tells us about the MADAS programme
Gilles Pison, INED research director and editor-in-chief of the monthly bulletin Population & Societies, is also coordinator of the MADAS programme for assessing census and survey data on adult mortality in sub-Saharan Africa.
(Interview in october 2015)
What do we know about mortality trends in sub-Saharan Africa?
In contrast to developed countries, demographic levels and trends cannot be measured on the basis of civil registries in Africa because they are not complete; not all births and deaths are reported. Demographic indicators are estimated from censuses and national surveys, but is that data reliable?
Here we can learn from the Millenium Development Goals (MDGs). The fifth of those goals is to reduce maternal mortality by three-fourths from 1990 to 2015—certainly a commendable goal, but in many countries, especially ones with high maternal mortality, the data collected is not sufficiently reliable to enable us to determine whether that goal has been reached today or not.
INED, together with the Agence Nationale de la Statistique et de la Démographie de Sénégal (ANSD) and the Institut de Recherche pour le Développement (IRD), has been working for the last 4 years on a research programme in Senegal called MADAS to assess the quality of the country’s censuses and national surveys. MADAS is funded by France’s Agence Nationale de la Recherche (ANR).
What research methods were used in the MADAS programme?
MADAS has made use of three existing population and health observatories in Sénégal, in the rural areas of Bandafassi, Mlomp and Niakhar. Populations on these sites were counted for the first time several decades ago and have been regularly monitored to the present time. Every time interviewers came through they visited each household, checked the list of members who were present on the previous visit and collected information on births, deaths, marriages and migrations that occurred in the intervening period.
The point of the study was to compare census or survey data with the information collected by the rural observatories. Fine comparison at the level of the household or individual pinpoints differences and enables researchers to study errors and bias—in age reporting in the census, for example, on father’s and mother’s survival, information that can then be used to estimate adult mortality, etc.
And what bias or errors were found?
Demographic survey interviews often collect information on sibling survival. Respondents list their siblings and indicate their age or age at death as well as year of death. This is the primary source for estimating adult mortality today. If a respondent reports that his or her sister died in her childbearing years (age 15-49), the interviewer then asks whether she was pregnant when she died, died in childbirth, or within 42 days of giving birth. This type of information is used to estimate maternal mortality.
We compared respondent reporting on siblings with the real situations found by the on-site observatories. Women interviewed in Bandafassi omitted 9% of deaths of adult sisters and a high number of siblings who died at young ages. The same women respondents less often omitted recently deceased siblings and siblings born to the same father. Conversely, women who had emigrated from their home village made more mistakes. Women respondents also showed a tendency to underestimate the age of older sisters.
A revised questionnaire on siblings that reiterates questions to reduce omissions and a life event calendar to improve date and age reporting were tested at Niakhar. There are fewer omissions of deaths of adult sisters with the new questionnaire: 90% of those deaths were reported, as opposed to 76% with the earlier questionnaire.