Harmon (2019). When the Fed Speaks: Arguments, Emotions, and the Microfoundations of Institutions

Author:
Derek Harmon – University of Michigan, Ross School of Business

Interviewers:
Micah Rajunov – Boston University, Questrom School of Business
Elisabeth Yang – Yale School of Management

Article link: https://journals.sagepub.com/doi/10.1177/00018392211020660


1. How did you arrive at your theoretical puzzle? You talk about how the Fed itself believes that reinforcing taken-for-granted assumptions is helpful, and yet you question (and show) that this (perhaps taken-for-granted) belief doesn’t hold up.

Harmon: The theoretical puzzle honestly came about through both theorizing and luck. It all started with wanting to test a general proposition posited in an AMR paper that my colleagues (Sandy Green & Tom Goodnight) and I published in 2015. That paper theorized that we can understand the stability of our institutions by looking at the arguments people use. Likening an institution to a “game,” we distinguished between when people argue within the rules of the game from when they argue about the rules of the game. We theorized that when you observe more discussion within the rules of the game, the taken-for-granted assumptions of the institution are being implicitly reinforced and so the institution is maintaining itself. In contrast, when you observe more discussion about the rules of the game, the assumptions of the institution are being noticed, which makes the foundations of that institution more open to change.

Building on this theory paper, I developed a hypothesis that I thought one could empirically test—the more we observe a leader talking about the assumptions of an institution, the more likely it is that institution could change. Given my interest in studying the effects of language on collective audiences, I knew that testing such a proposition using an event study might be useful. So I started looking into research on empirical measures for collective change or collective uncertainty, and came across the VIX Index as the premier measure for market uncertainty used by finance scholars. With this as a potential outcome measure, I asked myself—what singular speech act had the persuasive power to move the entire VIX? The answer I came up with was the Federal Reserve.

Although this was how the paper developed, the setting I chose frankly contained some luck that made this theoretical puzzle more interesting, at least in my opinion. Recall that the hypothesis I wanted to test was simply about a leader talking about the assumptions of an institution. But note that this includes situations when they talk about changing those assumptions, which isn’t as interesting. That is, it seems obvious that when a leader talks about changing the foundations of an institution, people become more uncertain. However, another manifestation of this hypothesis that I thought was less obvious and more interesting was that this same effect can emerge when a leader is simply trying to reinforce these taken-for-granted assumptions. As it turns out, this is precisely what these Fed speeches were trying to do. Thus, the theoretical puzzle in this paper emerged from a combination of theorizing and finding an empirical setting that enabled me to test this theory in a very interesting way.

2. Your framing seems to bridge an institutional theory approach with a social constructionist / interactionist approach (Goffman & Garfinkel, with hints of performativity even). Was this part of your original framing, and how difficult was it to integrate these two perspectives?

Harmon: This is a neat question because my first round submission to ASQ was explicitly social constructionist. In the introduction of the original submission, I wrote—“In this study, I take a social constructionist view of strategic communication in markets.” Since that first submission, however, my thinking has developed. I still believe the paper bumps up against a social constructionist or interactionist perspective, but that formulation is too broad in my opinion. Instead, what I see the paper doing is archaeologically excavating one of the key thinkers that I believe sits at the epicenter of institutional theory—Harold Garfinkel—and integrating his micro-level perspective, which was applied to everyday interactions, to macro-institutions. Garfinkel describes how we all have background expectancies that enable us to move along in life and, when they are revealed, create uncertainty. This was the basis for his breaching experiments, where he asked his grad students to say or do things that are unusual given the normal course of things in social interaction. Lynn Zucker, who has inspired a lot of my work, notes Garfinkel’s influence to institutional theory in her 1977 ASR, but this point has since been largely ignored. If one wanted to point to a concrete micro-foundational element underlying institutional theory today, Garfinkel would be one of the first people I would encourage people to (re)read.

3. Often in qualitative research we are interpreting what the content means, as well as more discursive elements such as the intentions of the speaker. We don’t usually think of quantitative research as being interpretive. So we were intrigued by your inclusion of emotions in your theory and especially how you theorized emotions as “interpretative contexts” that shape market participants’ interpretation. Were emotions initially part of your hypotheses, or did these emerge more inductively throughout the process?

