Improving the measurement of household wealth to better understand global inequalities in education

POST2015_equity_borderBy Jeroen Smits, Associate Professor on Inequality and Development at the Economics Department of Radboud University in the Netherlands and Director of the Global Data Lab.

Over the past few decades, household surveys have become an important source of information on educational outcomes in low and middle income countries. These surveys enable many factors associated with unequal educational outcomes within countries to be studied. The World Inequality Database on Education (WIDE), established by the EFA Global Monitoring Report team, shows what is possible with survey data, revealing within-country disparities in educational outcomes between key social groups according to gender, ethnicity, religion, location, region and household wealth.

Measuring the economic situation of households in poor countries is particularly problematic. In rich countries, we tend to use income to measure a household’s economic status, but in poor countries income is difficult to measure and tends to fluctuate from year to year. For that reason, consumption expenditure is recommended as a preferable measure. However, this approach also has disadvantages, not least the cost of collecting the information.

Credit: Karel Prinsloo/ARETE/UNESCO
Many surveys measure household wealth using asset-based indices such as the characteristics of a house, a household’s possession of durables and its access to basic services. Photo: Karel Prinsloo/ARETE/UNESCO

Drawing on the pioneering work of Deon Filmer and Lant Pritchett (2001), many surveys measure household wealth using asset-based indices. These indices measure the characteristics of a house, a household’s possession of durables and its access to basic services. Households owning more expensive durables, living in a better quality house, and having access to basic services are considered to have a higher economic status than households with less expensive durables, worse housing and no access to services.

Such measures have gained prominence in household surveys such as the Demographic and Health Surveys (DHS) and the UNICEF Multiple Indicator Cluster Surveys (MICS). They are relatively easy to utilize and have intuitive appeal. However, despite their wide availability, these indices suffer from one serious disadvantage: they are not comparable across time and place. For each survey a separate wealth index is constructed on the basis of the assets available in the survey data. To make the scores usable, the lowest 20 percent of the population are defined as the poor and the upper 20 percent as the rich. WIDE also uses this method to display wealth variations in educational outcomes within countries.

The wealth variations displayed by WIDE are as such not incorrect. For the country-year combinations for which they are computed, they provide a valid picture of the differences between the lowest and highest 20% of the population. However they have little connection with the wealth quintiles used for other countries or for the same country in another year. For example, the poorest 20% of the population of Ethiopia in 2000 is poorer than the poorest 20% in Malaysia in the same year or in Ethiopia in 2010. Differences between countries and years in educational performance measured by these indices are thus only in part the result of differences in performance.

For a more comparable picture of household wealth, the same measure should be used for all countries and years. In the graph below, for 24 low-income countries around 2010, primary educational attendance is shown for the poorest households based on the two indices: The blue circles show the lowest 20% wealth group based on the relative wealth index used by WIDE and the red squares show the group with a value below 20 on the International Wealth Index (IWI), an absolute wealth index running from 0 to 100 (Smits and Steendijk, 2014).

School attendance rates of poorest quintile children, ages 7-12, based on national and international indexes of wealth, circa 2010

Note that different groups of people are compared in the graph. That is, the blue circle represents one fifth of households whereas the red square represents more than one fifth of households.

For the countries with low attendance rates, in almost all cases the rates based on IWI are 5 to 10 percent higher than those based on the relative index used by WIDE.  This reverses at the highest attendance levels, with rates based on the relative index being higher than the IWI rates.

The IWI based figures are comparable among countries, in the sense that all households are in the same wealth category (under the IWI-20 poverty line). The figures indicate school attendance rates among those households. In very poor countries such as Ethiopia, Burundi and Malawi a large part of the population falls under this poverty line; in relatively wealthier countries such as Indonesia, Colombia and Peru this is only a few percent of the population.

The blue figures, in contrast, are for groups that have the same size (20%) of households in all countries. In the wealthier countries, this group is richer than in the poorer countries, so parents are better able to send their children to school. Therefore, a better country performance does not necessarily indicate a better performing education system.

Given the increased focus on inequality in the post-2015 education agenda, it is important that in addition to relative wealth indices, the international community adopts a comparable wealth index like IWI as a basis for better measuring global inequality in education.



  1. Great article and very insightful but measuring household wealth won’t give us the bigger picture. Rather measure each nation’s allocated budget for education and the qualifications of teachers/ leaturers, as well as their quality of skills transfer

  2. Thanks for this, it is an interesting idea and I know one that many would agree with. It seems like IWI can be useful in cross-national research but it should not replace the household asset scores used in DHS and MICS. These are designed to provide policymakers at the country level with useful information on the degree of disparity within their country– urban/rural breakdowns, subnational regions, wealth, etc.

    If we look at Malaysia and Ethiopia, according to the logic of the IWI, we would implicitly be saying the Ethiopian government should be designing education interventions for the 60% of its population that is comparable to maybe the poorest 10-20% (obviously these are rough guesses) in Malaysia. I don’t think that’s realistic, and given political realities and scarce resources, could end up further marginalizing those most in need.

    On top of this, wouldn’t it be true that other common disaggregations such as urbanicity are not valid from country to country? A more valid measure would be an index somehow measuring “remoteness from concentration of public services” across countries.

  3. hiiii,
    your article is so insightful, we know that education level is so decrease as compare to developing countries and in middle class countries is not proper school and college are present that’s why education level is deceased day by day.
    Animation note

Leave a Reply