|
|
|
Mathematicians of all political persuasions are interested in studying the 2004 election results due to the low likelihood of exit polls in so many swing states differing so widely from vote counts. Statisticians can use the following datasets to pinpoint counties and voting districts that seem likely to have vote counting errors and statistians can use these data sets to produce convincing evidence for Freedom of Information Act type requests for a closer investigation of voting equipment and ballots:
We have put together a team of statisticians, mathematicians, and programmers to begin the process of assembling all the data from the 2004 election into a database for analyses. We plan to make the data publicly available on the Internet for anyone who would like to analyze it, plus have our own team do analyses.
In addition to statistical analyses of patterns of election results, we can mathematically combine the patterns of the following data sets with each other to narrow down the search to the most likely counties/precincts to need recounts in America using demographic data as well:
FYI, Precinct level data from North Carolina
1. Tables of election results versus "expected votes" based on registration by party by voting machine type data just like in http://uscountvotes.org and do statistical analyses.
2. Control data. i.e. Comparisons with other states with with similar demographics and voting machine patterns. Then analyze and compare. Unfortunately, voter registration numbers by party are unavailable in AR, AL, and OH has only Ohio primary vote by party and Ohio voter reg. numbers that are not accurate enough for 2004. State election officials may be able to make those numbers available for 2006, at your urging. Here is PA.
3. We can compare counties with themselves for different methods of counting ballots using data for each county broken out by ballot type (early, absentee, provisional, and precinct) and ballot counting method for each type. Depending on the state, precinct data becomes available about a week after the election. Each county/precinct's results can be compared to its own results for different ballot types statisticians may be able to analyze the patterns accross counties. WE NEED HELP OBTAINING THIS DATA. If you know anyone who is collecting such data please tell them that we shall be ready soon to receive it. We need to know the Presidential, US Senate, US House election results for each precinct and for each ballot type and what method was used to count each type of ballot. (It can be an arduous task to purchase it from districts or counties and put it into a uniform form.)
4. The above data can be compared with results from other local district or state races or especially in Florida Senate Races that show a pattern that is different than the presidential race. FL04 by district comparison President vs. District. These patterns can be studied over past years to see trends.
5. Error rates when there are recounts, plus spoiled/uncounted ballots from prior election recount corrections can be indicators, when they are available.
6. Earlier election results by county/precinct in former elections going back to 1992 and earlier ( Berkeley did a statistical study that purported to show that touchscreen counties in FL are likely to have been miscounted (look under Voting). Here is a good critique of the Berkeley study. Here is an informal trend study by Rhosgobel Rhadaghast However, studies at the county level do not have the statistical power needed and precinct level studies are needed.)
7. Exit polls are unlikely to be inaccurate. WE NEED BY COUNTY EXIT POLLS BY THE 2006 ELECTION. By county exit poll comparisons may be available for 2004 by aggregating multiple polls and adding county codes to existing polls to find large enough sample sizes in some counties (No one has mentioned that this is possible for 2004 to me yet though.)
8. We can flag counties/precincts which are under investigation for reported vote loss/miscounts by new BBV, CommonCause, or verifiedvoting projects, by working with groups like Vote Protect - Nationwide Election Incident Reporting and Voters Unite - Machine Malfunctions
9. Overvotes, spoiled, or uncounted ballots, i.e. Votes cast but not counted for the race under consideration.
10. Historical trends to see if counties/precincts do not follow them. For instance, Ida Briggs, used this type of analyses to pinpoint voting districts in New Hampshire that look like they needed to be recounted, and consequently New Hampshire will have a recount.
11. Comparison with similar analyses of Election Primaries.
12. Bev Harris used turnout figures to pinpoint precincts in Volusia county that need closer examination. Turnout rates can be calculated and compared accross precincts by subtracting total ballots from registered voters.
By 2006 we plan to develop, test, and refine measures of analyzing elections and organize to produce many of these measures within a day of the election for some states. Before a candidate thinks of conceding an election, he will have this analysis available.
Please Subscribe to our mail list if you can support us financially or would be able to volunteer or are just interested in following what we do.