Monday, February 4, 2013

Planet of the Bride of the Son of the Return of Cranky Blogging

You know what? Gerrymandering is real and has real consequences, but I suspect that despite that most news outlets would be better off just ignoring it. People love to blame everything on gerrymandering, and (although again it is real and has real consequences) it's just not that big a deal.

Which gets to the NYT Sunday Review piece by Sam Wang, which examines the strange results of the 2012 House elections -- Republicans retained a very solid majority of seats despite Democrats winning more votes -- and more or less attributes it all to gerrymandering.

I suspect my real complaint here is with the New York Times even more than Wang, but this just won't do. Since the election, we've had a steady stream of pieces by political scientists showing that neither the 2012 gerrymander nor gerrymanders in general are responsible for the bulk of the 2012 results. Instead, it's mostly about the "natural" effects of districting. Basically, over time Democrats have come to live in more and more compact areas. It's not true what you sometimes might read that districting can't overturn the effects of where people live, but it is true that to do so districting would have to violate other things that people care about, things such as other political boundaries and communities of interest. As John Sides sums it up, "Across a very wide range of counterfactual scenarios, the geographic concentration of Democrats will produce votes-seats discrepancies without deliberate partisan gerrymandering."

What really annoyed me about Wang's case against gerrymandering is that he makes a point of bashing "ugly" districts -- you know, the ones that have odd shapes, such as the one that gave gerrymanders their name in the first place. The piece is illustrated with some of them. What's wrong with that? If districts are going to be "fair" to Democrats (that is, get the same ratio of seats as votes), then because of where Democrats and Republicans live it's precisely the value of having "pretty" (compact, regular-shaped) districts that's going to be violated. Under current conditions, compact, regular districts strongly tend to favor Republicans. One can argue for them anyway, but anyone who cares more about partisan "fairness" shouldn't also be playing up the importance of "pretty" districts. Well, to my tastes no one should care about compactness; there's simply no reason, in my view, to care about the shape of districts, all else equal. But at the very least, anyone complaining about ugly districts should know the very predictable effects of compactness.

Which gets to a larger point about Wang's piece. He says that "[g]errymandering is not hard." On a technical level, it's true that drawing lines to achieve a specific goal is a lot easier than it was thirty or even ten years ago (although it's worth pointing out this mainly applies to immediate effects; population shifts and political changes over ten years can be large and unpredictable, so we don't really know how the lines drawn for 2012 will affect the 2018 and 2020 elections).

But it's still true that gerrymandering is difficult because there are so many competing, and contradictory, goals in redistricting. The best explanation of this I've seen is in Bruce Cain's The Reapportionment Puzzle. In particular, in addition to what I mentioned above, are the natural enemies of redistricting: parties who want the most efficient use of their votes in order to maximize seats, and politicians from those parties who want the largest possible majorities, even if it "wastes" votes for the party. But beyond that, there are simply a lot of competing goals, and it's rare for state legislators to be able to ignore them and focus only on a simple partisan goal.

But back to the 2012 redistricting: there's plenty of information out there about this, and the Times should do better.

26 comments:

  1. I'm not sure I understand this claim, "If districts are going to be 'fair' to Democrats (that is, get the same ratio of seats as votes), then because of where Democrats and Republicans live it's precisely the value of having "pretty" (compact, regular-shaped) districts that's going to be violated."

    What metric are you using to judge how compact the set of districts within a state is? At the time districting is performed, the number of districts is already set. The total area covered by the districts is also set: the total area of the state. Making one district smaller (more compact) must inherently make another larger (less compact). I can understand describing individual districts as being more or less compact, but how do you define one set of districts as being more compact than another?

    As for the second, the impossibility of creating regularly-shaped districts that produce a D/R balance similar to state voting proportions, I'm highly suspicious of this claim. I live in Pennsylvania, where the Democratic population is concentrated into Pittsburgh and Philadelphia, and we have a few small districts that cover these people, and then a bunch of large ones that cover the less-populous and more-Republican parts of the state. However, one could also divide these cities up by using, as one simple example, the center of each city as the meeting point of many districts expanding outwards as triangles (like centering a sliced pizza in the middle of each city). These districts could be regular triangles (i.e., a regular shape), and could be oriented to create many districts with a modest Democratic majority by taking a narrow portion of a city and a broader portion of the surrounding area (just like how a slice of pizza is wider at the crust than the tip).

