Reasoning about Interference Between Units (with Jake Bowers and Costas Panagopoulos, forthcoming in Political Analysis)
If an experimental treatment is experienced by both treated and control group units, tests of hypotheses about causal effects may be difficult to conceptualize let alone execute. In this paper, we show how counterfactual causal models may be written and tested when theories suggest spillover or other network-based interference among experimental units. We show that the “no interference” assumption need not constrain scholars who have interesting questions about interference. We offer researchers the ability to model theories about how treatment given to some units may come to influence outcomes for other units. We further show how to test hypotheses about these causal effects, and we provide tools to enable researchers to assess the operating characteristics of their tests given their own models, designs, test statistics, and data. The conceptual and methodological framework we develop here is particularly applicable to social networks, but may be usefully deployed whenever a researcher wonders about interference between units. Interference between units need not be an untestable assumption; instead, interference is an opportunity to ask meaningful questions about theoretically interesting phenomena.
RItools: Flexible randomization inference in R (PDF) Demonstrates the use of the RItools" package for R to create and test hypotheses using randomization inference. Users can describe the randomizatino procedure, their model of effects, and test statistics, and the function RItest provides confidence intervals an p-values. The paper provides background on randomization inference and applied examples using RItools.
Returning to the Cradle of Democracy: Citizen responses under election and sortition. (PDF) The hallmark of modern democracies is the competitive election. This institution is seen as the primary connection between leaders and the population. This has not always been the case. Sortition, the random selection of leaders from the population, served as the primary institution of democracy in ancient Athens. How would citizens in a modern democracy react to the use of sortition to select leaders? This study employs a survey experiment in which subjects read about a local development grant, overseen by either an elected or randomly selected committee. I find that sortition encourages more citizens to seek leadership positions, though other forms of participation remain unchanged. I also find that despite a stated preference for election, subjects see the two committees as equally capable and responsible, even when confronted with corrupt acts and closed door meetings.
RISE: Randomization Inference for Spillover Effects (also available as source documents) Presented at ACIC 2011. This work speaks to two audiences. First, experimenters afraid that interference between experimental units violates the assumptions of their statistical tools (specifically, the Stable Unit Treatment Value Assumption, or SUTVA for short). Second, this research offers new tools to researchers studying the effects of networks themselves. This introduces the basics of the work and demonstrates the technique on some simulated data, for which we know the truth. It also demonstrates new software for flexible randomization inference.
Collaboration for Social Scientists, or Software is the Easy Part (with Paul F. Testa and Nils B. Weidmann) The Poltical Methodologist (2011) 18(2).
In this article, we consider how to improve two different modes of collaboration: synchronous and asynchronous. When working synchronously, contributors focus on the same portions of the research at the same time. Of course, virtually any research project will require collaborators to spend time working on either different portions of the project or working on the same sections but at different times. We label this form of collaboration asynchronous. Asynchronous collaboration requires more careful attention to dividing labor, and we spend more time providing software solutions in this domain. These suggestions are based on what has worked for us. These suggestions are grounded in experience, and we think they are useful techniques for any team to adopt. We have also found that software is the easy part of any collaboration while the personal and intellectual parts of collaboration are both more difficult and more fulfilling than playing with software tools. Hopefully, adopting some of these techniques may help your team get past technical details faster and down to the real business of producing research.
A simulation study of knowledge (with John Ostrowski). In this paper, we argue that simulation studies of knowledge (in which researchers construct a counterfactual world of fully informed citizens) should employ models that are the most accurate at predicting responses in observed data. Previous studies used only linear models, while we consider a variety of linear and non-linear machine learning techniques. While the best linear model is as accurate as the best non-linear model, the non-linear models predict relatively little change under a fully informed population, indicating that political knowledge may not be as important to attitude formation as we previously thought.
ACE in the hole: a constructive critique of classical twins studies. While there has been a steady stream of research linking genetic predispositions to political behavior, this research has not been well integrated into the broader political behavior literature. In part, this is because of the model most frequently employed in classical twin studies. The use of the so-called “ACE” model requires strong assumptions and does not directly engage the models and outcomes of other political science research. This working paper attempts to layout a path where by the logic of the natural experiment embodied in twin studies can be used to simplify the analysis in a way that more directly engages traditional political science studies. A version of the paper was presented as a poster at Polmeth 2011.
Finding common interests in the U.S. Senate. An introduction and sketch of an analysis of the U.S. Senate to detect groups of Senators sharing common interests. Previous studies have used hierarchical methods of grouping senators, which will inevitably find party to be the high level grouping factor, excluding opportunities for shared interests across party lines. This paper develops a technique from the educational assessment literature to find common ``skills’’ using cosponsorship activities as the unit of analysis.
Stochastic vs. Deterministic Attitudes. A research design proposal for testing whether survey responses are given deterministically or stochastically. That is, do people have fixed positions at the time of the survey or do they sample from a distribution to produce a response? This design addresses the difficulty of adjudicating between these two positions in a healthy population by repeatedly surveying patients suffering from short-term memory loss.
