Measuring gerrymandering: A closer look at three key states

Eric J. Ostermeier
Humphrey School of Public Affairs Eric J. Ostermeier

A few months ago, I annoyed several Democrats (judging by the comment thread) by arguing that the success of Republicans in gerrymandering the map of congressional districts after the 2010 census was not “entirely or primarily” the explanation for Republicans success in 2012 maintaining control of the House (and by a significant margin), despite the fact that nationwide Democrats got more votes in all House races combined.

I’m still defending the “entirely” part. But some smart recent work by Eric Ostermeier over at Smart Politics has me reconsidering on the “primarily” part.

Ostermeier has so far focused on three states, all of which had House districts redrawn in 2011 by an all-Republican lineup, all three of which were carried by Obama in 2012.

Despite the fact that a significant majority of Obama-supporters — presumably mostly Democrats or Dem-leaners — turned out last year in those three states, Republicans managed to send large majorities of their nominees to the House.

Here are the three states followed by the portion of the presidential vote Obama won in 2012, followed by the Republican-Democratic breakdown of each state’s House delegation chosen in the same election.

  • Ohio/51 percent/12-4.
  • Pennsylvania/52 percent/13-5.
  • Michigan/54 percent/9-5.

Totaling the three, that’s 34 Republican House members to 14 Democrats coming out of three states that Obama carried. It’s impressive, so impressive that it’s hard not to conclude that the Republican map magicians in a few key states contributed significantly to the House majority.

But it’s too soon to leap all the way to “entirely or primarily,” for several reasons.

Some offsets

There are the offsets. Democrats captured full control of the redistricting in Illinois and managed to turn a House delegation with an 11-8 Republican majority in a 12-6 Dem advantage. Of course, Obama carried Illinois, but not by a 2-1 margin (which is the margin of Dem control of the Illinois House delegation). And Dem presidential nominees had carried Illinois in the past six elections.

Reason No. 2 to be careful about over-attributing the results to the decennial redrawing of the map would be to compare the first result of House races under the new map to the last results under the old map. On that basis, the Dem map-makers in Illinois produced a stunning result. By contrast to Illinois, in the three Ostermeier has featured so far, the Republicans made net pickups of no more than one seat in each state over the composition of the House delegation from the previous election, conducted under the previous, 10-year-old district map.

Considering the overwhelming tendency of House incumbents to be reelected in every cycle, it doesn’t seem reasonable to rely over-much on mapmaking to explain why most of the incumbents held their seats (although it is certainly possible for an incumbent to get into reelection trouble because of a change in the district boundaries, even including those cases in which a new map puts two incumbents into the same district).

The next problem is an assumption that voters vote a straight party ticket. Many do, but many don’t. If the results of the presidential election was such a great predictor of races for Congress, how would we explain the (fairly astonishing) success of Democratic Senate candidates in states like Montana and the Dakotas? All three states were bright red in all recent presidential elections. As those races make clear, plenty of voters split their votes between a presidential nominee of one party and the Senate candidate of the other.

Of course, a Senate race is not shaped by a redrawing of district maps every 10 years according to the preferences of whichever party has control of the state, and that insight falls further apart when you consider that many states (including Minnesota) often have divided partisan control of the state government, which requires the parties to either compromise on a map or turn the job over to the less-partisan courts.

More analysis coming

Forgive me if I have thunk myself into a knot here. I just got off the phone with Ostermeier and I know he plans to work up some other states for subsequent analysis. There is not one definitive way to measure gerrymanderly impact. Our system, as evolved, provides whichever party gets control of the decennial mapmaking in a state an opportunity to maximize the number of House seats its own party will win.

If you are a Democrat and you felt a surge of pleasure at what the Illinois Dems were able to do with the last mapmaking exercise, it could undermine your outrage at what the Repub mapmakers were able to do in Ohio, Pennsylvania and Michigan. Over the course of 10 years, demographics will take over and the perfect partisan advantages of a particular map will erode. After 2020, everything will be up for grabs. And by the way, while 2010 was a great year for Republicans, and Republicans often do better in midterm elections, 2020 will be a presidential election year, when turnout is always higher and Dems tend to do better, unless something happens between now and then to change that old polysci rule of thumb.

None of this was foreseen by the Framers, who hoped for and imagined politics without national parties and who, by the way, did not require states to divide themselves into congressional districts. In the early history, there were at-large seats, multi-member districts, all kinds of stuff.

You can also learn about all our free newsletter options.

Comments (9)

  1. Submitted by Paul Brandon on 07/17/2013 - 09:16 am.


    Riffing on self selection.
    There is a demographic effect.
    The populations who tend to vote Demo are more likely to live in cities, and this tendency is increasing. So Demo votes will tend to cluster in high population density areas where this effect has not been completely compensated for in redistricting (would you make one high rise apartment house a district?).
    So to this extent the Repubs are benefiting from an existing phenomenon rather than causing it.

  2. Submitted by Peder DeFor on 07/17/2013 - 10:40 am.

    City concentration

    From what I’ve read, the use of statewide percentages isn’t all that useful here. One of the issues that Dems face is that Dem voters are very concentrated in the cities. In Pennsylvania, for instance, there were several voting precincts that went 90%+ for Obama. The same isn’t true in rural areas. In other words, even if you used some non-partisan computerized system to allocate House districts, Dems would have a disadvantage simply due to geographic compression.
    Which isn’t to say that there isn’t gerrymandering and that the GOP didn’t use it in 2012. There are some solutions to this, some radical and some fairly simple. Iowa has used one of the simple ones and we should urge other states to go at least as far as they do.

    • Submitted by Paul Brandon on 07/17/2013 - 06:54 pm.


