Tuesday, October 8, 2013

The Complicated Relationship Between Redistricting and Polarization

Ezra Klein yesterday offered a helpful list of "The 13 Reasons Washington is Failing." I recommend it, but I do want to quibble with his characterization of redistricting. As Ezra says,

Whether it affects the partisan balance or not, it clearly packs candidates into less representative districts — which makes it easier for them to ignore popular will.

In fact, that's not at all clear. Sometimes redistricting does precisely the opposite.

There are typically two major strategies a state legislature will pursue when conducting a redistricting:

  1. Protecting incumbents. When this is done, districts are drawn to make incumbents safer -- Republican districts become more conservative, Democratic districts more liberal. This could contribute to polarizing the legislators who represent those districts; they are less likely to be punished for extreme behavior. This strategy is often pursued when a state has split party control, and protecting all incumbents makes everyone happy.
  2. Enhancing the majority party's share of seats in the chamber. When this is done, many districts are redrawn such that they become more competitive. This gives the majority party a better chance at expanding its numbers. You don't need all your members winning with 80% of the vote if they can still win with 65% of the vote and give you a shot at winning a few more seats. Notably, this form of redistricting contributes to depolarization -- legislators become more likely to be punished for extreme behavior. This strategy is often pursued when a state government is under unified party control.

As should be clear, these strategies work against each other. To the extent that you're protecting incumbents, you're not increasing your party's seat share, and vice versa.

And yet we frequently hear claims that the 2012 redistricting did both at the same time. In the post linked above, Ezra is claiming that the redistricting made Republican seats safer. And here's an NY Times piece claiming that it gave Republicans greater seat shares.

Now, of course, some states may be pursuing the first strategy while other states are pursuing the second. And, in fact, the political science evidence on this front (see here, for example) suggests that the effect of redistricting on polarization is marginal, if it even exists. But if it does exist, let's try to be clear about what redistricting is doing, and not simply blame it for whatever we currently don't like about politics.


  1. This comment has been removed by the author.

  2. Nice post Seth. This is one of the points that the journalist/layperson typically doesn't consider. Every time I speak to a journalist about this, they are surprised by this, but they immediately see the logic. I think the other issue that is important to note on the question of representatives is that the term is ambiguous. Two districts with the same mean of voter preferences, can have wildly different distributions underlying it which dramatically changes the types of policy that legislator might/can/will support (e.g., tight unimodal vs. extreme bimodal) depending on the subconstituencies s/he is beholden to.


  3. Excellent--this SO needs to be said. Tremendous confusion out there.

    FWIW, I have a paper coming out in LSQ that shows (among other things) that there is no relationship between a seat-maximizing gerrymander and responsiveness/competition. I suspect that's because a seat-maximizing plan only makes competitive seats on one side of the aisle.

    But your overall point is completely right: Rs would have a hard time gaining more seats through redistricting AND drawing a lot more solidly safe R seats.

  4. Sorry, I can't find any joy here. Redistricting is clearly the root of Republican intransigence. When you isolate a segment of society for clearly economic reasons, something will go sour. Redistricting
    should be done by computer, no human hand.

    1. This is not something that can be "done by computer," as there is no objectively correct way to do it. Humans have to decide what values are important when making new districts. Equal populations? Compactness? Politically competitive? Preserving local community lines? These goals often work against each other. Computers can't sort that out for us.

    2. I that what Mr. blogspot.com is getting at is having some sort of stochastic element more than using an actual computer—it could involve a big bingo hopper, or a bunch of D&D dice. But something that procedurally-but-not-deterministically draws the lines so that, while the two teams may agree to the rules, they can't completely predict the outcome. This doesn't get rid of the problem, of course, but does at least abstract it.

  5. I wondered about this myself - how much, if any impact there was in 2012 in PA. I tallied the vote counts:
    Rep Candidates
    Dem Candidates
    Other Candidates

    Of the 18 house seats 13 went to Rs and 5 to Dems.

