Exit Polls Made Easy
Something's fishy, and it ain't just Ohio

by Gary Beckwith
January 31, 2005

It's been said that exit polls are unreliable and that they can be interpreted many ways. But a look at the numbers provides some solid information anyone can understand, that is beyond interpretation.

These three facts can easily be derived from the exit poll data:

  • In 43 out of 51 states (inculding DC), the "official" vote totals differ from the exit polls in the direction favoring Bush.

  • Bush's victory exceeds the mathematical margin of error in 15 states. Kerry's victory does not exceed the margin of error in any state.

  • Exit polls were accurate in 2000.

In short, the gap between the exit polls and official vote (nearly all in the direction favoring Bush) is extremely large. Just how large?

The question that statisticians would ask is, "What is the probability that this could happen by random chance?"

Inquiring minds would like to know, and it's not just statisticians. It's the patriotic folks who care about the state of democracy in the United States of America.

Chances of exit polls being off in Bush's favor in 43 out of 51 states

Exit polls are not 100% accurate, but their errors should happen randomly. They should be off favoring Bush and Kerry fairly evenly. But they were off in Bush's favor in 43 out of 51 states.

According to statistician "Tia," the chances of this happening by random chance are 1 in 2.9 million*.

Do I smell smoke?

Chances of Bush's victory exceeding the margin of error in 15 states

The huge gap between the exit polls and official results is clearly demonstrated by looking at how many states were outside the Margin of Error.

The margin of error (MOE) is mathematically calculated with standard formulas. It is related to the size of the sample - the larger the sample, the smaller the MOE. Under normal circumstances, the actual vote count should fall within the MOE 95% of the time.

The difference between Bush's final vote percentage and his exit poll percentage exceeded the exit poll margin of error (MOE) in 15 states. There was not a single state which deviated beyond the MOE for Kerry.

This table shows all 50 states, how far off the exit polls were from the actual count, which direction they were off, and whether or not they deviated beyond the MOE.

StateMOEDeviation*>MOE?Favors Bush?
AK3.31%-3.73%YESYES
AL3.70%-3.90%YESYES
AR2.67%-1.53%YES
AZ2.32%-2.28%YES
CA2.28%1.04%
CO1.99%-1.47%YES
CT3.39%-3.23%YES
DC3.55%-0.48%YES
DE3.60%-4.67%YESYES
FL1.87%-3.02%YESYES
GA2.55%-1.35%YES
HI4.48%1.10%
IA2.00%-0.94%YES
ID4.23%-2.82%YES
IL2.68%-1.79%YES
IN3.29%-1.43%YES
KS3.91%2.13%
KY3.11%-1.01%YES
LA2.45%-1.83%YES
MA3.35%-3.26%YES
MD3.16%-0.43%YES
ME2.25%-0.12%YES
MI2.02%-0.77%YES
MN2.14%-2.74%YESYES
MO2.15%-1.12%YES
MS3.54%-2.81%YES
MT3.95%-0.30%YES
NC2.15%-4.24%YESYES
ND3.93%2.09%
NE3.57%-3.65%YESYES
NH2.33%-4.71%YESYES
NJ2.56%-1.63%YES
NM2.26%-1.70%YES
NV2.17%-0.72%YES
NY2.62%-3.71%YESYES
OH2.26%-3.16%YESYES
OK2.55%-0.57%YES
OR3.07%0.91%
PA2.28%-3.04%YESYES
RI3.52%-3.43%YES
SC2.40%-4.64%YESYES
SD2.59%1.29%
TN2.37%1.31%
TX2.45%1.49%
UT3.54%-3.85%YESYES
VA2.64%-2.13%YES
VT3.82%-4.70%YESYES
WA2.17%-1.25%YES
WI2.12%-2.31%YESYES
WV2.41%-1.78%YES
WY3.82%-1.21%YES
Avg2.85%-1.73%
Med2.59%-1.70%
Total1543
KEY
Official total favors Bush
Official total favors Bush, outside MOE
Official total favors Kerry
Official total favors Kerry, outside MOE (none)

If Bush's official count differed significantly from the exit polls in just a state or two, we wouldn't be raising our eyebrows. But the table clearly shows a pattern that defies scientific explanation.

What are the chances of this happening?

The odds of Bush exceeding the MOE in any given state is 2.5% (1 out of 40). The odds of Bush exceeding the MOE in any two given states is 1/40 X 1/40 = .0625% (1 out of 1600). Now that you can see how quickly the numbers start to skyrocket, consider what the chances of this happening in 15 states might be.

Microsoft Excel has a built in formula for calculating the chances of events outside the margin of error. With our 15-state scenario, the chances of this are 1 in 1.091 trillion.*

First I smelled smoke. Now something smells fishy. Is someone barbequing some fish?

Exit polls were correct in 2000

TV news programs, when they even address the issue of the exit polls, keep repeating that the exit polls have been wrong in the last few elections.

