Thursday, November 21, 2002

In defense of gerrymandering: Are term limits a bad answer to a non-existent problem?

This is a new weblog. I have been intending to write an introduction of sorts for a while, but it was proving to be difficult. I figure it will be easier just to jump right in. I'll fill you in on the details later...

Peter Beinart of The New Republic lays out a case for term limits as a solution to the evils of gerrymandering. Matthew Yglesias agrees, and subsequently advocates for the Hare voting system, because non-partisan redistricting commissions "ain't gonna happen."

But wait a second, what's wrong with gerrymandering safe congressional districts in the first place?

According to Beinart:

[Right-thinking people] hate the fact that only about 10 percent of seats in the House of Representatives were competitive this election. Bipartisan gerrymandering following the 2000 reapportionment produced hundreds of safe Democratic seats, hundreds of safe Republican seats, and not much else. No wonder so few Americans voted, the right-thinkers sigh in disgust: In most House districts there was really only one candidate to vote for.


If every state followed Iowa's lead [and "handed redistricting over to nonpartisan technocrats"], American politics would be a lot more democratic, a lot fairer, and a lot more fun.

My instinct is that the main reason Beinart and other pundits/bloggers/whatever, oppose gerrymandering is because of the "fun" issue: elections would be more interesting to follow/write about if there were more competitive House seats. I can certainly sympathize with this argument; it's what I thought at first. But consider this: is there any benefit to gerrymandering other than to the congresspersons whose jobs are made safe?

I think there is. To illustrate: Let's assume that ideally, we'd want something like the system suggested by Matthew Yglesias. But also, in line with various comments on the above post, we also like the idea of a many constituents-one representative system, because keeping track of multiple statewide candidates is bothersome and probably inefficient. What would be a good way of solving this problem? Say for instance we have a state that is 40% Democrat and 60% Republican and has 10 seats in the House. We'll assume for simplicity's sake that everyone votes, that there are no independent voters, and little ideological variation in statewide candidates. We can always relax these assumptions later if need be. (My educational background should be very obvious by now.) Under the Hare voting system, the people of the state will elect 4 Democrats and 6 Republicans. But now say we want to go one step further and assign one representative to each constituent. How should we do this optimally? One way would be to attempt to give everyone their first choice (i.e. who they voted #1 in the election), or maybe second or third choice if any one congressperson was more popular than average. This will create non-contiguous, overlapping, "virtual districts." But think for a moment about what these "virtual districts" are likely to look like. Since many people can't be expected to be familiar with more than a couple of House candidates, and because people also vote their local interests as well as their partisan ones, the first or second choice of many voters will be the candidate that lives or operates in a locality nearest them. So these virtual districts will most likely be at least concentrated geographically, if not strictly contiguous. (You could even increase the geographic concentration by using the distance from the legislator as a criteria for deciding who gets their first choice, in the case of more popular than average candidates). Furthermore, this solution is Pareto optimal, meaning you cannot make anyone better off (by assigning a representative higher up on the voted preference ranking) without making someone else worse off.

So far, a pretty good system. Everybody is represented by one of their top choice legislators, and every Republican is represented by a Republican, every Democrat by a Democrat. The partisan make-up of the congressional delegation even reflects the partisan make-up of the state, wonder of wonders. However, someone always has to spoil the party, so lets impose geographically contiguous districts. We could first do this in a way which preserves partisan representation. We draw borders and define 6 contiguous Republican districts and 4 Democratic ones. These R and D districts overlap each other, but no D district overlaps another D district. (Each person lives in two districts, an R one and a D one, but has only one representative, determined by party affiliation.) Are we happy now? We are until some bozos come along and say "all politics is local" and now insist that we have 10 distinct contiguous districts. So we do this with computers; we draw borders around each representative according to some algorithm that maximizes the number of original constituents contained within each district and voila! What do we end up with? That's right: odd-looking, gerrymandered, safe congressional districts. And there was much rejoicing.

Does this make any sense?

Alternatively, I'll argue the case in the other direction. Suppose we have a two horridly odd-looking, gerrymandered, safe congressional districts. One regularly elects a Democrat with 80% of the vote, and the other elects a Republican with 70% percent of the vote (note: fewer absurd assumptions than the previous example). Now a non-partisan redistricting committee comes along and makes the two districts more compact, more competitive, and hence, more "fun." Whatever the election results are, the actual citizens in these districts are worse off. Before redisctricting, 75% of the population of the two districts were represented by someone they personally voted for, and who shares their views. After redisricting, probably only 55-60% of the population is represented by someone they voted for. Furthermore, the two districts could conceivably go to two Republicans or two Democrats, which wouldn't represent the partisan make-up of the population as well as 1 and 1 would.

Finally, as "fun" as it would be, do we really want 435 competitive House districts? I think I've rambled on long enough, so I don't want to get into the specifics of the argument for why this would be a bad idea, so I'll leave it as an exercize for the reader.

Now. After all that. I certainly don't want to give the impression that I think partisan gerrymandering is a good and beneficial thing. In the first example above, a Republican statehouse could create 10 districts, each containing a 60% Republican and 40% Democrat population, and thus sending 10 Republicans to the House rather than 6. That would be unfortunate (for the Democrats at least). However, in an uncertain environment, this is not really in the self-interest of each individual representative, who would rather have a more comfortable margin. In this regard, California's "bi-partisan" redistricting effort, widely assailed as the most egregious example of incumbent protection, is actually the most optimal.

To sum, I don't understand the rationale behind asserted claims that gerrymandering is inherently either anti-democratic or unfair. But I would be happily enlightened in this regard.