Designing choropleth maps: What projection to choose?


TL;DR: Using equal-area projections is important for reliably interpreting data in choropleth maps. We asked social science researchers, interested in the Global South, to design world choropleth maps according to their research goals. They ranked the map projection as the most important choice for fairly portraying the world. The recently proposed Equal Earth projection was the most preferred overall.

This post is based on the IEEE VIS short paper “Mapping the Global South: Equal-area projection for world choropleth maps” by Gabriela Molina León, Michael Lischka, and Andreas Breiter. You can read the preprint on arXiv.

Choropleth maps are among the most popular visualizations we find online. Also known as thematic maps, they visualize relative data that correspond to geographical regions. Nowadays, we often see them presenting data related to the COVID-19 pandemic, such as the daily positive rate and the ICU bed capacity in hospitals. These maps help us perceive spatial patterns and quickly detect extreme values.

Three world choropleth maps showing three different datasets.
World choropleth maps using the Equal Earth projection.

However, little is known about the importance of the projection in a choropleth map. Map projections are mathematical functions that transform our 3D planet into a 2D plane. But these functions cannot preserve every aspect of the globe without distorting the size, the angles, or the distances between world regions. Therefore, multiple types of projection exist according to the property they preserve:

  • equal-area projections preserve the relative areas without distortion,
  • conformal projections preserve the angles between lines on the globe, and
  • equidistant projections preserve the distances.

There are hundreds of projections and each projection distorts more or less these three properties of the globe. This is why compromise projections also exist. They preserve no property but try to find a balance between the different types of distortion. Preserving one of these properties should be prioritized according to the goal of the map designer. Each projection can help emphasize a certain world region.

A history of projections: Mercator vs. Gall-Peters

The most famous projection — called Mercator— significantly distorts the size of the countries. This conformal projection created by Gerardus Mercator was meant to help sailors on planning the course of their ships. It makes sure that a straight line on a Mercator map is also a straight line in the real world. Back in the sixteenth century, this was a great achievement for navigation planning. However, it required distorting the areas of the countries. The farther a country is from the Equator, the more distorted its size is. This is why Greenland seems to be almost as big as Africa, although Africa is actually fourteen times bigger.

Unfortunately, this information seems to have been forgotten by later map makers that used Mercator for maps that had nothing to do with navigation. A variant of this projection is still the standard for facilitating navigation on web maps such as Google Maps and OpenStreetMap. The size exaggeration of Europe and North America has led many people to believe that Mercator emphasizes colonizer or high-income countries and discriminates against the Global South. Researchers have previously argued that Mercator’s popularity has caused people to believe that Europe is bigger than what it actually is. However, no conclusive evidence has been found.

The area distortion can easily be avoided by using an equal-area projection. There are actually dozens of projections that preserve the relative size of the countries. Some of the most known are Gall-Peters, Mollweide, and Eckert IV.

In 2017, Boston Public Schools changed all their maps to use the Gall-Peters projection. The district authorities argued that their previous Mercator maps represented a Eurocentric view of the world. Their effort to “decolonize the curriculum” attracted a lot of media attention. Although Gall-Peters is an equal-area projection, it distorts the shapes of the countries strongly. Robinson, a well-known cartographer, famously described the projection as if “the land masses look like wet, ragged, long winter underwear hung out to dry on the Arctic circle.”

Two world maps, one using the Mercator projection and the other using the Gall-Peters projection.
The Mercator projection and the Gall-Peters projection. Mercator was designed to facilitate marine navigation while Gall-Peters was created to compare country sizes accurately. Source: d3-geo-projection.

The debate around projections has been going on for centuries and is still active. You may even wonder what your favorite projection says about you. Surprisingly though, there are no clear criteria to help you select a projection for map-based visualizations. Even if you limit your search to equal-area projections only, there are dozens of them and no clear winner. This is one of the reasons that motivated the research paper this post is based on.

Choropleth maps of social science data

At our university, my co-authors and I work together with social science researchers that investigate welfare. The researchers collect data about social policy programmes all over the world. One of their main goals is comprehensively including the Global South (low- and middle-income countries) in their analysis because social science research has focused mostly on high-income countries so far. The researchers want to fairly represent all countries in their work, making an effort to systematically include the Global South.

We designed visualizations together with the social scientists. During a series of co-creation workshops, we discussed how to explore and present their data. As part of the Global Dynamics of Social Policy project, the researchers collect data on family and education policy, among others, to create e.g. social policy profiles for each country. Their goal is to analyze how social programmes have evolved during the last century.

