Gerrymandering: Tough to Avoid

One of the most attractive scapegoats for the polarization of Congress is the practice of gerrymandering, the deliberate drawing of House districts to achieve a desired result in congressional elections. Because the districts are (for the most part) drawn by state legislatures, it's easy to see why this is a tempting argument: it places blame squarely on the politicians in Carson City or wherever that the people dislike so much. (Don't take it personally, Nevadans--I used you as the example because by several metrics you're the most gerrymandered state in the Union. It's only partially your fault, which I'll get to. For the most part I'll be dumping on Pennsylvania.) Pennsylvania is actually one of the less gerrymandered states for its size (it's apportioned 18 congressional districts, so there's a lot of room for creativity):

Pennsylvania House districts, 113th Congress (2013 - 2015).

You can guess pretty accurately where the cities are in Pennsylvania--Philadelphia is the three districts in the southeast corner, the big angular one in the northeast contains the city of Scranton, and the 14th district in the west is essentially an attempt to stuff as much of Pittsburgh as possible into one district. (Mind you, it's worked: the 14th district has a PVI of D+4, while the two surrounding districts are R+6.) The 14th district is the result of a gerrymandering technique known as "packing", where the supporters of one party (in this case, Democrats) are squeezed as tightly as possible into one district to diminish their influence in neighboring districts. This is the opposite of "cracking", which breaks apart a voting bloc to ensure that it doesn't make up a large proportion in any district.

One metric you could look at to determine how gerrymandered the state is is to look at the shapes of the districts and how it compares to a circle. (A quick way to quantify this is to look at the perimeter-to-area ratio. A circle has the lowest perimeter-to-area ratio: it's pi, 3.14159.... But as you add more sides, the perimeter grows much faster than the area does.) You could also just qualitatively look at it and see how many districts look like jagged snakes, which is exactly how the term "gerrymandering" was first coined: a Massachusetts newspaper spotted a district, drawn by Gov. Elbridge Gerry, that looked uncannily like a salamander.

Another metric that's probably more relevant to the politics of gerrymandering is to compare the party control of House districts to the proportion of the statewide popular vote they received. By this measure Pennsylvania actually rates as much more gerrymandered than the squiggles on the map would indicate--while Democrats received a slim majority of the statewide popular vote, they won only 5 of the 18 (28%) districts.

Giving blame where it isn't due

Gerrymandering appears to be one of the stereotypically sleaziest practices in state politics, but it can't be blamed for everything. We can see this by comparing the results of the current system to a different system designed to remove partisan bias from district drawing. My personal favorite is the shortest splitline method, which uses a mathematical algorithm that divides out districts by population distribution. C. G. P. Grey has a great video on YouTube that goes into more depth, but here's the general gist:

1. Find the shortest line that divides the state into two equally populous groups. If you need an odd number of districts, divide the number of districts you need in two and take the ratio of the two adjacent integers. For example, if you needed five districts, 5/2 = 2.5. The two adjacent integers are 2 and 3, which tells you that you'd need to divide the area into two groups that have a 3:2 population ratio. Again, you'd use the shortest line possible that does this.
2. Repeat step 1 as necessary for each group until all of your districts are drawn.

 The Center for Range Voting, which first proposed this method, has drawn maps for several states. Unfortunately, they haven't drawn one for our example, so we took the liberty of drawing roughly what Pennsylvania's districts would look like using the shortest splitline algorithm:

2010 redistricting of Pennsylvania under the shortest splitline algorithm.

The thickness of the lines represents the order in which they were drawn--thickest first, down to thinnest. We've also estimated, based on the results in each county, which party would have won each district based on the results of the 2012 House election:

2012 House results in Pennsylvania based on shortest splitline redistricting.

Interestingly enough, the situation hasn't improved all that much for Democrats. They've retained their seat around Scranton, their Pittsburgh voters now take up two districts instead of one, and Philly has four districts instead of three. But that still leaves them with only 7 districts to the Republicans' 11 in a state that should be divided evenly between the two parties. The problem is simply one of population distribution--when Democratic voters are crammed together in densely populated cities and Republican voters are spread out, it becomes very difficult to split them across districts in a way that mirrors the state.

The problem can't be resolved under the current voting system (a mixed-member proportional representative system would fix most of the problems). The problem is that whatever you may think of gerrymandering, there's going to be some "natural gerrymandering" in any state with a large city, simply because of the voting blocs the parties have decided to appeal to: it's easiest to "pack" Democratic voters, and it's easiest to "crack" Republican ones. While a mathematical algorithm would remove intentional partisan bias from the districting process, it doesn't mean that the resulting districts would reflect the balance of the state overall.

My point is this: while we can recognize that gerrymandering is a bad thing, there is no cure-all for it that wouldn't exacerbate the problems that gerrymandering causes. We could theoretically hire a bunch of expensive consultants with expensive computers in expensive briefcases to craft districts so that the partisan balance best reflected the votes of the entire state--but the easiest way to ensure that would be to create, in Pennsylvania's case, 9 safe Democratic seats and 9 safe Republican seats. And that, in the end, is one of the major problems of gerrymandering: the seats are too safe, and while the parties in Pennsylvania get represented accurately in Washington, it's not necessarily so for the voters. Gerrymandering, to be honest, is not much worse than the other political problems we have now--we'll have to look elsewhere if we want to play the blame game fairly.