1

Paper

Data Mining for Theorists Kenkel, Brenton Signorino, Curtis Uploaded 07-26-2011 Keywords empirical implications of theoretical models basis regression adaptive lasso bootstrap functional form misspecification Abstract Among those interested in statistically testing formal models,two approaches dominate. The structural estimation approach derives a structural probability model based on the formal model and then estimates parameters associated with that model. The reduced-form approach generally applies off-the-shelf techniques---such as OLS, logit, or probit---to test whether the independent variables are related to a decision variable according to the comparative statics predictions. We provide a new statistical method for the comparative statics approach. The decision variable of interest is modeled as a polynomial function of the available covariates, which allows for the nonmonotonic and interactive relationships commonly found in strategic choice data. We use the adaptive lasso to reduce the number of parameters and prevent overfitting, and we obtain measures of uncertainty via the nonparametric bootstrap. The method is "data mining" because the aim is to discover complex relationships in data without imposing a particular structure,but "for theorists" in that it was developed specifically to deal with the peculiar features of data on strategic choice. Using a Monte Carlo simulation, we show that the method handily outperforms other non-structural techniques in estimating a nonmonotonic relationship from strategic choice data.

2

Paper

Copula Functions for Approval Ratings and Endogenous Political Events Quiroz Flores, Alejandro Uploaded 07-27-2011 Keywords copula functions approval ratings endogenous bivariate distribution Abstract Empirical investigations of US presidential approval ratings often control for the exogenous materialization and duration of particular events such as wars or political scandals. Although the materialization of some of these events might be exogenous, their duration is not. Once an event takes place, a President has sufficient power to reduce its duration if the event reduces approval ratings, or increase its duration if the event is politically beneficial. In other words, the duration of these events is endogenous. In order to address this potential problem, this paper uses a copula function to jointly estimate presidential approval ratings (with a Normal distribution) and the duration of significant political events (with a Weibull distribution). The estimation of this bivariate Normal-Weibull distribution will help us test whether and to what extent American presidents manipulate political events in order to increase their approval ratings. Estimation results from the bivariate Normal-Weibull distribution suggest that there is a positive, two-way association between presidential approval ratings and the duration of political events.

3

Paper

Comparative Effectiveness of Matching Methods for Causal Inference King, Gary Nielsen, Richard Uploaded 07-27-2011 Keywords Causal Inference Matching Propensity Scores Abstract Matching is an increasingly popular method of causal inference in observational data, but applications of it are often poorly executed. We address this problem by providing a graphical approach for choosing among the numerous possible matching solutions generated by three methods: the venerable "Mahalanobis Distance Matching" (MDM), the commonly used "Propensity Score Matching" (PSM), and a newer approach called "Coarsened Exact Matching" (CEM). In the process of using our approach, we also discover that PSM often approximates random matching, both in real applications and in data simulated by the processes for which PSM theory was designed. Moreover, contrary to conventional wisdom, random matching is not benign: it (and thus PSM) can degrade inferences relative to not matching at all. We find that MDM and CEM do not have this problem, and in practice CEM usually outperforms the other two approaches. However, with our comparative graphical approach, focus is on choosing a matching solution for a particular application, which is what may improve inferences, rather than the particular method used to generate it. The easyto- follow procedures we describe thus enable researchers to improve the application of any one of these methods, to choose among them and from the various matching solutions generated by any one method, and ultimately to increase the validity and extent of causal information extracted from their data. Link to paper: http://gking.harvard.edu/files/psparadox.pdf

4

Paper

Racing Horses: Constructing and Evaluating Forecasts in Political Science Brandt, Patrick Freeman, John R. Schrodt, Philip Uploaded 07-27-2011 Keywords forecasting political conflict scoring rules model training forecast density verification rank histogram probability integral transform Abstract We review methods for forecast evaluations and how they can be used in political sciences. We examine how forecast densities are more useful summaries of forecasted variables than point metrics. We also cover how continuous rank probability scores, probability integral transforms, and verification rank histograms can be used to calibrate and evaluate forecast performance. Finally, we present two illustrations, one a simulation and the other a comparison of forecasting models for the China-Taiwan (cross-straits) conflict.