    What seems to be the actual problem is that you can't create districts that will produce representatives with the same proportion of D/R as the voting within the state by making compact, regular districts around the cities. But that's a very different claim than the set of districts being generally more compact and more regular, which appears to be the claim that you are making.

    ReplyDelete
    Replies
    1. "Compact" is a term of art in redistricting. What does it mean? Well, that's more complicated. Alternatively, one might consider a circle or a square the most "compact" area. We could talk about minimizing the ratio of perimeter to area. We could talk about minimizing the number of times the pen changes direction in drawing the boundaries.

      At the end of the day, though, what "compact" mostly means is "districts that look pretty."

      But, as to whether compact (so defined) districts would reflect that state D/R balance, in many places, those pizza slices become so unbearably thin and tenous as to make them impossible. Geography plays a big role here: NYC/NY state is simply impossible to pizza-slice, and I have a tough time imagining it in SF, LA, or Seattle. Atlanta? Yes. The inland cities in Ohio? Sure. But, all these coastal/river cities are pretty constrained geographically. The other thing I'll note is one of my favorite redistricting stories, from the bad old days of malapportionment. The state in question used county lines for congressional lines. One member lost one county and gained another. In fact, his three counties before were in a line (in that county 1 and 3 didn't share a border). His new district kept two of those counties, but dropped #3 for one that bordered both counties 1&2. By any definition, the new district was more compact. This redeistricted member called it "the most egregious gerrymander in history." (Wish I could remember more details, so I can't give a better cite than my memory on this one)

      Delete
    2. The idea that "compact" is both a term of art and nearly undefined is pretty funny. Which isn't to say you're wrong about either, of course.

      If we accept that square (or, more practically with oddly-shaped states, boxy near-rectangles) are what we're looking for, we could still have considerably more balance in most states. States being dominated by cities like New York are the exception, not the rule. So long as your cities don't need to be divided up into more than four pieces to achieve balance (which they wouldn't in most states, like PA and OH) we can still use right angles and rectangles, just by placing the vertices in the middle of cities instead of centering rectangles around the cities.

      If we care about the more mathematical question of "minimizing the ratio of perimeter to area" and "minimizing the number of times the pen changes direction in drawing the boundaries," then it should be possible to simply apply these constraints to the actual states and see what we'd get. Has this been done? None of the many articles (such as here, the Monkey Cage, etc.) I've read have gone down this more geometrically-defined route, so I suspect that this isn't the sort of thing they're talking about. Note that we wouldn't just care about the absolute-maximal compactness solutions, but presumably the question of whether it's impossible to make strongly-compact districts (by one of these definitions of compactness) which also produce balanced results.

      I'm willing to accept that NYC, SF, LA, and Seattle are sufficiently dense and stuffed with Democrats as to require that balanced districts would have to be aesthetically-displeasing, slivers of the city bulging out grossly across the rest of the state. But the analyses I've seen that the House would still be Republican don't rely just on poor districting around a handful of our most populous and Democratic cities, they assume that Pittsburgh, Philadelphia, and other moderately-sized cities from Atlantic/Midwestern states will be densely packed too, and I haven't yet seen a good argument that those states can't produce districts that are both pretty and balanced.

      Delete
    3. Cain, McDonald and Hui (http://weber.ucsd.edu/~tkousser/CainHuiMacDonald.doc) had a bunch of people (I think mostly students, but students who had knowledge of redistricting software!) redraw California, with instructions to maximize certain criteria: competitiveness, compactness, etc. There still ends up being no real solution for all the wasted Democratic votes in SF & LA. The concentration of Dems in urban areas really is stunning: Jose Serrano would routinely win his Bronx district 96-4. That's not gerrymandering...it's just that there are so few Rs living there.

      In the end, Republicans live in areas that are moderately red, whereas Democrats live in areas that are more deeply blue.

      Delete
    4. When did compactness become a priority virtue among competing criteria? Has it always been this way?

      Delete
    5. As I said, I can believe that LA and SF (along with NYC and Seattle) are simply too dense. Do you know of similar work that has been done in midwestern states? In particular, do you have reason to believe that not being able to fix the districts around a half-dozen highly-dense, highly-Democratic cities is sufficient to explain House Republican control? Because all of the analyses I've seen have included imbalance in states like PA (only 5/18 representatives in the House are Democrats in a majority-Democrat state) in trying to claim that gerrymandering isn't responsible for Republican control of the House. And Pittsburgh, and the like, aren't anywhere near as big or dense as SF and LA.

