Journal on Policy and Complex Systems
charged for prostitution ( Richey , 2018 ). Historically , over 90 % of those arrested for prostitution in the United States are sellers and fewer than 10 % are buyers ( Demand Abolition ’ s CEASE Network Boston ). When comparing the effectiveness of supply-focused versus demand-focused intervention strategies , the empirical evidence suggests that demand-focused law enforcement is more effective in reducing commercial sexual exploitation of children . There is very little evidence that targeting supply yields more than a temporary suppression or displacement of prostitution ( Hunt , 2013 ; Shively , Kliorys , Wheeler , & Hunt , 2012 ).
Related Work
The U . S . Department of Justice
commissioned a national study examining the effectiveness of efforts to reduce prostitution and sex trafficking across the United States ( US ). This study found that the multifaceted nature of these intervention programs makes it hard to isolate the effect of a single component of the programs , such as a demand-suppression strategy . Some programs combine supply and demand intervention efforts so , when there is a positive change , it is hard to be certain which part of the strategy was more responsible for the measured outcome ( Shively et al ., 2012 ). For example , as a result of a comprehensive field experiment that included targeting demand through reverse-sting operations in Jersey City , New Jersey , prostitution declined by 75 % ( Weisburd et al ., 2006 ). The study controlled for displacement effects , finding that the reduction in prostitution was not because of displacement effect to other areas . However , because of the comprehensive nature of the experiment , it was hard to identify the effect of targeting demand versus other elements of the program on the reduction of prostitution . Computational policy models can overcome these limitations .
Policy models permit experimentation in an artificial society , which has many advantages compared to extant methods . Existing methods are limited in their ability to break down how specific elements of the program , under what conditions and with what spatiotemporal dynamics at what cost worked or did not work because of the multifaceted nature of the program ; we cannot change a single parameter and see the effect while keeping everything else equal ( Gilbert , Ahrweiler , Barbrook-Johnson , Narasimhan , & Wilkinson , 2018 ). In addition , experimenting in the real-world policy domain is costly , time-consuming , and resource-intensive . Computational policy models can provide a deeper understanding of the policy domain .
Policy modeling has an important role to play in two areas of policy process : ex-ante policy design and appraisal and ex-post policy evaluation ( Gilbert et al ., 2018 ). Ex-ante policy design and appraisal refers to the assessment of the relative merits of competing policies in meeting policy objectives , leading to better policy design . Ex-post policy evaluation examines whether a policy meets its objectives and deter-
102