About

I work on prediction and inference in politics and policy. My technical focus is on measurement problems involving information extraction, entity linkage, and latent variables.

I finished a Ph.D. in political science at MIT in fall 2018. Previously, I studied public policy at NYU for an M.P.A., and I built digital campaigns for issue-advocacy groups.

Code

I write Python and SQL for work. Graduate training was in R. I use JS for D3, Dash, or Shiny.

dgo implements dynamic estimation of latent characteristics using multilevel Bayesian models. It estimates subpopulation means from individual-level item response data. Written in R with Devin Caughey and Christopher Warshaw, across a number of projects, including our paper below.

I wrote qsurvey while staffing MIT PERL. It’s a toolkit for R users working with the Qualtrics platform: designing surveys and analyzing results.

From my time at the MIT Election Lab, you can find one example of a Python package for data management and another generating consistent documentation through Jinja templating.

Publications

“The Ideological Nationalization of Partisan Subconstituencies in the American States,” with Devin Caughey and Christopher Warshaw, in Public Choice:

Since the mid-twentieth century, elite political behavior in the United States has become much more nationalized. In Congress, for example, within-party geographic cleavages have declined, roll-call voting has become more one-dimensional, and Democrats and Republicans have diverged along this main dimension of national partisan conflict. The existing literature finds that citizens have only weakly and belatedly mimicked elite trends. We show, however, that a different picture emerges if we focus not on individual citizens, but on the aggregate characteristics of geographic constituencies.

Using biennial estimates of the economic, racial, and social policy liberalism of the average Democrat and Republican in each state over the past six decades, we demonstrate a surprisingly close correspondence between mass and elite trends. Specifically, we find that: (1) ideological divergence between Democrats and Republicans has widened dramatically within each domain, just as it has in Congress; (2) ideological variation across senators’ partisan subconstituencies is now explained almost completely by party rather than state, closely tracking trends in the Senate; and (3) economic, racial, and social liberalism have become highly correlated across partisan subconstituencies, just as they have across members of Congress. Overall, our findings contradict the reigning consensus that polarization in Congress has proceeded much more rapidly and extensively than polarization in the mass public.

“Voter ID Laws: A View from the Public,” with Paul Gronke, William D. Hicks, Seth C. McKee, and Charles Stewart III, in Social Science Quarterly:

Voter identification (ID) emerged as a policy innovation during elite debate over lessons learned from the 2000 election, but as proposals for new strict ID laws have swept the states, it has resonated among the mass public as well. As we show, although there were clearly partisan differences over the necessity and form of photo ID laws from the beginning, the public began the decade of the 2000s by and large favorably disposed toward such laws. This favorable disposition included all relevant demographic groups. However, the popular consensus on voter ID laws has frayed. We document the decline of this consensus and explore the ways that partisanship now shapes two views of strict photo ID laws.

For Republicans, favorability about the effects of photo ID laws appears to have attained the status of party orthodoxy. For Democrats, attitudes about the effects of photo ID laws are much more variable. The greater diversity in Democratic opinion resembles the political reality of a party that is more heterogeneous than its opposition.


© 2019 James Dunham, design credit ankitsultana