[EM] Redistricting, now with racial demographics

Fri Jul 17 09:07:27 PDT 2009

Kathy and Frank,

Deviation from the ideal population per district usually is very small for 
congressional districts (generally far less than .5%), but for state House 
and Senate districts the deviation is generally around 9% (not 0.9%...but 
a whopping 9%) with a several states in the high teens and even 20-30 % 
deviation. You can find a handy table on the NCSL web site here:
http://www.ncsl.org/LegislaturesElections/Redistricting/RedistrictingPopulationDeviation2000/tabid/16636/Default.aspx

However, any redistricting formula can only go so far, and cannot assure 
fair representation nor competitive elections because of the very nature 
of single member winner-take-all elections.

Terry Bouricius

----- Original Message ----- 
From: "Kathy Dopp" <kathy.dopp at gmail.com>
To: "Raph Frank" <raphfrk at gmail.com>
Cc: <election-methods at lists.electorama.com>
Sent: Friday, July 17, 2009 11:37 AM
Subject: Re: [EM] Redistricting, now with racial demographics

On Fri, Jul 17, 2009 at 3:11 AM, Raph Frank<raphfrk at gmail.com> wrote:

> From the link:
>
>> (i) has equal population districts to within 0.5%; and
>
> Sounds reasonable, as the error in the census is likely higher.
> However, the Supreme Court might object. They have ruled that equal
> population is essential.

I don't know. I suppose this needs legal research and research into
what exists now.

>
>> (ii) utilizes natural and geographic boundaries and barriers in the 
>> creation of district boundaries; and
>
> This requires a definition of both.

Yes. Perhaps, but natural geographic boundaries are crucially
important to pay attention to when drawing districts because huge
mountains or large impassable rivers, etc. are real barriers that
affect transportation, often affect existing political boundaries and
therefore affect how easy a district would be not only to serve, but
to administer.

>
>> (iii) utilizes existing government boundaries (in particular election 
>> administration boundaries such as county boundaries) in the creation of 
>> district boundaries; and
>>
>> (iv) minimizes the sum of all the perimeters of all the districts 
>> (produces compact districts); and
>>
>> (v) minimizes the ratio of the number of uniquely administered 
>> districts to the number of election jurisdictions (to reduce election
>> administration complexity.)
>
> Since the number of districts is constant, I am not sure you need a 
> ratio here.

Yes, but the number of "separately administered districts" that are
split by the number of jurisdictions can be very large, especially
with gerrymandering or with any redistricting plan that does not
consider this important issue.  Election officials will be much more
supportive of districting plans that minimize the complexities of
election administration like this simple ratio helps to do.

>
>> Let each political party draw up district maps and the "winning" map 
>> would be the one that:
>
> You could extend it to anyone.

Yes.  And I like that idea.

>
>> 1. has a minimum sum of perimeters, and
>> 2. has the minimum ratio of the number of uniquely administered 
>> districts to the number of election jurisdictions (for administrative 
>> simplicity.
>
>> (these two conditions can be equally weighted), and that meets the 
>> other three conditions.
>
> How do you equal weight these? Condition 1 gives a length and
> Condition 2 gives a ratio.

That's true.  Hmmm.  I suppose that there are several methods of using
these two measures to give a score to a submitted plan.  Both numbers
could be normalized using the same scale (say 0 to 1 or 0 to 10) in
comparison with the same measures for all the other submitted plans
for instance.  I can't think of a better way currently, but there may
be some.

>
> You could convert 1 into a ratio by saying something like "Ratio of
> sum of perimeters to the perimeter of the state".

Oh. That is interesting and something like that might work well
because it provides a concrete measure for comparison.  Still, it
would still need some adjustment to work with the measure of
administrative complexity because the two ratios do not have the same
range on the same scale.

>
> Another option is that you could redefine the rule as:
>
> the "winning" map is the one that:
>
> - has the minimum ratio of the number of uniquely administered
> districts to the number of election jurisdictions

(but only one map might meet this condition, in which case no further
test would be needed and that map would win, thereby undoing our
concern for compact districts.)

>
> and
>
> - where the sum of the perimeters is at most 5% larger than the valid
> map with the lowest sum of perimeters.
>
> and
>
> - where the boundaries follow "valid boundaries" as defined prior to the 
> census

Does the census "define" mountain ranges, etc.?

Here in Utah two towns may be very close as the crow flies, but take
many hours to drive between in the winter (and sometimes in the summer
too) due to having to drive all the way around the mountain ranges --
and same thing can be true of "close" towns separated by rivers with
very infrequent bridges over them.

>
> and
>
> - where the population of the lowest population district is at least
> 95% of the population of the highest population district
>

I don't know if that would be legally acceptable and is common
practice or not.  Perhaps you've researched this more than I have.

>
> Are election admin areas defined as part of the map, i.e. does the
> person submit maps for all elections (State+local+Federal) + how they
> should be administered?

Election administration areas are defined differently in each state
already and generally do not change much over time. In most states
counties are the jurisdictions, but in some states in New England
townships administer elections, in LA parishes (equivalent to
counties) do.

>
> Otherwise, I think that by defining admin areas, you could defacto
> gerrymander, as the best maps would have to follow those boundaries.

That is what the ratio measure is for, to try to minimize election
administration complexity.

>
> Also, the natural and geographic rule could be abused. Boundaries
> should definitely be decided prior to any census, so that there is
> some randomness.
>

Yes. I suppose so, but that could be prevented and it is crucial to
take the geographic and natural boundaries into account

Cheers,

Kathy

-- 

Kathy Dopp

The material expressed herein is the informed  product of the author's
fact-finding and investigative efforts. Dopp is a Mathematician,
Expert in election audit mathematics and procedures; in exit poll
discrepancy analysis; and can be reached at

P.O. Box 680192
Park City, UT 84068
phone 435-658-4657

http://utahcountvotes.org
http://electionmathematics.org
http://kathydopp.com/serendipity/

Realities Mar Instant Runoff Voting - 18 Flaws and 4 Benefits
http://electionmathematics.org/ucvAnalysis/US/RCV-IRV/InstantRunoffVotingFlaws.pdf

History of Confidence Election Auditing Development & Overview of
Election Auditing Fundamentals
http://electionarchive.org/ucvAnalysis/US/paper-audits/History-of-Election-Auditing-Development.pdf

Voters Have Reason to Worry
http://utahcountvotes.org/UT/UtahCountVotes-ThadHall-Response.pdf
----
Election-Methods mailing list - see http://electorama.com/em for list info