Determinants of Nutrient Demand: ANonparametric Analysis

Abstract

The influence of socioeconomic variables on nutrient intake is studied using nonparametric procedures that admit estimation of multivariate functions. The analysis indicates a nonlinear relation between intake, age, education, and income. Specifically, intake rises with income reaching an inflection point beyond which it is essentially flat. Socioeconomic variables influence intake primarily at lower-income levels. Nonparametric procedures prove useful in avoiding ad hoc specifications that would fail to uncover these findings. Understanding the link between the nutritional well-being of individuals and their socioeco- nomic status is important for influencing consumer behavior through public policy. Interest in understanding this relationship dates back to the 1940s and the development of the minimum cost diet plans (Senauer, Asp, and Kinsey). Adrain and Daniel conducted one of the first comprehensive studies of the relationship between nutrient intake and socioeco- nomic status. Since then, numerous researchers have reexamined this relationship using newer data and more innovative techniques. Senauer, Asp, and Kinsey and Morgan provide a complete survey and appraisal of this literature. They indicate that while research in this area has progressed considerably, the influence of some key socioeconomic variables needs further investigation. In particular, Morgan and Davis suggest a need for more interdisciplinary research to better understand the relationship between nutrient intake and income, household size, and education. The literature contains a diversity of findings with respect to these variables (Morgan). The variation in findings are reminiscent of similar analysis in the economic development literature, where researchers have obtained a wide range of estimates for income elasticity of nutrient intake, even among the very poor. This has led to an important debate regarding the effectiveness of increasing incomes in alleviating malnutrition in poor countries (Behrman, Deolalikar, and Wolfe). Recent studies by Strauss and Thomas and Subramanian and Deaton have shown that the low elasticity estimates are partly due to model misspecification. Measurement error may also cause the differences in these findings.