Total Survey Error in Practice

18Classification Error in Crime Victimization Surveys: A Markov Latent Class Analysis

Marcus E. Berzofsky¹ and Paul P. Biemer^2,3

¹ Division for Statistics and Data Science, RTI International, Research Triangle Park, NC, USA

² Social, Statistical, and Environmental Sciences, RTI International, Research Triangle Park, NC, USA

³ Odum Institute for Research in Social Science, University of North Carolina, Chapel Hill, NC, USA

18.1 Introduction

Many countries rely on crime victimization surveys to assess the volume of crime and to monitor trends in victimization rates among their populations (Kesteren et al., 2014). Obtaining accurate data in crime victimization surveys is challenging due to variations in how specific types of victimization are defined and understood by respondents, how well they are recalled, the sensitivity and social stigma of some victimizations, satisficing, and other respondent burden issues (see, e.g., Koss, 1996; Lynch et al., 2002; Rand and Catalano, 2007; Travis et al., 1995; Yan, 2008).

Although there have been efforts over the years to improve the quality of victimization data in surveys, concerns persist that victimizations may be substantially underreported for some crimes and that crime statistics, generally, are inaccurate (Lynch, 2014). Evidence to support these concerns is usually provided from studies comparing survey reports with official police reports. However, it is also well known that such comparisons may understate the true error in the ...