24 January 2013

The (Poorly Executed) Geography of Abortion Access

I love maps. I love geography and cartography and information visualization and GIS and everything about maps.

But sometimes they aren't the best way to look at something. Most maps give you a look at an effect measured by area, not by person. Sometimes that's good. Often it's distorting.

After the 2000 election I kept seeing maps like this, and invariably some Rightist would crow about how such a vast majority of America was red, as if square miles counted or elections were a tool for determining the opinion of parcels of land instead of citizens.

Here's a Lefist version of the same error.
The Daily Beast | Michael Keller & Allison Yarrow | The Geography of Abortion Access

Four decades after Roe v. Wade, the U.S. has 724 remaining abortion clinics. From the dearth of clinics through the country’s center to the states requiring in-person counseling, see some of the findings from our abortion map.

The Panhandle-Dakotas Divide

The clearest trend on the map is the dearth of clinics through the center of the country—from northern Texas through Kansas, Nebraska, South Dakota, Wyoming, and North Dakota. Roughly 400,000 women of reproductive age (between 15 and 44) live more than 150 miles from the closest clinic in this region. The county farthest away from an abortion clinic is Divide, N.D. All of these states except
The problem with this is that it's essentially indistinguishable from a population density map. You could relabel this "distance people must travel to get to ______" and virtually anything with about 500 to 1000 locations in the US would fit in that blank.

And that's how it should be. You know why there's a great big gap in the western Dakotas? Because nobody lives there. What do you want this map to look like, if not this? Should clinics be placed on a uniform cartesian grid? That would be insane.

This either needs to be re-done as a cartogram, or not done at all, because as is it provides negative information.

Site placement is a hard problem. Literally, it is computationally impossible to find optimal solutions for many types of problems. But there are numerous techniques for very good approximations. I would be far more interested in this as a piece of work if Keller, Yarrow, or whoever actually coded this took the effort to compute an ideal placement of abortion clinics based on population data and then compared it to the actual placement, rather than just giving me this population density proxy.

As is often the case, Randall Munroe has already made this point better than I can.

PS 400,000 women sounds like a lot, but there are over 61,000,000 women in that age range, so we're talking about less than three quarters of one percent of reproductive-age women. Add in the data point that less than half of women think "abortion should be generally available" (and probably much less than half in the low density/rural parts of the country we're talking about here) and this looks like a non-finding to me.

PPS The Bastiat Institute shows you how that first map really ought to look:

(Although this is for 2012, not 2000, the principle stands.)

No comments:

Post a Comment