Princeton Election Consortium

A first draft of electoral history. Since 2004

The mailbag

January 5th, 2013, 4:00pm by Sam Wang

Gerrymandering makes for interesting mail! Here are some excerpts from activists, a journalist, political scientist, and a few redistricters.

Journalist Paul H.Rosenberg writes:

I still don’t grasp how your methodology is supposed to identify non-geographic partisan gerrymandering as a distinct phenomenon. If it did, that would be super. But I just don’t follow… doesn’t bother me — at this stage, anyway — to ignore other concerns, such as minority representation. I just want to understand the logic dividing geographic impacts on redisctricting from all the rest. Lumping all the rest under “potentially partisan” makes perfect sense to me.

Here’s my take. The general idea is that the US is diverse in many ways: population density, partisan preference, and ethnicity/interest groups. Given that diversity, we can get a sense for what these factors could do to the partisan makeup of a Congressional delegation, by sampling districts around the country to see how they would add up to State X’s vote totals.

My approach is built on the idea that one can account for the general relationship between all this diversity and seat counts. It includes all the factors above. What is left is factors that are peculiar to a state. Since all large states have cities within them, that is not a factor that can account for a particular state’s outcome being different.

Probably the right way to resolve this is to do the same calculation using 2002-2010 House election results. I will do that in the near future using the same approach. If there’s a big jump in the index I have designed from 2010 to 2012, that would indicate that something happened in 2011. Which brings us back to redistricting.

Paul also suggested separating states into three tiers of urbanization. That would certainly address the urbanization-traps-Democrats idea. I have found a certain lack of clarity as to the relative size of these effects. My current guess is that gerrymandering’s overall effect is about 2-3 times as large as that of urbanization. But that’s strictly a gut feeling.

Several people described some close-up views of redistricting. Brent Benson, who writes the blog Mass. Numbers, reports:

I had a first-hand look at the redistricting process in Massachusetts as my wife participated in the process as a State Representative. In my opinion, the Democratic supermajority in the Massachusetts state legislature bent over backwards to not gerrymander, going along with your proposition that U.S. gerrymandering is not symmetrical. I actually didn’t see much emphasis on incumbent protection in MA at the Congressional level, to the point that Barney Frank retired rather than face running in a district in which he didn’t feel comfortable and John Tierney came within a a few votes of losing his seat. That being said, all of the Congressional Dems won in MA.

Before anyone makes light of that last sentence…in six races where Massachusetts candidates had opponents, the vote share was 66% D, 34% R (and across all districts, it was 75% D, 25% R). Based on these numbers, a 9-0 split is well within expectations.

S.A. gave a view from Ohio:

Here in Ohio, they tried to institute an “anti-gerrymandering” issue using a citizen’s commission to replace the legislature-drawn districts. It failed miserably because it was unfunded and it was confusing.

As a numbers guy, I thought, “why not solve the gerrymandering problem numerically?” In other words, make some simple, numerical rules, easy-to-measure, easy enough to program, and let the legislature do its gerrymander thing with tight handcuffs on? No “split-line” confusion or anything, use existing precincts, etc.

Illinois State Representative Mike Fortner writes:

As a member of the minority in Illinois, I didn’t have much impact on the map here, but I am an avid supporter of redistricting reform and competed in and won the 2011 Ohio Redistricting Competition. Even though I am a Republican, the Democrats in Ohio used one of my entries as their proposal for Ohio congressional districts.

Since then, I have been developing a neutral redistricting model based on graph theory. It is based on geographic criteria, but some have claimed that any purely geographic model would be inherently biased in favor of Republicans. This is where your article peaked my interest, since it implies that any geographic bias is small compared to the effects of partisan gerrymandering. Your statistical model would seem to be one way to more thoroughly test that hypothesis.

I have started to apply my model to some of the states, but would like to confirm that it is sufficiently neutral. If you (or a colleague) are interested in testing the contention that there are sufficiently neutral criteria for creating districts, then I would offer a collaborative effort by supplying plans to test.

