Are we measuring what we want to measure?

The social context of survey interviews is likely to be important in survey measurement in developing countries, where respondents expect to benefit from participation. In the recent literature on survey measurement, however, there are few attempts to analyze the impact of the respondents' social context on response error, and they tend to be limited to developed countries. This paper follows the narrow path traced by these attempts. The opportunity for this study is offered by a set of 134 unplanned re-interviews collected during the fieldwork operations of a household panel survey in rural Malawi. Personal benefit was the main reason some respondents were willing to be re-interviewed, since the survey compensated them with an additional gift for the second interview. By comparing the answers to the first and second interview given by the re-interviewed respondents, this paper therefore assesses how the search for personal benefit (which captures some aspects of the respondents' social context) biased the results.