Examining measurements of fear of crime: cognitive processes and psychometric properties

Abstract The present study examines how different operationalisations of fear-of-crime questions influence the possibility of measuring fear of crime as latent traits. To address this aim, the study comprised two related parts. In the first part, cognitive interviews aimed to gain deeper insights into the response process to questions about fear of crime, and how different phrasings and formats may influence that process. The second part aimed to evaluate the possibility of using survey data to form a measure of fear of crime. Cognitive interviews (n = 17) were analysed for response processes. Rasch analyses were conducted on the Swedish Crime Survey (n = 4264) and the City of Stockholm Crime Survey (n = 7333). Response comprehension was generally adequate, yet some terms were interpreted differently. Respondents found it difficult to isolate the fear of crime from other motives in the consequential items. Affective outcomes, when stated as intensity questions, imposed a lower response burden compared to when stated as frequency questions, as well as compared to consequential-framed. Psychometrically, unidimensional solutions were hard to achieve for both affective and behavioural/consequential outcomes. The evidence suggests heterogeneity across population groups. Measuring fear of crime as a latent trait depends not only on how fear-of-crime questions are operationalised, but also on how respondents interpret and respond to those questions and on the measurement properties of the resulting responses. By combining evidence from cognitive interviews and psychometric analyses, this study provides a broader understanding of both the opportunities and limitations of measuring fear of crime through self-report surveys. Overall, advancing the measurement of fear of crime requires simultaneous consideration of conceptualisation, response processes, and psychometric properties, rather than refinement of survey questions or psychometric models in isolation.