Social Desirability Bias: Experimental Evidence on Reporting Parental Practices

We analyze social desirability bias in the reporting of parenting practices through survey questions. We develop a method to experimentally identify this bias by purposely inducing social desirability in questions on feeding practices through a random information provision on best practices. Our results show a treatment effect of -0.160 standard deviations in the reporting of children ultra-processed food consumption, in line with the presence of social desirability bias. We find a larger bias for women, less educated individuals, caregivers that believe child development is not malleable to parental investment, and those with risk preferences above the median. Although the Marlowe-Crowne scale positively correlates with our experimental measure of social desirability bias, we show that an heterogeneous effect analysis by this variable does not fully remove the issue.