| It is interesting to note the role-reversal of partisans. Those content with the first map argued for a minimalist interpretation of the Iowa Code -- it's within spec, so let's move on! Those promoting a second effort invoked augmenting the Code with common sense prescriptions -- the Iowa Way, if you will.
While the Legislative Service Bureau (correctly) pointed out in rebuttal that the federal census provides no formal classification of rural/urban persons and offers no formal definition of political subdivisions corresponding to "communities of common interest," arguing that the judgment of pointy-headed D.C. bureaucrats trumps local concerns is a case nobody wants to make. Census, Schmensus: If the initial effort in 2011 packs the population into three thriving, fecund districts while corralling the low-density, aging population into one district with almost half of the state's counties (a possible outcome in 2011), expect to hear these arguments yet again from parties representing the plan's opponents.
Commentators have noted that the underlying driver for the successful campaign to reject the initial proposal stemmed from state legislator dissatisfaction with the accompanying state senate/house configurations. Probably quite true, however, the key point is that almost any map produced will have its detractors, so the name of the game is to meet, or better yet, to exceed Code specifications with an otherwise judicious blend of interests as the best hedge against likely arguments.
In my view, the LSB made a tactical error by asserting in its first proposal:
the Congressional plan selected had the lowest absolute mean deviation of any of the plans drawn that also had districts that were composed of conveniently contiguous territory that were compact in form.
Now let's look at the comparative statistics from the second report:
(Note: last entry should read 2nd Proposal/House)
No matter how suspect the intentions of the first proposal's critics, the second plan displays significant improvement in population equality without real deterioration in compactness standards (an argument for nominal improvement may even be attempted). Note similar trends for senate/house maps. Furthermore, while both plans are in compliance with the Code, it is difficult to argue against order-of-magnitude improvement from 130.2 persons/district to 47 persons/district. This also established a precedent that minimization of population statistics is desirable, if possible.
Generally speaking, the blanket statement made in the original report was unwise because redistricting is a combinatorial optimization problem that is targeted with heuristic methods, which, by definition, do not guarantee convergence to the optimal solution in a finite amount of time.
Most citizens are not interested in the finer details of algorithms that find solutions to these types of problems. Fair or not, the average person is suspicious when a proposal is advertised as 'best,' yet improvements are found under pressure. This undermines the authority and legitimacy of the plan-makers. For this reason, I expect that in 2011, every reasonable effort will be made for successful approval of the initial plan, with much attention focused on minimization of population variance statistics. For example, one district in the neighborhood of the allowable 1% deviation from the ideal population is highly unlikely.
In 1992, a (Republican-leaning) group called Iowans Against Gerrymandering sought access to the source code and associated database procured by the LSB in service of its redistricting mission. The Iowa Supreme Court found:
The databases were created for and purchased by the General Assembly for use in legislative redistricting and the computer program developed by the vendor interrelated the different forms and sources of information. The database was encrypted, or coded, as required by contract with the original creator of the software. The information or raw data used in the databases was public information and was otherwise available to the plaintiff. The Supreme Court held that the computer databases were a trade secret of the vendor who prepared them.
As noted, the data is (or will be) available to the public. ArcView GIS is the front-end tool currently used to display and manipulate spatial information. A plug-in to ArcView from a Maryland vendor generates plans for initial consideration. Not only is it fairly straightforward to emulate a similar environment without spending a penny, the potential exists to improve on any automated drawing procedure. There are at least three applicable algorithms, which is probably more than what the plug-in supports.
The LSB has a narrow window for producing the initial plan, on the order of two months, but there is nothing that prevents a blog like this one from embarking on a group project to do the same, whether reviewing individual submissions or enabling automated procedures to exhaust possibilities in conjunction with 'public hearings' sponsored by the blog. Well, there is one caveat -- partisans may not like some of the contenders. On the other hand, this may produce the configuration of your dreams, or, some lucky Bleeding Heartlander may serve on the next TRAC!
Finally, a few comments about plans already proposed by various bloggers and commentators. For convenience, all three are reproduced below. K refers to Krusty Konservative's proposal, discussed in detail here.
His discussion is a specific comparison to nathaniel90's configuration, N, presented here. On Friday, new Bleeding Heartland user ghbraves, now known as G, offered a third option in comments.
Some tedious housekeeping is required:
The following source data are available:
July 1, 2008 estimate of statewide population: 3,002,555. Not usable, no associated breakdown by county available.
Revision of July 1, 2007 estimate: 2,983,360. No associated breakdown by county available. Note that the revision adjusts the population downward. Generally, revisions are made to provisional estimates once final figures for births, housing permits and state tax returns are available.
