Wisconsin 2010 Senate True Vote Analysis
June 16, 2011
Updated May 6,2012 to include unadjusted exit polls
Updated July 21, 2015 to include Cumulative Vote share
In the 2010 Wisconsin Senate race. Johnson defeated Feingold by 52-47%.
Using the 2008 Wisconsin presidential recorded vote (a conservative assumption) as a basis for returning voters, Feingold won the True Vote by 52.0-46.9%, a 110,000 vote margin. McCain returning voter turnout was assumed to be 66%, compared to just 60% for Obama voters.
In Wisconsin’s 72 counties, Feingold had 51.6% at the 10% mark confirming the True Vote Model). But his share declined to 47.1% at the final. The cumulative vote shares were counter-intuitive.
In the Top 15 counties which comprised 2/3 of the total vote, Feingold had 56.3% at the 10% mark. His share declined sharply to 48.8% at the final. The trend was not plausible. Republican shares should not increase in large Urban (Democratic) counties.
In 57 small rural counties, Feingold had 42.8% which increased slightly to 43.8% at thee final. The trend was plausible. It indicates that the GOP did not steal votes in the small counties where they dominated.
It’s common sense. The GOP knows that it must steal Democratic votes in the heavily populated counties to win elections because that’s where the votes are.
There is a positive relationship (0.31 correlation) between Feingold’s cumulative vote share at the 10% mark and county vote size. The correlation is just 0.15 at the final recorded vote. This is another confirmation that the recorded vote is bogus.
Counties………. Votes …….. 10%……. Final (Feingold)
Top… 15…….. 1,417,248… 56.29%… 48.79%
Other 57………. 743,584… 42.82%… 43.80%
Total…………. 2,160,832… 51.65%…. 47.07%
The sharply increasing Johnson cumulative vote share in Milwaukee and other counties defies the Law of Large Numbers.
Unadjusted exit poll are always forced to match the recorded vote. It is standard operating procedure. In order to force a match in the 2004 and 2008 presidential elections, the exit pollsters had to assume an impossible number of Bush voters returning from the previous election.
The returning voter mix should reflect the True Vote, not the recorded vote. According to the adjusted 2010 exit poll, 49% of recorded votes were cast by returning Obama 2008 voters and 43% by McCain voters. The ratio is consistent with Obama’s 7.5% national recorded vote margin. But Obama did much better than his recorded vote indicates.
Nationally, Obama’s true 18% margin was based on the unadjusted state exit polls (58-40%). But he had a 24% margin (61-37%) in the unadjusted National Exit Poll (17,836 respondents).
In Wisconsin, Obama had a 56.2% recorded share but led the unadjusted Wisconsin exit poll by 63-36% (2,545 respondents; 2.4% margin of error).
There is a 97.5% probability that Obama’s true Wisconsin vote share exceeded 61%. Assuming Obama had 61%, how could Feingold have had just 47% two years later?
In the 2010 WI exit poll, vote shares were not provided for returning third party and new (DNV) voters, They represented 3% and 5% of the recorded vote, respectively. In order to match the recorded vote, Johnson needed to win these voters by an implausible 60-35%. In 2008, Obama won new voters by 71-27%; returning third party voters by 66-20%.
Note that in Oregon, Obama’s 57% recorded share was matched by Ron Wyden. Like Feingold, Wyden was a progressive Democratic senator running for re-election.
Johnson needed an implausible 19% of Obama voters to match the recorded vote.
WI Exit Poll Party-ID (37D-36R-27I) understated the actual Democratic registration split (43D-41R-16I). Johnson needed to win Independents by an implausible 56-43% to match the recorded vote.
A comparison of demographic changes from 2004 to 2010 yields interesting results. The 2010 numbers are suspect since they are based on the the 2010 recorded vote.
Note that the exit poll was forced, as usual,to match the recorded vote: Johnson defeated Feingold by 52- 47%. Feingold had an implausibly low 44% of white voters and 81% of blacks. Vote shares for Latino, Asian and Others are missing. This is a “tell”.
In the second table, the vote shares were adjusted to plausible rates. Feingold was a very popular Democrat who must have at least tied Johnson among whites. In addition, Democrats usually get 90-95% of blacks and 70% of other minorities. These minority vote shares were included in the second table as they should have been in the first place.
With these changes, Feingold is a 52-47% winner, exactly reversing the recorded vote.
|Final Exit Poll||Mix||Feingold||Johnson||Other|
Johnson needed 70% of voters who decided in the final week to win.
Implausible Feingold declines from 2004 to 2010:
Females: 53% > 50%
Party ID: 38R/35D > 37D/36R
Independents: 62% > 43%
Labor: 66% > 59%
Milwaukee County: 68% > 61%
Suburban/Rural: 51% > 43%
Vote shares are displayed for various scenarios of a) returning Obama and McCain voter turnout and b) Feingold’s share of returning and new voters.
The True Vote Base Case analysis assumes a 1.0% annual voter mortality rate. The percentage mix of returning 2008 third-party (other) voters could not have been the 3% indicated in the WI exit poll. That would mean there were 65,000 third-party voters but there were just 44,000. Therefore, the model assigned the 1.5% excess of Other voters to New/DNV (first-time voters and others who did not vote in 2008).
Feingold was the winner in all scenarios of returning Obama and McCain voters. But it is important to keep in mind that the WI exit poll gave Feingold just 84% of returning Obama voters. It is difficult to accept the premise that nearly one of six Obama voters defected to Johnson.