Christopher Ingraham at the Washington Post writes about how "data journalists" can create drastically different visualizations of the same data. Specifically, he criticizes a choropleth map recently published by Pew Charitable Trusts that shows population change from 2014 to 2015 by county according to the Census Bureau's yearly population estimates.
He raises some good points: When showing data that can be both positive and negative, it makes sense to use a diverging color scheme. But I think a main limitation of both of these maps is the measure they have chosen to map: the percent change in population in each county.
Percent change can certainly be informative, but on a map, it can tell you nothing about the magnitude of population change in any given area. Looking at Pew's map, you might think that population growth centers around Loving County, Texas. You might also look at WaPo's map and conclude that the increases in the blue counties are roughly balanced out by all the red counties where population has declined.
Neither of these things are true: Loving County's "boom" was actually just an increase of 25 people from 87 in 2014. The red, declining counties in WaPo's map are mostly rural and sparsely populated, while many of the blue, growing counties are the most densely populated counties in the country. The full story of how US population is changing cannot be told using "percent change in population" as your measure. In this case, doing so overemphasizes small changes in rural areas and obscures the larger changes.
A better way to visualize the magnitude of population change by county is to normalize the numbers to reflect the population change by land area. My version (below) shows the change in population divided by the land area (in square miles) of each county. In some ways it looks less exciting than Pew's and WaPo's, but it can tell a clear story that is less apparent in those versions. Unsurprisingly, we see that most population change is happening in metropolitan areas. But what you can see in this map that isn't obvious when looking only at percent change in population, is that there are a select few metro areas where population has declined: Chicago, Detroit, Cleveland, and Pittsburgh seem to account for much of the population decline from 2014 to 2015. Unlike in Pew's map, Loving County's 25 new residents no longer dominate the picture. And unlike WaPo's version, it is now clear that population gains outweigh the declines.
He raises some good points: When showing data that can be both positive and negative, it makes sense to use a diverging color scheme. But I think a main limitation of both of these maps is the measure they have chosen to map: the percent change in population in each county.
Percent change can certainly be informative, but on a map, it can tell you nothing about the magnitude of population change in any given area. Looking at Pew's map, you might think that population growth centers around Loving County, Texas. You might also look at WaPo's map and conclude that the increases in the blue counties are roughly balanced out by all the red counties where population has declined.
Neither of these things are true: Loving County's "boom" was actually just an increase of 25 people from 87 in 2014. The red, declining counties in WaPo's map are mostly rural and sparsely populated, while many of the blue, growing counties are the most densely populated counties in the country. The full story of how US population is changing cannot be told using "percent change in population" as your measure. In this case, doing so overemphasizes small changes in rural areas and obscures the larger changes.
An Alternative Approach
If you want to see the magnitude of population change in a given area, you could just plot the raw numbers for each county. But there is a problem with that approach. Counties differ in area for many reasons that are unrelated to population. For example, San Bernardino County added 18,000 people over an area of 20,000 square miles while the Bronx packed an additional 14,000 people into an area of just 42 square miles. The purpose of map visualizations is to show how data varies across space, so it doesn't make sense to give so much more visual weight to San Bernardino County's population change than to that of the Bronx based on nothing more than where the county lines were drawn.A better way to visualize the magnitude of population change by county is to normalize the numbers to reflect the population change by land area. My version (below) shows the change in population divided by the land area (in square miles) of each county. In some ways it looks less exciting than Pew's and WaPo's, but it can tell a clear story that is less apparent in those versions. Unsurprisingly, we see that most population change is happening in metropolitan areas. But what you can see in this map that isn't obvious when looking only at percent change in population, is that there are a select few metro areas where population has declined: Chicago, Detroit, Cleveland, and Pittsburgh seem to account for much of the population decline from 2014 to 2015. Unlike in Pew's map, Loving County's 25 new residents no longer dominate the picture. And unlike WaPo's version, it is now clear that population gains outweigh the declines.
Population change from 2014 to 2015 by county
Comments
Post a Comment