[EL] Gerrymandering and partisan composition of Congress

Tue Feb 14 16:15:10 PST 2017

The plans produced by the algorithm aren't actually random samples from the
feasible space of compact (however defined), equal population, and
contiguous plans, so no proper inference can be drawn from them. We apply
the algorithm to a small unit toy example to show how it fails to produce
two equally compact arrangements of districts with the same probability
here:

https://papers.ssrn.com/sol3/papers2.cfm?abstract_id=2583528

In the Florida litigation, the defendants presented example district of
where the algorithm even failed to produce compact districts. The district
court judge found the evidence produced by the algorithm so unconvincing,
there was no mention of the method in his ruling.

In the larger sense, just because an algorithm does random things, doesn't
mean that it randomly samples. I suspect given the behavior of the toy
example that in a large state simulation, the algorithm tends to
concentrate urban precincts into districts, and is thus biased towards the
finding that Democrats are inefficiently concentrated for redistricting
purposes.

============
Dr. Michael P. McDonald
Associate Professor, University of Florida
352-273-2371
www.electproject.org
@ElectProject

On Tue, Feb 14, 2017 at 6:20 PM, Nicholas Stephanopoulos <
nicholas.stephanopoulos at gmail.com> wrote:

> Like Rick and Justin, I think that comparing actual plans to randomly
> simulated plans can be very informative under the right circumstances. For
> example, if there exist randomly simulated plans that comply with all
> federal and state criteria at least as well as the actual map -- but that
> are much more symmetric in how they treat the major parties -- that's
> powerful evidence that the actual map's asymmetry cannot be justified.
>
> However, I agree with Justin's words of caution, and I'd add a couple more
> to them. First, Jowei and David's plans are drawn using equal population,
> compactness, and contiguity as criteria. They do *not* take into account
> compliance with the Voting Rights Act or other state requirements such as
> respect for political subdivisions and respect for communities of interest.
> The plans therefore are not lawful in many cases. Second, there's a
> fundamental normative question lurking as to what the right benchmark is
> for comparison: the "typical" randomly simulated plan or actual partisan
> symmetry. The argument for the latter is pretty intuitive; we wouldn't
> normally say that a biased map is fair just because the usual output of a
> computer simulation is a biased map. And third, even if we're setting the
> benchmark using a computer simulation, it's unclear why the *average*
> simulated plan has any normative significance (or even if that average plan
> can be identified with any confidence given the near-infinite universe of
> possible plans). If there exists a simulated plan that's lawful and that
> treats the major parties symmetrically, why shouldn't *that* plan be the
> relevant yardstick?
>
> So I too congratulate Jowei and David on an excellent paper, but warn
> against jumping too quickly to conclusions as to whether the current
> Congress is or isn't severely gerrymandered.
>
> Nick
>
> On Tue, Feb 14, 2017 at 4:26 PM, Justin Levitt <levittj at lls.edu> wrote:
>
>> I admit that I've only skimmed the Chen-Cottrell paper that Rick Pildes
>> mentions here, and I look forward to reading far more carefully.
>>
>> But to Rick's description below, it may be worth adding a few short
>> thoughts, particularly given inevitable shorthand descriptions of the paper
>> in the media.
>>
>> Most important, this paper doesn't mean that there's no point in
>> addressing congressional redistricting (either process or result), or that
>> the benefits and detriments of the status quo are basically a wash.   (And
>> though some might draw that conclusion from the summary, I don't think
>> that's the implication that Rick would draw either -- and it's not the
>> implication that I think the paper authors have suggested in the paper
>> itself.)
>>
>> First, though I understand why net partisan gain is a compelling focus,
>> it's not the only measurement for whether redistricting is being performed
>> acceptably (or "best," which is a different question).  Political
>> scientists value the overall balance of Rs and Ds, and many citizens who
>> aren't political scientists do too.  But there is more to value out of
>> districts than just random final assignment of R or D.  Democrats in states
>> in which the Republicans "gained" seats due to gerrymandering may not view
>> those seats as fungible even given comparable states in which the Democrats
>> "gained" seats due to gerrymandering.  People (including, but not limited
>> to, minorities long excluded from the political system) may prefer
>> equitable opportunities to elect their particular candidates of choice.
>> Communities may feel like they've got a representative delivering pork back
>> home.  Particular communities benefit from not having their incumbents
>> "targeted" for district dissection, and that impact is magnified in
>> Congress by the seniority system.  And on and on.
>>
>> Second is the "compared to what" question suggested by the headline, and
>> explicitly acknowledged as the "non-gerrymandered counterfactual" by the
>> study authors.  The headline should really say that gerrymandering has
>> "little to no effect" on the partisan composition of Congress *compared
>> to a randomized system of drawing districts based on criteria designated by
