Small-area estimates (SAE) of poverty and malnutrition in Cambodia are produced at
commune level by combining survey data with auxiliary data derived from the 2008 General
Population Census of Cambodia (Census2008). A model for predicting log average per
capita household expenditure is estimated from the 2009 Cambodia Socio-economic Survey
(CSES2009) based on the Cambodia National Institute of Statistics calculation of expenditure
in each of the households sampled in CSES. The model is applied to household-level census
data to estimate poverty incidence, gap and severity. FAO and NIS have used CSES2009
to derive estimates of caloric intake in the form of kilocalories consumed per capita for each
sampled household; and a survey based model for kilocalorie consumption is also applied to
household-level census data to investigate the feasibility of predicting kilocalorie consumption
per household; when compared with a kilocalorie cut-off norm this could potentially be
used to estimate undernourishment at commune level. We find however that there remains
considerable unmodellable uncertainty in the kilocalorie data, so that, since the small-area
estimates of undernourishment at commune level are not sufficiently reliable, they have not
been included in the report. Models for predicting standardized height-for age and weight-forage
are estimated from both the 2008 Cambodia Anthropometric Survey (CAS2008) and the
2010 Cambodia Demographic and Health Survey (CDHS2010), each being applied to childlevel
census data to estimate prevalence of stunting and underweight; the separate estimates
from each source are combined using inverse-variance weighting to produce a single set
of estimates for each of stunting and underweight. Estimates of wasting, though desirable,
are not produced here because of the inadequacy of predictive models for weight-for-height
from both CAS2008 and CDHS2010. The small-area estimation procedure used in this study
does not produce direct measures of poverty, caloric intake or child malnutrition at the local
level. Rather the procedure applied here is able to estimate welfare outcomes – based on a
statistical model estimated in the relevant household survey. These estimates of wellbeing are
measured with error, and the degree of imprecision will vary as a function of a wide variety
of factors, most notably the degree of disaggregation at which indicators of wellbeing are
being estimated. In this study it was found that estimates at the level of a commune– which
comprises on average around 1700 households – are generally reasonably precise. Estimates
at village or enumeration area level are far less precise. The precision of estimates varies with
the specific indicator of wellbeing, and precision is generally better with consumption poverty
estimates than with estimates of caloric intake and child malnutrition, because there are fewer
survey variables that can be matched with the census in the latter models. Comparisons are
made with the poverty estimates derived from the 2009 Commune Database (CDB2009), and
with the earlier small-area estimates of poverty and malnutrition detailed in Fujji (2003). For
Cambodia as a whole, an overriding consideration from the current study is the generally
strong positive link between poverty and child malnutrition.