Page 33 - IJPS-2-1

Bruno Yempabou Lankoande

of this cutoff point also helped to keep the sampling errors at an acceptable level for the analysis and
to attenuate the effect of heaping for five years prior to the survey (Bicego, 1997).
A major drawback also related to the estimation of adult mortality by place of residence using
sibling survival data is the difficulty to apprehend siblings’ place of residence. The DHS surveys do
not collect information on the place of residence of siblings at the time of survey or at the time of
death. To address this issue, the place of residence of interviewed women was tested as proxy for the
place of residence of their siblings. Misclassification of siblings’ place of residence may thus lead to
misinterpretation of differences in urban/rural mortality levels, particularly when migrations flows
are important (Bicego, 1997).

3 Results

3.1 Growth Balance Method (GBM)
Estimates of adult mortality obtained by place of residence and by sex in 2006 from the GBM, and
from the census official report are presented in Table 2. The completeness of death reporting was
higher in urban areas compared to rural areas (80.5% and 73.5% respectively), suggesting that data
quality issues are more prevalent in rural areas. The estimates were also higher than those published
in the census report (except for women in rural areas). Disregarding these methodological differenc-
es, the value of the probability 45q 15 was higher in rural areas than urban areas according to both
sources. In the estimates presented here, the urban-rural differentials in mortality were lower among
men with a ratio of urban to rural mortality of 0.7. This ratio rose to 0.9 in women. As expected, the
levels of male mortality were higher than those of female mortality.

Table 2. Estimates of 45 q 15 (per 1000) from the GBM and results published in the census’ official report by sex and according to urban/rural
residence in Burkina Faso in 2006
Men Women
Urban Rural Urban/rural Ratio Urban Rural Urban/rural Ratio
GBM 270.2 366.1 0.7 215.2 248.6 0.9
Census report 220.0 321.5 0.7 183.6 280.3 0.7

3.2 Orphanhood Method
Figure 1 presents trends in adult mortality obtained from orphanhood data collected in the 1993,
2003, and 2010 DHS according to place of residence and sex of parents. For men, only one estimate
was obtained from each dataset (because the estimation of male mortality requires that reports from
two adjacent age groups be combined). For women, two estimates were derived from each dataset
(based on 5–9 and 10–14 year-olds).
Overall, in urban as well in rural areas, adult mortality rates seem to have fallen since 1988 to
2005 for men, and since 1987 to 2007 for women. Quite surprisingly, urban residents seem to have
experienced a slight mortality increase in the late 1990s and the early 2000s. This counterintuitive
result is likely due to the issue raised by the approximation of parents’ place of residence (those who
are at risk of dying) by their children’s place of residence. Over the whole period considered, values
of 45q 15 were on average higher in urban areas compared to rural areas. In men, these inequalities in
mortality seem to have widened over time in favor of the rural settings. In women, the gap of ur-
ban/rural mortality was most pronounced in the first half of the 2000s.
By using parents’ place of residence in the estimation, thanks to data collected in the EMUIB sur-
vey (Figure 2), the mortality differentials were reversed in favor of urban areas in men, and the gap
widened in favor of urban areas in women. In men, when children’s place of residence is used as
proxy for their parent’s place of residence, values of 45q 15 were far higher in urban areas than in rural
areas. However, a lower mortality in urban areas was obtained when parents’ place of residence at

International Journal of Population Studies | 2016, Volume 2, Issue 1 27