Book Description
Height-for-age z-scores (HAZs) and stunting status (HAZ<−2) are widely used to measure child nutrition and population health. However, accurate measurement of age is nontrivial in populations with low levels of literacy and numeracy, limited use of formal birth records, and weak cultural norms surrounding birthdays and calendar use. In this paper we use Demographic and Health Surveys data from 62 countries over the period 1990–2014 to describe two statistical artifacts indicative of misreporting of age. The first artifact consists of lower HAZs for children reported to be born earlier in each calendar year (resulting in implausibly large HAZ gaps between January- and December-born children), which is consistent with some degree of randomness in month of birth reporting. The second artifact consists of lower HAZs for children with a reported age just below a round age (and hence implausibly large HAZ gaps between children with reported ages just below and just above round ages), which is consistent with survey respondents rounding ages down more than they round ages up. Using simulations, we show how these forms of misreporting child age can replicate observed patterns in the data, and that they have small impacts on estimated rates of stunting but important implications for research that relies on birth timing to identify exposure to various risks, particularly seasonal shocks. Moreover, the misreporting we identify differs from conventional age-heaping concerns, implying that the metrics described above could constitute useful markers of measurement error in nutrition surveys. Future research should also investigate ways to reduce these errors.
