« Sky Captain and The World of Tomorrow | Main | Kerry and the Draft »

September 21, 2004

Taking on Gerrymandering

Matt Yglesias proposes an interesting solution to the problem of gerrymandering. I've never seen this particular brand of voting scheme before, so I wanted to take a look at it and see if it might be a viable solution to gerrymandering.

I should first explain why I think gerrymandering is a problem. For non-politically oriented readers, gerrymandering is a process of designing voting districts guaranteed to vote as the designers want. For example, let's take a look at Colorado, since I live there. We have seven Congressmen: five Republicans and two Democrats. Yet Rasumussen currently shows Colorado as a toss-up. In 2000, President Bush won Colorado, but only with 50.75% of the vote. How can we have such a strongly Republican set of Congressmen if the state is so narrowly divided? Obviously, the answer is gerrymandering, but what does that actually mean?

Named for Elbridge Gerry of Massachusetts, a signer of the Declaration of Independence and one of the less well-known founders, gerrymandering is a relatively simple process. Let's say that I want to maximize the number of Republicans in the House. I take a look at the map and see that the vast majority of Democratic votes come from certain areas. I can then do one of two things: spread those voters out so that they're paired with areas that vote heavily Republican, weighting the distribution so that the Republican-heavy areas will outweigh the Democrat-heavy areas; or I can put the Democrat-heavy areas in as few districts as possible, so that the Democrats will have a few guaranteed seats, but the vast majority of Democratic voters are penned into those few districts. Colorado tends to do this naturally, as Denver is the most heavily-Democratic part of the state, and it holds such a large share of Colorado's population, there has to be a district almost entirely devoted just to Denver. The flip side of this is districts like Colorado's 3d, which by eye looks to take up almost half the state's landmass. In any case, the final result is that the Congressional delegation from Colorado is not truly representative of the voters of Colorado.

This isn't necessarily a problem (indeed, looking at the map, I'm not sure that it is in Colorado) if the districts are drawn reasonably. Cities do tend to have more Democratic voters, while rural areas tend to have more Republican voters. Because Representatives are supposed to represent a specific bloc of voters, districts based on geography may still be lopsided one way or the other. The problem occurs when we get districts like this. Here we see a district that exists solely for the purpose of electoral advantage. While you will always have a few people living on the borders of an election district, this district is nothing but border. The representative in this district may be representing people who tend to vote similiarly, but they probably have very little else in common. This circumvents the intent of the Constitution, which designed the House of Representatives as the voice of the people in the federal government, men chosen by the people to represent the people. Instead, most of the House is now representative who choose which people to represent, because the districts are so carefully designed to guarantee a representative of the right party or ethnic group.

This results in a Congress that is far more partisan than the country as a whole. Because the districts almost invariably will vote for a certain party, that means that the primary election is what decides who will represent the district. And primary voters are much more partisan than voters in the general election, which means that the representatives will tend to be more extreme. The difference can be seen in how the parties select presidental candidates as opposed to Congressmen: someone like Cynthia McKinney or Bob Barr could never win their party's nomination for President, but both served in Congress (and McKinney, sadly, will again), leading to asinine comments like McKinney's accusation that President Bush knew about 9/11 in advance. Worse, it leads to the situation we see today in Congress, where neither side bothers to compromise, the majority party simply rams through its preferred legislation. This is not good for the country, as it means Congress tends to spend lots of time on fringe issues while often setting aside issues that a majority of the people are more interested in.

To address this problem, Matt recommends the Hare System of Proportional Representation. This system asks each voter to fill out a ballot with numbers: #1 is your preferred candidate, #2 is your first alternate choice, #3 your second alternate choice, and so on. The number of votes required for selection is determined by a formula: number of votes cast divided by the number of people to be elected plus one, plus one. (NV/(NE+1)+1). People who garner votes beyond the minimum required are selected, but their additional votes are then harvested for #2 preferences. If those votes put other candidates over the minimum required, they are harvested for #3 preferences, and so on. Let's look at an example.

