Quick update on the app. Because so many of you have been using the app (and it was discussed on the DailyKos front page last weekend) and because the test data for NY is so large, we've blown thru my bandwidth quota for the month. I've upped it a couple of times, and we're over that, too. So far it seems the site still works, thanks to the grace of my host.
But, I've made a change so that once you load the date for a state, the app saves it on your machine in the same place as the drf.xml files. If you then refresh the browser and reload the state, it grabs the data from your machine, saving a lot of bandwidth and...bonus...making it a lot faster to load.
The Census Bureau has released its annual population estimates, so that means the usual players are in the field with their reapportionment projections. First up is Election Data Services. (You can check out their prior studies as well: 2007 | 2008.)
EDS now offers six different projection models. The column headers indicate the range of time used to calculate each projection.
State
2000-2009
2004-2009
2005-2009
2006-2009
2007-2009
2008-2009
Arizona
2
2
2
1
1
1
California
0
-1
-1
0
0
0
Florida
1
1
1
1
1
1
Georgia
1
1
1
1
1
1
Illinois
-1
-1
-1
-1
-1
-1
Iowa
-1
-1
-1
-1
-1
-1
Louisiana
-1
-1
-1
-1
-1
-1
Massachusetts
-1
-1
-1
-1
-1
-1
Michigan
-1
-1
-1
-1
-1
-1
Minnesota
-1
-1
-1
-1
-1
-1
Nevada
1
1
1
1
1
1
New Jersey
-1
-1
-1
-1
-1
-1
New York
-1
-1
-1
-1
-1
-1
Ohio
-2
-2
-2
-2
-2
-2
Pennsylvania
-1
-1
-1
-1
-1
-1
South Carolina
1
1
1
1
1
1
Texas
3
4
4
4
4
4
Utah
1
1
1
1
1
1
Washington
1
1
1
1
1
1
The biggest losers in this new batch of projections are, not too surprisingly, the sun-belt states of Arizona and Florida. Last year, Arizona was expected to gain two seats under every single projection model. Now, in a stark demonstration of how southern migration has slowed in the midst of the Great Recession, the three nearest-term projections all show it picking up just one seat. Meanwhile, Florida, which still looked to gain two seats according to longer-term projections in 2008, now grabs just one new seat under all models.
Other losers include Oregon, which was slated to grab a new district under four of five models last year - it's off the list entirely this time. North Carolina was in a much more marginal situation in 2008 (gaining a seat under two of five models), and it too drops from the list. Texas shows a teeny bit of wobble, as the longest-term projection now shows it picking up three rather than four seats, but it seems like the odds still favor four. California, on the other hand, stabilizes some more, with four of six models (including all the nearer-term ones) indicating it won't lose any seats (last time, only two of five did).
The biggest gainers? That would be Missouri, which isn't on this list at all - and for the Show Me State, that's a good thing. In 2008, all five models projected a one-seat loss, and in fact, in 2007, all three models did as well. Now EDS thinks Missouri won't lose any seats. Meanwhile, Washington state is brand-new to the list, gaining a tenth district acoss the board.
Polidata also has an analysis out. They only do one projection, based on the most recent year's numbers, which matches EDS's 2008-2009 projection in all respects. They also offer a list of which states barely hang on to their final seats and which states are oh-so-close to nabbing one more:
Rank
State
Makes/ Misses By
431
South Carolina
20,000
432
Washington
30,000
433
California
120,000
434
Texas
40,000
435
Missouri
10,000
436
Minnesota
10,000
437
Oregon
20,000
438
Arizona
50,000
439
Florida
150,000
440
North Carolina
75,000
441
Illinois
140,000
442
Ohio
130,000
443
New Jersey
110,000
444
Massachusetts
90,000
445
Louisiana
70,000
EDS has a similar chart with "last six/next six" on the final page of their PDF, with different iterations for each of their models. The bottom line is that right now, Missouri looks very lucky and Minnesota looks very unlucky. But given the small numbers involved and the fact that we're dealing with estimates rather than actuals counts, I would not be surprised at all if things changed by the time we get final numbers in from the 2010 Census.
P.S. Check out Dave's diary for some more discussion of these new studies.