5

Paper

Multivariate Matching Methods That are Monotonic Imbalance Bounding King, Gary Iacus, Stefano Porro, Giuseppe Uploaded 01-03-2011 Keywords Matching CEM Causal Inference Abstract We introduce a new ``Monotonic Imbalance Bounding'' (MIB) class of matching methods for causal inference with a surprisingly large number of attractive statistical properties. MIB generalizes and extends in several new directions the only existing class, ``Equal Percent Bias Reducing'' (EPBR), which is designed to satisfy weaker properties and only in expectation. We also offer strategies to obtain specific members of the MIB class, and analyze in more detail a member of this class, called Coarsened Exact Matching, whose properties we analyze from this new perspective. We offer a variety of analytical results and numerical simulations that demonstrate how members of the MIB class can dramatically improve inferences relative to EPBR-based matching methods.

6

Paper

The Split Population Logit (SPopLogit): Modeling Measurement Bias in Binary Data Beger, Andreas DeMeritt, Jacqueline Hwang, Wonjae Moore, Will Uploaded 02-28-2011 Keywords split populations binary data measurement error bias zero inflation substantive inference Abstract Researchers frequently face applied situations where their measurement of a binary outcome suffers from bias. Social desirability bias in survey work is the most widely appreciated circumstance, but the strategic incentives of human beings similarly induce bias in many measures outside of survey research (e.g., whether the absence of an armed attack indicates a country's satisfaction with the status quo or a calculation that the likely costs of war outweigh the likely benefits). In these circumstances the data we are able to observe do not reflect the distribution we wish to observe. This study introduces a statistical model that permits researchers to model the process that produces the bias, the split population logit (SPopLogit) model. It further presents a Monte Carlo simulation that demonstrates the effectiveness of the SPopLogit model, and then re-analyzes a study of sexual infidelity to illustrate the richness of the quantities of (empirical and theoretical) interest that can be estimated with the model. {\tt Stata ado} files that can be used to invoke the SPopLogit, as well as batch files that illustrate how to simulate commonly reported quantities of interest, are available for download from the WWW. The authors close by briefly identifying just a few of the many types of research projects that will benefit from abandoning logit and probit models in favor of the SPopLogit. NOTE: Files to implement SPopLogit and generate quantities of interest, as well as replication files for our Monte Carlo simulations and substantive application, are available at http://andybeger.wordpress.com/2011/02/03/split-population-logit/

7

Paper

Beyond LATE: A Simple Method for Recovering Sample Average Treatment Effects Aronow, Peter Sovey, Allison Uploaded 03-24-2011 Keywords compliance score instrumental variables LATE average treatment effect causal inference Abstract Political scientists frequently use instrumental variables estimators to estimate the Local Average Treatment Effect (LATE), or the average treatment effect among those who comply with treatment assignment. However, the LATE is often not the causal estimand of interest; researchers may instead be interested in the Sample Average Treatment Effect (SATE), or the average treatment effect for the entire sample. We first introduce the compliance score, a pre-treatment covariate that reflects a unit's probability of treatment compliance, to researchers in political science. We posit a maximum likelihood estimation technique for predicting compliance scores even in the presence of two-sided non-compliance. We then develop a new technique, inverse compliance score weighting, that, in conjunction with a standard IV estimator, will allow researchers to easily estimate the SATE. Finally, we estimate both the LATE and SATE for a randomized experiment designed to measure the effects of media exposure and reach striking substantive conclusions.