Harmon: I think there are two great questions here—the first being about the intersection of interpretative and quantitative research, and the second being about emotion as an interpretative context.

Regarding the first question, from my perspective, all research is interpretative. I think it’s easy to see this in qualitative research, since the methods require authors to talk about these interpretations as part of their data. However, I think the same standards should be considered for text-based quantitative research too. For example, when thinking about the construct validity of a text-based measure, we are making assumptions about what is meant by these words used in a particular context. Similarly, when interpreting quantitative results, we are making assumptions about how such actors might be interpreting these words in that specific setting. This is why much of my work using language uses both qualitative and quantitative approaches. I think it’s essential to qualitatively read and understand the texts in one’s setting to understand how they are used and interpreted in situ. This is because when the constructs are developed and, in my case, converted into a number for inclusion in a regression analysis, we can be more confident in a stable construct meaning and interpretation of findings. 

Regarding the second question, emotions were always part of the paper. In fact, my original submission to ASQ contained the same three hypotheses that are in the final version. However, the reviewers and AE during the review process helped me hone the role emotions were playing. In the original submission, I conceptualized emotions as a signal that financial markets paid attention to, and argued that the market triangulates these signals and comes to an overall evaluation. However, I think the way I was measuring emotions (i.e., using holistic word dictionaries) made it less likely that such emotions were a clear/sharp signal per se, but instead, more of a broad frame or context within which the other content (i.e., the reaffirming of taken-for-granted assumptions) was being interpreted. From my perspective, this is a nice example of how the quantitative measurement of a text-based variable directly informed the theory used to describe its effects.

4. With any paper, a deep understanding of the context is important. In this case we were particularly impressed by your in-depth understanding of the study context. We were curious how this affected the research process (e.g., coming up with a theoretical puzzle, including control variables and interviews to support them, supplemental analysis, etc.)?

Harmon: A deep understanding of an empirical context can be achieved in many ways. I think people tend to assume interviews are the most direct method to do this, but I think it depends on what it is one is trying to understand. For example, to develop an understanding of central banking and the history of their public communications, I read countless biographies and books about the history of the institution, along with every single speech ever delivered by any Fed official. I think this approach was far more useful to get an understanding of the specialized, technocratic discourse used by central bankers than any interview I could have conducted.

Having said that, I did conduct several key interviews that increased my confidence in my empirical findings. Perhaps the most useful were the ones with Ben Bernanke and Janet Yellen, both of whom have been wonderful and more than happy to talk. The key empirical issue with my research design was omitted variable bias—the idea that some variable not controlled for was leading the Fed to talk more about their assumptions and simultaneously increasing market uncertainty. To evaluate the likelihood that this was occurring, I needed to understand the speech-writing process (e.g., when speeches were written and edited, how topics were chosen, and what they were thinking about when developing these speeches). Since factors surrounding the day of the speech were most likely driving my outcome variable, I wanted to know if the speeches were also being edited very close to the speech date too, making endogeneity a bigger concern. It turns out, however, that most speeches were written one to six months in advance, making it less likely that an omitted variable could account for both effects.

But I also learned other useful things in these interviews I hadn’t anticipated. For example, even though my paper was about the market effects of the Fed talking about assumptions, I didn’t have a clear explanation for what led them to talk about these assumptions in the first place. Without at least a reasonable explanation, I think the lingering worry about endogeneity would be higher (even though it would never go away). During one interview, I recall Bernanke saying to me, “Well, you know what drives most of that sort of talk, right? It’s the local audience we’re speaking to.” The light bulb went on for me. Of course, the Fed knows that the broader markets are listening, but they are also tailoring most of their speeches to the audience that they are physically talking in front of. This insight not only led me to control for the local audience in my analyses, but also led me to confirm that yes, the local audience is indeed a strong predictor of the likelihood of reaffirming these assumptions.

5. How was the process of turning your dissertation into an ASQ article? When did you know when to submit to the journal?

