A twist on the heisenberg principle: Or, how crime affects its measurement

Abstract

Evidence suggests that surveys of victimization are affected by large and systematic sources of bias which reduce the validity of comparisons over time and among geographical areas. This paper argues that the bias is especially severe because errors of measurement are correlated with the level of ‘true’ victimization. Evidence relevant to four hypothesized sources of bias is considered. First, it is hypothesized that lifestyle characteristics which are associated with victimization are also associated with respondent inaccessibility, resulting in the exclusion of victims from surveys. Second, coverage and response rates are lower in high crime areas, in part due to mutual avoidance by interviewers and respondents. Third, the social context influences rates of reporting and the classification of incidents of victimization. Finally, the bias introduced by variations in survey procedures is more severe when concepts are ambiguous and ill-defined. The hypotheses and evidence pertinent to them suggest that the measurement of trends and differences in victimization may be subject to large and fluctuating sources of error. Possible research strategies for investigating the sources of bias are suggested.