8

Paper

Agnostic Notes on Regression Adjustments to Experimental Data: Reexamining Freedman's Critique Lin, Winston Uploaded 09-02-2011 Keywords Covariate adjustment Randomization inference Neyman's repeated sampling approach Sandwich estimator Social experiments Abstract Freedman [Adv. in Appl. Math. 40 (2008a) 180–193; Ann. Appl. Stat. (2008b) 2 176–196] critiqued OLS regression adjustment of estimated treatment effects in randomized experiments, using Neyman’s model for randomization inference. This paper argues that in sufficiently large samples, the statistical problems he raised are either minor or easily fixed. OLS adjustment improves or does not hurt asymptotic precision when the regression includes a full set of treatment-covariate interactions. Asymptotically valid confidence intervals can be constructed with the Huber-White sandwich standard error estimator. Even the traditional OLS adjustment has benign large-sample properties when subjects are randomly assigned to two groups of equal size. The strongest reasons to support Freedman’s preference for unadjusted estimates are transparency and the dangers of specification search.

9

Paper

Modeling Electoral Coordination: Voters, Parties and Legislative Lists in Uruguay Levin, Ines Katz, Gabriel Uploaded 04-20-2011 Keywords electoral coordination number of parties Bayesian estimation multilevel modeling strategic voting Abstract During each electoral period, the strategic interaction between voters and political elites determines the number of viable candidates in a district. In this paper, we implement a hierarchical seemingly unrelated regression model to explain electoral coordination at the district level in Uruguay as a function of district magnitude, previous electoral outcomes and electoral regime. Elections in this country are particularly useful to test for institutional effects on the coordination process due to the large variations in district magnitude, to the simultaneity of presidential and legislative races held under different rules, and to the reforms implemented during the period under consideration. We find that district magnitude and electoral history heuristics have substantial effects on the number of competing and voted-for parties and lists. Our modeling approach uncovers important interaction-effects between the demand and supply side of the political market that were often overlooked in previous research.

10

Poster

A New Approach to the Study of Political Participation Levin, Ines Uploaded 07-26-2011 Keywords political participation mixture modeling Abstract In this poster I present a new statistical procedure for the study of political participation based on mixture modeling and simultaneous consideration of involvement in multiple political activities. In this model, the relationship between underlying utilities and participation probabilities is regulated by a parameter that captures individual propensities toward political participation. First, I present the results of a simulation study used to evaluate the properties of the methodology. Then, I present the results of an application to survey data from the 1990 American Citizen Participation Study (Verba, Schlozman, and Brady, 1995), where the method is used to test hypotheses regarding the impact of resources and civic skills on political participation. Lastly, I propose a new dual-process explanation of the way participation decisions are made, which is consistent with empirical results, and is rooted in recent findings in cognitive psychology.

11

Poster

Bounds for Logistic Regression Coefficients with Nonignorable Missing Outcomes Kenkel, Brenton Uploaded 07-27-2011 Keywords partial identification bounds missing data measurement error Abstract I develop a new method to estimate logistic regression coefficients when there is nonignorable missingness or measurement error in the outcome variable. The estimator finds the set of all coefficient vectors that could be obtained under any assumption about the missing outcomes.

12

Poster

Latent Variables and Rolling Panels: A New Approach to Modeling Campaign Effects Therriault, Andrew Uploaded 07-27-2011 Keywords panel data latent variables campaign effects advertising persuasion public opinion voters elections Abstract Election panels which reinterview participants in rolling cross-sectional surveys offer new opportunities to study campaign effects, but also present unique methodological challenges. I develop an original approach to modeling this data, and demonstrate how its application leads to much stronger evidence for informing and persuasion effects from campaign ads than that found in existing research

