“The Weight of Too Much Liberty”: A Predictive Spatio-temporal Model of Police Response to Partisanship in England

In liberal democracies, the expectation is that basic democratic rights - freedom of expression and of association and peaceful assembly - will be passively observed, if not actively defended, by law enforcement. Anecdotal evidence from England, however, suggests certain groups, all ‘Leftist’, are increasingly unable to exercise these rights for fear of police reprisal. Due to recent reports, this paper empirically investigates how partisanship impacts police response. We primarily rely on the ACLED’s observational protest data, which already includes spatial and temporal measures. Using machine learning, observations are supplemented with a binary and a categorical measure denoting police response and severity, respectively. Protestor partisanship is manually assigned using expert surveys and the analysis of group manifestos and political communications. As police response is an inconsistent variable, this paper leans on non-parametric Bayesian estimation methods to fill gaps created by limited or missing data. Through Spatio-temporal Gaussian Process Regression, county-level models of predicted response are rendered. In short, this paper demonstrates the space-specific likelihood that the Left will be over-policed and, by that corollary, that the Right will be under-policed over time.