Maps for Social Good

The Economist had an interesting article about the increasing use of maps by non-profits and other socially-oriented organizations. These maps are used as a more intuitive way to put across data than the more typical charts, graphs, and spreadsheets of numbers.

While this is cool and has many advantages, I do wonder about this kind of thing:

“I remember the first supermarket-commission meeting,” says Jennifer Kozlowski, special assistant for the environment to David Paterson, the governor of New York. “Some of the maps in the report mapped obesity-related deaths and access to produce markets. It was as clear as day that something needed to be done.” In January Mr Paterson announced the Healthy Food/Healthy Communities Initiative, including $10m in grants and loans for supermarket projects in under-served communities.

Maps that show the coincidence of two variables (such as obesity-related deaths and produce access) may be powerful yet misleading if viewers interpret the data as causal. For instance, the produce-obesity link could go in another direction: people have food preferences such that they want and are willing to buy junk food, but not fresh vegetables; this makes them obese and also means there is not sufficient demand for produce markets in their area. I'm not saying I believe this alternative explanation, but merely showing that alternatives do exist and that the policy implications differ depending on which is true. If it's an issue of preferences, not only availability, simply making fruit & veg more accessible will not lead to a big increase in their consumption. While it feels quite logical to assume that lack of produce availability leads to obesity, the map can't demonstrate that. However, the map can be a strong persuasive tool to convince others of the legitimacy of a particular interpretation of the data.

This isn't a new phenomenon, or unique to maps; data used for political purposes is generally spun in some fashion. But it's funny (in a grim way) that statistical analyses have the reputation of being manipulable ("lies, damned lies, and statistics") while something like a map is more or less believed to be objective truth. Just looking at the example map in the article, clearly somebody had to decide what constitutes "higher rates of childhood obesity" to create those orange circles. (The gradations for number of parks, or park density, or whatever is meant by more or fewer parks also were chosen by someone, although that's at least not a dichotomous variable.) How do we know that the cut-offs weren't rigged to produce the most dramatic possible map? How do we know that other important variables aren't creating a spurious relationship? Are the mapped obesity rates adjusted for demographic variables?

It's possible as maps become more used in making political arguments, people will become more skeptical about what they represent and sophisticated in their thinking about them. But it seems like the better maps are at making things look obvious and concrete, the lazier people will be in analyzing what the data are and more importantly are not saying.