      Delete
    6. I don't know of any work for other states. My assumption for other states is based on the simple guidelines of Democrats being denser in urban areas than Republicans are in rural areas. Unfortunately, the data I have at my fingertips really lumps in suburban with urban (census tends to do that), so it just looks like urbanites are 55/35 D/R, and ruralites are 50/40 R/D. But, I'm pretty sure that if I had the data to separate out suburban from urban, that'd shift it even more.

      Delete
    7. I agree that it would shift even more and I don't disagree that Democrats are more densely packed into cities nation-wide. But your assumption completely ignores how atypical NYC, SF, and LA are. To go back to Pittsburgh, it has a population density of 5,600 people per square mile to NYC's 27,000. It has an area of 58 square miles to LA's 500. This is going to make a huge difference in how ugly balanced districts would be. Extrapolating from a handful of super-dense cities to the rest of the cities is simply not a valid line of reasoning. You can't take a group of professional basketball players and use them as your basis that everyone will look silly sitting in a compact car, and you can't extrapolate a general question of geometry and population density based on the examples of highest density.

      I'm not taking issue with you in particular, Matt Jarvis. But these arguments about gerrymandering have expanded from "gerrymandering isn't the whole story," which has been well documented, to "If districts are going to be 'fair' to Democrats (that is, get the same ratio of seats as votes), then because of where Democrats and Republicans live it's precisely the value of having 'pretty' (compact, regular-shaped) districts that's going to be violated" without the facts to back it up.

      Delete
  2. "population shifts and political changes over ten years can be large and unpredictable, so we don't really know how the lines drawn for 2012 will affect the 2018 and 2020 elections)."

    The Pennsylvania GOP gerrymander of 2002-2012 was often ridiculed as a "dummymander" after the Democrats gained a majority of the state congressonal delegation in 2006 and 2008. The GOP tried to maximize the number of Republican congressmen rather than to make the Republican congressmen absolutely secure, and the result was that some of them lost in the Democratic "wave" years.

    And yet, if you take the longer view, the "dummymander" was a success. It worked reasonably well in 2002, 2004, and 2010. That's three out of five elections--not that terrible a record.

    So yes, too greedy a gerrymander can backfire in "wave" years for the other party. But the fact remains that in most years it will work.

    ReplyDelete
  3. Are there any good proposals for how to do House elections that avoid turning the geographic distributions of voters into a persistent advantage for one party? If there's a way to get rid of the every-decade redistricting circuses, that would be great too.

    ReplyDelete
    Replies
    1. Get rid of districts. For larger states use proportional representation and party lists.

      In smaller states, House elections could be statewide and at-large. In a state like Wisconsin, for example, with 8 seats, let each voter rank eight choices of candidates. Any candidate who receives more than 12.5% gets a seat automatically. Once a candidate is seated, ballots that placed them as #1 are redistributed by their #2 place choices, until all eight seats are filled.

      Delete
    2. I like IRV a lot, so I'm well-disposed towards this. Is this basically the idea?
      http://en.wikipedia.org/wiki/Proportional_representation#Single_transferable_vote_in_a_multi-member_constituency

      Delete
    3. My general feeling on this is that individuals who represent specific, relatively narrow, interests has been a big strength of the US system, and that would be lost with PR/party lists. Others, of course, disagree that it's a strength.

      My other feeling is that what you want to encourage is a balance between party influence and the representative relationship of each Member with his or her district. A PR/list system basically eliminates the latter. IMO, it's going to be much more difficult for ordinary citizens to have much influence in that sort of system.

      Delete
    4. That might actually be one of the big unexplored issues out there, Jonathan. Namely, in some quarters the idea that "ordinary citizens" should have much influence is not overwhelmingly popular. And I think you can find this phenomenon in plentiful proportion on all sides of the spectrum, although the GOP strain of it is more immediately obvious these days. After all, one can even invoke Winston Churchill, that endless well of out-of-context proverbs, on this. Once when asked for the greatest and most damaging argument against democracy, he responded, "five minutes conversation with the average voter."

      On the other hand, I am not sure that automatically translates into being anti-democratic. One could make the argument (and I susupect many would if pressed to the wall) that democracy consists in the electorate choosing who is to run the government, not in the members of the electorate having much say in what the government does or how it is run. And that leaves aside all the issues of who should have the right to membership in the electorate.