“With 70 percent of precincts reporting, we are confident in calling Norm Coleman as the winner in the Minnesota Senate election,” the local news reporter concluded. It was November 5th, 2002, Election Day, and I was in the basement of the in St. Paul Hilton hotel. I was surrounded by fellow staffers and volunteers who had rallied around Walter Mondale, after the death of Sen. Paul Wellstone. Now, as we watched the map on the television screen, the picture was clear: Despite a huge swell of support in urban areas, Republican Norm Coleman was riding a tide of support in the outer suburbs to victory.
Returning to my junior year of college a few months later, I was convinced that the political beliefs of Americans were becoming increasing conservative. As I started to investigate this phenomenon, I became confident that this was occurring largely because of the changing nature of the places where people live: as more and more people moved into suburbs, and exurbs, I reasoned, the economics and social networks of those places would lead almost inexorably toward conservatism. At the same time, people were leaving suburbs for a renaissance in city centers. This fascination became my thesis: I spent 8 months tracking down exactly how American cities had changed, hoping to find explanation for my experience campaigning in Minnesota.
I found that on one hand, suburban areas and medium sized metropolises were experiencing rapid growth in housing stock and commercial space, and people in those places also became more conservative over time . On the other hand, urban cores also grew (reversing a 20 year trend), and, unlike suburban areas of growth, people in the newly growing central cities were increasingly liberal. At first I saw these two trends as yet more evidence of the effect of location on ideology. As I investigated the change in Democratic and Republican support in Florida across the 2000 and 2004 presidential elections, however, I was unable to find a significant increase in conservative support in exurban counties nor a significant increase in liberal support in urban counties. While this particular analysis was inconclusive, future research has the opportunity to provide a better understanding of ongoing social and political shifts, potential benefits and harms of targeted redistricting, and implications for representation of demographically changing districts.
I also have important experience outside of academia. As a professional programmer and campaign manager, I led teams in designing dazzling, interactive websites, developing new dynamic databases, and implementing new approaches to mobilizing voters. I worked independently and with software development teams across the globe. On one project, I had colleagues in Russia, England, and Australia. It was critical that I effectively communicate problems, possible solutions, my actions, and new ideas across languages, cultures, and time-zones. In addition to communicating research findings in person and in writing, I am a frequent presenter at local and national software developer groups. Most recently, I presented an introduction to the Scheme programming language to a local group of programming professionals, demonstrating language features using a software package I wrote myself and released under an open source license.
As a web programmer, much of my work was for non-profits and political campaigns. For example, I worked with Chicago Artists Resource to redesign their website to be more user friendly, easier to navigate, and better promote their exciting programs. Working with Planned Parenthood, I developed on-line phone banking programs, allowing volunteers to reach out to more people. I also developed polling location look up services, allowing voters to find out where to cast a ballot on election day as well as find Planned Parenthood endorsed candidates. Volunteering my time for Episcopal Group Homes, a supervised living center for adults with developmental disabilities, I am developing a comprehensive website to distribute news to families and donors, solicit donations, and interact with state and county social work personnel. Virtually all the software I have written has been released for free under an open source license, allowing other groups to pick up and use these same tools for free. The common thread of these experiences is marrying technology with community outreach.
In the summer prior to entering graduate school, I joined the research team of Dr. Ben Hansen of the University of Michigan and Dr. Jake Bowers of the University of Illinois. In this work, I supported Drs. Hansen and Bowers in their final preparation of a Journal of the American Statistical Association paper on randomization inference, as applied to Gerber and Green’s 1998 Get-Out-The-Vote field experiment. Specifically, I organized and prepared data and documentation (including writing several supporting computer programs) so that any researcher could quickly reproduce this paper from the ground up.
Through this work I am a coauthor of RItool, a package for R, which provides a framework for using randomization inference. My most important work on that project has been to integrate a sparse matrix library. The overall test statistic proposed by Bowers and Hansen requires many independent linear regressions, best represented as a very large matrix multiplication problem. One of my more recent contributions has been to extend the package to Stata, allowing this work to be used by more scholars. Currently, I am implementing ways to extend RItools to small data sets. This will allow researchers working with smaller samples to analyze their results without assumptions that require large samples, all in an easy to use package. Our primary goal with RItools is to create a package that others can use to strengthen their research.
As a result of my work with Drs. Bowers and Hansen, I was invited to attend PolMeth, the yearly conference of the Political Methodology Section of the APSA. As I was not a graduate student proper, this was a special honor. It was an opportunity to meet fellow political scientists, hear about the latest methodological research, and learn more about the process of presenting research. The presentations at the conference confirmed my intuition: there is a need for computationally oriented political scientists to provide novel substantive and methodological research. As Christopher Achen observed to me, these tasks are linked: new research demands new methods.