We decided to design choropleth maps for exploring data patterns across world regions. However, the social scientists expressed their discomfort with the common use of the Mercator projection in map-based visualizations of social science data. Multiple data portals that serve as their data sources, such as the websites of the WHO and the World Bank, use Mercator or other non-equal-area projections for choropleth maps. Furthermore, they do not provide any information on the projection used.

An equal-area projection is in general preferred because these maps are meant to visualize relative data. Distorted areas can cause people to mistakenly interpret the geographical distribution and density of the data. Therefore, without knowing the projection used, it is hard to know whether one can compare the countries reliably.

Our survey

We conducted a survey with 20 social science researchers to learn more about their goals and how those goals would influence their map design choices. During the first phase of our project, we noticed that there were no clear criteria on how to choose one projection for world choropleth maps (there is guidance for projections per region).

We asked the social scientists to design two maps: the first one according to their personal research goals (Task 1) and the second one according to their common goal of including the Global South (Task 2).

The researchers had diverse goals. Six researchers were investigating the diffusion of specific social policies across the world. Two focused on how colonialism influenced inequality and social protection. Their interests covered topics such as health care and migrant workers. Eight of them had a specific world region they were interested in, such as Africa or Eastern Europe.

The researchers created the choropleth maps through computational notebooks, such as this one. We gave them the ability to design the maps according to four parameters: the projection, the center, the scale (or zoom level), and the color scheme.

Five world maps, each using one of the map projections we selected.
The five equal-area projections the social science researchers could choose from. Source: d3-geo-projection.

For the projection, we gave them five options based on previous work: Equal Earth, Mollweide, Gall-Peters, Eckert IV, and Hammer (images above). All projections were equal-area. In contrast to the others, Equal Earth was recently designed as an aesthetic alternative to Gall-Peters given that it preserves better the shape of the continents. We did not reveal the projection names to avoid biasing the researchers.

Projection choices

In the map designed for their personal goals (blue bars in the figure below), 50% of the researchers preferred the Equal Earth projection. One researcher wrote that this projection was “aesthetically pleasing” but that it is probably not representing the country sizes accurately. The second most preferred was Mollweide as it seemed the most “familiar”.

For the second map, we asked the researchers to design according to the goal of comprehensively including the Global South in their work. The projection preferences for this task were not so clear. Gall-Peters was the most chosen projection but only by five people. Seven participants chose based on which projection seemed to be equal-area. Since we did not reveal the projection names, some people believed that not all projections were equal-area. In their justification, several people mentioned Mercator directly or indirectly, by saying that countries farther from the Equator should not be shown bigger than they are.

Overall, the researchers chose the Equal Earth projection more than any other. People chose based on what they felt looked more aesthetically appealing and that best matched their conceptions of reliable sizes. An interesting pattern was that a few researchers argued that the nicer-looking projections were probably not equal-area. Since they were all equal-area, it is safe to say that Equal Earth was the winner.

We also gave researchers the opportunity to control other parameters: the map center, the scale, and the color scheme. In the personal map, the researchers often centered the map on the world region their research focused on; sometimes they zoomed on it. In the Global South map, most participants centered their map on a region that they associated with the Global South such as Africa or South East Asia. Others argued that in order to avoid Eurocentrism, it is important to avoid centering the map in Europe.

Researchers ranked the design parameters according to the importance they had for their maps. In both maps, the projection was ranked as the most important parameter. They argued that the projection helped them have the best overview of the world and to distinguish small countries. They considered it the main tool to fairly portray the Global South. Interestingly, they also argued that the projection was more important than the map center to avoid Eurocentrism.

Animation showing each world map in sequence.
World choropleth map with each of the five projections selected.

What the projection says about your map

The design choices for choropleth maps varied according to the goals of the social scientists. Overall, using equal-area projections was not only important for the correct interpretation of the data but also for comprehensively including the Global South. If these goals are important to you, consider trying the Equal Earth projection.

Next time you create a choropleth map, it is worth testing what projection fits your goal better and reflecting on how your choice will influence the viewer’s perception of the world. Based on our results, providing information about the projection could increase the trust of the viewers and help them interpret the data. Nevertheless, more work is needed to learn about the choices of people with different goals outside of the social sciences. If you want to know more about our findings, you can find the pre-print of our paper here.



Gabriela Molina
Multiple Views: Visualization Research Explained

PhD student at the University of Bremen researching in InfoVis and HCI | Español, English et al.