How To Save American Democracy
A simple proposal
I think we can all agree that gerrymandering is bad. What many people do not appreciate is how serious the situation is now. I wrote about this in November – with computer technology and sufficient ruthlessness, Republicans have the ability to press a button and effectively turn off democracy. (Democrats too, though not as strongly.) Requirements that districts be compact or otherwise follow a rule of reason are not enough – you can get very close to the efficient frontier of gerrymandering with compact districts. If we do not fix gerrymandering now, we face one-party rule and autocracy forever.
The solution to this is proportional representation. There are a few main types of system, the most common of which are party lists. Each party gets seats in proportion to its total vote share, with some systems allowing you only to elect some number of candidates on the party list, and others allowing voters to have some influence over the exact candidates who are selected.
There are some disadvantages of list elections compared to our default of first past the post single member districts. It strikes me as good that members of Congress represent a distinct local constituency. One of the things they do, and for many the most important thing, are constituent services – basically acting as a prod to bureaucracy to do things. Only a few of them can be in leadership at any given time, and you don’t want to be totally useless. An effective requirement that the congressmen be geographically diverse also strikes me as a good thing. Finally, having many elections featuring many different possible wings of a party seems much healthier than a world where all elections are nationalized, and you can only vote the party’s platform up or down. That is, in essence, the U.S. presidential election – I do not regard that as a particularly healthy way of expressing the people’s preferences.
An alternative method to list elections would be to have voters have multiple votes, and then elect multiple members from a district. However, I do not think this is a good solution. First off, ranked-choice voting is indeed a significant enough mental burden on the population so as to encourage lots of errors. We saw this in the New York mayoral race, for instance – there were many ballots which were clearly filled out in a way that misunderstood how to optimize one’s vote. And, fundamentally, we cannot relax the problems of not having any local interest without giving the incumbent party the ability to manipulate the outcomes. Basically, anything in which there are discrete changes in the voting power as a function of votes gives scope for gerrymandering, because gerrymandering is exactly the same as bayesian persuasion. A national list election with proportional representation is like having a system of multi-member districts, except there is only one district. If you divided up the country into districts who each elect four, there would be considerable scope to get candidates just over the threshold and increase a party’s voting power. Thus, multi-member districts do not eliminate gerrymandering, and may even strengthen it.
Last of all, we exist now in a two party system. You will have a really hard time convincing the incumbent parties to give up their power to third and fourth and fifth parties. You’re going to have a hard time convincing incumbent congressmen to risk their seats, or to make their nomination contingent on the national party rather than on the local contacts they have accumulated over the years. Any system of proportional representation will have to be incentive compatible.
So here’s my proposal: keep your districts. In each election, the top two vote getters both go to congress, but with a catch. Each of them only gets a fraction of a vote equal to their proportion to each other. If there are only two candidates, A and B, who get 60 and 40 percent of the vote, then they go to congress with six tenths and four tenths of a vote, respectively. And if there are three candidates, A, B, and C, who have 50, 30, and 20 percent of the vote, then A and B go to congress with 62.5% and 37.5%, respectively.
This has a lot of advantages. First, it would be fully compatible with a two party system, and in fact likely strengthen it. Incumbents would love it, because it would allow them to stay in office longer. I view this as a good thing – right now, you do not want house leadership to be from a seat which is remotely competitive, and that skews what can get done. It also forces parties to run a candidate in every election, which I think is good. No group of people is ever fully without representation in Congress.
Unlike other systems, we basically eliminate gerrymandering, especially after accounting for the endogenous exit of third party candidates. There are no discontinuous thresholds at which representation changes. Even party-list proportional representation can be vulnerable to changes in the thresholds at which parties get representatives – back when Russia was not as totally repressive as it is now and required electoral dirty tricks, the Yabloko party, which consistently got around 6% of the vote, was kicked out of the Duma by changing the threshold from 5 to 7%.
There is one way to gerrymander under the system. If a party controls at least 67% of the vote in a district, then they can have their voters randomize their choice of the candidate they vote for. This is much, much harder to pull off than regular gerrymandering, though, and carries considerable risk — if you don’t pull off the randomization correctly, then you risk giving up a considerable portion of your voting power. The difference between getting 70% of the vote and 65% of the vote, if your voters are splitting, is 15% of the voting public having their votes totally wasted. In all but the very safest districts, I would expect it to be a straight, clean fight.
Big schemes to change American politics are a dime a dozen. I happen to think this is a good one, but it is still unlikely to be implemented. What I want people to take away from this is that gerrymandering is much worse than you think. If we cannot agree to ban gerrymandering, then democracy can be ended in America. The only check, frankly, is the inviolate nature of state boundaries, which means that the election for President cannot be deliberately gerrymandered. (Yes, the deck is stacked, but it is better it be stacked by accident).
Now maybe it all doesn’t matter, the legislature is powerless, and we are simply going to trade dictators (if we’re lucky). But if we want to have a legislature which represents the people, then we should implement this system.
That's pretty creative, and probably more realistic than a full-blown switch to PR. My sense is that political scientists who've worked through similar weighted proposals also figured it may create problems for how legislative power actually works in practice. There's some literature showing that the relationship between a legislator's formal vote weight and their real influence in coalition-building can be highly nonlinear (e.g., a representative with 0.35 of a vote might have functionally zero power if majorities never need them). So the proportionality you achieve on paper might not translate into proportional influence in practice. The anti-gerrymandering properties are real, but it's also not the biggest problems with our system IMO.
I love the idea. And here's a comparable case.
The reason we don't use such systems, is she to their being computationally heavy. Until computers, this wouldn't have worked. And we don't really want parliament leaders to do complex calculations when seeking majorities.
But now it's quite doable. The chief whip's phone can do it.
In the game of Bridge, clubs use system of scoring called "match points"/"pairs" whose only reason to exist was the difficulty to calculate otherwise.
Nowadays, one can use better scoring with computers. But most clubs still use the other system. But this is because it's effectively a different game strategically and everyone is used to it....