13

Poster

WhentheSTARs Align:What IOs AreMoreConducivetoDemocratization Chyzh, Olga Uploaded 07-31-2011 Keywords democracy spatial dependence diffusion international organization spatial regression m-STAR Abstract The scholars of democracy have long noted the tendency of democratic states to cluster in time and space. While most theoretical explanations of this phenomenon posit causal mechanisms related to spatial interdependence (e.g. diffusion, socialization), very few studies have conducted adequate empirical tests of these theories. This methodological oversight is due both to the scarcity of available statistical techniques that allow for testing these types of effects, as well as to the methodological sophistication of the existing techniques. Yet the value of empirical inferences is largely dependent on correct model specification. I develop several hypotheses linking state democracy level to membership in international organizations (IOs) that vary in scope, institutional capacity, and centralization. I test these hypotheses using several alternative approaches that allow to correct or explicitly model spatial and temporal dependence. I start with more common approaches, such as the use of a lagged dependent variable, fixed effects, and panel corrected standard errors, and then re-estimate the results using a multi-parametric spatio-temporal autocorrelation model (m-STAR). In this final model, I test my hypotheses using overlapping IO memberships in different types of IOs, as well as geographic contiguity as the spatial weights. I argue that while the lagged dependent variable, fixed effects, and panel-corrected standard errors show more desirable qualities than a naïve model, the m-STAR provides for the most adequate testing, from both a methodological and a theoretical perspective. Unlike the former three techniques that treat spatial and temporal dependence as a nuisance, the M-STAR allows for explicit modeling and estimation of contemporaneous spatial effects. Its ability to estimate spatial effects occurring within the same time-period as the unit-level effects makes this model particularly useful at evaluating the hypotheses posited in this paper, as well as such phenomena as diffusion and socialization more broadly.

14

Poster

Moments in Time: Studying European Conflict using a Change-Point Model Nieman, Mark Uploaded 07-31-2011 Keywords change-point constructivist time regime structural break Europe system conflict Abstract Constructivist theories provide insights into understanding systemic violence in Europe by accounting for preference formation in multiple time periods. Thus, constructivism explains how the influence of a set of independent variables on a dependent variable can change over time. By allowing preferences to change, constructivist theories accounts for changes in the direction and statistical significance of weakly exogenous explanatory variables over multiple time periods. Owing to this, different rationalist theories may be appropriate to explain different time periods. To test this, counts of militarized interstate dispute involving European states from 1870 to 2001 are analyzed. Bayesian MCMC change-point models provide an effective tool for identifying multiple time periods and generating unbiased and efficient estimates of explanatory variables. Because change-points are calculated probabilistically, the determinates of structural breaks are also examined by testing for Granger causality. Results generate support for constructivist theories because the influence of variables and statistical significance change depending on the time period. These results differ from tradition models which ignore structural breaks in the dependent variable. Lastly, Granger causality tests indicate that changes in systemic power and democratization are determinates of these structural breaks.

15

Poster

Stronger Instruments by Design Morgan, Jason Keele, Luke Uploaded 07-31-2011 Keywords 2SLS instrumental variables matching non-parametric Abstract There is growing interest in natural experiments in political science. Natural experiments are often analyzed with instrumental variable estimators reflecting a belief that combining the power of natural random assignment with an instrumental variable approach will solve many of the research design problems endemic to social science. Here, we highlight how weak instruments can interact with the assumption of random assignment of the instrument. When the instrument is not randomly assigned, weak instruments produce bias that is not alleviated by additional data. We demonstrate how matching combined with a reverse caliper can be used to strengthen an instrument within a subset of the overall study. We start by presenting an alternative non-parametric instrumental variable estimator first proposed by Rosenbaum (1996) that allows us to combine matching with an IV estimator. Unlike the standard 2SLS IV estimator, this non-parametric approach provides accurate confidence intervals and consistent causal estimates even when the instrument is weak. A further advantage of this non-parametric method is the opportunity it provides to probe the random assignment assumption with a sensitivity test. We provide substantive examples of the proposed approach with a reevaluation of a recent paper that uses rainfall as an instrument for voter turnout in US counties (Hansford & Gomez 2010).

