Narcissus and Social Science

I remember learning in elementary school about the Greek myth of Narcissus — the man who was famed for his beauty and who fell in love with his own reflection. (Actually, I should come clean that I probably learned about Narcissus because he was in the Percy Jackson books, not because I was some Greek mythology buff.) As I remember it, Narcissus sees himself in a pool of water, becomes entranced by his own visage, and then ends up drowning. Looking at Wikipedia suggests there are many variants of the story, but the central idea of a man who was deeply self-focused (indeed, narcissistic) stays constant across the different retellings.

I’ve been thinking about dissertation topics and design, and Narcissus’s story popped into my head as I was reflecting about what I’m hypothesizing. To put it briefly, I want to take Daniel Carpenter’s idea that reputation determines regulatory power and apply it to DOJ’s Voting Section, the division that enforces all federal voting rights laws. The essence of Carpenter’s argument is that regulatory agencies have different audiences, and the agency’s reputation among those audiences is important for its ability to exert power. During budget hearings, the agency wants to demonstrate to Congress that it does its job effectively, that it is filled with apolitical experts, and that its mission is worthwhile. When the President seeks to advance some policy goal, the agency likely wants to further the President’s agenda (or at least, to be able to point to material actions it is taking to do so) so as to avoid getting on the bad side of the President.

Of course, the agency will capitulate to these demands based at least in part on how credible a threat each of these institutions poses. A strong President, in the Neustadtian sense, can credibly signal to his or her adversaries that they should not cross the President, or they will suffer political costs: “When one man shares authority with another, but does not gain or lose his job upon the other’s whim, his willingness to act upon the urging of the other turns on whether he conceives the action right for him. The essence of a President’s persuasive task is to convince such men that what the White House wants of them is what they ought to do for their sake and on their authority.” (For an insightful thread on Neustadt’s theory as applied to Trump, see here.) Similarly, during a congressional hearing, the agency may not be receptive to demands for changed action unless Congress seems like it could actually pass a law to alter the agency’s regulatory authority or to cut its budget. Take, for example, how in the immediate aftermath of the 2008 Financial Crisis, Congress and President Obama were threatening to create a new regulatory agency that would cut authority from the Federal Reserve. Due to the credible threat of legislative change, Carpenter documents how the Federal Reserve began stepping up its enforcement practices — basically (so the theory goes) to demonstrate to Congress and the President that legislative change wouldn’t be necessary.

Agencies, from a constitutional perspective, are subservient to the elected branches of government — Congress and the President are the ultimate overseers. One way of thinking about the reputation theory, then, is that maintaining a powerful reputation acts as a shield against political interference with the agency’s turf. We might expect that really popular agencies — say, the CDC, USPS, or the Census Bureau — are more immune from political interference. Indeed, if we think about the last year, major news stories involved each of these agencies: the coronavirus pandemic, mail-in voting, attempts at putting a citizenship question on the Census. In each example, political interference was costly for the President — Congress quickly responded with oversight and the media shined a spotlight on alleged interference. And arguably, much of the attempted interference failed. Bureaucrats in the Census Bureau pushed back on the political appointees, and the administrative workaround for gathering citizenship data failed. USPS didn’t have major failures in ensuring ballots got to where they needed to. (From a reputational perspective, it’s noteworthy that the USPS Board of Governors literally published an op-ed in USA Today defending the agency’s performance.) CDC perhaps is another story, in the sense that interference was quite effective, but, on the other hand, it’s pretty discrediting to an Administration and its political appointees to have headlines in the New York Times such as: “CDC Testing Guidance Was Published Against Scientists’ Objections.” Reputational analysis helps us understand intra-bureaucratic behavior, such as how line bureaucrats who might care more about establishing a reputation among expert communities (say, scientists) than among political appointees will end up leaking details to point the finger up the food chain for perceived policy failures.

Reputation theory thus also suggests agencies will be responsive to external audiences in certain ways. Bureaucrats at the CDC will care about their credibility in front of scientific audiences; lawyers at DOJ will care about how esteemed legal academics or Big Law practitioners in DC perceive their conduct; securities regulators at the SEC will care about how financiers view their regulations. From an instrumental perspective, building reputational capital enables agencies to get support from important interest groups when they take controversial steps or if they get in the media spotlight. But from a more identity-based perspective, many of these bureaucrats might feel a normative pull toward certain ways of behavior. It is right, they might think, for a lawyer to follow what they believe the law requires rather than what their politically appointed boss tells them to do. It is right, they might say, to listen to scientific consensus and to privilege certain trusted scientific sources over others, even if it points in a politically unfeasible direction.

