[EL] Simulated Districting Plans and Majority-Minority Districts

[Michael McDonald]

Automated redistricting is hard (in fact, mathematically, NP-Hard). As
introductory computer science texts warn, just because you program a
computer to do random things does not guarantee random results. Automated
sampling algorithms are from a mathematical perspective suspect since their
biases are unknown, and can never be known for any practical application
given the current limits of computing power since there are a staggering
number of feasible redistricting plans. Therefore, no valid inferences can
be drawn from automated redistricting algorithms. Chen and Rodden have a
"simple" algorithm that attempts to draw districts for equal population,
contiguity, and compactness (and actually fails on the latter, as
demonstrated in court). Bringing in additional criteria like voting rights
districts only makes the automated algorithm more complex. If the bias
properties of the Chen and Rodden algorithm are unknown, bias would only
likely be acerbated further by bringing in additional criteria.

We've demonstrated the bias of the Chen and Rodden algorithm in a toy
example found in here, and we are working on expanding the critique to an
analysis of all automated algorithms, including the more complex MCMC
algorithms and the deceptively-impressive super-computing solutions (even
drawing millions of plans are but a drop in the ocean of all feasible
plans).

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

For anyone thinking about using the Chen and Rodden algorithm in court, you
should review Rodden's oral testimony in the Florida litigation where
defendants produced as rebuttal evidence districts drawn by the algorithm
that were clearly non-compact. This serves as a road map for any litigation
team rebutting evidence from an automated redistricting algorithm. I'd also
recommend for fans of the work, Nolan McCarty's rebuttal expert report for
defendants, which demonstrated Chen and Rodden's findings are dependent on
whether or not one chooses to normalize presidential vote for the analysis.
The Florida district court discounted the Chen and Rodden evidence so
heavily there is no mention of it in the decision, or in the subsequent
Florida Supreme Court decision.

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

On Thu, Sep 14, 2017 at 12:43 PM, Kogan, Vladimir <kogan.18 at osu.edu> wrote:

> Many thanks to Richard Pildes for bring attention to the * amicus *brief
> by political geographers. I have a question I’m hoping someone might be
> able to answer:
>
>
>
> I’m a big fan of Jowei Chen and Jonathan Rodden’s work (and indeed assign
> their “unintentional gerrymander” piece), but it seems like the automated
> districting approach (and the other two alternatives described in the
> brief) produce plans that satisfy only contiguity, compactness, and equal
> population requirements, but do not speak to the Voting Rights Act issues.
> Drawing a majority-black district can also result in inefficient “packing”
> of Democrats, and can potentially explain why a given set of maps is
> outside of the range produced by simulations. Is there any automated
> districting algorithm out there that specifically incorporates
> VRA/majority-minority districts?
>
>
>
> Thanks!
>
>
>
> Vlad Kogan
>
>
>
> [image: The Ohio State University]
> *Vladimir Kogan*, Assistant Professor
> *Department of Political Science*
>
> 2004 Derby Hall | 154 N. Oval Mall, Columbus, OH 43210-1373
>
> 510/415-4074 <(510)%20415-4074> Mobile
>
> 614/292-9498 <(614)%20292-9498> Office
>
> 614/292-1146 <(614)%20292-1146>
>
> http://u.osu.edu/kogan.18/
>
> kogan.18 at osu.edu
>
>
>
> _______________________________________________
> Law-election mailing list
> Law-election at department-lists.uci.edu
> https://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/20170914/1a939a60/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 3605 bytes
Desc: not available
URL: <http://webshare.law.ucla.edu/Listservs/law-election/attachments/20170914/1a939a60/attachment.png>