Call Us: US - +1 845 478 5244 | UK - +44 20 7193 7850 | AUS - +61 2 8005 4826

application to child nutritional status

The setting and data
To illustrate the use of stochastic dominance for difference-in-difference evaluation we use a unique,
large dataset of child nutrition from arid and semi-arid lands (ASALs) in Kenya. These areas are
characterized by extensive livestock production and highest incidences of poverty in Kenya. Over 60% of
the population live below the poverty line and levels of access to basic services are very low. Infant
mortality rates are high, in some districts more than double the (already high) national average. Child
malnutrition levels in Kenyan ASALs are generally declining but are still above emergency threshold
levels, worsened by recurrent droughts, high poverty rates, and HIV/AIDS. In the North Eastern Province,
for example, 23.2 per cent of children under five suffer from acute malnutrition and infant and underfive mortality rates are rising (UNICEF 2008).
The data we use were collected by the Kenyan government under the second phase of the Arid Lands
Management Project (ALRMP II), a World Bank-financed community-based drought management
initiative that provided additional financial resources to 28 arid and semi arid districts in Kenya from
2003 to 2010. The project sought to improve the effectiveness of emergency drought response while at
the same time reducing vulnerability, empowering local communities, and raising the profile of arid and
semi-arid areas in national policies and institutions.
Since one of the objectives of ALRMP II was to reduce child malnutrition the project’s monitoring
strategy included the collection of information on child nutritional status. The specific anthropometric
indicator collected was the Mid-Upper Arm Circumference (MUAC) measurement for children younger
than 60 months. MUAC is a reliable and relatively cheap-to-collect indicator for child nutrition status. It
is also closely correlated with clinical and other anthropometric indicators of nutritional status (Shakir
and Morley 1974; Shakir 1975). In addition it is considered more appropriate than other measures for
children in pastoral areas (Mude et al. 2009).
We use MUAC Z-scores rather than absolute MUAC measures as they allow a direct comparison across
age and gender of children. Z-scores for weight-for-age or height-for-age are routinely used to measure
child nutritional status. For some reason, perhaps inertia from when MUAC Z-scores were difficult to
calculate, even recent studies (Ritmeijer 1998) and the current 2006 WHO Child Growth Standards for
emergency nutrition programs still use raw MUAC measures in centimeters, despite clear evidence that
Z-scores are the preferable measure