Department of Political Science

A time series of political power configurations of the Indic world system 500 BC — AD 1800 is examined by way of Poisson process analysis, phase transition analysis, Markov process analysis, mathematical modeling and simulation, information theory, analysis of autocorrelation, periodogrammetry, and time-spectral analysis (including epoch superposition). The Indic system displays an unusual propensity toward bipolarity and unipolarity. The behavior of the Indic world system at first sight resembles a Poisson process in which the age of a configuration is irrelevant to its stability. However, the uneven stability of Indic configurations reveals a first-order Markov process at work, within a new sort of longue durée: extremely long-term rules of political change, durable over 2000 years, which seem to route all power structure transitions to, from, and through bipolarity and unipolarity. No progression is found, but rather temporal symmetry. However, upon spectral analysis, several periodicities, both long (300-400 years) and short (1, 2 and 3 generations) are found in the data, and proposed for historical examination.

Americans have differing opinions on whether greater regulation of firearms results in improved public safety. One area that seems to enjoy broad support is to limit firearm access in specific locations. “Bar Ban” laws—which prohibit firearms where alcohol is served—represent one such approach, yet their effectiveness remains largely unexamined nationally. This paper provides the first comprehensive evaluation of the impact of Bar Ban laws on shootings near alcohol-related establishments. Using a geospatial panel dataset of over 1.6 million alcohol vendors active across the United States between January 2019 and January 2025, we analyze the relationship between restrictions on carrying a gun where alcohol is served and gun violence. Results show that shootings occur close to alcohol-serving establishments: across 263,464 shooting incidents, the median distance to the nearest alcohol vendor was 222 meters. To assess the effectiveness of Bar Ban laws, we employ a stacked difference-in-differences design examining monthly shooting exposure rates of alcohol vendors across counties in five states (Hawaii, Maryland, New York, New Jersey, South Dakota) that adopted Bar Ban laws during our study period. Our findings reveal a policy puzzle: while spatial analyses confirm that shootings routinely occur near alcohol establishments, Bar Ban laws targeting these locations show no discernible effect on reducing shooting incidents. Thus, states adopting these restrictions saw no substantial change in county-level shooting exposure rates near alcohol vendors.

The difficulty of collecting social conflict data has caused its study to focus on recent events predominantly in the West. This paper shows how to use artificial intelligence to generate social conflict data regardless of language, location, or date. Digitizing tertiary books, lightly cleaning their text representation, and submitting that text to a large language model (LLM) produces accurate date, location, and event descriptions. These capabilities and results are demonstrated with books on Latin America after 1492, Imperial Russia, and Tokugawa Japan. Using LLMs requires a larger fixed cost than working a team of research assistants, but it produces results more quickly for most sources. It is less accurate for the Tokugawa Japan source; whether it or human translation is cheaper depends on the cost of correcting the LLM’s work. These results represent the floor of accuracy for artificial intelligence, suggesting researchers should soon use it for creating social conflict data from tertiary sources.

Abstract: Interethnic marriage is commonly employed as an indicator of social cohesion. However, intermarriages are a reflection of both preferences and opportunities. If we are to interpret intermarriage rates as indicators of people’s willingness to cross group boundaries, we must find a way of controlling for exposure to out-group members in local marriage markets. In this Note, I exploit census data from Zambia to demonstrate how this can be done. The findings, which reveal significant differences across estimates that do and do not control for local exposure to out-group members, underscore a significant weakness in common approaches. The findings also point to important substantive implications for understanding changes in social cohesion in Zambia—and likely other African societies—over time.

Meaning-making systems underlie perceptions of the efficacy of threat-mitigating behaviors. Religion and science both offer threat mitigation, yet these meaning-making systems are often considered incompatible. Do such epistemological conflicts swamp the desire to employ diverse precautions against threats? Or do individuals—particularly individuals who are highly reactive to threats—hedge their bets by using multiple threat-mitigating practices despite their potential epistemological incompatibility? Complicating this question, perceptions of conflict between religion and science likely vary across cultures; likewise, pragmatic features of precautions prescribed by some religions make them incompatible with some scientifically-based precautions. The COVID-19 pandemic elicited diverse precautions thus providing an opportunity to investigate these questions. Across 27 societies from five continents (N = 7,844), in the majority of countries, individuals’ practice of religious precautions such as prayer correlates positively with their use of scientifically-based precautions. Prior work indicates that greater adherence to tradition likely reflects greater reactivity to threats. Unsurprisingly given associations between many traditions and religion, valuing tradition is predictive of employing religious precautions. However, consonant with its association with threat reactivity, we also find that traditionalism predicts adherence to public health precautions—a pattern that underscores threat-avoidant individuals’ apparent tolerance for epistemological conflict in pursuit of safety.

With the precipitous decline in response rates, researchers and pollsters have been left with highly nonrepresentative samples, relying on constructed weights to make these samples representative of the desired target population. Though practitioners employ valuable expert knowledge to choose what variables must be adjusted for, they rarely defend particular functional forms relating these variables to the response process or the outcome. Unfortunately, commonly used calibration weights-which make the weighted mean of in the sample equal that of the population-only ensure correct adjustment when the portion of the outcome and the response process left unexplained by linear functions of are independent. To alleviate this functional form dependency, we describe kernel balancing for population weighting (kpop). This approach replaces the design matrix with a kernel matrix, encoding high-order information about . Weights are then found to make the weighted average row of among sampled units approximately equal to that of the target population. This produces good calibration on a wide range of smooth functions of , without relying on the user to decide which or what functions of them to include. We describe the method and illustrate it by application to polling data from the 2016 US presidential election.

Understanding the role of emotions in protest is a growing field of research, but existing research does not address the role of emotions once protests start. By applying computer vision models to the expressed emotions of 37,558 faces in 7,824 geolocated protest images across twelve protest waves in ten countries, this article contributes to the study of emotions and protest. Most importantly, it measures emotions within protest waves, not before them. It also investigates emotions’ temporal effects, measures multiple emotions simultaneously, connects emotions directly to actual protests, and analyzes data across multiple countries. The results suggest that anger, disgust, fear, happiness, sadness, and surprise occur simultaneously throughout a protest, though happiness peaks on the first day. Emotions sometimes correlate with protest size in unexpected directions, and the coefficient signs differ by country. The most consistent finding is that models without lagged terms outperform those with lags, suggesting emotions and protests covary more than the former causes changes in the latter.