Constructing Internally Consistent BMI Z-Scores for Adults and Children to Examine Intra-household Outcomes

Abstract

The ultimate goal of development is to improve health and well-being at the individual level. Yet, survey data are routinely collected at the household level. To bridge this gap the existing literature has used various, not fully satisfactory direct and indirect methods to impute individual level outcomes from household data.

This paper proposes a new direct statistical method that constructs BMI Z-scores for all ages. It develops age-gender specific Body Mass Index (BMI) growth curve distributions using generalized LMS estimations in a GAMLSS framework which allows flexible modelling of all parameters. These growth curve distributions are then used to construct BMI Z-scores that are consistent for children and adults and that can thus be used to examine individual-level well-being and intra-household allocation patterns.

Compared to existing methods that estimate intra-household allocations, BMI Z-scores present several advantages. One, the BMI measures health outcomes rather than caloric or monetary health inputs. Two, it is easier to collect and is measured with less error. Three, the BMI Z-score methodology also enables a range of previously unmeasurable within-household comparisons and it opens up the use of existing sources of survey data to answer new research questions in health and development economics. An empirical example for Tanzania provides one simple application of the BMI Z-score method, showing how household-level measures significantly overestimate individual measures of well-being and health.