Written by Laura Rodriguez Takeuchi, Research Officer in ODI’s Growth, Poverty and Inequality Programme.
In the Millennium Development Goals (MDGs), countries committed to reduce the under-five mortality rate by two thirds between 1990 and 2015. Despite progress, it is likely that the target will not be met. UNICEF estimates that in 2012, 6.6 million children died before their fifth birthday, at a rate of around 18,000 per day. About half of those deaths occurred in just five countries: India, Nigeria, Democratic Republic of Congo, Pakistan and China. There are no paranormal phenomena or secret agents involved, but understanding how these numbers are derived is a daunting task.
Civil registration and vital statistics systems – administrative records of births and deaths – are essential to monitor a country’s progress in improving the lives of its people. But keeping such records involves strong administrative capacity, not least transportation, literacy, organization and the management of thousands of local government employees spread across a country. Household surveys are a second best approach, and it looks like between the two, we have a very comprehensive picture of child mortality in the world. At a first glance, with the exception of Oceania, data on child mortality are widely available (136 out of 161 developing countries have data to track this MDG). If the world has not achieved the goal, it is not for a lack of data that can inform and lead action. Or is it?
Recently, I started digging into the methodologies to estimate child mortality from household surveys and I came across this document: ‘Manual X. Indirect techniques for demographic estimation’. Interesting name, and a laudable effort by UN agencies to set some guidelines on best practices in measuring child mortality from household surveys. In particular, it sets the most appropriate procedures and parameters to use when little data are available and child mortality needs to be estimated indirectly. Just two answers are needed from women of reproductive age during the survey: the number of children ever born and the number of children surviving. On the basis of this information, the proportion of children who have died per group of women is converted into an estimate of a probability that a child will die before age five for all women. This is problematic for at least two reasons – first, the quality of reporting of this data can be variable; and second, these estimates rely on a ‘model’ of population fertility and mortality patterns (model life tables) which derive mainly from European historical experience. There have been some methodological updates – for instance, ‘Time since first birth’, has proved to be more sensitive to different risks associated with women’s socioeconomic conditions and age, and large numbers of children born outside marriage. But although increasingly used by UNICEF and others, it has yet to be fully adopted. Secondly, model life tables have been created with data from a wider range of developing countries, but their representativeness in sub-Saharan Africa remains limited.
In a world that is calling for a Data Revolution, it is worrying that this method is still at the core of some countries’ knowledge about the survival prospects of their children. According to data from the UN Interagency group for child mortality estimation, less than one-third of the 75 `countdown countries’ have civil registration and vital statistics systems and only 10 of those countries have a child mortality estimate derived from such data since 2009; furthermore, 11 countdown countries base their most recent child mortality on an indirect methodology and 26 countries do not have any child mortality estimate that is less than 5 years old.
An updating of the methodologies is necessary, but newer and more sophisticated techniques will not fully solve existing data gaps. For example, although DHS and recent MICS surveys, the main surveys used to track many MDG indicators, have now moved beyond the indirect methods and include a more complete source of information (full-birth history), survey estimates will always rely on the quality of the underlying sample and the method used, and a benchmark to assess the robustness and reliability of the estimates is needed. Take Nigeria. Here, survey data are abundant, but under-five mortality estimates widely differ depending on the specific survey and method used. For the year 1998, for example, we find the following estimates:
- 225 deaths per 1,000 live births (using 2003 DHS and a full-birth history method);
- 186 deaths (2008 DHS, full birth history);
- 173 deaths (Malaria Indicator Survey);
- 157 deaths (2011 MICS and indirect method).
The difference between the highest and lowest estimate for 1998 is close to 1.5 million children. How can policymakers make sense of those widely differing numbers and determine an appropriate policy response? Efforts in basic data collection, such as vital registration systems, are required to draw a more accurate picture of how countries are faring.
Obtaining reliable measures of progress will be crucial to transform the post-2015 goals into reality. In the excitement over new technologies and the prospects they offer, developing and strengthening vital registration systems should not be forgotten. Vital registration systems are not only a complete source of information, but could be the basis to assess the robustness of new survey-based methodologies. Both could complement each other, for example using surveys to frequently monitor progress, and to disaggregate child mortality estimates by different categories of women and households, such as ethnicity, education level and wealth, while data from vital registration systems are used for the benchmark, mid-point and final reporting on the targets.
If the post-2015 goals are to make a difference, a clear roadmap on how to measure the targets is needed. This should consider new technologies for data collection and different types of data but without leaving behind the basics. Hopefully, we won’t need more X-files to track progress post-2015.
 The group depends on the method used. It can either be defined by the woman’s age, by the time since first marriage or by the time since first birth.
If you’re interested in getting involved in these activities, please email email@example.com.