Measuring inequality from above

A satellite image shows differences in night lights across the globe

The sources of increasing inequality in recent times have become one of the most debated issues in economics. However, underlying these debates are discussions on how to accurately measure inequality. Doing so poses a difficult challenge, even before moving onto studying cause and effects. In Barcelona School of Economics Working Paper 1252, “Measuring Inequality from Above,” Jose G. Montalvo, Marta Reynal-Querol, and Juan Carlos Muñoz-Mora explore the potential of using luminosity at night to construct measures of inequality. They present a new methodology that aims at improving the consistency of inequality measurement.

The authors outline two key issues that come up when building inequality statistics. The first is which indicator to use. A variety of options exist, but the focus is placed on the well-known Gini index. The second is what source of basic data to apply this indicator to? There are various problems with existing data sources, and the choice of data can cause large variations between results. For example, consumption data is sensitive to memory recall and often under-estimates the consumption of services. Consumption surveys are also prone to under-estimating the number of high-income earners. In addition, existing statistics are calculated using different methodologies, which produce indicators that do not necessarily coincide.  In the case of Spain, the Gini index reported by the World Bank is almost ten percent greater than that reported by the National Statistical Office. 

Therefore, the authors present a measure that can be generalized to a global scale and produce methodologically homogeneous measurements. This is particularly important when estimating inequality, as it enables the authors to compare inequality worldwide.

Collecting data via satellite

Recently, in addition to traditional data sources such as national accounts and income or consumption surveys, academic research has increasingly used satellite imagery to estimate economic activity. Previous work has established the usefulness of this data source. It has been shown that light density at night is a useful proxy for population and GDP growth. 

Night light measured by satellites has some advantages over traditional data sources. The approach can be used for measurements for very small areas, where it is difficult to find estimations of GDP or enough individuals in national surveys to produce a reliable estimation. The measure proposed by this paper captures data at an impressively fine detail. There are more than 2.2 million populated pixels in the average country, with 81 individuals per populated pixel. The approach is also particularly useful when trying to measure economic activity in war areas or places where there is a high level of social tension that makes other approaches unfeasible. Finally, night light is measured in the same way around the world and, therefore, simplifies comparisons at a global scale. 

The authors develop the use of this innovative data source by using satellite imagery to calculate inequality measures instead of simply economic activity. Their objective is to show that these data can also be useful when measuring a broader set of economic variables. They calculate economic inequality at high levels of geographical disaggregation and decompose inequality to examine between and within-group trends.

The paper provides some important methodological contributions to the use of satellite night light data. The first contribution is a new procedure to calculate inequality using luminosity and population, measured using small pixels. Second, it is well known that saturated night light, which is mostly used in the literature, is top-coded, and different satellites can alter the measurement. The data is top-coded since, in big cities, where economic activity is greatest, satellites cannot distinguish between the brightest areas. The method proposed includes fitting optimal correction terms to saturated night light in areas where lights are very intense.

Inequality across the world

The methodology consists of using the average nightlight per capita as a proxy for income per capita. The income of each individual is calculated as the average night light in a very small geographic unit (i.e., pixel), calculated using data from satellite imagery. This is then weighted by the corresponding population size. 

They then rank the pixels and calculate the Gini index for each local area. The authors name their inequality index the Measuring Inequality From Above (MIFA) Gini Index. Their new measure performs well when compared to existing alternatives. It performs particularly well for countries with a low proportion of pixels that record zero light, high population density, and high income per capita. These are some of the areas that have previously been hard to estimate accurately because of the top coding within satellite data.

Their results are shown at a global level below:

Map shows Gini Index at the global level
Figure. Average Gini from Above

As with the standard Gini index, lower numbers mean more equal societies, and higher scores represent a greater level of inequality. According to the measure, the most unequal countries are in Africa, where the measurements are quite high. In most developed countries, the MIFA index varies from 0.3 to 0.5.


Utilizing their newly developed income per capita proxy, the authors build another commonly used indicator, the Theil statistic. The Theil index is an alternative inequality measure to the Gini coefficient, which has the benefit that it can be decomposed into a within and between-group component. They then use this statistic to calculate within and between regions/ethnic groups’ inequality levels across the world. 

The authors compare the ratio of within or between-group inequality to total inequality. They find that between ethnic groups, inequality represents an important proportion of total inequality in Africa. In fact, the six countries with the highest proportion of between ethnic groups inequality over total inequality are from Africa: Central African Republic (82%), Chad (75%), Kenya (64%), D.R. of Congo (64%), Tanzania (63%) and Gabon (54%). The inequality between regions is also an important source of inequality in Africa. 

The highest proportion of between-regions inequality is found in Guinea-Bissau, Mauritania, Uganda, Burundi, Central Africa Republic, and Gambia. They find that between regions, inequality in Latin America is quite important, while the inequality between ethnic groups is less relevant. That is also the case in Asia: the inequality between regions is more severe than the inequality between ethnic groups.

The methodology presented was designed intentionally to be applicable on a global scale, which facilitates these comparisons. The authors hope that future research will further investigate the relevance of such inequality decompositions. In particular, by focusing on the ethnic dimension, explain conflict, public good provisions, and other important social topics.