>> the study authors*.  There is no state in the country -- and there is no
>> local jurisdiction in the country that I'm aware of, though I'd welcome
>> information to the contrary -- that currently draws districts using the
>> simulation comparison described below, and in many states, the baseline
>> simulation comparison would likely not be lawful.  The study did not, I
>> take it, conclude that gerrymandering presently has little to no effect on
>> the composition of Congress compared to a system in which Republicans (or
>> Democrats) controlled the process across the board, or compared to a map
>> drawn according to federal law and the state laws in place in each
>> individual state, or compared to some other proposal for how districts
>> might be drawn or should be drawn, including the desire to avoid splitting
>> meaningful communities along arbitrary computer-drawn lines.
>>
>> Neither of these caveats is a critique of the paper, which is both
>> impressive and valuable, and should be read on its own merits.  But they're
>> important caveats, I think, about the broader conclusions that can be drawn
>> from the paper itself.
>>
>> Justin
>>
>> --
>> Justin Levitt
>> Professor of Law
>> (on leave through spring 2017)
>> Loyola Law School | Los Angeles
>>
>> On 2/14/2017 7:43 AM, Rick Hasen wrote:
>>
>> Gerrymandering Has “Little to No Effect” on the Partisan Composition of
>> Congress <http://electionlawblog.org/?p=91074>
>>
>> Posted on February 13, 2017 12:52 pm
>> <http://electionlawblog.org/?p=91074> by *Richard Pildes*
>> <http://electionlawblog.org/?author=7>
>>
>> That is the finding of an important new paper
>> <http://www-personal.umich.edu/%7Ejowei/gerrymandering.pdf> co-authored
>> by one of the leading social science experts on districting, Professor
>> Jowei Chen.  In recent years, a debate has been taking place over whether
>> it is particularly aggressive Republican gerrymandering in the 2010 round
>> of redistricting or increasing geographic sorting of voters by partisan
>> affiliation that explains the Republican “advantage” in the House — the
>> fact that Republicans gain a larger percentage of House seats than their
>> nationwide share of votes in House elections.
>>
>> Chen and David Cottrell frame their inquiry as an effort to answer how
>> many seats each party would control in the complete absence of
>> gerrymandering.  I won’t explain their full methodology here, but it
>> basically consists of doing hundreds of computer simulations to measure the
>> election results in differently designed districts, in which the building
>> blocks are election-return results from the 2008 presidential election, all
>> the way down to the Census block level.  The computer is then told to start
>> randomly at different points in the state and design equally populated,
>> geographically continuous, and compact districts.  The simulations do not
>> take partisan or racial information into account.  This method of using
>> thousands of computer simulated districting plans based on objective
>> criteria is increasingly being offered by experts, including Chen, in
>> litigation.
>>
>> Their bottom line finding is that if congressional “districts were drawn
>> randomly with respect to partisanship and race, Republicans would only
>> expect to lose a single seat in Congress to the Democrats.”
>>
>> They do find that there are modest partisan gains from gerrymandering in
>> individual states.  But the gains to each party cancel out, in their
>> analysis.  Thus, they find that Republicans gain about five seats in states
>> in which they controlled the redistricting process in this cycle.  In
>> states Democrats controlled, they gained about three seats.  And once race
>> is taken into account through the way the requirements of the VRA
>> pre-clearance process demanded preservation of VRA districts, the Democrats
>> gained another 1.75 seats compared to what a process based just on
>> contiguity, compactness, and equal population would tend to produce.
>>
>> This is certainly not the last word on this important subject.  Any
>> complex study of this sort poses many methodological issues.  And their
>> findings for congressional districts do not necessarily mean that
>> gerrymandering has not made a significant difference for state legislative
>> elections.  But this study provides one of the most important counters to
>> the argument that partisan gerrymandering plays a major role in the current
>> composition of the House.
>>
>> Further debates and discussions of this issue, including in the media,
>> are going to have to take account of this important new analysis.  It is
>> consistent with what at least some other social scientists, using different
>> approaches, have also concluded about the limited effects of gerrymandering
>> on the composition of the House.
>>
>> Posted in Uncategorized <http://electionlawblog.org/?cat=1>
>>
>>
>>
>>
>>
>> _______________________________________________
>> Law-election mailing list
>> Law-election at department-lists.uci.edu
>> http://department-lists.uci.edu/mailman/listinfo/law-election
>>
>
>
>
> --
> Nicholas O. Stephanopoulos
> Assistant Professor of Law
> University of Chicago Law School
> nsteph at uchicago.edu
> (773) 702-4226
> http://www.law.uchicago.edu/faculty/stephanopoulos
>
> _______________________________________________
> Law-election mailing list
> Law-election at department-lists.uci.edu
> http://department-lists.uci.edu/mailman/listinfo/law-election
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://webshare.law.ucla.edu/Listservs/law-election/attachments/20170214/2d2e67ed/attachment.html>