The choropleth problem

Before a reader draws me up short on Monday’s link to an interactive map showing explosive growth of unemployment in the US, I should acknowledge the choropleth problem. James Fallows introduced the issue, and the word, in a blog post about the same map Tuesday.

The problem is that geography does not equal population. A choropleth map depicting social trends (unemployment or election results) can mislead if its geographical units (states or provinces) vary widely in population. (The word derives from Greek terms for “area/region” + “multiply.”) Fallows gives the example of the razor thin 2004 US presidential election, in which the Democratic candidate outpolled the Republican, but a state-by-state choropleth gives the impression of a Republican landslide, because lower density Republican states take up most of the room. Maps that resolve to a county-by-county level (as opposed to state-by-state or province-by-province) greatly reduce but do not eliminate the distortion.

Cartograms are maps that attempt to solve the choropleth problem by distorting their geographical units to reflect the numerical value being measured. The county-by-county cartogram of the same election (at left) distorts boundaries but reflects the results more accurately. More here, here, and here. My earlier, much-complained about post about altitude maps depicting crime in San Francisco, relied on a special category of choropleth known to cartobuffs as a prism map.