Assessing mode effects in online and telephone victimization surveys

Abstract

Victimization surveys not only represent a vital data source for estimating the amount of crime within a social entity but also provide victimological researchers with valuable information, e.g. about the social distribution and consequences of victimization experiences. To assure adequate interpretation of victimization survey data, knowledge about the data-generating process in general and the measurement of victimization experiences in particular is essential. In 2010, a new round of the International Crime and Victim Survey (ICVS-2) was conducted to explore the different effects of telephone and online victimization surveys. To understand the processes behind these, this study follows a theory-driven approach. Based on a four-step cognitive model of survey response, hypotheses about expected mode effects on the response behavior were developed. For hypothesis testing, a hierarchical modeling strategy based on logistic regressions was applied to the ICVS-2 data. As hypothesized, no effects of survey mode on the response process could be identified for most victimization items. After controlling for sampling frame and unit non-response, the effects of survey administration mode on the propensity of reporting victimization vanished. The sensitive sexual victimization item represented the only exception. In this case, the remaining mode effect could be fully accounted for by systematically differing item non-response rates between modes.