16

Poster

Estimating the Effects of Unemployment on Voter Turnout Incantalupo, Matthew Uploaded 08-01-2011 Keywords design-based inference causal inference voting participation turnout unemployment inequality Abstract Unemployed Americans face numerous hardships: lost income, financial uncertainty, family struggles, strained or broken social ties, and issues with both mental and physical health. Job loss is a profound personal experience, but unlike many other consequences of hard economic times, it has not been strongly linked to changes in political behavior. In this study, I identify a causal link between involuntary job loss and reported voter turnout in recent elections. Under low unemployment, job loss has a negative effect on voter turnout. During a period of high unemployment, such as the one caused by the current economic recession, job loss has a positive effect on voter turnout.

17

Poster

Testing Theoretical Structures of Mass Preferences Jackson, Natalie Uploaded 08-01-2011 Keywords public opinion path analysis multidimensional scaling ideology mass preferences Abstract This project applies path analysis and multidimensional scaling models to complex, interrelated theoretical concepts to investigate the causal origins of policy preferences in the mass public. Ideology, in the sense of the liberal-conservative continuum, has often been used to explain policy preferences in the mass public with considerable success, but the causal origins of ideology are unclear due to the complexity of the concept. In this project, a theoretical model is developed that posits that ideology is created from, and therefore caused by, culture, as defined by and operationalized in the Cultural Theory framework developed by Douglas and Wildavsky (1982). However, the theoretical relationship between culture and ideology is different for those who consider themselves "liberal" or "conservative" (the ideologues) than it is for those who consider themselves "moderate" or non-ideologues. Ideologues will demonstrate a strong direct relationship between ideology and preferences, whereas moderates' preferences will be more directly related to their culture than ideology. Additionally, the concepts of culture and ideology should be more spatially similar for moderates than for ideologues since moderates are less likely to make strong distinctions between political views and their overall worldviews. This poster applies path analysis to determine which direct and indirect relationships are strongest between culture, ideology, and preferences. Multidimensional scaling analysis is then used to examine the spatial configuration of the constructs for moderates and ideologues.

18

Poster

A Split Population Model for Middle-Category Inflation in Ordered Survey Responses Bagozzi, Benjamin Uploaded 08-01-2011 Keywords split-population models ordered dependent Variables survey data Europe public opinion Abstract Recent research find that, for social desirability reasons, uninformed individuals disproportionately give ``neither agree nor disagree'' type responses to survey attitude questions, even when a ``don't know'' option is available (Sturgis et al. 2010). Such ``face-saving don't knows'' inflate the indifference (i.e. middle) categories of ordered attitude variables with non-ordered responses. When this inflation occurs within one's dependent variable, estimates from ordered probit/logit models are biased and inefficient. This poster develops a set of mixture models (the middle-inflated ordered probit with and without correlated errors) that estimate and account for the presence of ``face-saving'' responses in middle-categories of ordered survey response variables, and applies these models to (1) simulated data and (2) a commonly studied survey question measuring support for EU-membership among EU-candidate countries. Findings suggest that, when middle-category inflation is present in one's ordered dependent variable, the estimates obtained from middle-category mixture models are less biased than---and in some cases substantively distinct from---the estimates obtained from ``naive'' ordered probit models.

19

Poster

Hookworm Eradication as an Instrument for Schooling in the American South Henderson, John Uploaded 08-01-2011 Keywords hookworm education participation rockefeller sanitary commission instrumental variables matching permutation inference sensitivity analysis Abstract I exploit an historical natural experiment to assess whether more schooling causes greater vote participation. Specifically, I leverage the Rockefeller Sanitary Commission’s campaign to eradicate hookworm infection in the early-20th century American South as a plausibly-exogenous instrument for primary and secondary education. I evaluate two county-level interventions from the public health campaign: (a) exposure to the campaign and (b) pre-campaign hookworm incidence. Due to the presence of possible confounders, I use pair (genetic) and dose (optimal) matching techniques to strengthen the exogeneity of both instruments. I then use Rosenbaum permutation inference to assess the inclusion strength of the campaign exposure instrument, and I employ a simultaneous sensitivity analysis to evaluate robustness to remaining bias. Throughout, I find a robust and positive effect of education on participation.