July 1, 2007 estimate (Excel/.xls): 2,988,046. This spreadsheet provides the most recent estimates for county populations.
County population projections for 2010 (Excel/.xls).
It is important to note that as the decennial census is still in progress, no determination of success can be made with certainty. The main point here is that small changes do make a difference. ghbraves remarks:
I imagine that all of our respective pop. estimates won't be too different. As for me, I simply took the most recent county population figures and educationally estimated a logically appropriate increase in population.
I don't think it'll be a huge difference, though
and Krusty explains:
In fact, I could not find my original notes so the numbers I plugged into their model are slightly different from the numbers I must have used in January.
Let's start with Krusty. Adding up the column of population figures associated with his plan yields 2,983,805, a mystery figure. It is a bit larger than the 7/1/2007 revision, which offers no associated county-level allocation but less than 7/1/2007 estimate by 4,241 persons. Applying the known data linked above to Krusty's proposal results in its failure -- specifically, excess population in District 4.
In a bizarre twist, Krusty assigns another mystery figure, a total of 2,982,383, to nathaniel90's configuration. In fact, nathaniel90 is comfortably within specifications using the figures he obtained from the ISDC. In other words, if the redistricting were held today, using the most recent county population figures, nathaniel90's plan would be promoted for further consideration while Krusty would be close but not a kontender.
Model G fares quite poorly, but as ghbraves has stated, it is intended to represent county growth and decline rates beyond 2007.
Krusty also lays claim to greater compactness. Compactness is a very controversial issue, with over thirty standards in the literature. The software used by the LSB features twelve optional measures. The Iowa Code specifies two standards for use in Iowa redistricting: the average of the absolute differences between district lengths/widths (smaller score is better) per plan in conjunction with a plan perimeter score, the total amount of perimeter used to construct districts in a map (smaller is better).
If this seems puzzling, consider the following. The most compact district is a circle, the shape that uniquely minimizes a perimeter for any given area. Creating circular districts is near-impossible, of course, so regular n-sided polygons are considered next. While the Code mentions hexagons, realistically, the shape of Iowa's counties and the contiguity requirement makes the square the natural model for compactness. So, the absolute value of a district's length minus width is the deviation from squareness.
As an example, arrange nine "counties" in a 3x3 matrix, assigning each county unit length. Now, remove each corner county and assign one to each remaining branch. The two shapes, a square and a cross, possess identical |L-W| values, but the cross generates greater perimeter, hence the inclusion of a test to reward parsimonious usage/penalize for excess.
To assess compactness, I created ArcView shapefiles for the three plans, to ensure uniform application of the standard (measurement of natural boundary perimeters, identification of extreme points of lat/lon, etc). For the |L-W| measurement, it's a toss-up between Krusty and nathaniel90, with ghbraves closely behind.
For the perimeter standard, ghbraves makes most efficient use of perimeter, followed by nathaniel90 and Krusty. The differences are not of sufficient magnitude for any one configuration to claim dominance. This, of course, makes desmoinesdem's father the real winner with his observation that "reasonable minds can differ." The take-home point is that the compactness measure, as defined, has little discriminatory power and is unlikely to be a focal point in 2011. That Iowa even defines a measure is simply a guard against extremely suspicious meanderings -- you're fine unless your districts have fingers.
As cited above, the state has some "official forecasts," which are actually made by a Washington, D.C. firm. desmoinesdem comments:
A big question mark will be Dallas County, which grew at a ridiculous pace between 2000 and 2007. Presumably the collapse of the housing market will limit further growth from 2008-2010, but who knows?
Here you will find the historical trend and forecast for Dallas County on one graph. Note that the trend is damped, while ghbraves uses the more bouyant figure of 65,000. Perhaps it is not surprising that model G improves using the official projections for 2010 but still fails. However, the forecasts will almost certainly not be the actual census figures, so the real point is that small differences do have an impact on viability.
The principle difference between a forecast and a revision is that the latter uses actual state data on births and deaths, from housing permits and from income tax returns as soon as they are available. A forecast is different as the trend is based on predicting the macroscopic net migration rate as well as the natural change (birth - death), all with national/state/local components of trend.
I wondered in comments how model N would fare in the future, especially with boomer aging. There are some recent negative spikes in annual natural change in certain counties like Adams, Appanoose, Audubon, etc. nathaniel90's model does fail using these forecasts, as his District 4 falls below the 1% threshold while District 3 continues to grow beyond it. The Dumb Luck award goes to Krusty, as the relative decline of District 4 to his District 3's growth brings his model back from the dead.
Sometimes, life is unfair. |
![[Bleeding Heartland Logo]](http://www.bleedingheartland.com/upload/logo.jpg){kind=logo}
