Explaining patterns of malnutrition among children in Uganda

Jing Hao, Purdue University

Abstract

Child malnutrition is one of the most serious problems threatening people's lives in Uganda. This thesis assesses the factors associated with patterns of child nutritional outcomes in Uganda, as indicated by Z-scores for growth. In particular, I examine whether agricultural performance helps to explain patterns of malnutrition among children less than five years of age. The study uses data collected from the Uganda Demographic Health Survey (2006 UDHS) and the Uganda National Household Survey (2005/06 UNHS). Normalized Density of Vegetation Index (NDVI) data observed by NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) is also used in the analysis. To study the factors associated with child stunting (indicated by a low height-for-age Z-score (HAZ)) and wasting (indicated by a low weight-for-height Z-score (WHZ)), two regression approaches are used: an ordinary least square (OLS) regression model and a logistic regression model. Explanatory variables of interest are incorporated in five groups: (i) those corresponding to the child; (ii) those corresponding to the child's mother; (iii) those corresponding to the child's father and household; (iv) district-level variables on agricultural performance derived from the UNHS; and (v) remotely-sensed vegetation data(NDVI). As a check of robustness of the results, regressions are run on subsamples of the data from rural and urban areas, and among internally displaced persons (IDPs). In addition, an in-depth analysis of the relationship between NDVI anomalies and OLS residuals from the child growth regressions is presented. Results suggest that, among all the explanatory variables considered, child and mother variables play the most important role in explaining a child's nutritional status. Important factors include child age, whether a child is a twin, whether a child has anti-tuberculosis vaccination or anemia, and a number of characteristics of the mother, including her BMI and age, her pregnancy and breastfeeding status, and her level of education. DHS variables for household wealth, the household dependency ratio, and the sex and age of the household are also significantly correlated with child Z-scores. The explanatory power of UNHS variables nested at the district level is generally poor. Using NDVI values as proxies for agricultural potential, we find some evidence that average NDVI values corresponding to recently preceding harvest months are positively correlated with a short-term measure of child wasting (WHZ). Overall, factors associated with child malnutrition appear to differ across sub-regions. This underscores important spatial and idiosyncratic features of child malnutrition in Uganda, and limits our ability to draw general conclusions and policy recommendations regarding child health.

Degree

M.S.

Advisors

Shively, Purdue University.

Subject Area

Agricultural economics|Nutrition|Public health|Sub Saharan Africa Studies

COinS