Harmon: I think I was fortunate in several ways. First, my advisor, Peer Fiss, allowed me to develop my dissertation with this paper as a standalone chapter that was basically written as a journal article. This set the stage for a seamless transition between my dissertation and job market paper, which minimized the amount of time spent adjusting things for bureaucratic reasons. Second, I was also really fortunate because this was my job market paper, which meant that I received just an incredible amount of terrific feedback from really smart people during job talks. I don’t think I have ever been more confident, or will be so confident ever again, about every aspect of a paper. That didn’t mean I thought the paper was bulletproof (that doesn’t exist), but I had answered every question anyone had ever asked about these ideas, from both a theoretical and empirical perspective, and was comfortable with the answers. Given these two factors, I finished my job market process in February of 2016, revised the paper one last time, and submitted the manuscript to ASQ that June.

I think it’s important to note, however, that this process was unlike any paper I’ve submitted since. Given that this was my job market paper, I had explored every single possibility and critique I could think of. Yet this level of exhaustiveness is simply not practical or often possible as one develops a larger pipeline of projects, has more teaching and service responsibilities, and faces all the other commitments that go along with being a faculty member. As a result, over the years my “bar of perfection” has lowered and I will now submit things more quickly.

6. From your conversations with colleagues, policymakers, and the media about this paper, has your findings ever been misinterpreted?

Harmon: One thing that can get misinterpreted is the level of analysis at which my findings are occurring. What I theorize and find is that reaffirming an institution’s taken-for-granted assumptions increases market uncertainty—that is, collective-level uncertainty. The way this is measured is via the VIX Index, which you can think of as the variation across option traders’ pricing. Thus, reaffirming these assumptions increases the variance across individual trader’s pricing decisions, thereby increasing uncertainty in the overall market.

Note that this finding says nothing about whether the option traders themselves are more uncertain. I think it’s easy to make this mistake, however, because it’s natural to think about how this sort of talk might also increase our own uncertainty. I was curious about this, so I decided to conduct an experiment to test the idea. I used one of Bernanke’s actual speeches as a sample text, and manipulated the degree to which that speech reaffirmed assumptions (low vs. high). Along with some additional background information, I asked participants 1) what they would price an option at and 2) how uncertain they were about that pricing. When talk about these assumptions was high, not only was the standard deviation of the option prices significantly higher (i.e., higher collective uncertainty), but the individual participants in this condition were also more uncertain themselves (i.e., higher individual uncertainty). I didn’t report this experiment in the final version of the paper, but I think it’s useful finding. It suggests that individual-level uncertainty in this setting might undergird market uncertainty, and that the findings may be more general than I originally claimed.

7. Your paper utilizes a novel argumentation theory by Stephen Toulmin. How did you first become interested in these ideas?

Harmon: This is a story I love to share with students because it highlights a practice that I’ve found helps me stay motivated in my career. As a PhD student at USC, I went to the library every Friday afternoon and sat in the section where they had all the physical academic journals across all disciplines. I walked around and randomly selected 10 journal issues, sat down with a cup of coffee, and sifted through them. I mainly read the table of contents to see what the journal was about, but if any title piqued my curiosity, I went to its abstract and kept reading. One day I picked up an issue of some British journal of communication and it had a paper using the Toulmin model of argument to study courtroom argumentation. The simplicity of this model fascinated me, so I checked out all of Toulmin’s books and started reading them. At the time, Sandy Green (who was an assistant professor at USC) saw me reading one of these books and said “Hey, if you want to know more about Toulmin’s work, talk to Tom Goodnight.” Tom was a professor of communications at USC and was good friends with Toulmin when he—believe it or not—worked at USC (before he passed away). I promptly talked with Tom, corralled him and Sandy into writing the AMR paper I mentioned at the beginning of this blog post, and the rest is history…

The reason I bring this story up is because I have always found it useful to build into my schedule planned time for exploration. Far too often, both as a student and junior faculty, I get lost in my current projects. Getting lost can sometimes be important, of course, but these projects often seem to just drag on and I sort of lose my broader motivation. Developing a practice where I have planned or scheduled exploration forces me to consistently open my mind (and excitement) back up. Not only has this helped me find new ideas to study, but it has placed in perspective the projects I’m currently working on, helping me get through some of those tough days we all naturally face.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: