The long-awaited update is finally uploaded to Dave's Redistricting App. In summary, these are the states with revisions.
2008 presidential data has been added for Kentucky, Massachusetts, and West Virginia.
Revised 2008 presidential data has replaced the previous data for Georgia, Idaho, Maryland, Nebraska, Ohio, Oklahoma, and Tennessee.
I'll go over the revisions after the fold.
I have not yet had a chance to review the uploaded data so if anyone catches anything that looks like an error, please let me know. It may be a few days before I have a chance to review all of the uploaded data myself.
Meanwhile, this is what I'll be working on next.
Arkansas: I have a full 2008 precinct data set but it needs a good bit of prep work to translate it into the DRA voting districts. I simply have not had a chance to get to it. Unfortunately for Arkansans, it's not a swing state in this election nor does it have any competitive congressional races, so it's low on my priority list.
Florida: I want to add the minor party figures. This is now the only state I've done in full where they're missing. It's not high on the priority list, so I don't know if this will be in the next round of updates or in the one after that.
Iowa: I finally got a full statewide set of precinct shapefiles on Tuesday. This should now be a fairly quick state to complete. The main issue with Iowa is that rural election precincts are divided into multiple voting districts for census purposes, so there was no viable way for me to do it without 2008 precinct maps.
New Jersey: I had hoped to have the long-overdue NJ fix ready for this update. I've done a lot of work on New Jersey as it is. Tragically, my underlying data set got corrupted (short version: the formulas I use to cross check municipality totals with official results somehow got scrambled into the data itself). Anyway, I hope to rebuild my data set and wrap up the revision late next week.
Virginia: To my knowledge, the discrepancies are not major, and mostly involve missing early votes. I've collected the data I need for a revision. I just haven't had a chance to revisit the state. It's high on my to-do list though since it's a swing state. I'll probably do it before I go back to New Jersey, even though the NJ errors are far more extensive.
Now, here's a recap of the revisions.
Georgia: The previous file on DRA was a 'hack' where absentees were distributed based on voting age population. The new data distributes absentees by the same formula that I've used for a handful of other states where early vote data is not provided by precinct. My formula distributes absentees based on the proportion of each candidate's countywide election day votes received from each precinct. This revision also corrects errors in some of the underlying election day data for a number of precincts.
Update: Georgia is fixed!
Idaho: I added minor party votes and fixed the precinct numbering for Burley.
Maryland: There were two major types of error in the replaced data. (1) Wherever precincts were split into multiple census VTDs, the original precinct figures were duplicated for all the consituent VTDs. That's why, for instance, there were voting districts in Prince George's County with zero people and thousands of votes. (2) Where precincts were merged between 2008 and 2010, the previous data simply dropped the highest numbered precincts in the given municipality and reordered the rest. So, all the precincts after the lowest-number merged precinct were wrong.
Nebraska: The data previously on DRA was a big, huge mess. I've replaced it with the statewide official canvass. The previous data incorrectly assigned precincts for a number of counties. It excluded about 140,000 early votes and added the minor candidate votes to the McCain figure. A number of precincts were also missing any election figures, particularly in Douglas County (Omaha), which is why the old data incorrectly had McCain winning NE-02.
Ohio: The replaced data did not account for precinct splits, mergers, or boundary changes. The new data set corrects about 800 precincts according to my notes. It also adds in roughly 60,000 votes that were missing from the old data set. For the record, the only votes that are not included in the revised data are the federal ballots for Columbus (606 Obama, 139 McCain, 8 Other) and Toledo (124 Obama, 17 McCain, 3 Other).
Oklahoma: The previous DRA data was missing about 190,000 votes. There were also mixed up precinct figures in all the main cities. Note that Oklahoma didn't have any third-party candidates or qualified write-ins on the 2008 ballot, so there are no Other votes.
Tennessee: I've added minor candidate vote figures. I've also added about 150,000 missing Obama/McCain votes. The data previously on DRA missed precinct figures that were reported on multiple lines (especially around Memphis and, to lesser extent, Nashville). Also, the precinct figures in several counties had been scrambled, especially for Hamilton County (Chattanooga) and for Knox County (Knoxville).