Central Bankers are rarely “responsive” to the public through the democratic process. In the United States, most members of the (unelected) Federal Reserve Open Market Committee (FOMC) have no other government experience or aspirations, and in spite of their clear importance to the economy, the body releases relatively little information to the public about how or when they make their choices. Many eminent economists have argued that the FOMC should make decisions based on known rules, instead of using their discretion (Friedman 1982, Taylor 2012). Others argue that transparency can go too far (for instance, see the review in Mishkin 2004).

The FOMC and many central banks internationally have chosen to become more transparent over the past generation. One of the major components of the transparency effort is a decision to release the transcripts of the FOMC deliberations with a 5 year lag in 1992.

In the mid-1970s, in the wake of the Vietnam War and Watergate, the U.S. government passed a series of transparency laws, which should have forced the FOMC to release the transcripts at that moment. However, the FOMC managed to circumvent these laws, did not they finally complied. As first pointed out by Meade and Stasavage (2008), this provides for a natural setting for empirically studying the effects of transparency: a large and salient policy change, which was not driven by other factors of the economy.

In our paper (Egesdal et al. 2015), we focus on a particular consequence of increased transparency, which is its effect on the deliberations of the FOMC. We do this by comparing the sometimes-public deliberations to the always public meeting summaries, which have been released soon after the meetings. Our argument is based on the insight that the summaries are written with the public in mind, and so the relationship between the language in the summaries and deliberations is informative about if transparency changes the way that the FOMC deliberates. Unsurprisingly, we find a large shift after the reform: the language in the deliberations and meeting summaries significantly converge.

There are three broad explanations for this finding: that the meeting summaries changed to be more reflective of the deliberations (because now their accuracy would be verifiable), that transparency changed the content of deliberations themselves (since that behavior was now observable), or that the relationship between the actual deliberations and the transcripts changed. The latter explanation contradicts the historical evidence on the process of transcription. We develop new methods to distinguish between the former stories, and discover that most of the change was driven by changed behavior at the meetings themselves.

Almost all of the effect was driven by very few words, including “growth,” “market,” and “inflation” (whose usage has been consistently high in the summaries, and increased in the transcripts after the reform), and “think” (whose usage dramatically decreased in the deliberations). That said, just because language changes does not mean that meaning changes. We introduce a systematic method to account for semantic similarity, using published dictionaries designed to define technical economics language (Black et al. 2012). Words that appear in each other’s definitions, and that commonly appear in other words’ definitions as well, are likely to have similar meanings. Accounting for these linkages allows us to be confident that most of the effect of the transparency reform was to change meaning, not just the words themselves.

That the language of deliberations changed is backed up by the qualitative history of the episode: FOMC members used prepared remarks instead of speaking extemporaneously following the reform. Changes in discourse may be due to career concerns. Even though the FOMC members are likely to leave government, there is a (small) chance that they will be asked to lead the FOMC, or at the very least gain access to lucrative book or speaking deals after they leave. In order to access these opportunities, they change their behavior in order to speak about the economic topics that the public cares about, even if privately the FOMC member would have spoken about different issues.

In addition to our descriptive analysis, we also hope that others find our methods useful. Text data is becoming more and more ubiquitous, but is difficult to use in a systematic way. One approach is to use “black-box” machine learning tools to analyze the data, but there are high entry barriers to learn how to use the tools, and understanding how they work is difficult. Our method, which we describe more fully in the paper, uses straightforward tools from linear algebra to describe the evolution of text over time, both overall and decomposed into pieces.

BIBLIOGRAPHY

Black, J., Hashimzade, N., & Myles, G. (Eds.). (2012). A Dictionary of Economics. OUP Oxford.

Egesdal, M., Gill, M., & Rotemberg, M. (2015). How Federal Reserve Discussions Respond to Increased Transparency. Working paper

Friedman, M. (1982). Monetary policy: Theory and practice. Journal of Money, Credit and Banking, 98-118.

Meade, E. E., & Stasavage, D. (2008). Publicity of Debate and the Incentive to Dissent: Evidence from the US Federal Reserve. The Economic Journal, 118(528), 695-717.

Mishkin, F. S. (2004). Can central bank transparency go too far? NBER Working Paper w10829

Taylor, J. B. (2012). Monetary policy rules work and discretion doesn’t: A tale of two eras. Journal of Money, Credit and Banking, 44(6), 1017-1032.

The preceding post comes to us from Michael Egesdal, doctoral candidate at the Department of Economics at Harvard University, Michael Gill, doctoral candidate at the Department of Government at Harvard University, and Martin Rotemberg, Visiting Assistant Professor of Public Policy at the Harvard Kennedy School. The post is based on their working paper, which is entitled “How Federal Reserve Discussions Respond to Increased Transparency” and available here.