Page 36 - IJPS-2-1

Monitoring adult mortality by type of residence in the absence of death registration: a perspective from Burkina Faso

areas compared to rural areas in 1999 and 2003. The same pattern is observed in the 2010 DHS
among older respondents (30–49 year-olds). Second, the average number of siblings declines with
the respondent’s age in rural areas, and it remains stable in urban areas. These results are surpris-
ing, because fertility is higher in rural areas in Burkina Faso, and over the years, it has declined both
in urban and rural areas (Shapiro and Gebreselassie, 2009). The patterns observed here are like-
ly caused by rural exodus of women or a pronounced omission of women’s siblings in rural areas. As
reported by Stanton et al. (2000), as well as Masquelier and Dutreuilh (2014), it is likely that older
respondents disproportionately omit to report all of their siblings. These omissions seem to be more
marked in rural areas, and if related to adult death, it is likely that adult mortality will be underesti-
mated, particularly in these settings. This could be explained by the low level of education among
interviewed women and misunderstanding of local language by fieldworkers in rural areas (Johnson,
Grant, and Khan, 2009).

Figure 4. Average parity of respondents’ mothers by respondents’ age and place of residence according to each round of survey (DHS data).

3.4 Summary

Figure 5 presents estimates of differentials in urban/rural mortality derived from the different sources
and methods. Estimates derived from the orphanhood method applied on DHS are discarded from
this plot, since those obtained from the EMIUB survey are to be preferred.
Among men as well as women, an urban penalty is unlikely from 1989 to 2006, if one remembers
that sibling survival data overestimate adult mortality in urban areas and underestimate adult mortal-
ity in rural areas. It is only since around 1994 that adult mortality appears to be particularly lower in
rural areas. These estimates were derived from sibling survival data of the DHS conducted in
1998/1999. A cursory evaluation of data quality indicates that this survey is not as reliable as the
others. For example, the percentage of deaths which were reported without any information on “age
at death” and “years since death” was respectively 2.7% and 0.2% in 2003 and 2010 respectively,
while it was as high as 8.5% in 1998/1999 (INSD, 2000b; 2004; 2012).
4 Discussion

At first glance, the picture of differentials in urban/rural mortality derived from the different methods
is inconsistent. While the orphanhood method produced mixed results, the growth balance method
tends to conclude to an urban advantage in the recent period, and sibling histories data suggest that
urban areas suffer from a penalty that is vanishing overtime. This highlights the crucial need to re-
strain from drawing any firm conclusion on differentials based on a particular technique of estima-
tion in isolation. The different estimates have to be screened in order to come up with a better mea-
surement of urban/rural differentials in adult mortality. Despite technical issues related to the estima-
tion (issues of assumptions, data quality, and selection effects), and the administrative problem in
defining urban areas, a systematic analysis shows that an urban disadvantage in terms of mortality is
unlikely in Burkina Faso.

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