States gaining based upon July 1, 2009 estimates projected to April 1, 2010: AZ +1; FL +1; GA +1; NV +1; SC +1; TX +4; UT +1; WA +1.
States losing based upon July 1, 2009 estimates projected to April 1, 2010: IL (1); IA (1); LA (1); MA (1); MI (1); MN (1); NJ (1); NY (1); OH (2); PA (1).
Thus, changes in seats compared to the previous projections based upon 2008 estimates would be: AZ gains only 1 seat while MO stays even and OR stays even while WA gains 1 seat.
Check it out: Washington would gain a seat; Arizona only gains 1; Oregon doesn't gain any.
New York State now has partisan data in Daves Redistricting! Thanks to jeffmd for the ton of work he did to make this happen. Please read below the fold before trying this!
Last night I uploaded upgraded software that can handle partisan data. The state of Maryland is the first state (and only so far) to include that data. There are some caveats (below the fold). Also, keep checking here for ways you can help get more partisan data.
[Update: there was little glitch so partisan didn't show up, but it should work now! Thanks!]
[Update 10/13 3:41 PDT: the Hispanic data for California, Kentucky and Wisconsin was incorrect. I've now uploaded the correct data. Thanks to CalifornianInTexas and nico for noticing.]
My time coming, any day, don't worry about me, no
...
California, preaching on the burning shore
California, I'll be knocking on the golden door
Like an angel, standing in a shaft of light
Rising up to paradise, I know, I'm gonna shine
David asked me to do a quick example of how I do my redistricting, so I decided to use a relatively 'easy' state, Wisconsin, since it has townships. In the same way I wouldn't drive a car or ride a bike without brakes (yes, fixie riders, the laws of physics hate you too), I don't want to redistrict without having both sets of information.
Before you start, take a look at the state, and try to think how much detail you're going to want to go into. I usually do townships/municipalities when available, but sometimes precincts. For Wisconsin, I'm going shooting for 7-1, by making the 1st and 6th more Democratic. Wisconsin doesn't have precinct-level census data, so we'll have to stick with townships. I really do want to split the city of Milwaukee though, so we're going to do some estimating in just a little bit.
So now, here's a quick overview of the process ahead:
1. Collect political and population data
2. Aggregate data together into one spreadsheet
3. Prepare a GIS shapefile (we'll get into that) for the precinct and precinct-equivalents we're going to use.
4. Import data from the spreadsheet into shapefile
5. Redistrict!
This redistricting map has really been a long time coming, but trying to match up the 36,000 census blocks of NYC and the 20,000 blocks of Nassau County (among many, many others) with a precinct and creating a consolidated population-political data file tends to take a while. But unquestionably, I also think this is my best map yet. I used 2008 population estimates by city/town, projected forward to 2010. I also worked painstakingly hard for population equality, which I think is evident in that the total deviation for all 28 districts is 86, or 0.012% of a single district.
At a whopping 26-3, our NY delegation is already pretty maxed out. With the loss of a district in 2010, there's a good opportunity to consolidate and create what could possibly be a 27-1 delegation. I chose to do this by eliminating John McHugh's district and then employing some old-fashioned cracking of Nassau County to defeat Peter King. Side goals included strengthening Massa, Arcuri, Murphy, Hall, and McMahon, while preserving VRA districts for Meeks, Towns, Clarke, Velazquez, Rangel, and Serrano.
I've just uploaded a change that allows the application to use voting districts in addition to block groups, as the building blocks for drawing redistricting maps. This is a step toward getting partison data (specifically the 2008 presidential vote) into the app, because voting district partisan data is available for many states. That will require a bit more work, but will be coming soon.
I've added 6 new states, all using voting districts: Alabama, Colorado, Illinois, Maryland, Mississippi and Virginia. I added voting districts to the existing state Georgia, so it supports both. If you have saved DRFs, you'll need to keep using block groups on those files.
Johnny beat the annoucement that New York is ready, with a great map. (Nice going!)
I've also made a couple of improvements to the software, prompted by working with New York:
- Thinner block lines, so you can see the tiny blocks.
- The slow work of Assign Old CDs now happens when you click the button instead of on load. (It took about 4 minutes for NY on my not too slow machine.)
- Zoom: Scrollbar replaces slider.
