Long-term time-series forecasting of social interventions for narcotics use and property crime

Abstract

This paper presents a policy analysis based on a multivariate long-term model of narcotics-related behaviors and social interventions. We first examine the forecasting performance of the long-term model. Next, we design a simulation study to investigate the long-term impacts of hypothetical policy changes in methadone maintenance on narcotics use and property crime. The data used for model development were based on retrospective self-report information on various narcotics-related behaviors collected from methadone maintenance patients in Southern California. For time-series analysis, we aggregated each subject's longitudinal addiction history to provide group-level data which consisted of 99 bimonthly periods for the following five variables: abstinence from narcotics, daily narcotics use, property crime, legal supervision, and methadone maintenance. Post-sample forecasting performance is compared between the long-term time-series model and a more common time-series model which captures only short-term relationships. Overall, the results demonstrate superior performance by the long-term model, indicating its adequacy to explain the system dynamics among the five variables. For the simulation study, four hypothetical conditions of methadone maintenance are investigated, and the resultant impacts on narcotics use and property crime for each methadone maintenance condition are predicted using the long-term time-series model. The analyses lay out quantitative pictures of the effects of hypothetical changes in methadone maintenance that are consistent with a priori expectations. Finally, monthly average changes in cost of methadone maintenance were estimated for each simulation setup.