Work in Progress


Empirical Strategies to Counter Non-Ignorable Non-Response When Estimating Coronavirus Prevalence

Understanding the prevalence of coronavirus infections is important for understanding and responding to the trajectory of the outbreak. Unfortunately, the tendency to test the sickest people and the variation in testing rates across geographic areas makes it difficult to credibly estimate prevalence. While large-scale randomized testing is ideal, it is very expensive and imperfect compliance can make it vulnerable to non-response bias. This paper explores how to estimate prevalence using first-stage instruments that affect the probability of being tested but not the outcome of the test. First-stage instruments are indispensable when evaluating randomized testing with less than perfect compliance. They also can improve inference based on location-based testing.

Normalizing Covid Positivity Rates across States

State-level Covid positivity rates cannot be directly compared due to differences in testing rates across states. This note presents a normalizing method based on a two-stage model of testing.

Supreme Court

Policy and Legal Preferences of Supreme Court Justices: Change and Continuity, 1950 – 2019

Estimating the preferences of Supreme Court justices is crucial to understanding their behavior and testing theories about the role of the Court in the political system. Ideally, these estimates will be comparable over time, comparable to estimates of other political actors and will allow the possibility that justices value non-ideological factors. This paper updates @Bailey2013 to present Supreme Court ideal points up to 2019 that satisfy these requirements, based on the use of bridging observations across institutions and over time. This paper also presents updated results on the influence of legal factors such as precedent.


Campaign Finance

Do Campaign Contributions Lead to Policies That Favor the Wealthy? An Examination of Taxing and Spending in the American States

Understanding if and how campaign contributions affect policy is important for many policy and normative debates.  In this paper, I use data on gubernatorial spending and state level policy from 1978 to 2000 to assess three competing perspectives on money in politics: the wealth bias perspective, the minimal effects perspective and the neo-pluralist perspective.  The results are most consistent with the neo-pluralist perspective, as increased campaign spending appears to have systematic effects (contradicting the minimal effects thesis) and that these effects are not in the direction of policies benefiting the wealthy (contradicting the wealth bias perspective).  Campaign spending is associated with higher spending in areas where spending has broad public support and is associated with lower spending where that is not the case.