A growing number of researchers studying intimate partner violence (IPV) employ aggregate measures of relevant attitudes to serve as proxy measures for norms around IPV. However, there is a lack of consistency in how these measures are constructed and how their validity is confirmed. The first aim of this study is to demonstrate and validate innovative techniques for exploring social norms proxies in quantitative data and identifying the relative appropriateness of different available reference groups. The second aim is to demonstrate how such an approach can contribute to IPV research. The analysis employed data from the 2016 Tanzania International Men and Gender Equality Survey, including 1008 men and 1008 women ages 15-49 years. An attitudinal score measuring acceptance of IPV and two measures for individual-level descriptive and injunctive norms were constructed. The intraclass correlation coefficient (ICC) was used to assess the extent of clustering for the attitudinal score within several sets of groupings. Bivariate multi-level Ordinary Least Squares regressions estimated the predictive effect of an individual's group norm proxy on their descriptive or injunctive norms. Attitudinal clustering was most significant for reference groups defined as males only across villages and males only across a combination of villages and marital status, with ICCs of 0.229 and 0.236, respectively. Men's social norms were found to be correlated with reference groups comprised of both men only and women only, though men's norms were substantially more correlated with the attitudes of men in their reference group than with women's. Results highlight the importance of critically examining the validity of proxy measures for social norms prior to their inclusion in analysis. Findings also underscore the importance of collecting attitudinal data from men to better understand norms around IPV.
Keywords: attitudes; descriptive norms; injunctive norms; intimate partner violence; social norms measurement.