Abstract
Socially desirable responses to survey questions may be universal, but scales to capture the phenomenon are unvalidated in low-education and resource-limited settings. We therefore conducted a validation of the 16-item Balanced Inventory of Desirable Reporting (BIDR) short form in a two-round health survey of 1291 12-20 year-olds in rural Burkina Faso in 2017 and 2018. Confirmatory factor analysis (CFA) of the scale's designed dimensionality found poor fit (CFI = 0.50, TLI = 0.42, RMSEA = 0.10, SRMR = 0.08). Exploratory factor analysis of Wave 1 data suggested a novel 11-item, 2-factor structure, with all but two of the original scale's Self Deceptive Enhancement items discarded. CFA in Wave 2 using this novel structure gave poor fit indices (CFI = 0.62, TLI = 0.51, RMSEA = 0.10, SRMR = 0.07), test-retest reliability was low (ICC(A,1) = 0.06, Pearson's r = 0.06, R(2) = 0.004) and internal consistency was unsatisfactory (α and ω < 0.70) across waves for both scales. Measurement invariance was confirmed for age but not gender. This failure of BIDR implementation may reflect issues with item translation and delivery, locally appropriate content or use of reverse-coding in a low-education sample. It is possible, but less likely, that it reflects non-universality of the SDR construct. Our work highlights the importance of validating instruments in new study populations.