Update: In comments someone asked for an executive summary, my main conclusions. Here they are:
- Kerry did much worse than the "vote straight Democratic ticket" option did in a wide variety of counties.
- There is no hard, obvious evidence of malfeasance, such as vote totals exceeding number of registered voters.
- Bush did slightly worse than the "vote straight Republican ticket" option in only two counties - Greene and Taney - and both of these are heavily pro-Bush.
- I'd like to know if there are benign reasons for these results or if there are grounds for a recount.
While a lot of folks seem hot to trot on chewing up Ohio's numbers, I thought I'd look at Missouri's votes this weekend. Missouri has a Republican Secretary of State and all of the signs - the afternoon exit polls as well as the early returns - pointed to a mild Kerry victory. So I wanted to see what went wrong. I spent some time on the Missouri Secretary of State's website this weekend (and as many county board of election websites as I could find) to compile the following numbers...
- The number of votes for Kerry, and the number of votes for Bush, for each county
- The number of straight-ticket Democratic Party votes, and the number of straight-ticket Republican Party votes
These numbers are not available for every county - obviously the Bush/Kerry numbers are available for each county, but the Democratic/Republican straight ticket vote is not.
Where possible, I checked that the number of Presidential votes cast was no more than the number of voters registered - this appeared to be true everywhere I could find the numbers. Also where possible, I retrieved the precinct-by-precinct numbers and compared them to the county-wide totals. Again, these seemed to match, with one exception. Lincoln County reports its totals as 8367 Kerry, 11316 Bush; but adding up the precincts, you get 5211 Kerry, 8216 Bush. This is a little strange, but its possible that they forgot to link to a few of their precincts on the website. The net effect on the margin is also very small (the margin expands in favor of Bush by 56 votes when you use the precinct-by-precinct numbers.)
Then I computed the following statistics:
- Kerry Projected: the number of votes Kerry would get if he got the same share of the Presidential vote that Democrats got of the straight-ticket vote.
- Bush Projected: the number of votes Kerry would get if he got the same share of the Presidential vote that Republicans got of the straight-ticket vote.
- New Votes for Kerry: the additional margin Kerry would have gotten, on top of the margin he already has, if he and Bush got the projected votes.
- New Votes for Kerry as Percent of Total Vote: the additional margin Kerry would have gotten, divided by the total number of Kerry-and-Bush votes
In mathspeak:
- KP = (K + B) * D/(D + R)
- BP = (K + B) * R/(D + R)
- NV = (KP - K) - (BP - B)
- PT = NV / (K + B)
Here KP = Kerry Projected, BP = Bush Projected, K = Actual Kerry Votes, B = Actual Bush Votes, D = straight-ticket Democratic votes, R = straight-ticket Republican votes, NV = New Votes for Kerry, and PT = New Votes for Kerry as Percent of Total Votes.
Note: the reason I've ignored third-party votes and ballots without votes for President above is because there is no option to vote straight ticket for third parties, and also, in several counties the total number of ballots is not available. However, the third party vote is very small - out of 2,700,000 votes cast for President, about 16,000 went to a third party. I don't expect that replacing "K + B" in the above formulas with "total number of ballots cast" would change the numbers below very much.
Here are the results:
County | Kerry | Bush | Dem | Rep | Kerry, Projected | Bush, Projected | New Kerry Votes | Percent of Total |
ST LOUIS COUNTY | 294402 | 244388 | 172693 | 111316 | 327613 | 211176 | 66424 | 0.12 |
ST LOUIS CITY | 115414 | 27460 | 75483 | 9484 | 126926 | 15948 | 23025 | 0.16 |
KANSAS CITY | 102233 | 35529 | 60468 | 12078 | 114826 | 22936 | 25187 | 0.18 |
GREENE | 46626 | 77841 | 21887 | 38056 | 45449 | 79026 | -2354 | -0.02 |
JEFFERSON | 46057 | 46624 | 21280 | 13360 | 56935 | 35745 | 21757 | 0.23 |
CLAY | 44522 | 51033 | 17726 | 16717 | 49177 | 46378 | 9310 | 0.1 |
BOONE | 37480 | 37515 | 11363 | 8251 | 43660 | 31703 | 12133 | 0.16 |
BUCHANAN | 17799 | 19812 | 7368 | 5099 | 22228 | 15383 | 8858 | 0.24 |
CASS | 16681 | 27248 | 5948 | 7877 | 18899 | 25029 | 4438 | 0.1 |
COLE | 11753 | 24752 | 4338 | 4102 | 18761 | 17741 | 14018 | 0.38 |
TANEY | 5585 | 13550 | 2193 | 5932 | 5164 | 13970 | -841 | -0.04 |
TOTAL | 738552 | 605752 | 400747 | 232272 | 829638 | 515035 | 181955 | |
STATEWIDE TOTAL | 1253950 | 1452793 | | | | | | |
The two numbers that jump out at me right away are the total New Kerry Votes (181,000) compared to the statewide margin for Bush (about 200,000). Only 11 of 116 counties are reported here, but they total up to about half the total number of votes statewide. One of the biggest counties not reported here is Jackson County (the county surrounding Kansas City; it voted for Bush 94,083 - 78,711.)