I heard from political scientists as well. Eric McGhee from PPIC and The Monkey Cage had useful comments, focusing in part on incumbency. An excerpt:

First, I think your analysis is more similar to Nicholas Goedert’s Monkey Cage post than you might think…both you and Nick get a total redistricting effect of about 12 seats for Republicans under a counterfactual where every state uses an independent commission. Second, I don’t think you can add this first effect to the 6 or 7 seats that you and I found for the 2011 redistricting. The 2011 effect is (presumably) due to increased Republican control….

[Finally], I hate to climb on my old hobbyhorse, but none of this analysis incorporates incumbency. It’s easy to discount the incumbency advantage because party is so important these days (indeed, the district presidential vote is a much, much better predictor than it used to be). Nonetheless, independent of district partisanship, there’s about a 10 point difference between a seat with a Democratic incumbent and one with a Republican incumbent….

One final thought: I’ve never argued that partisan gerrymanders don’t exist. Certainly, parties often *try* to gerrymander. Instead, I’ve argued (based on a lot of good poli sci evidence) that the intended results often don’t materialize, that there are many constraints that frustrate a party’s attempts at rigging the lines in the first place, that incumbency is often a big part of the story and further complicates efforts at lasting effects, and that any sort of national impact is usually difficult to find.

The first two points are good. Incumbency…this is connected to the question of building competitive districts, which is a priority for many redistricters.

But “Many constraints that frustrate a party’s attempts” to gerrymander? The evidence I’ve been laying out plainly contradicts this. Looking at Pennsylvania, North Carolina, Ohio, and Michigan, where a 50.6D-49.4R %vote split led to a 43R-18D seat split, clever people can find a way around those constraints.

Rob Richie from FairVote wants to find a way to toss the whole system nationally.

[Regarding nonpartisan redistricting commissions], progress has been slow and erratic – with the latest example being a massive rejection of reform in a statewide ballot measure in Ohio last year and a landslide vote to uphold Maryland’s gerrymandered congressional map. It’s hard to build a reform movement around independent redistricting…the message really is that one set of elites can do a better job structuring your representation than another group of elites….

So then the question becomes what might happen that could get us to a point where a national law could be imposed? What has a better chance of mobilizing people to fight for it — to have a coalition of key constituency groups that get something directly out of it? Although obviously a heavy lift, I think a statute to do fair voting has a much better chance of doing that, as it really empowers both individual voters and constituency groups in a way that is quite different than any single member district process. And it sure gets you to a whole different place if you win it.

Finally, I thank Corey Kane (soon to be Washington correspondent for the Houston Chronicle) for conversations on this subject some time back. He has a recent piece on gerrymandering here.


Tags: 2012 Election · House · Politics · Site News

24 Comments so far ↓

  • Leading Edge Boomer

    Iowa has been using an independent commission with real power for several cycles now. Both Ds and Rs recognized that a rational redistricting process would be good for both parties.

    California is a newcomer to this process.

  • William Ockham

    I have a couple of suggestions for ways to disentangle competing factors. First, the effects of gerrymandering out to be most notable in election years ending in 2 and least notable in election years ending in 0. Mobility, new voters, and voter death all contribute to changing districts. However, the amount of change is highly variable. Some districts are very stable and others have high levels of voter turnover. Identifying these distinctions might help to calibrate the analysis.

    Second, thirty years ago when I was in grad school, redistricting was viewed as an incumbent protection racket. the primary goal of each side was assumed to be the creation of safe districts and the goal of many “good government” types was a lot of competitive races. The difference in the states that Sam Wang has highlighted is that the state legislatures went for maximum short term partisan advantage. That translates into a few very safe seats for the other party and a lot of thin majorities for you. That seems to me to be a rational strategy in only a few scenarios. If you think demographics are definitely on your side statewide, this could work in the long term. If you really believe that the incumbency advantage is big enough to solidify your grip, then it makes sense. Otherwise, this seems like a very big risk.

  • Crazy Pete

    On that last point, I think it will be hard to get people on board for ending gerrymandering unless it is indeed a national, simultaneous change. Otherwise it will be seen as “unilateral disarmament”, as with efforts to get large states like California to apportion electoral votes proportionally.

