Methodological challenges in the redesign of the 2015 Survey of Prison Inmates

Abstract

The Survey of Prison Inmates (SPI) is being redesigned for its intended implementation in 2015 by the Bureau of Justice Statistics (BJS). The goal of SPI is to provide estimates which describe the U.S. state and federal inmate population across topics such as criminal history, mental and physical health status, and socioeconomic status at the national-level, as well as at the level of particular subdomains (e.g., gender) and jurisdictions (i.e., states with large prison populations such as California, Texas, and Florida). This paper discusses the methodologies and associated challenges encountered during the SPI redesign, including the determination of an optimal first-stage allocation scheme, utilization of both implicit and explicit stratification to select prison facilities across subdomains, and estimation of the required number of inmates to be sampled in order to achieve the desired precision. Using a redesigned instrument and historical experience from the 2004 SPI to approximate design effects, a simulation study was conducted to determine the proportion of outcomes that would meet the study's precision goals while minimizing respondent burden on inmates and facilities.