply adds currency to checkable deposits to time deposits, etc. These aggregates are thus potentially flawed by the fact that they are not consistent with economic theory, aggregation theory, or index number theory. With regards to economic theory, the simple sum aggregation procedure assumes that all assets within a particular aggregate are perfect substitutes — a characteristic that we know to be false.
An alternative to the simple sum aggregates are the Divisia aggregates — or the monetary services indexes — initially derived by William Barnett and available through the St. Louis Federal Reserve. In contrast with simple sum aggregates, these aggregates are consistent with economic, aggregation, and index number theory.
Two questions naturally arise. First, do these aggregates provide more information than the simple sum counterparts. Second, are these aggregates preferable to the interest rate.
In a recently completed working paper that I have written, I address these questions by re-examining some empirical work that suggests that monetary aggregates are unimportant and that the interest rate is sufficient for predicting movements in the output gap. Using these monetary services indexes rather than the simple sum aggregates, I find that many of the conclusions are reversed. One major conclusion that is reversed is regarding the IS equation referenced in the earlier. Specifically, Rudebusch and Svensson published a paper in 2002 that estimates the following backward-looking IS equation:
y(t) = b1*y(t-1) + b2*y(t-2) +b3*r(t-1) + e(t)
where y(t) is the output gap at time t, r(t-1) is the lagged real interest rate, b1, b2, and b3 are parameters, and e(t) is a disturbance term. They find a negative and statistically significant relationship between the real interest rate and the output gap. In addition, they suggest that when money terms are added, they are not statistically significant.
My paper addresses this by adding real money balances measured by the monetary services index counterpart to M1, M2, and MZM. For the entire sample, 1961 – 2005, I find a positive and significant impact of real balances on the output gap. The real interest remains negative and significant. However, when estimating the results for the subsample that begins with the Volcker-led Federal Reserve (a benchmark used by those who think monetary aggregates are not useful), I find that the monetary service index counterpart to M2 and MZM has a positive and significant impact on the output gap. What’s more, for these equations, the real interest is no longer statistically significantly different from zero. In other words, one cannot reject the null hypothesis that the real interest rate has no effect on the output gap.
Thus, whereas my last post claimed that models with a single interest rate that measures monetary policy are a weak reed on which to develop one’s theory of the monetary transmission process, this post makes clear that there exists a better alternative to the interest rate. The alternative is a familiar one — monetary aggregates. However, the monetary aggregates are not the simple sum variety, but rather they are ones which are consistent with economic, aggregation, and index number theory.
[[[I have yet to upload my working paper. If anyone is interested in reading a copy of the paper, feel free to send me an email: firstname.lastname@example.org]]]