Uncovering measurement issues in collective efficacy: A Rasch-based evaluation

Survey items measuring collective efficacy are commonly included and used to construct composite scores in crime studies, serving as the basis for comparisons across groups, residential areas, and time periods. This study evaluates the measurement properties of a set of twelve items designed to measure collective efficacy and two theoretical subdimensions ( Social cohesion and trust , and Informal social control ) in the City of Stockholm Crime Survey. The analysis is based on pre-existing data from three random samples (2014, 2017, and 2020) of Stockholm residents aged 16 to 79, comprising a total of 47 664 individual responses. The Rasch partial credit model was used. The findings reveal significant unidimensionality issues in the current item compositions, questioning the use of these 12 items to measure collective efficacy and their division into the theoretical subscales. Beyond issues of unidimensionality, for the Social Cohesion and Trust , the five-point agreement response scale showed disordered thresholds, and several items displayed a notable lack of invariance across age groups. For Informal Social Control items, the primary concerns were related to DIF and item misfit, particularly for isc1 (children skipping school). Furthermore, evidence of local dependency was observed between isc4 and isc5 (fighting and car theft). Smaller subsets of items demonstrating more promising measurement properties were identified. Three items from Social Cohesion and Trust capture perceptions of community behaviors; two items from Informal Social Control reflect residents' willingness to intervene in cases of children's misbehavior; and two additional items from Informal Social Control reflect residents' willingness to take action against observed violence or crime. Yet, those smaller item subsets are limited by the number of items, resulting in low reliability and failing to fully capture the multifaceted nature of collective efficacy. This study provides valuable insights into the measurement properties of items intended to assess collective efficacy. There is a lack of clarity about what the sets of items in the City of Stockholm Crime Survey truly can measure, and it is possible that previously reported differences in collective efficacy across neighborhoods, groups, or time partly reflect measurement artifacts rather than substantive variation. A significant strength of the study lies in its rigorous psychometric evaluation and the limitations it reveals regarding the validity of the existing scale. In addition to the specific findings on collective efficacy, the present study also demonstrates the value of applying the Rasch model in social sciences and crime research.