  • Pat B

    You made the New Yorker. “Building a Better Democracy by Ending Gerrymandering” by Steve Coll. January 10, 2013 Daily Comment.

    • Sam Wang

      Oh, that’s good. Thank you!

      This is a good kick in the pants for me to finish the series – and to prove a point regarding the districting vs. redistricting argument. I have a way to kill that, but I need to polish up the graphs.

  • Mike Clinch

    Here’s how bad it is: In Ohio, there’s two districts where a Democrat and a Republican run unopposed – a majority-minority VRA district in Cleveland and John Boehner’s district.

    In the other 14, the highest GOP percentage in the districts they won was 63.7 percent. The lowest percentage of Democrats in districts they won was 67.8%. It’s almost as bad in Pennsylvania, where all races were (theoretically) contested. There, the highest GOP percentage in the districts they won was 65.9 percent. The lowest percentage of Democrats in districts they won was 60.5%.

    This demonstrates the strategy of “packing and cracking”, Where as many Democrats as possible are “packed” into a few districts which vote overwhelmingly “blue”. Where Democrats aren’t abundant enough to make up all of a district, such as in Cincinnati or in the Dayton/Springfield area, they are “cracked” and split between several rural districts, so that Republicans outnumber them. The efficiency of these two gerrymanders is brutal. Equal numbers of Democrats and Republicans would only be elected by an 8 point swing towards Democrats in every Ohio district, and a 7 point swing towards Democrats in Pennsylvania. Even if these swings took place, Republicans will still outperform Democrats in both states.

    Data from

    District Republican Democrat
    8th 100
    12th 63.7 36.3
    10th 60.2 36.9 2.9
    15th 61.8 38.2
    2nd 59.1 40.9
    4th 58.7 36.2 5.1
    1st 58.3 37.1 4.6
    5th 57.6 38.9 3.5
    7th 56.7 43.3
    14th 54.3 38.5 7.2
    6th 53.4 46.6
    16th 52.2 47.8
    13th 27.5 72.5
    3rd 26.9 67.8 5.4
    9th 23.5 72.6 3.9
    11th 100

    District Republican Democrat
    10th 65.9 34.1
    18th 64.0 36.0
    5th 62.9 37.1
    9th 61.6 38.4
    4th 59.7 34.4 5.8
    7th 59.5 40.5
    11th 58.5 41.5
    3rd 57.4 41.1 4.2
    6th 57.0 43.0
    15th 56.7 43.3
    8th 56.6 43.4
    16th 55.5 38.9 6.1
    12th 51.8 48.2
    17th 39.5 60.5
    13th 31.1 68.9
    14th 23.1 76.9
    1st 15.0 85.0
    2nd 9.4 89.4 1.3

    • mediaglyphic

      love the “cracking and packing” descriptive. I wonder if this strategy doesn’t leave the reds open to attacks from OFA. Since many of the “cracked” districts will be susceptible to smaller shifts in perception than the “packed” districts. From a gaming perspective, maybe the reds have painted themselves into a corner. Or am i being too optimistic about changes in voting patterns in response to information?

    • Mike Clinch

      Mediaglyphic, it’s true that the Republicans are much more likely to lose districts under their gerrymander than they are to gain more, but that’s because it is virtually impossible for them to gain enough votes in the “packed” Democratic districts. When I estimated a 7 to 8 point swing to gain parity in either ohio or Pennsylvania, I ranked the districts from the lowest to highest Republican voting percentages, and assuming a 50:50 vote in the state, how far would the swing have to be uniformly in every district to achieve parity. That number turns out to be about 7 points. If the swing was 10 points, (to 60:40), the democrats would go from parity to 12 of 16 districts in Ohio and 14 of 18 in Pennsylvania.

      On the other hand, there would have to be an 11 point republican swing in Pennsylvania to swing even one more Democtratic district to Republican, and a 23 point swing in Ohio.

