Beginning in the 1970s, US Representative and Chairman of the House Banking, Finance, and Urban Affairs Committee Henry B. Gonzalez began advocating for increased transparency from the Fed. In 1993, a Freedom of Information Act (FOIA) request prompted by Gonzalez’s efforts revealed that transcripts of key Fed policy-making meetings — previously believed to have been destroyed — had been preserved. By the end of the year, the Fed, under pressure, began releasing the transcripts.
Proponents of transparency argue that openness about operations and decision making promotes better policy and outcomes. “It’s the concept of sunshine as disinfectant: when people expect their work and decisions to be scrutinzed in the public realm, they are more likely to work harder and better,” Professor Andrea Prat explains. Opponents, including former Fed Chair Alan Greenspan, have invoked the candid advice defense of executive privilege. “If President Obama calls me into his office to consult him, I’m more likely to give him an open and honest opinion if I know our conversation is secret,” Prat says. No universal standard exists among central banks, though the tide seems to be turning toward transparency: the European Central Bank recently announced that it might begin to release its meeting transcripts; so has the Bank of England.
The release of Fed transcripts created an opportunity for Prat, working with Stephen Hansen of Universitat Pompeu Fabra in Barcelona and Michael McMahon of the University of Warwick in Coventry, to investigate the impact of transparency on a deliberative governmental body charged with carrying out important policy decisions.
The Federal Open Market Committee (FOMC), whose meeting transcripts were the subject of the 1993 FOIA request, is the Fed’s main policy-making body. Analyzing FOMC meeting transcripts before and after 1993 provided the perfect natural experiment. “Meeting transcripts prior to November 1993 reflect members’ belief that their deliberations would never be disclosed,” Prat says. “And it’s clear that later they were absolutely surprised to learn the transcripts had been preserved.”
To paint a quantitative picture of the FOMC before and after transparency, the researchers analyzed the transcripts using a popular machine-learning algorithm for Big Data developed by computer scientists called latent Dirichlet allocation (LDA). Without any direct input from humans, LDA quickly identifies topics in text documents to find patterns that a researcher working manually might fail to identify. LDA analyzes each topic in the text as a distribution of words, where the same word has different weights in different topics. For example, in a discussion about the World Cup, the word “rain” might occasionally come up, since weather would affect the soccer field. But it would come up much more often in discussions about the weather. By identifying words’ frequency as well as their relationships to other words, and not simply their mere presence, LDA can reflect the probability that a conversation is about the World Cup rather than the weather. “So when we create an analysis of an FOMC member, I can say not just that she talked about inflation and investment, but 73 percent of the conversation is related to inflation, 21 percent to investment, and the rest to other topics,” Prat says.
The researchers found that after transparency, members tended to make fewer statements, ask fewer questions, and cover the same topics as the chair more often than before transparency, suggesting greater conformism. But they also found another effect. At the end of each meeting day, the chair summarizes the discussion — a summary that is typically a clear predictor of policy recommendations the chair will make. Using a method similar to how Google measures the importance of web pages, the researchers could see which members were most influential with the chair. High influence scores were awarded to members who introduced new topics that the chair subsequently discussed in his summary. “If everyone is talking about a topic and the chair picks it up, that’s not really being influential,” Prat explains.
The researchers find that after transparency, the less experienced members of the FOMC become more, not less, influential. “We take that as a signal that despite the conformist effect, a discipline effect also happens,” Prat explains. “As a less experienced member, you know that your career depends on your contributions as reflected in the transcripts. Presumably, these members are putting more work into preparing about topics they think are truly important. They might speak up less, but when they do, it’s more important. The chair picks up on that.”
“Greenspan’s concerns about conformism were correct in one sense, but transparency as a whole tends to be a good thing because it increases the informativeness of deliberations,” Prat says. “That suggests the Fed should seek transparency while finding way to reduce the risk of conformism. Perhaps junior members should speak first.”
Andrea Prat is the Richard Paul Richman Professor of Business in the Finance and Economics Division at Columbia Business School.
Read the Research
Hansen,Stephen, Michael McMahon, and Andrea Prat. “Transparency and Deliberation within the FOMC: A Computational Linguistics Approach.” Center for Economic Policy Research, Discussion Paper No. 9994, May 2014.