Better to be indirect? Testing the accuracy and cost-savings of indirect surveys

• Published in: World development 2021-06, Vol.142, p.105419, Article 105419
• Main Authors: Alix-Garcia, Jennifer M., Sims, Katharine R.E., Costica, Laura
• Format: Article
• Language: English
• Subjects: Accuracy; Antipoverty programs; Errors; Households; Indexes; Indirect survey; Polls & surveys; Poverty; Respondents; Survey design; Surveying; Targeting
• ISSN: 0305-750X; 1873-5991
• DOI: 10.1016/j.worlddev.2021.105419

•We compare reports of assets obtained indirectly from knowledgeable local informants to those collected directly from households.•We test whether these reports are accurate, both for specific characteristics and when aggregated into asset indices.•There is high correlation between household and informant reports.•Indirectly reported asset indices perform well in simple regressions and in a very basic targeting exercise.•In our context, indirect surveying reduces the survey budget by 50%. We test the validity of indirect surveying as a method to collect household data. We compare household and informant reports of assets, develop poverty indices from both, test their performance as regression covariates, and examine errors in reporting and targeting resulting from using indirectly reported variables. Informant-based targeting indices are highly correlated with household measures and can be reasonable substitutes for self-reported indices in simple regressions. They can also be used to assign a simulated anti-poverty program with similar error rates to related methods. In our setting, eliminating direct household surveys would have reduced survey costs by 50%.