      So right now, the Republicans have maxed out their advantage in both states, but given the red, rural nature of the Republican districts, there might not be much change for the rest of the decade.

  • Olav Grinde

    Wow, talk about this idea going mainstream! Now even Joe Scarborough is talking about the GOP’s gerrymandering.

    “…we actually got a minority of votes nationwide in House races. It was just gerrymandering from 2010 that gave us the majority.”

    • Joel

      The REDMAP project is pretty much the GOP’s public declaration of their gerrymandering intent (and the effectiveness of this approach).

  • mediaglyphic

    i saw scarborough admit this on the sunday morning shows yesterday. I wonder what this admission really means? Clearly Republicans have known this fact for a time. Perhaps OFA et al, are reducing the effectiveness of the redistricted red majority and guys like Joe the Scar are running scared.

  • Matt McIrvin

    What do you think of the efforts in blue states with Republican-controlled legislatures to split electoral votes by (gerrymandered) district? It sounds as if the Electoral College might bite us after all: if this stands, the Presidency could be permanently Republican (or at least conservative) until the EC is abolished.

    • Wheeler's cat

      Matt, cher, I think it’s not going to work out for them. Because of the Internet and sapients like Dr. Wang. The GOP survives on information starvation. The Internet is the antithesis of info starvation. Thus the most successful tactics become exposed and mocked. Mockery is the weapon of the sapient. Even if red phenotypes believe they are entitled to force choke the rest of humans to get their way, they are thin skinned and subject to the base perception of them being mocked. In a way, the flattening of info provided by the Internet is the antidote for the Stupid Half.

  • Olav Grinde

    I am sure that it will come as no surprise that the GOP is now pursuing its successful agenda. They’re moving full-speed-ahead to change Electoral College rules, state-by-state, in order to reap new fruits from their gerrymandering.

    Virginia is a striking example!
    Here is an article on recent developments.

  • Ashbel Green

    Sam — Can you address the Republican plans to change the allocation of electoral votes from winner-take-all to a system based on congressional districts in Virginia, Pennsylvania, Ohio, Michigan and Wisconsin. I have read the claim that such a change — if in place in 2012 — would have allowed Mitt Romney to win despite losing by 5 million votes. This seems chilling, if slightly difficult to believe.

  • Ashbel Green

    Yes, but most states aren’t looking at this. Only states controlled by Republicans that voted for Obama are looking at it — Virginia, Pennsylvania, Ohio, Michigan and Wisconsin. All those states could realistically make the change given which party controls the legislature and the governor’s office. Under this scenario, Obama would have lost votes in all those states — but not gained any in Texas, North Carolina, Arizona, Georgia.

    • Sam Wang

      That’s right. Such a nationwide analysis does not hit the point. The whole push is in swing states. In this respect it is like Bush v. Gore in 2000, and voter-ID laws: the goal is to win near-tied situations.

    • Olav Grinde

      So what would have been the 2012 Electoral College count had the suggested rule changes been in place in Republican-controlled swing states?

      I don’t recall seeing recalculated numbers…

  • Wheelers cat

    Gerrymandering is a tactic only. Integrated base would be a strategy.
    Good tweet here, I’m paraphrasing.
    CKMacLeod: they (GOP) are like a steamship out of coal, ripping up the decks and furnishings to feed the boilers, desperately scanning the empty horizon for a nonexistent landfall.

  • Tom Pfauth

    Statisticians ought to be familiar with the spatial analyses functions that GIS can perform. Surely there exist relatively simple spatial algorithms that divide a state’s population on the ground into polygons which contain the same number of people (or nearly the same) and also have the shortest possible perimeters. If states were required by law or constitution to make use of such an algorithm to accomplish redistricting, the results could be checked by anyone and partisanship would be minimized.

  • Bill de Lara

    Let’s assume that a democratic-republican ratio of the popular vote in all districts is 52-48. Can we not create fair districts acceptable to both by expanding or contracting a district until the overall ratio is achieved for that district? In the end, all the districts will reflect the ratio of the popular vote. How can any fairminded person be against this?

Leave a Comment