How does recall bias in farm labor impact separability tests?

Abstract

In the agricultural household literature, empirical tests of separability between production and consumption decisions commonly exploit theoretical predictions of household labor allocation. Many of these studies rely on data that asks respondents to recall labor usage over the entire growing season. Two recent field experiments in Tanzania and Ghana show that such labor use data, collected at the end of the growing season, is a systematically unreliable measure of actual labor allocation. In this study, we examine how inaccurate measures of labor influence the reliability of market failure tests based on separability. In Ghana, we find no statistical evidence that recall bias influences the reliability of the separability test. In Tanzania, we find that recall bias increases the probability that such tests fail to reject separability. Thus, we find partial evidence that classic tests based on typical household survey labor data may erroneously conclude that markets are adequately functioning.