adopt two approaches in estimating a childhood nutrition model. In the first, child nutrition enters
as a continuous dependent variable, allowing me to estimate the average marginal effects on
childhood nutrition (proxied by three anthropometric measures). The second approach allows me to
estimate how the probability of being undernourished changes with variations in the explanatory
variables, making use of a limited dependent variable approach – after converting child
anthropometry into a binary variable indicating whether or not the child is undernourished.
The ultimate purpose of this exercise is to inform those involved in the design of nutrition policy in
South Africa as in what measures can be expected to be most effective. In addition, I intend to make
interested researchers aware of the complexities involved in an attempt to promote more accurate
studies that will serve to further refine child nutrition policy.
The rest of this paper is devoted to this dual task. First, I motivate the importance of policy designed
specifically to target childhood nutrition. I then proceed to discuss the types of factors that influence
childhood nutritional status, which provides the rationale for the structural model, analysed in the
following section. I test the relative strength of specific variables on childhood nutrition in an
analysis of South African survey data. The final section concludes with some recommendations
based on the paper’s findings.