This study investigates whether geographic polarization—a strong correlation between political preferences and population density at national, state, and metropolitan scales—is driven by voters actively sorting themselves based on partisan neighbors. Using new voter registration data from records analysis, we conduct a simulation study to test the role of residential mobility.
This article argues that classroom simulations of real-world electoral institutions are a powerful pedagogical tool for political science students.
Supervised machine learning is increasingly used in political science, but these models require costly manual labeling of documents. Active learning is a framework that targets which documents human coders label rather than selecting them at random, with the goal of minimizing the amount of labeled data needed to train a model.
The Supreme Court’s recent decision that partisan gerrymandering claims are nonjusticiable ends a long search for a judicial standard. That shift increases the importance of legislative solutions. A history of successful congressional action and existing statutory requirements show that the legislative route is more practical than commonly believed and allows consideration of a broader suite of districting objectives.
Many bilateral relationships that require mutual agreement produce observable networks that are symmetric (undirected), while the underlying, substantively interesting ties are asymmetric (directed). The latent directed network—who initiates, who agrees, and who benefits—is often the object of scientific inquiry but is not directly observed.
A new computational algorithm was developed to generate legislative district maps that are contiguous, population-balanced, and relatively compact. The design explicitly aims to avoid the forms of bias identified in prior automated redistricting methods while improving computational efficiency so it can handle medium and large jurisdictions.
Political scientists often report quantities of interest derived from model coefficients—predicted probabilities, first differences, and marginal effects—rather than the coefficients themselves. These derived quantities do not automatically inherit the small-sample properties of coefficient estimates. An important consequence is that unbiased coefficient estimates are neither necessary nor sufficient for unbiased estimates of quantities of interest. This phenomenon is labeled transformation-induced bias.
Robust standard errors are widely used to correct standard errors for model misspecification. However, when robust and classical standard errors diverge, that divergence itself signals substantial misspecification. Assuming the misspecification is nevertheless small enough to leave everything else unbiased requires considerable optimism. Even if that optimism is warranted, relying on a misspecified model—whether or not robust standard errors are used—will bias estimators of all but a few quantities of interest.
Grant and Lebo (2016) offer an optimistic view of fractionally differencing short time series, while Keele, Linn, and Webb (2016) caution that estimates of fractional integration will be highly uncertain in short series. This study tests these competing recommendations by simulating fractionally integrated processes and comparing alternative modeling choices.
The literature recommends ARFIMA modeling to detect fractional integration and then (a) fractionally difference the data or (b) estimate a fractional error-correction model when variables are cointegrated. However, prior work also shows ARFIMA can give misleading indicators of fractional integration in series with fewer than 1,000 observations. The core question here is whether a simpler approach—the autodistributed lag model (ADL) or its equivalent error-correction model (ECM)—can still recover useful immediate and long-run effects without first diagnosing or correcting for fractional integration.