In Colorado, based on current population estimates and voting habits, let's assume 1,923,083 votes are cast in November. Since we're selecting seven representatives, the number of votes required for election is 240,386. Again using the Rasmussen figures, let's assume that Bush voters vote for Republican candidates, Kerry voters for Democratic candidates, etc., in the current percentages. That would be 884,618 votes for Republicans, 865,387 votes for Democrats, 57,692 votes for Naderites, 38,462 votes for Libertarians, 19,231 votes for Constitutionalists, 19,231 votes for Greens, and 38,462 votes for other candidates. That means three Republican representatives with 163,460 votes left over and three Democratic representatives with 144,229 votes left over. Assuming Naderites and Greens next choices are Democrats, and Libertarians and Constitutionalists prefer Republicans, the Republicans end up with 221,153 votes to the Democrats' 221,152. Neither has enough to clear the threshold level, so the choice of the final representative would come down to the choices of the 38,462 votes who went away from the known parties. But the best case for the Republicans would be a loss of one of their seats, leaving them a 4-3 majority in their Congressional delegation, and possibly even putting them at a 3-4 disadvantage.

Proportional voting might also result in minor candidates reaching office. People like me, who prefer to vote third party, but who generally vote major party to make sure that their vote counts, would be free now to put (for example) Greens at the top of our ballots, backing them up with (for example) Democrats. That creates the possibility of a minor party candidate clearing the threshold level. For example, this year Republicans aren't necessarily thrilled with their presidential candidate. If one in four put a Libertarian at the top of their ballot, that would be sufficient to put a Libertarian in Congress. If the vote reports noted the total #1 votes all candidates received, minor party candidates could try to build momentum towards just such a result. It would still be an uphill battle, but the odds of getting a minor party Congressman under the current system are so long they're not even worth calculating. Under this proposal, their chances would still be small, but at least they might be extant.

There is a downside to this proposal that Matt doesn't note, however. Representatives would no longer actually represent anyone. Since we're talking about a statewide ballot, Congressmen would simply know that at least the threshold number of voters preferred him or her, but not which ones. Who would the House of Representatives represent? The original intent of the House was to allow the people a voice in the federal government, as I already noted. While it's difficult to do without a big contribution, you can meet with your Congressman far more easily than your Senator or the President. Under this system, you would no longer have a Congressman (or, more precisely, you'd have as many as your state had, but none of them would be directly linked to any of their constituents). That, and the requirement to run statewide campaigns, would seem to be tailor-made to further increase the influence of special interest groups at the cost of individuals. While I'm not as excited about special interests as many (they are, after all, generally made up of...people), I don't know that this is necessarily a good idea. Personally, I'd rather see the House expanded to meet the original Constitutional minimum of one representative per 30,000 people, as that would give the people a real opportunity to reach their representative (albeit at the cost of individual representatives having very little pull). Be that as it may, I'm not convinced that proportional voting would be the best way to select Congressmen, at least not if we really want them to represent people at all, rather than parties.

References & Notes:

My example balloting in Colorado assumes that all voters vote straight party line, so that a Republican would place all seven Republican nominees at the top of his ballot, a Democrat all seven Democrats, etc. This is a simplification; those who want to explore the various permutations are welcome to do so in the comments.

Posted at September 21, 2004 06:12 PM

Andrew Olmsted

Trackback Pings

TrackBack URL for this entry:
http://andrewolmsted.com/cgi-bin/MoveableType/mt-tb.cgi/845

Comments

You ought to look at Ireland. That's how they do it there.

Posted by: Scott Kitterman at September 21, 2004 06:45 PM

I'm kind of fond of the straight line test. Any point in an electoral district must be reachable by a straight line drawn from any other point in the district without crossing the district boundary at any point with the sole exception being district boundaries that coincide with state borders. A +/- fudge factor in terms of population levels from one district to the next might be needed to make this work. I'd also like to see the 17th Amendment superceded and go back to state appointment of the US Senate.

Posted by: JSAllison at September 22, 2004 08:19 AM

You say "Because Representatives are supposed to represent a specific bloc of voters," which is incorrect.

Why did Matt choose Hare over Droop, or Hare over Imperiali? The mathematical basis of your endorsement seems lacking.

The rational choice would seem to be to develop a determinative computer-generated system. The number one value should be compactness of district, but where space is divided not as depicted on maps, but based on the location of roads (for cars) and other forms of transport.

