An Improved Method of Automated Nonparametric Content Analysis for Social Science

Some scholars build models to classify documents into chosen categories. Others, especially social scientists who tend to focus on population characteristics, instead usually estimate the proportion of documents in each category—using either parametric “classify-and-count” methods or “direct” nonparametric estimation of proportions without individual classification. Unfortunately, classify-and-count methods can be highly model-dependent or generate more bias in the proportions even as the percent of documents correctly classified increases. Direct estimation avoids these problems, but can suffer when the meaning of language changes between training and test sets or is too similar across categories. We develop an improved direct estimation approach without these issues by including and optimizing continuous text features, along with a form of matching adapted from the causal inference literature. Our approach substantially improves performance in a diverse collection of 73 datasets. We also offer easy-to-use software that implements all ideas discussed herein.

A Bias-Corrected Estimator for the Crosswise Model with Inattentive Respondents

The crosswise model is an increasingly popular survey technique to elicit candid answers from respondents on sensitive questions. Recent studies, however, point out that in the presence of inattentive respondents, the conventional estimator of the prevalence of a sensitive attribute is biased toward 0.5. To remedy this problem, we propose a simple design-based bias correction using an anchor question that has a sensitive item with known prevalence. We demonstrate that we can easily estimate and correct for the bias arising from inattentive respondents without measuring individual-level attentiveness. We also offer several useful extensions of our estimator, including a sensitivity analysis for the conventional estimator, a strategy for weighting, a framework for multivariate regressions in which a latent sensitive trait is used as an outcome or a predictor, and tools for power analysis and parameter selection. Our method can be easily implemented through our open-source software cWise.

Identification of Preferences in Forced-Choice Conjoint Experiments: Reassessing the Quantity of Interest

Forced-choice conjoint experiments have become a standard component of the experimental toolbox in political science and sociology. Yet the literature has largely overlooked the fact that conjoint experiments can be used for two distinct purposes: to uncover respondents’ multidimensional preferences, and to estimate the causal effects of some attributes on a profile’s selection probability in a multidimensional choice setting. This paper makes the argument that this distinction is both analytically and practically relevant, because the quantity of interest is contingent on the purpose of the study. The vast majority of social scientists relying on conjoint analyses, including most scholars interested in studying preferences, have adopted the average marginal component effect (AMCE) as their main quantity of interest. The paper shows that the AMCE is neither conceptually nor practically suited to explore respondents’ preferences. Not only is it essentially a causal quantity conceptually at odds with the goal of describing patterns of preferences, but it also does generally not identify preferences, mixing them with compositional effects unrelated to preferences. This paper proposes a novel estimand—the average component preference—designed to explore patterns of preferences, and it presents a method for estimating it.

Human Rights Violations in Space: Assessing the External Validity of Machine-Geocoded versus Human-Geocoded Data

Political event data are widely used in studies of political violence. Recent years have seen notable advances in the automated coding of political event data from international news sources. Yet, the validity of machine-coded event data remains disputed, especially in the context of event geolocation. We analyze the frequencies of human- and machine-geocoded event data agreement in relation to an independent (ground truth) source. The events are human rights violations in Colombia. We perform our evaluation for a key, 8-year period of the Colombian conflict and in three 2-year subperiods as well as for a selected set of (non)journalistically remote municipalities. As a complement to this analysis, we estimate spatial probit models based on the three datasets. These models assume Gaussian Markov Random Field error processes; they are constructed using a stochastic partial differential equation and estimated with integrated nested Laplacian approximation. The estimated models tell us whether the three datasets produce comparable predictions, underreport events in relation to the same covariates, and have similar patterns of prediction error. Together the two analyses show that, for this subnational conflict, the machine- and human-geocoded datasets are comparable in terms of external validity but, according to the geostatistical models, produce prediction errors that differ in important respects.

Choosing Imputation Models

Imputing missing values is an important preprocessing step in data analysis, but the literature offers little guidance on how to choose between imputation models. This letter suggests adopting the imputation model that generates a density of imputed values most similar to those of the observed values for an incomplete variable after balancing all other covariates. We recommend stable balancing weights as a practical approach to balance covariates whose distribution is expected to differ if the values are not missing completely at random. After balancing, discrepancy statistics can be used to compare the density of imputed and observed values. We illustrate the application of the suggested approach using simulated and real-world survey data from the American National Election Study, comparing popular imputation approaches including random forests, hot-deck, predictive mean matching, and multivariate normal imputation. An R package implementing the suggested approach accompanies this letter.