      Anyway, just pointing out that the issues may even be more basic than the level under discussion.

      Delete
    5. FairVote has a complete proposal for a US congressional map that does just what Neil mentioned: multi-member districts elected under Single Transferable Vote. You can see the plan at fairvoting.us

      An candidate-based PR system like STV used in medium-sized districts might actually increase the extent to which members of Congress represent specific interests. While the geographic interests they would represent would be broader, the political interests they would represent would be narrower. In a five seat district, for instance, one member would represent the far right, another the moderate right, another the center, another the moderate left, and another the far left.

      Similarly, as long as you stay away from party list forms of PR there's no reason why a PR system has to eliminate the influence of ordinary citizens. To the contrary--when voters in every district have a realistic chance to elect someone who actually represents their political views, the influence of the average voter on their representation will be significantly increased.

      Delete
    6. What about mixed-member proportional representation - MMP? I.e., local district elections (draw the lines as ugly or pretty as you like), balanced by a party-list second ballot? You get to balance local and national interests that way. Or is that not true in practice?

      Delete
    7. On Devin's point...

      Meaningful individual influence on the system really can't happen through voting. It just can't.

      Voting is fine for what it is, and it's certainly a necessary and important part of any democracy, but there are just severe limits as to what voting can ever really do.

      So while voting fairness is certainly important, it can't be the whole ball of wax in designing a system.

      Delete
  4. My biggest pet peeve when people complain about "ugly" districts is when a district's boundaries are ugly because they follow a municipal or other political boundary that's ugly.

    ReplyDelete
  5. Then there's the old example of dividing East L.A. into 5 pieces that were each attached to a large white bloc, a method that diluted Hispanic electoral power for decades.

    With respect to rural vs. urban districts, I am unclear on the rules that dictate the allowed variation in population between districts. I have the impression that the number of voters in rural districts in PA are substantially fewer than in urban districts. Is that true?

    ReplyDelete
    Replies
    1. Voters: possible. It would depend on the turnout differential.

      People: according to census numbers at the time of the drawing (ie, the census taken nearly 2 years before those lines went into effect), essentially ZERO variation (at the congressional level) and VERY LITTLE variation (at the state legislative level) is allowed. Now, we all know that census data is, itself, imperfect. And 18months after the census? Certainly not perfect. Yet, the Supreme Court's reading is that there should be almost no variation in districts based on these inaccurate numbers.

      Delete
  6. Yes, the point should be that the existence of ugly districts shows that "inevitability" positions (by pundits or political scientists) are hypocritical. There's nothing unrealistic about making sure districting is representative by drawing many districts in a more "ugly" way. We already have ugly districts for questionable, seemingly quite illegitimate purposes. There's no reason it shouldn't be easier to argue for some ugly districts for laudable, more legitimate purposes.

    ReplyDelete
  7. One consideration is whether you want elections to serve geographical communities of interest, in which case compactness is usually important, or something else. Ideological communities of interest would be best served by scrapping the current system altogether and going to proportional representation.

    ReplyDelete
  8. Great Britain the same pattern of left-wing voters being concentrated in urban areas and right-wing voters being concentrated in in rural areas, but the district lines are highly favorable to Labour and not the Tories.

    ReplyDelete
    Replies
    1. In 1951 Churchill's Conservatives came to power even though Labour got more votes.

      http://en.wikipedia.org/wiki/United_Kingdom_general_election,_1951

      Delete
  9. General note on this thread: thanks to Matt Jarvis for his excellent contributions.

    Other general note: over twitter today, Michael McDonald responded to this by saying that new research soon would should relatively less of a "where people happen to live" effect and relatively more of a partisan gerrymandering effect. I'll look out for it. There's good theoretical reason to not be surprised by relatively small partisan gerrymandering effects, but that doesn't mean it can't happen.

    ReplyDelete
    Replies
    1. It's just very annoying that you commented on Wang's article without going through the trouble of reading his research supporting it, which shows the results McDonald is also talking about. In fact, Wang already replied to your points before you made them. The entire quantitative analysis he's making is precisely designed to separate natural effects from political gerrymandering effects.

      Now,you may well criticise his methodology, or question his results, but to write a whole piece on his paper flatly ignoring the actual argument made in it? That's just unprofessional.

      Delete

Note: Only a member of this blog may post a comment.