Personally, I find reputation theory to be extremely persuasive (if a bit difficult to operationalize — how do you measure “reputation”?). I’ve been thinking about why I find reputation theory so persuasive. There are roughly two other broad camps for thinking about regulatory agency behavior: “public interest theory,” which essentially says that agencies do what they do because they think it benefits the public; and “capture theory,” which posits that agencies, politicians, and regulated interests are all self-interested rent-seekers who simply want to extract money (or other kinds of resources) through regulation. (A third theory, “ideational theory,” suggests that ideas and ideology are important factors, but I think ideational theory can’t really explain when action happens — it’s much more relevant for thinking about the form of policy.)

For me, the public interest theory is not well defined (and, indeed, isn’t really a theory — no one calls themselves as “public interest theorist of regulation”). And though it offers a reason for regulation — like the ideational explanation — it doesn’t necessarily help us understand when regulation actually happens. Moreover, though I think regulatory actors are often driven by a pursuit of the public interest, the “public” they are responsive to is going to be shaped by their regulatory mission: EPA staffers probably care much more about “environmental publics” (think environmental advocacy groups or imagined future generations) than about, say, oil lobbyist publics. On the flip side, the Chicago School “capture theory,” as I’ve discussed previously, is a pretty bleak view of regulators, politicians, and regulated interests — it assumes that all actors essentially care only about their own self-interest, and their goal is profit (or at least, “utility”) maximization. These factors are important, but I don’t think they can explain everything. There is an element of ideological interest and so-called “public value” held by bureaucrats: they get into their line of duty not just due to job security or the “perks” of a government job, but because they care about what they do. I think that’s true of teachers, as I’ve written before. I also think that’s true of many bureaucrats in more cause-oriented agencies (say, DOJ’s Civil Rights Division).

In essence, though, a lot of these arguments about why I think reputation theory is a better explanation than public interest or capture theory boil down to: it just seems more correct. I could go through and find some empirical evidence to justify my arguments — I could point to case studies where capture theory or public interest theory fail, and the opposite for reputation theory. But, as Carpenter writes, “To call something a ‘case study’ assumes the goal of extracting universal knowledge about a population from a singular entity.” Just phrasing the endeavor clearly reveals how problem-filled the endeavor is.

Much empirical social science, I’m coming to believe, hinges on the question of representation. What, in other words, is a “representative” sample or case? And how do you know what’s “representative”?

For me, when I think about “government bureaucracy,” I imagine the Voting Section at DOJ. I imagine highly trained lawyers, statisticians, and support staff who believe deeply in their agency’s mission. I imagine people who move between cause-oriented NGOs (the ACLUs, NAACP-LDFs, and MALDEFs of the world) and the government. And so I don’t think it’s surprising that I imagine that the agency is driven by reputational demands from these external actors who ostensibly maintain moral purity and call upon the governmental agency to act in certain cause-oriented ways despite, say, top-down pressure from a presidential administration that might be skeptical of its actions.

Beyond representative samples, I also think the causal logic naturally coheres with how I see the world. A professor once said to me that I care too much about how others see me. I think she was probably right. I think I often feel “reputational pulls,” in the sense that I feel a need to engage with those who I think of as morally or intellectually pure — and that I feel a need to be up to snuff with them. (In hindsight, I think this was probably why getting raked across the coals with our human rights paper last year was so affecting — it was a clear demonstration that I wasn’t, in fact, up to snuff in comparison to an audience that I admired.)

As I’ve been thinking about designing my dissertation, Narcissus feels relevant because I think many of my choices — the topic, the theoretical approach — feel like they’re just reflections of how I see the world. The nice thing about positivist science, of course, is that it’s still anchored to some reality. I’m going to have to find the historical records, the other scholarly accounts, and the interviewees who can confirm or push back on my theory. (Unlike this blog, where I can just say what I want — ha!) The question is whether I can do so in a way that approaches objectivity — or whether I’ll basically be writing to confirm my own bias.

In one of my introductory political science classes, two-and-a-half years ago, my professor summed up a paper’s argument with a neat arrow: “X -> Y.” I raised my hand and said, essentially, that I thought there were times when Y could cause X. She looked quizzically at me and then drew a reverse arrow. At the time, I thought my comment was smart — I had poked a hole in the theory! In hindsight, I actually think that comment suggested a misunderstanding of social science. The world is really complicated, and it’s probably always possible to draw arrows in many different directions. What makes social science interesting is when you can demonstrate a simple relationship that explains much of a phenomenon.

