In 2014/15, police forces across England, Scotland and Wales received over 300,000 calls relating to missing persons, a figure that appears to be increasing. Despite this growing figure and the workload it places on police forces, there has been a lack of research into the area of missing persons. For most forces, the level of police response that a missing person case requires is set by the risk classification that has been assigned to the case. These levels are ‘Standard Risk’, ‘Medium Risk’ and ‘High Risk’. This project examines the appropriateness
of these risk classifications based on how they are assigned and the effect they have on the time it takes to resolve a case. The data used comes from Lancashire Constabulary and contains all missing person reports that were made to the force in 2015. Logistic regression is used to investigate the individual risk factors that best predict a ‘High Risk’ classification and examine how these differ to the risk factors that the police believe indicate higher risk. The
main body of the analysis focusses on modelling the time to resolution for missing from home cases as predicted by their risk level and other explanatory variables using event history analysis methods. Kaplan-Meier estimates are used to model the probability of no resolution by risk level. Cox Proportional Hazards Models are then used to determine which factors alongside risk level are significant in their prediction of time to resolution. These models are then extended to account for the effects of repeatedly missing persons with the inclusion of a frailty term. This project concludes that risk classification does have a significant effect on the time to resolution and puts forward the notion that the large amount of cases being classified as ‘Medium Risk’ has absorbed police time and resources and in turn taken these away from the more complex ‘High Risk’ cases which see a slower time to resolution than ‘Medium Risk’ after the first 24 hours of a missing person case being created. Recommendations and possible extensions to the analysis are provided.