The effect of “sunshine” on policy deliberation: The case of the Federal Open Market Committee

Abstract

How does an increase in transparency affect policy deliberation? Increased government transparency is commonly advocated as beneficial to democracy. Others argue that transparency can undermine democratic deliberation by, for example, causing poorer reasoning. We analyze the effect of increased transparency in the case of a rare natural experiment involving the Federal Open Market Committee (FOMC). In 1994 the FOMC began the delayed public release of verbatim meeting transcripts and announced it would release all transcripts of earlier, secret, meetings back into the 1970s. To assess the effect of this change in transparency on deliberation, we develop a measure of an essential aspect of deliberation, the use of reasoned arguments. Our contributions are twofold: we demonstrate a method for measuring deliberative reasoning and we assess how a particular form of transparency affected ongoing deliberation. In a regression model with a variety of controls, we find increased transparency had no independent effect on the use of deliberative reasoning in the FOMC. Of particular interest to deliberative scholars, our model also demonstrates a powerful role for leaders in facilitating deliberation. Further, both increasing participant equality and more frequent expressions of disagreement were associated with greater use of deliberative language.

Keywords:

• Deliberation
• Federal Open Market Committee
• Transparency
• Publicity

Notes

1 This to reverse “a culture of secrecy in Washington.” See http://https://www.whitehouse.gov/open.

2 The terms are sometimes used in similar if not identical ways (e.g., Gutmann & Thompson, 1996; Stasavage, 2004).

3 Chambers also notes the benefits of public deliberation and detrimental effects of conducting it in private.

4 There are more references to deliberation in the November 1993 meeting than in any other from 1978 to 2007.

5 The Fed is the only major central bank releasing transcripts except Japan, which imposes a 10-year lag (Warsh, 2014).

6 See Eijffinger et al. (2015). Relevant replacements included: Hoskins to Jordan; Angell to Yellen; Mullins to Blinder.

7 Schonhardt-Bailey provides evidence for such a pattern on the FOMC (2013).

8 For some, a less bounded view of deliberation might also rely on emotional appeals or personal storytelling, and a focus on reason could penalize those less adept at reasoning. See Bickford (1996); Sanders (1997); Young (2003).

9 Steiner et al. (2004) similarly describe reasoned argumentation.

10 Based on a search for any of the following: deliberation OR deliberative OR “to deliberate” OR “we deliberate” OR deliberates OR deliberating OR deliberated.

11 Across meetings, we compared the median sentence containing reasoning terms to the median sentence for the entire meeting. If the Meade and Stasavage view is correct, the medians would be very different. For our time period they are almost identical.

12 Although most of this work has focused on citizen deliberation, deliberation in political institutions has also received attention, e.g., Bessette (1994); Mucciaroni and Quirk (2006); Steiner et al. (2004).

13 Rosenberg (2007) and Mucciaroni and Quirk (2006) also measure aspects of the deliberative process.

14 They find that publicity interferes with respect. We do not observe change in respect in the FOMC due to sunshine.

15 Completely reasonably so, because that is not her research problem.

16 Grimmer and Stewart (2013) call this a “dictionary method” of text analysis. Golub, Kaya, and Reay (2014) apply a similar method to the FOMC transcripts.

17 We considered dictionaries in the General Inquirer and Diction language programs. We searched online for sources offering guidance on the structure of arguments and inference and for lists of premise and conclusion terms for writers.

18 Schonhardt-Bailey (2013); Steiner et al. (2004).

19 For example, amicus briefs for Metlife v. FSOC, Civil Action No. 1:15-cv-45a.

20 Such terms include affirm, approve, deduce, demonstrate, establish, examine, “seeing that.”

21 No list could be exhaustive, but clusters of other potential terms produced trends similar to what we report below. Of course, subtle aspects of reasoning may not be easily detected by any particular candidate words.

22 Certain sensitive or private information is redacted by the FOMC. Among the limited political science works using these records are Schonhardt-Bailey (2013) and Bailey and Schonhardt-Bailey (2006). Also see Meade and Stasavage (2008), Meade (2005), Thornton (2006), and Chappell, McGregor, Vermilyea (2005).

23 This meeting was the first extensive discussion of transcripts in the FOMC. The date might be put a bit later when the FOMC formally decided to release transcripts with a lag. That would not change our conclusions.

24 Accessed at: http://www.betteratenglish.com.

25 “This is Life with Lisa Ling,” “Parts Unknown,” and “Somebody’s Gotta Do It” Accessed at: http://www.cnn.com.

26 Which of course are of secret conversations. Accessed at: http://www.nixonlibrary.gov/virtuallibrary/tapeexcerpts/.

27 The presidential debates of 1960, 1976, 1980, 1984, 1992, 1996, 2000, 2004, 2008, 2012, accessed at the American Presidency Project. http://www.presidency.ucsb.edu/debates.php.

28 Accessed at: http://intelligencesquaredus.org/debates. Debaters hope to persuade audience members.

29 Accessed at: http://www.cftc.gov/PressRoom/SpeechesTestimony/index.htm.

30 Also see Patrikis (2015), “a vigorous, unscripted debate… became instead choreographed performance, with committee members reading from prepared scripts.” Patrikis left the Fed in 1998.

31 For example: I’m OR I’ll OR I’d OR I’ve OR You’re OR You’ll OR You’d OR You’ve…

32 Use of outside contractors to speed transcript preparation for member review to allow quicker release of meeting minutes was announced May 4, 2004. Of course, “laughter” is not a standardized event.

33 Descriptive statistics for all variables can be found in Appendix B.

34 There is no coefficient estimated for Arthur Burns, whose effect is captured in the intercept term.

35 FOMC members cite a different (but highly correlated) measure of inflation and use many “real economy” indicators. FOMC practices changed over the study period. We believe these measures suffice for present purposes.

36 Accessed at: http://www.nber.org/cycles.html.

37 The measure of disagreement and the measure of inequality mentioned next are both presented in more detail in [redacted]. Tests of human/computer coding accuracy are consistent with an error rate of less than 5%.

38 Our data also show an increase in equality in the Greenspan era, but no separate effect of sunshine.

39 There is no additional effect of sunshine when interacted with Greenspan’s service length.

40 Inequality and disagreement might be affected by sunshine. Models excluding each variable and both together had no change in the main findings. We put each variable on the LHS, leaving the other RHS variables the same. The coefficients on sunshine were not significant in either case. Results available on request. The speaking rates of the specific subset of members spanning the “sunshine moment” were unchanged by that event.