[EM] More on Gerrymander prevention

Joe Weinstein jweins123 at hotmail.com
Wed Mar 20 15:47:10 PST 2002

Re: [EM] More on Gerrymander prevention

Of course one can specify an automated redistricting scheme, and it may be a 
very good one too.  And, from a Realpolitik viewpoint, the scheme’s actual 
adoption or nonadoption may have little to do, pro or con, with its inherent 

The essence of anything like rational districting calls for the following.

   First, define a class of a priori allowed districting schemes.  (This 
calls for describing precisely just what information is needed in order to 
specify a scheme.)

   Second, define a measure of merit which can be precisely computed for any 
such allowed scheme.

Both these definitions require that, given the official ‘districting 
computer system’ - henceforth termed just ‘the computer’ - any allowed 
scheme can readily be specified to the computer, the computer can readily 
verify its allowability, and the computer moreover can rapidly compute the 
scheme’s score, i.e. measure of merit.

As I noted recently, and I think it bears emphasis, it may be IMPRACTICAL to 
guaranteeably and provably get the (or an) OPTIMUM allowed scheme.   It 
might take the computer - even of current vintage - too long to check 
terribly many cases.

However, it should be VERY practical to get an EXCELLENT scheme - one which 
no party at interest would have any moral right to dispute.

In fact, there may be zillions of allowed N-district schemes, and the number 
of allowed schemes may grow exponentially with N, but that need not matter 

De facto, all that is required is that available computing power could 
rapidly enough process up to X electronically submitted proposed districting 
schemes, where X is any agreed-on reassuringly large number, say X=10,000 
(or even X=1,000).

The mechanism is simple:  let V be the total number of voters, and let each 
voter be allowed to join (i.e., sign a petition or roster of) one 
‘nominating group’ comprising at least V/X fellow voters.  Presumably, some 
such groups will represent political factions or special interests; but at 
least a few will represent good-government citizen organizations.

Each such nominating group may nominate, i.e. electronically submit, a 
districting scheme.  After the computer has scored each nominated scheme, a 
highest-scored scheme must be adopted.

If such scheme is unique, the choice is automatic; if there is a tie, the 
tie is broken randomly, or possibly by action of a court, or of a special 
commission.  Conceivably the special commission may be highly politicized, 
maybe deliberately by law as the price of getting basic rationality into the 
districting process.  But, so long as good-government citizen groups submit 
superior schemes, the final choice will have to be quite good.

By the way, in order to ensure accountability of the official computation, 
each nominating group would be sent the specifications of the winning 
scheme, so that their own computer can verify that this scheme has at least 
as much merit as their own nominated scheme.

In all this, the main problem is to define a rational defensible measure of 
merit, i.e. to define exactly what makes (or should make) a scheme 'good' - 
or, if you prefer other terminology, of high 'social utility'; and what 
makes (or should make) one scheme 'better' (of higher utility) than another.

For instance - despite the implication of recent postings by Rouse and 
DemoRep - a long skinny (and even twisted) district may be better than a 
square or near-square rectangle.  Let me close by discussing this issue in 
more detail.

     Here in the USA there is a bias toward rectangularity and squareness of 
land plots, most often achieved through a north-south by east-west grid.  
This bias owes to the nation’s arguably most influential single federal law, 
the 1787 Northwest ordinance, which ordained survey and gridded division of 
the then-unsettled lands (roughly 200 miles and further west of the Atlantic 
Coast).  The gridding works well enough in the essentially flat and anyhow 
largely homogeneous ('isotropic') Midwest, but makes little sense further 
west in the Rockies and Sierra Nevada and other mountains.

    What's really at issue in districting is not geometric shapes per se but 
rather what these shapes are meant to achieve.  Namely, we want districts to 
be relatively homogeneous internally, compared with inter-district 
differences.  This is the basic principle for a statistically sound and 
useful experimental design:  to divide the population under investigation 
into strata which are relatively homogenoeus internally, compared with 
inter-strata differences.  The same principle is (or should be) also used to 
distinguish so-called 'bioregions' from mere arbitrary divisions of the 

    Compact district shape is but a tool to promote easy intra-district 
communication.  In itself, such communication is a worthy means for 
promoting political accountability, and it is also an important means for 
fostering or maintaining the desired homogeneity and abiding commonality of 
interest of the district population.

    For this reason of relative ease of communication, at least near long 
high mountain ranges like the Andes or the Sierra Nevada, a long skinny (and 
maybe a bit twisted) Chile-like or Owens-Valley-like district may be 
precisely what one needs.

Joe Weinstein
Long Beach CA USA

MSN Photos is the easiest way to share and print your photos: 

More information about the Election-Methods mailing list