In one of my statistics classes, my professor emphasized the importance of simple models. If you have n data points, then as your statistical model approaches n independent predictor variables, your model will get “better” in the sense that it’ll fit the data closer. But you don’t want to predict your data perfectly. Otherwise, in his words, you’d be playing “connect the dots.” Data are imperfect and noisy in the real world. The tricky work of statistics is to figure out how to throw away the noise in favor of the signal. And usually that means keeping your model simple — that way, even if it fits individual samples poorly, on the whole, it won’t be biased when tested against multiple samples.

My professor’s “X -> Y” causal theory was meant to be easy because we wanted to know how much that theory could predict. In higher education, we love to talk about “complicating” certain ways of thinking. But in this example, I actually think complicating the model paradoxically decreases its relevance. If we think back to the public interest, capture, and reputation theory debates, we can think about what “complicating” means. We can “complicate” the notion of public interest, and posit that the theory is actually that an agency is responsive to many different publics’ interests — say, Congress, the President, courts, technical publics, broader public opinion, and so on. At that point, isn’t a “public interest” theory just the reputational model, except “interest” seems worse defined than “reputation”? What if we “complicate” the notion of regulatory “rent-seeking” behavior so that we include not only monetary costs, but also factors that clearly play a role in self-interest: esteem, sense of identity/self, a feeling of being a do-gooder, praise from audiences, etc.? Doesn’t the capture model basically also become the reputation theory, except we’re trying to psychoanalyze people’s “self-interest” rather than focusing on their reputation? Public interest theory starts from the proposition that regulation is primarily meant to advance “the public interest,” but it’s clear we can stretch that notion to become anything. Similarly so with capture theory — it starts from the idea that everyone is “self-interested,” but if we think about all the kinds of “self-interest,” it becomes quite the elastic theory.

Political scientists love 2x2 charts — we have two different variables that each have two categories, and so we have four different possible configurations. Those four categories are supposed to predict some behavior. It’s always really coarse, almost silly business. But I think they choose these configurations because they’re constraining — and, indeed, might make social science more reliable (in the sense of not being mere reflections of the social scientist’s perspective). They focus attention on specific, relatively consistently measurable attributes, and they seek to show that those alone can lead to some outcomes. Deliberate simplicity, I think, separates the political scientist from the historian, who seeks to revel in the full complexity of any era. There’s value to both, I think. But I don’t think I really understood the value of the political scientist’s endeavor until quite recently.

* * * * *

Rose: I had a call with a friend while walking along the Thames. It was really cold, but it was nice. Also, I had many calls this week, which are always a pleasure.

Lights along the Thames

More snow!

Same

I’m also realizing that I really love reading for my dissertation — it’s fascinating stuff. I think I actually really like the process of academic research. But, alas, the purpose (actually trying to say stuff) isn’t nearly as fun.

History of Ideas Season 2 has started, and, predictably, I’m loving these talks. The latest episode was about Jeremy Bentham, and he talks about this idea of utilitarianism as a kind of “acid” — it cuts away the crap, so that you focus on what’s the real, underlying justification for an action. That idea sounds really simple, but I found it profound — and, as you can probably tell, it undergirds this entire post.

Bud: It’s pretty exciting that Taylor Swift’s first re-record album is coming out soon! I thought her version of “Love Story” was pretty similar to her old version, though, so I’m curious how much the new Fearless will sound different from the old.

Thorn: I feel like I’ve been self-owning a lot recently; I’ve spent a decent amount of this week feeling dumb and sheepish. But (as I was getting at earlier), it’s nice to have things that pull you out of your own mind — whether that’s reading interesting books, listening to podcasts, going for walks, or watching TV.

Future topics:

* LSE’s history — a professor yesterday at my reading group made a comment about how LSE is run like a business school, and how LSE’s orientation toward social science is “utilitarian,” in the sense that it emphasizes applicability rather than knowledge for its own sake. I’ve felt this a little bit and want to explore it a bit more.

* LASA’s history — this is my high school, a magnet school in Austin, Texas. I’ve been interested in education politics and policy for a while, but listening to Nice White Parents (a great podcast by the way) made me feel like I should try to dive down and think a bit more about my own experiences with public education.