Other interesting facts to notice:
- The counties tallied above represent about one-half of the total statewide Bush-Kerry vote (1,344,304 / 2,706,743).
- The New Kerry Votes is a substantial part of the total in every county except two: Greene (which contains Springfield, Missouri's third biggest city) and Taney. In each of these counties Bush actually does a little worse than the Republican share of the straight-ticket vote. Greene County provides a precinct-by-precinct breakdown, which I tallied separately from the official topsheet results; the numbers match up.
- In heavily pro-Kerry areas (St. Louis City and Kansas City) Kerry underperforms the Democratic share of the straight-ticket vote by 16% and 18%, respectively.
- In mildly pro-Kerry St. Louis County Kerry underperforms the Democratic share of the straight-ticket vote by 12%.
- In mildly pro-Bush Boone, Buchanan, Clay, and Jefferson counties, Kerry underperforms the Democratic share of the straight-ticket vote by 16%, 24%, 10%, and 23% respectively.
- In heavily pro-Bush Cass, Cole, Greene and Taney counties, Kerry underperforms the the Democratic share of the straight-ticket vote by 10%, -2%, 38%, and -4%, respectively. (The -2% and -4% indicates, as discussed before, that in Greene and Taney it's actually Bush who does a little worse than the Republican share of the straight-ticket vote.)
- I also performed a correlation of Kerry's underperformance as a percent of the total ballots (PT, in the above) versus the Democratic share of the straight-ticket vote (D/(D+R)). The correlation statistic is 0.483; squared is 0.23, which is a pretty weak correlation.
- The total number of votes for Kerry and Bush in these counties range from 538,790 in Missouri's largest county, St. Louis County, to 19,135 in the much smaller Taney County.
Taking all these facts together, it appears to me that in a vast swath of counties - very big, medium sized and very small, strongly pro-Kerry, mildly pro-Kerry, mildly pro-Bush and strongly pro-Bush - Kerry significantly underperformed the Democratic straight-ticket vote. I've also included some counties which I know have interesting characteristics - St. Louis city is heavily minority; St. Louis county contains heavily minority North St. Louis county, heavily working-class South St. Louis county, and upper-class West St. Louis county, so while it is mildly pro-Kerry, there are big pockets that are strongly pro-Bush and strongly pro-Kerry; Jefferson county is something of a "yellow-dog" county; Boone county includes the University of Missouri at Columbia, which makes it lean a bit more Democratic than it would without the university. But in all these places, Kerry underperformed the Democratic share of the straight-ticket vote by 10%, and sometimes much more. In other words, people who we might think of as "Independents" voted for Bush very heavily all over the state - with the two exceptions noted, Greene and Taney counties.
This conclusion seems very strange. I know that not all partisans vote a straight ticket - so labeling non-straight ticket voters as "Independent" is a bit misleading; it actually includes some very reliable partisans, who will vote individually for every Democrat (or Republican) on the ballot, in the "Independent" camp. Even so, the fact that Independents went for Bush in such tremendous numbers in such a wide variety of counties seems suspicious; especially given that in Greene county, a very heavily pro-Bush county, Independents favored Kerry slightly.
I'm not suggesting malfeasance but I'm curious to see if anyone has some idea what would explain these numbers. Also, I wonder if these numbers speak to some kind of strategy ("energize the base", "improve the message"?) that would make the counties vote more in line with their straight-ticket breakdown in 2008, and thus turn Missouri blue next time around. And I wonder why Bush did worse (albeit only a little worse) than the Republican straight-ticket vote in such strong pro-Bush counties as Greene and Taney.
Also - I'm wondering if there are any grounds for a recount in some of these counties? Keep in mind that if Missouri, New Mexico and Iowa flipped to Kerry, Kerry would win the election outright, Ohio be damned.