In other words, it would be minimizing the time it would take you to get to the center of your district.

There has to be a second value, past compactness, to make the process determinative. Otherwise I see a gazillion different computer-generated compact maps, each different one derived with different algorithms and/or starting points.

Posted by: Josh Narins at September 22, 2004 10:09 AM

I'm kind of fond of the straight line test. Any point in an electoral district must be reachable by a straight line drawn from any other point in the district without crossing the district boundary at any point with the sole exception being district boundaries that coincide with state borders.
--JSAllison

This is a nice, but somewhat naive idea. By requiring the distrincts to be convex (which is what the 'straight line test' does) you're forcing all of the (non-state) district boundaries to be straight lines. Moreover, since the earth is not flat, you'd probably end up having to specify exactly what you mean by straight line as well (shortest geodesic *might* work.)

The rational choice would seem to be to develop a determinative computer-generated system. The number one value should be compactness of district, but where space is divided not as depicted on maps, but based on the location of roads (for cars) and other forms of transport.
--Josh Narins

Depending on what you want districts to do, that's also not necessarily a good plan. Minimizing district diameter (more on this in a second) is liable to get districts that have a large rural area and a nibble of a city. I'm not sure that it's possible (in the general case) to show that a particular districting plan is numerically optimal in that sense.

It's unclear what you mean by 'compactness'. Possible examples include minimal boundary length, minimal total diameter, or minimal variation in population density per district.

Moreover, any 'deterministic' algorithm for districting is going to have to do some arbitrary descisionmaking, even in some really simple cases.
For example, let's say that there's a circular state with two congressional districts and constant population density. How do you decide which way to cut the state? This type of problem really doesn't get much better with more complicated population densities.

Personally, what I find interesting in the discussions of gerrymandering is that people generally object to redistricting for political gain, but generally do not propose what the goals for a redistricting plan should be. It's a bit preposterous to suggest that a particular algorithm for redistricting (or for voting) is a solution to a particular problem in, or an improvement of a process when the goal of the process is unclear.

Posted by: Incognito at September 22, 2004 02:30 PM

I think the important thing to consider is that while grouping people together by residence seems the 'natural' way to do it to most people in the Anglosphere, it is not the only way to do it. Speaking for myself, I think that I have more in common with someone who believes in the same kind of politics or ideology that I do than with someone who just happens to live in the same county. Grouping people by party or candidate support is just another way of approaching a task which is fundamental to representative government.

Posted by: A D Moore at September 22, 2004 04:24 PM

You say "Because Representatives are supposed to represent a specific bloc of voters," which is incorrect.

I do so love argument by assertion. Now, I will concede that the Constitution does not say that Representatives have to represent anyone more specifically than the people of a particular state, so it does appear correct that such a new scheme would pass Constitutional muster. But clearly Representatives are supposed to represent specific blocs of voters, just not necessarily in the way we've done it historically.

Why did Matt choose Hare over Droop, or Hare over Imperiali? The mathematical basis of your endorsement seems lacking.

Since Matt and I are not actually the same person, might I suggest you're more likely to get an answer to your first question from Matt? As for my 'endorsement,' I'm not sure where you get the idea I'm endorsing an idea when I specifically state that I'm not convinced by it.

My own personal view is that district boundaries ought to follow preexisting boundaries (counties/town lines) and form generally contiguous districts where the people being represented have something in common. But, let's face it, reasonable districts tend to be much like art or pornography; we know them when we see them, but we generally have difficulty defining how to identify them.

Posted by: Andrew at September 22, 2004 09:09 PM

I have one objection to the idea of creating more congresspeople. I roughly calculated we would need about 9300 congresspeople to represent the current population. This would be ok with me if they used the present salary and benefits of the existing congress as a cap. That is, keep the cost of congress constant as the number of congresspeople expands. This would have the added benefit of getting most of these people back in the real world once they had used up there savings to be a congress person.

One other problem is the shear lack of space to get this number of people located in Washington. This could be alleviated by having congress meet over the internet I suppose.

Posted by: Dad at September 23, 2004 09:30 AM

Post a comment




Remember Me?

(you may use HTML tags for style)