      There’s no reason on principle why you could not equalize congressional districts by population; city districts would just be very small in area. The problem is political; this would increase drastically the number of Democratic voting districts in the House.

  3. Submitted by Eric Paul Jacobsen on 07/19/2013 - 02:40 pm.

    Proportional Representation

    That’s the only proven solution to gerrymandering of all kinds. Create larger voting districts and let each send five delegates to the legislature, according to their proportional share of the vote, and gerrymandering will be eliminated both as a strategy and as a problem. The solution is laughably simple: Represent people, not districts. Ranked-choice voting won’t solve the problem by itself, but it’s a step in the right direction, and I hope it will make the implementation of a truly proportional system easier when we finally get around to it.

    Gerrymandering is an undemocratic means of winning elections, and it’s done by manipulating an undemocratic structure – single-delegate districts. Blaming Democrats or Republicans for playing by a time-worn strategy that’s been proven to work – for them, but not for democracy – is pointless, even though we all agree it’s a rotten strategy. Instead, eliminate the underlying undemocratic structure.

    Douglas J. Amy has been arguing in favor of proportional representation for years. (Read REAL CHOICES, REAL VOICES: THE CASE FOR PROPORTIONAL REPRESENTATION ELECTIONS IN THE UNITED STATES, 1993.) I wish more political scientists were like him. Most political scientists with a public face seem to me to be perpetual play-by-play commentators on the US-American political system who care nothing about how antiquated and clunky this system is, compared to others that have made many of our problems obsolete. If our engineers were like our political scientists, the internal-combustion engine would never have come into use, and they’d all be commenting on Detroit’s latest steam-powered vehicles.

  4. Submitted by Donald Larsson on 07/17/2013 - 10:53 pm.

    A “Founder” and the original Gerrymander

    Just a technical note here. Whether or not the Framers envisioned party politics and congressional districts, it was a Founder who gave the practice its name–Elbridge Gerry, who signed the Declaration of Independence and Articles of the Confederation, but originally opposed adoption of the Constitution without an attending Bill of Rights, which he wound up helping to draft. He also, among other things, served as James Madison’s Vice President. The “gerrymander” label came from a newpaper editor during Gerry’s term as Governor of Massachusetts. The editor noted that a new state senate district looked like the mythical salamander (as opposed to the actual lizard) and dubbed it a “Gerrymander.” You can see the resuting political cartoon at

  5. Submitted by Donald Larsson on 07/17/2013 - 10:56 pm.


    Real salamanders are amphibians, not lizards. My apologies to any herpetologists I may have offended.

  6. Submitted by Paul Udstrand on 07/18/2013 - 08:50 am.

    Predictions vs. Effects

    Statistical analysis can get very dicey when your looking at complex phenomena. The effects of Gerrymandering are undeniable, look at Bachmann’s district. However, those effects are not necessarily predictable, look how close Bachmann came to losing. I think Erik is making the classic mistake of assuming that effect produces predictability. Weather forecasters do this all the time.

    All election outcomes are the product of votes, which are by definition complex decisions. Few people vote the same way their entire lives, or even from one election to the next. And the reasons for a given vote vary greatly. Most Americans don’t love the candidate or party they vote for. This complicates predictions. Aside from the votes cast there is no single factor that determines every election, but that doesn’t mean that statistically there can’t be dominant factors in a given election. It’s hard to sort out variables like that from one event to the next. I don’t think there’s any doubt that Gerrymandering preserved the Republican majority in the House in the last election, they only held on by what? 18 seats? That doesn’t mean Gerrymandering will be a deciding factor in every election however.

    Modern statistics looks at Odds Ratios, you give statistical weight to a given set of variables. Sometimes the weight between one variable and anther is small, but it’s still significant. With something like Gerrymandering you could have a very small but significant Odds Ratio, it’s not necessarily a huge difference, but it’s enough to make the difference. With Gerrymandering it’s not always clear cut, and it’s not necessarily the same in every election.

    • Submitted by Paul Brandon on 07/18/2013 - 09:26 am.

      Modern statistics

      look at many things, including odds ratios. The Bayesian approach (fairly common now) also takes into account prior knowledge to use conditional probabilities as a basis (Nate Silver’s book has a good summary). Weather forecasters have improved their predictions over the past 20 years by using meteorological principles to fine tune their statistical predictions.
      The main point is the aggregates are much more predictable than individuals; the larger the aggregate the more accurate the prediction (at least as long as the aggregate represents the same population as the individual).

      You are right that phenomena involving determination by multiple variables are less predictable than ones involving only a single variable, but that’s the real world we have to deal with.

      • Submitted by Paul Udstrand on 07/22/2013 - 10:09 am.

        Weather forecasts have improved?

        It’s little off topic but I don’t think forecasts have actually improved in the last 20 years. They do a lot of computer modeling that doesn’t seem to be able to handle chaotic variables. I think if you were to actually take a certain variables, like cloudiness, precipitation, and temps within ten degrees or so, most days you would do just as well flipping a coin. Beyond 48 hours forget it. I know meteorologists claim a high degree of accuracy, but I don’t buy it. As a photographer I’ve tried to use their predictions to schedule outdoor photo shoots and it’s pretty much useless. I think Bud Kraehling did better.

        Circling back to Gerrymandering, statistical analysis isn’t just about variables and equations. Statisticians still have to give different variables certain weights. The thing that has made Nate’s predictions accurate isn’t just the equations he uses, but the weight he gives to different variables.

        The problem with Gerrymandered districts is that you can’t give them the same universal weight as a variable. For instance I don’t think there’s any doubt that Bachmann’s district itself kept her in office, but the district alone was obviously not sufficient to keep other Republicans in office. So you have an effect, but it’s not completely generalizable. If you average that effect out you may find a null effect over all that ignores local realities.

Leave a Reply