November 12, 2004

Exit Polls Were 99.9% Right

Sheldon Drobny on Votergate 2004:
"As a former C.P.A and auditor, I have used statistical sampling throughout my career with great confidence. With electronic record keeping, it's easy to create a program to falsify the books. But there are ways to uncover that. Auditors have developed statistical ways to cut right through corruption in companies. You don't even need a paper trail. These statistical approaches can be used with almost 100% accuracy to uncover fraud.

With the votergate 2004 it's a numbers game just like it is with corporate accounting, even easier. All you're talking about is one number-- total votes for each candidate.

There's a huge difference between polling what WILL happen and polling something that has already happened. The reliability of polling something that has already happened is highly reliable vs. predictive polls, like Gallup or Zogby, which is very risky. The reliability can be, not plus or minus 4 percent as we see with predictive poplls, but rather a much more reliable plus or minus one half or one tenth of one percent with exit polls, because those are based on asking people who already voted. I would even say that if the exit polling were done in the key precincts of Florida and Ohio, which it was, then these results should be practically "bullet proof."

It is important that people know how accurate random sampling of historical events can be in order for them to understand how unlikely it is that the exit polls were wrong. So if you want to fight the battle correctly, you must get more statisticians and forensic accountants involved as well as the lawyers. These statisticians can show with great credibility the probability of manipulation within the computer programs used for counting the ballots. They do this kind of work all the time to uncover fraud based upon computer manipulation in commercial and corporate activities. And these types of expert analyses are admissible in a court of law. "

No comments:

Pages

Blog Archive