Wisconsin Recalls: Exit Polls and the True Vote Model

26 Aug

Sept 7, 2011

In each of the five Citizen Exit Polls conducted in two Wisconsin districts, the Democrats did much better than the official count (67.8% vs. 52.4% on average). Why the large discrepancies? Are the polls to be believed? This analysis provides a possible explanation, keeping in mind that it is based on a limited number of exit poll locations.

The Wisconsin Recall True Vote Model

It is important to understand the difference between the Wisconsin Citizen exit polls and corporate sponsored state and national exit polls in prior elections. The Citizen polls had a very simple aim: to compare how respondents said they voted to the official count.

State and National Exit Polls: Forced to Match the Recorded Vote
Corporate state and national exit polls are designed to determine how various demographic groups voted. The National Exit Pool (NEP) is the consortium of six media giants that funds state and national exit polls.
The NEP uses stratified sampling to select precinct locations that are representative of the voting population.

It is standard operating procedure to force Final state and national exit polls to match the recorded vote. It is also standard policy for the NEP to keep “unadjusted” precinct exit poll data hidden from the public, claiming the need to protect voter “privacy”. It’s a canard; exit poll responders do not reveal their identity. There is no excuse for suppressing the release of unadjusted precinct exit poll data.

Recall Exit Poll Discrepancies
In three strong Democratic recall locations, Democratic exit poll shares (78.8%) were significantly higher than the vote counts (66.9%). Shorewood was 10.9% higher, Pardeeville 8.5% and Baraboo 15.8%. The True Vote Model (TVM) closely matched the recorded votes. Overall voter participation was 52%. If the vote counts were correct, 61% of Democrats and 33% of Republicans responded. Republican voters may have been reluctant to respond in these heavily Democratic locations. If that was the case, then the exit polls overstated the True Democratic vote.

In the two strong Republican locations (Butler and Menomonee Falls), the Democrats had 43.1% in the exit polls compared to 31.5% in the count(16.5% higher in Butler, 11.0% in Menomonee). In the TVM, the Democrats did 11.0% and 8.3% better, respectively. Overall just 33% of voters participated. If the vote counts were correct, then 45% of Democrats and 28% of Republicans responded – an implausible Democratic participation in these strong GOP locations. Therefore, it is likely that the votes were miscounted and the exit polls were close to the True Vote.

The aggregate Democratic True Vote share was 55.2%, a very close match to the 55.5% aggregate share calculated based on equal Democratic and Republican response in each location.

The discrepancies could also have been due to a combination of vote miscounts and differential response. Assuming equal Democratic and Republican response rates, Democratic shares were 73.0% (6.1% higher than recorded) in Democratic locations and 35.4% in Republican locations (3.9% higher).

Republican Exit Poll Response Required to Match Recorded Vote
The total Democratic exit poll response rate was 57%; the Republican rate was 30%. Democratic response exceeded Republican response in each of the five locations. What if the overall Republican response was also 57%? What would the response have to be in each location for the exit poll to match the recorded vote (52.4% Dem, 47.6% Rep)? Using the Excel Solver algorithm, the required Republican response was derived (Democratic response was held constant to the actual exit poll). Republican response was constrained to exceed Democratic response in Menomonee Falls and Butler. The required response rates were not plausible (see the table below). Republican response exceeded Democratic response in 4 of the 5 locations (including Democratic strongholds Pardeeville and Shorewood). In Pardeeville, 100% of Republican voters responded.

Exit Poll Refusal Rates
We define Exit Poll refusals as the difference between the total number of non-responders and the number of non-responders required to match the recorded vote. The refusal rate is the ratio of refusals to the recorded vote. The average Republican refusal rate in Democratic locations was 18.8%: Baraboo 16.6%, Pardeeville 12.9%, Shorewood 22.5%. In Republican locations, the rate was 5.6%: Menomonee Falls 5.9%, Butler 5.2%. Republican voters were four times more likely to refuse an exit poll interview in Democratic than in Republican locations.

Based on the True Vote Model and the exit polls, it is very likely that the District 8 and 14 recall elections were stolen. The Republicans control the state senate by a 17-16 majority, but the Democrats should be in control by 18-15.

True Vote Model
The TVM requires an estimate of the True Vote in the previous election in order to determine a plausible mix of returning voters. In 2008, Obama had a 56.2% recorded vote share in Wisconsin, but he had a whopping 63.3% in the unadjusted exit poll. The margin of error was 2.4% (including a 30% cluster effect) for the 2545 respondents. It is obvious that using a miscounted recorded vote from the previous election as a basis for forecasting or post-election analysis will produce a fraudulent result in the current election.

These were the base case assumptions used in the TVM:
1) Equal percentage turnout of Obama/McCain voters.
2) Zero net defection: Democrats win 95% of returning Obama voters, Republicans win 95% of McCain voters.
3) New voters broke for the Democrats and Republicans in the same proportion as the 2008 recorded vote.

For each exit poll location, a sensitivity analysis table displays Democratic vote shares based on nine scenario combinations of returning Obama and McCain voters. The base case scenario is the central cell of the 3×3 table.

2004 Exit Polls
In 2004, there were 1480 exit poll precinct polling locations nationwide. More than 76,000 voters participated. To explain the average 6.5% exit poll discrepancy, the exit pollsters hypothesized that 56 Kerry voters participated for every 50 Bush voters. They provided no evidence for this. They did not consider that the discrepancies may have been due to Election Fraud. The so-called reluctant Bush responder (rBr) theory was refuted by the exit pollster’s own data which showed that exit poll response was higher in Bush strongholds than Kerry strongholds.

Link to the source data tables:


Tags: , , ,

3 responses to “Wisconsin Recalls: Exit Polls and the True Vote Model

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: