Association for Public Policy Analysis & Management

Having good quality data is fundamental for doing applied research and for making public policy decisions. Yet, recent literature has found problems in the quality of the data collected, particularly in developing countries. Using a survey that asks questions regarding possession of assets of the wife and of the husband living in the same household from families participating in the Mexican conditional cash transfer program PROGRESA, I find discrepancies in the possession of assets reported between the spouses. For example, when asked about the possession of a washing machine, the information reported by the spouses did not coincide in 24% of the households. Psychological evidence has found that the accuracy of the information provided by the individuals is linked with self-esteem; and individuals with low self-esteem are more likely to change how the situation appears as a coping mechanism. Thus, self-esteem can be an important variable to identify data abnormalities. The results show that the self-esteem of the husband explains the differences in the possession of assets reported, but not for the wife’s self-esteem. Specifically, one standard deviation increase in the self-esteem of the husband reduces an index regarding the differences of assets reported by each spouse by .10 standard deviation. To address the problem of omitted variable bias, two novel bounding techniques are implemented: Oster (2016) and Krauth (2016). My results suggest that the estimate presented is robust to omitted variable bias.