Page 37 - IJPS-2-1

Bruno Yempabou Lankoande

Figure 5. Estimates of adult mortality in Burkina Faso by place of residence and sex according to different sources and methods, from 1989 to 2006.

A common limitation of the application of the growth balance method to measure adult mortality
in sub-geographical units is the issue of migration. The method could be adjusted to take into ac-
count internal migration flows between urban and rural areas, but such data are rarely available
(Bhat, 2002; Hill and Queiroz, 2010). In the case of urban/rural differentials in mortality in Burkina
Faso where migration flows are mainly from rural to urban areas, discarding migration effects will
result in an underestimation of mortality in urban areas and an overestimation in rural areas (Moul-
trie, Dorrington, Hill et al., 2013). The approach adopted here by limiting the estimation of the com-
pleteness of the reporting of deaths on people aged 35 years and more, seems to yield consistent re-
sults. Estimates do not seem to be affected by migration since it is possible to get a straight line in
each case by plotting the partial births against the partial deaths (Figures A2 and A3 in the Appendix).
In addition, the completeness of death registration was high enough, both in urban (80.5%) and rural
areas (73.5%). This limits the effect of adjusting for completeness on the final mortality estimates in
the two settings (Moultrie, Dorrington, Hill et al., 2013). In summary, estimates derived from the
growth balance method for each place of residence are acceptable.
The EMUIB survey allows bypassing a major problem encountered when applying the basic or-
phanhood method in sub-national geographic units using DHS data, as respondents and parents do
not always share the same place of residence (Moultrie, Dorrington, Hill et al., 2013). Indeed, the
urban disadvantage observed in orphanhood estimates in DHS data (based on reports from children),
is likely due to the effect of the practice of child fostering in Burkina Faso. A common practice in
the country is to move children from one family to another for schooling or housework (Dabiré,
2001; Serra, 2009). Based on the definition adopted by Grant and Yeatman (2012), the DHS data
indicate that the prevalence of child fostering varies greatly with place of residence in Burkina Faso.
For example, in 2010, only 7.6% of children under 15 were fostered in rural areas whereas this fig-
ure rose to 17.2% in urban settings. In addition, orphanhood prevalence was higher in fos-
tered children compared to non-fostered ones in both urban and rural areas. By using children’s place
of residence as a proxy for their parent’s place of residence, a fraction of deaths occurring in rural
areas was therefore transferred to urban areas, and those of the urban areas were allocated to rural
areas. The problem raised by the misclassification of parent’s place of residence is likely to play
against urban areas because flows from rural to urban areas are more important. A great share of ur-
ban growth in Burkina Faso is still explained by internal migration (Guengant, 2009). I can conclude
that mortality levels based on DHS data are overestimated in urban areas, and underestimated in ru-
ral areas, yielding a spurious “urban penalty.” When data on the survival of young adult’s parents
were used to estimate the difference in mortality between urban and rural areas, the urban disadvan-
tages vanishes, as shown in Figure 2, even though the difference in favor of the urban advantage is

International Journal of Population Studies | 2016, Volume 2, Issue 1 31