But this table reveals that in the 2000 election, the final counts in the vast majority of states (39 out of 50) were within one percentage point of their exit poll. In contrast, in 2004, only 9 states were within one percentage point of their exit polls.

Number of states
that deviated by
20002004
1% or less399
2% or less320
3% or less28
over 3%616

Now I think they burned the fish.

Possible Explanations

That Mr. Bush sure is a lucky guy. He effectively won the lottery and got struck by lightning... on the same day!

It doesn't take a mathematician to look at these tables and tell that something beyond random chance is going on here.

There are the exit polls, and the official results. There is a huge impossible gap between the two. As many statisticians have stated, the unexplained gap needs to be investigated.

Why don't we start a little investigation right here and now.

There are not many possible explanations for the huge gap. One is that Mr. Bush did win the lottery and get struck by lightning on the same day — a chance in a trillion. Another is that the exit polls were just wrong. In one such attempt to explain the exit poll problem away, last week the overseer of the exit polls published a report saying that for an unknown reason, more Kerry voters answered the exit polls, and this skewed the results. Then a consortium of PhD experts issued a report citing huge flaws in the report, stating there is no evidence to support this theory whatsoever**.

The only other explanation is that something happened to the votes between the time they were cast by the voters and the time the official results were announced.

Let's consider what happens to our votes after we walk away from the voting booth and until we hear the official results announced on TV. We all have images of armored cars carrying ballots to some kind of central counting station, where they're kept under lock and key by non-partisan people who preserve our democracy.

Now for a reality check. The days of non-partisan election officials is over. We've seen it in Florida with Katherine Harris, we've seen it in Ohio with Ken Blackwell, and it's happening in Georgia and other states. Highly partisan Republicans who are responsible for running fair elections are also running the campaign of one of the candidates at the same time. Probably not exactly what the founding fathers had in mind.

Now let's look at how those votes are counted, while under the supervision of "non-partisan" election overseers.

Approximately 30% of the votes cast in the last election vanished into thin air the moment each ballot was cast; not by a magic trick, but by design. It was decided that it's best to put the votes into a computer's memory card, with no printed receipt or hard copy track record of the vote cast — and no way for it to be verified or recounted. This was decided by the manufacturers of the machines (we'll get to them shortly) and the Republican controlled Congress who blocked legislation requiring a paper receipt for every ballot cast.

For the vast majority of the rest of the votes, the paper ballots were fed into a machine or computer that counted the ballots and reported the totals.

These machines and computers represent the bridge between the voters and the official results. Any questions about discrepancies between exit polls and final numbers would obviously have to focus on the machines and companies that manufacture them and write the source code.

"The people who vote decide nothing. The people who count the votes decide everything."

It is not widely known, nor is it reported in the media, that the vast majority of these machines and computers were manufactured by a very small number of companies, all of which are owned and run by supporters of the Republican party.

These companies have refused to let anyone see the software code that counts our votes. They say it's "proprietary information."

The two private companies Diebold and ES&S alone were responsible for counting approximately 80% of the votes cast in the 2004 election. Diebold's CEO Walden O'dell was a major contributor to the Bush campaign. He organized a $1,000 per plate fundraising dinner for Bush and he's been to the Crawford ranch. In a 2003 fundraising letter O'dell wrote that he was "committed to helping Ohio deliver its electoral votes to the president next year." Not exactly the person you'd want to entrust to counting the votes. Then there's ES&S, which used to be owned by Republican Senator Chuck Hagel (he resigned right before his election, just in time for his own machines to be used to count the votes that put him in office). Could there be a more obvious conflict of interest?

One of three things happened: Either an event occurred that is so mathematically impossible that the chances are in the millions and trillions to one. Or an amazing computer glitch occurred on thousands of computers nationwide, even though they were not connected in any way. Or, there was purposeful manipulation of the results through the use of Republican backed electronic vote counting machines.

Would they really do that, in our beloved democracy? Given their track record, the chances are probably somewhat greater than, say, a trillion to one.

NOTES

MOE table:
The numbers have been converted to two-party percentages so as to total 100%. The purpose was to maintain comparability to other pre-election projection models. The adjustments are proportional to the exit poll and vote percentages. The post-adjusted numbers are in proportion to the original numbers. There is no bias for Kerry or Bush.

For example:
Assume Bush won a state vote 50%-49% with 1% going to third parties.
Bush's adjusted two-party percentage is 50/99 = 50.505%
Kerry's two-party percentage is 49/99 = 49.495%

These are the relevant statistics from the table:

MOE = Margin of error = 1/sqrt(N)
N=sample size
*Prob is calculated for each state using the Excel Function:
Prob = 1-Normdist(Poll, Vote, StdDev, true)

More complete table of this data is available here.

Exit poll raw data downloaded by Jonathan Simon at 12:22 AM Nov 3. http://www.scoop.co.nz/stories/pdfs/Mitofsky4zonedata/

Credits

Data compilation and calculation: "TruthIsAll"

**More information on election fraud: The Solar Bus Election Fraud and Reform Center