How to detect a faulty election;

Recently Texas' Republican Primary featured three potential Republican candidates: Rick Perry, Kay Hutchison and Debrah Medina.

The polls predicted that Medina would play spoiler, with a runoff election later between Rick Perry and Kay Hutchison. To my surprise, Rick Perry won with 51%. I don't know if this was within the margin of error of the Polls. Conspiracy theorists can cry foul: either the winner or an outside influence "fixed" the election.

Iraq had a recent election too. It is more subject to accusations of tampering than Texas. To ratify the election a statistical study with a statistically significant sample size that is truly random, should yield results that agree with the vote - again with a small margin of error. If differences are statistically significant, then the election should be ruled invalid and re-done. This is only theoretical, because of secret ballot provisions. If only volunteers are polled, the sample cannot be random. However, since 30 -120 is a statistically large sample for one variable, statistics ARE worthwhile.

I don't seriously think Texas' election was fixed, but we could profit from more Scantron ballots, and a paper trail for early voting. Iraq could profit from hand held counter clickers for the ballot counters.

If we go to electronic voting, we should do so in context of a discussion of Direct Democracy, and bear in mind that shipping a 12 B_illion count ballot box around the internet with utter electronic security fails to accomplish the goal.

Post Script: Another application of the statistical study of votes is that of the extant batch of ballots. A random study of ballots should agree with the total count. If it does not, the count is faulty.