Adjusting national statistics for inequity
Why measuring inequalities matters?
Measurement matters because it influences public action. When the nature and extent of poverty are unknown, it is unlikely that society will take strong action to reduce it. The same applies for inequity. Societies measure things that are considered important. When something is not being counted, it typically means that it does not count.
It is becoming abundantly clear that the world will not meet the Millennium Development Goals (MDGs) in 2015 because disparities within countries have widened to the point of slowing down national progress in terms of human development. Monitoring must bring this to the fore. The MDG indicator for measuring equity – i.e. the ‘share of the poorest quintile in national consumption’ – is seldom mentioned or used in analytical reports. Moreover, the indicator is problematic from the point of view of its accuracy, coverage and relevance.
The growing availability of disaggregated data, however, makes it possible to adjust key national statistics for inequity. Recent surveys – especially the Demographic & Health Surveys – generate information by wealth quintile (i.e. a fifth of the population). The quintiles are not based on income or consumption, which are notoriously difficult to measure. Instead, they are based on the possession of household assets that can be readily observed – such as a bicycle or a radio, electricity or water connections, size of dwelling and type of construction materials.
How to obtain data that are more relevant to the poor?
Adjusting a national statistic for inequity can be done by weighing the quintile-specific values so as to accord less importance to progress for the upper quintiles than to similar progress for the lower quintiles. The method is illustrated at the hand of quintile-specific data for under-five mortality, one of the more robust MDG indicators.
It is not unreasonable to give the bottom quintile a 30 per cent weight and the top quintile a 10 per cent weight – implying that progress for the lowest quintile will count for three times more than the same progress made for the highest quintile. Other quintiles receive intermediate weights (25, 20 and 15 per cent respectively) so that they add up to 100 per cent. These suggested values illustrate that national statistics can be adjusted to reflect disparities within countries. Other weights can be considered, including non-linear ones.
Take a country that manages to reduce its national Under Five Mortality Rate (U5MR) from 100 to 70 per 1,000 live births. Three possible scenarios can be considered, as depicted in the diagram below. First, the low-equity scenario implies that most of the gains accrue to the upper quintiles; implying that the gradient across quintiles becomes steeper. Second, the mediumequity scenario reduces U5MR by 30 points for all quintiles; implying that the gradient remains the same. Third, the high-equity scenario equalises the U5MR for all quintiles to 70; implying that the gradient across quintiles disappears.
Under each of the three scenarios, the quintiles face very different realities, yet it is not captured by the national U5MR statistic. The U5MR for the bottom quintile ranges from 120 to 70; that for the top quintile varies between 70 and 20. Nonetheless, the un-adjusted national U5MR statistic is the same under the three scenarios, namely 70 per 1,000 live births. Thus, it cannot be known from the national statistic how equitable progress has been. By using equityadjusted weights, however, the national U5MR statistic indicates how equitably progress has been. The more equitable the pattern of progress, the better the national statistic will be. The equity-adjusted national U5MR statistic in this example varies between 83 and 70; a difference of about one-fifth between the low-equity and the high-equity scenarios.
We were able to collect – from UNICEF (2008) and from recent DHS surveys – quintile-specific U5MR estimates for 63 countries, with a total of 132 observations. Data for most countries indicate a clear and consistent association between household wealth and the level of human development. They also show that in-country disparities are sizeable, confirming that national statistics conceal as much as they reveal. Of the countries in the sample with trend data, the majority witnesses widening disparities over time. The majority also sees a change in their ranking according to their equity-adjusted U5MR statistic.
It could be argued that the ratio between the bottom and top quintiles has greater intuitive value to express disparities than the equity-adjusted weighing of quintile-specific values. The former, however, does not inform about the other three quintiles and is therefore less representative of the entire distribution.
How to move from measurement to policy reform?
Bolivia and Namibia, for instance, have the same national (un-adjusted) U5MR. But the gradient across quintiles is much steeper in the former than in the latter. The equity-adjusted statistic makes it explicit that inequality is entrenched in Bolivia. Her equity-adjusted U5MR is about 10 per cent higher than that for Namibia. Hence, Bolivia ranks 4 places lower among the 63 countries in the sample with the equity-adjusted statistic. Some countries jump as many as 6 places.
League tables are imperfect and equity-adjusted values are subject to some statistical caveats. But they do catch the eye of political leaders. Changes in country rankings in an equityadjusted table are likely to trigger a much needed focus on disparities, which is vital for ensuring future relevance of the MDGs, or whatever they will be called beyond 2015.
Some general sources:
1. Gwatkin D. (2005) “How much would poor people gain from faster progress towards the MDGs for health?” The Lancet 365.
Note: HD Insights are network members' contributions and do not necessarily represent the views of UNDP.
Back to the list of past editions.