Counting Crowds: Measuring Contentious Political Action across the United States

• In Event: Conflict Modeling, Prediction and Forecasting

Abstract

We develop and validate a series of measurement models on the size and frequency of political crowds in the United States. Since January 2017, the Crowd Counting Consortium (CCC) has produced event data on contentious politics in the United States (i.e., marches, demonstrations, rallies, strikes, pickets, walkouts, caravans, and nonviolent direct actions) from a broad set of sources, including online print and broadcast media outlets, social media, and public submissions via an online form. Because these sources do not perfectly capture the size, type, and occurrence of these events, the goal of this project is to improve the estimation of the total number of events that occur. Our measurement models expand beyond the estimating just the number of protests to also includes information about violence, property damage, and arrests at each protest. The measurement models essentially link four related count processes together that account for the total number of protests, the number of protests at which violence occurs, the number of protests at which property damage occurs, and the number of protests at which arrests are reported. We make two additional modeling innovations. First, we develop a simple Multiple-Indicators Multiple-Causes (MIMIC) model which connects the different observed indicators of protest in a county, to county-level covariates and national shocks to protest activity. Next, we extend this model by introducing unique dynamics to each country, if the degree of protests within each county follows an autoregressive process where shared national shocks along with local shocks at each month.

Authors

• Kevin Reuning, Miami University

• Christopher J. Fariss, University of Michigan

• Jay Ulfelder