(An excerpt from the recently published edition of ‘Central Banking and Monetary Policy: What will be the post-crisis new normal?’ and the Suerf Bocconi Working Paper ‘Monetary policy committees, voting behavior and ideal points’)
There is a substantial literature using voting records of monetary policy committees to learn about decision making at central banks. This literature is motivated by the fact that central banks ought to take decisions independently, need to be held accountable for their decision making and ideally leverage on the pooling of opinions.
Monetary policy committees come, however, in a wide variety of forms often reflecting their historical development. However, even if one could redesign a central bank from scratch, it is not entirely obvious how an optimal design should look like (Reis, 2013).
The literature has studied various aspects of committees around the world. These aspects include appointment systems, career concerns, the inclusion of outsiders, size, etc. Some aspects are relatively well understood by now, but generalizing conclusions requires carefulness because most of these studies are in essence case studies.
One way forward is the study of a more diverse set of central banks. However to date, quite a few central banks are fairly reluctant to share transcripts and voting records, despite pleas for more transparency (Eijffinger, 2015).
The methods we put forward to study voting behavior of monetary policy committees are commonplace in the analysis of legislative bodies or judicial courts by until now underutilized in economics. While we see substantial advantages to this approach such as the flexibility to create quantities of interest and the careful incorporation of uncertainty, there are certainly drawbacks.
One objection against the approach outlined in this chapter is that it is too simplistic. We have heard on more than one occasion that a complex decision making process cannot and should not be reduced to mapping policy makers on one simple dimension. This objection is similar to objections made in the context of political science (see Lauderdale and Clark, 2014). However, as demonstrated in Eijffinger, Mahieu and Raes (2013a), even a single latent dimension generates low prediction error and hence leaves little room for statistical improvement.
We do not argue that spatial voting is superior to approaches commonly used in the study of monetary policy committees (e.g. estimating reaction functions). However, we find spatial voting models a neat addition to the toolkit of scholars investing committees.