    While I would agree that gerrymandering is not the sole cause of issues with representatives, it seems clear to me that there is certainly something fundamentally wrong if a party can garner more votes than another and still be out represented nearly 3-1 in house seats awarded.

    1. I agree that this is troubling, but gerrymandering need not be the cause of it. Democrats tend to have many "wasted" votes. That is, there are many urban districts in which the Democratic candidate gets 80 or 90% of the vote. There aren't that many similarly-lopsided Republican districts. So even if Democrats win more votes overall, a lot of those go to a small number of districts and don't affect the outcome. This is not due to gerrymandering so much as geographic residency patterns. Lots of Latinos and African Americans live in densely populated areas like South LA, upper Manhattan, etc., and it's hard to draw a district in such areas that doesn't end up at least 80% Democratic.

  6. I would argue that in PA, you first had a big Republican wave election that created a whole bunch of new Republican incumbents. Then you had an incumbency protection gerrymander that kept the newly elected Republicans (plus Dem Matt Cartwright) safe.

  7. Hi, this is Devin McCarthy of FairVote. Great to point out that gerrymandering is given an inordinate amount of blame for polarization.

    I would object to the claim that states couldn't increase their share of seats while still protecting incumbents, though. The strategies work against each other in theory, but in practice there is no difference between an 80% district and a 65% district for one party. Both are not competitive at all. No incumbent in a 65% district came close to losing in 2012. That 65% incumbent has no more incentive to moderate his behavior than the 80% incumbent.

    The 2011 redistricting did both increase the number of districts where Romney outperformed his national average and the number of safe seats for both parties. That's because most Republicans were drawn into districts that are around 60% Republican - safe enough that they have no real threat of losing in the general election, but not so safe that they waste many votes.

    That said, while gerrymandering can exacerbate the Republican bias and number of polarized districts, they would likely exist regardless of how lines are drawn as long as we keep using single-member winner-take-all districts, just because of where voters live.

  8. Actually, when a party practices gerrymandering to increase the number of it's own seats, it may be making moderate seats for itself, but it does so by making extreme seats for the OTHER party.

    So, what's going on is that Republicans engage in gerrymandering to enhance their numbers. This makes Republican office holders more responsive to the part of the public they represent, while making Democrats more extreme, because Democratic districts become less representative of the nation as a whole. (Boehner won his last election with 65% of the vote, Pelosi rarely gets less than 85%)

    Democrats then perceive Republicans as having become more "extreme", because Democrats measure this by how much Republican positions diverge from their OWN positions, which are actually becoming more extreme relative to the general public. Because Republican gerrymandering has made Democratic districts less representative of the general population.

    1. Yes, that's an important point! We rarely hear how the majority party's redistricting has made the minority's districts more extreme.

  9. "some states may be pursuing the first strategy"

    Yeah, your article was informative (you're right that consolidation and expansion aren't the same thing) but misguided: you can pursue both strategies, the GOP did pursue both strategies, and anyone else that wanted to both secure and increase seat counts would, as well, falsifying your 'it is one or the other' argument entirely. A distinction without a difference, as they say.

  10. Actually, you CAN do 1 & 2 at the same time. Allow me to illustrate with a simple example. Imagine you have a state that has 500 voters. Exactly 250 Democratic voters and 250 Republican voters, allocated to 5 districts as follows:

    District 1: R-60 D-40
    District 2: R-55 D-45
    District 3: R-50 D-50
    District 4: R-45 D-55
    District 5: R-40 D-60

    If you want to give Republicans a big advantage (and assuming enough Democrats live in contiguous areas) you could do this:

    District 1: R-62 D-38
    District 2: R-62 D-38
    District 3: R-62 D-38
    District 4: R-62 D-38
    District 5: R-2 D-98

    Viola! Observe: every district is more extreme than it was before. And yet every incumbent is also more protected than s/he was before.