About me
I am a political scientist (in training) at Princeton, where I study the career concerns of political elites using applied econometrics. I will go on the academic job market in 2027.
Career concerns in politics
“Career concerns” are implicit incentives provided by future labor market outcomes. In a career concerns analysis of political behavior, we consider the extent to which political actions can be instrumental toward career goals, especially through the mechanism of reputation-building.
In politics, career concerns are everywhere, alongside more traditionally political dimensions:
- incumbents face the prospect of reelection (or higher office);
- congressional party members can be promoted to positions of influence; and
- a journalist or pundit wants to grow his or her readership/viewership.
My agenda-level question: How responsive is elite behavior to career concerns, and how can we design institutions to maximize incentive alignment between elites and the public? The main obstacle to answering questions like this is that the “treatment” is the unit’s beliefs about the mapping between their actions and some future career outcome. Exogenous/as-if-random variation is rare, and where it does exist its interpretation is ambiguous.
As such, my applied research falls into three categories. First, I study how institutions affect career paths. For example, in one paper I show how term limits increased the tendency of state legislators to run for higher office. This work uses conventional “policy evaluation” methods, such as difference-in-differences analysis.
Second, I study how short-term actions and outcomes affect future career outcomes. This captures the “forward direction” of career concerns, demonstrating that actions do affect career paths. In one study, I examine the effect of incumbents contributing money to their party’s PAC on intra-party advancement. This setting includes the possibility of selection into treatment as well as dynamic effects.
Finally, I use methods from the economics of industrial organization to estimate rational-choice models of elite behavior. Just like methods for estimating ideology from roll calls and campaign contributions, I exploit assumptions about elite decision-making to identify estimands which are theoretically meaningful but model-dependent, in particular those associated with career concerns.
Methods, Done Wrong
While my research is not “methods,” quantitative research methodology is a passion. As such, I have developed some teaching materials, references, and essays related to methods. I am tentatively calling this Methods, Done Wrong.
I have sometimes been asked to recommend a single book on research methods in political science, but there really isn’t one. My sense of the “correct way” to learn methods is to start with basic statistics followed by a course on causal inference (a la Angrist-Pischke or Cunningham). Students then may pursue “special topics” in networks, Bayesian statistics, machine learning, text analysis, experiments, surveys, or qualitative methods. A lot of these are encountered and learned only through reading applied research. Any single book trying to cover all of these things would probably be a waste of everyone’s time.
A problem I encountered as a student and instructor is the universal presupposition of the linear model. Most students (including me) encounter linear regression before causal inference. As a researcher, I don’t think there is nothing wrong with a linear regression. As a student, thinking of causality with a linear model can be deeply misleading. Applied methods texts typically either don’t worry about this common misconception or else reinforce it. My approach is to attack it directly as a poor model of causality, only to later introduce it as a parametric model. Instead, we consider the rich possibilities of a non-parametric framework for thinking about causal relationships.
A second problem is that we political scientists seldom explore the epistemology of our own discipline. Implicitly many political scientists feel there is a hierarchy between “causal” evidence (controlled and natural experiments) and “non-causal” evidence (observational regressions, qualitative research, correlations). However, our theories are ultimately about causes and effects. A better way to structure knowledge is between robust and non-robust evidence for causal claims. But even this is problematic.
A robust claim is one which remains credible even if a handful of models are wrong. It is a valuable property for research in public policy and medicine. Interventions need to be safe and effective regardless of the reason why. On the other hand, evolutionary biologists cannot make interesting claims that are robust to natural selection being false. Physicists cannot make interesting claims that are robust to relativity being false. Indeed, controlled experiments in biology and physics yield data that are only interpretable under theoretical assumptions.
Even if we ultimately want robust evidence, we are rarely taught to question why robustness is valuable. Instead of taking robustness at its word, my approach to methods undermines the robustness criterion by forcing an explicit discussion of theory. This opens up a pathway to relating formal theory, quantitative evidence, and even qualitative evidence.
I hope that by teaching quantitative methods “the wrong way,” this project can help students think more clearly about causal relationships in their theories and take ownership of their methods choices.