Sunday, September 23, 2012

The Need for Exactness

Whether truthfully or not, stating the need for exactness is an effective tool for planting doubt in the public's mind about your opponents claims.  Some birthers and JFK conspiracy theorists will never be satisfied with the official explanation of their respective claims.

At the bi-Monthly Goo Goo Gathering that the Pittsburgh Coffee Party held, there was a lot of discussion of the PA Supreme Court's decision to have the lower courts reconsider whether the State's voter ID law is feasible.  The Republicans, who passed the law, publicly stated that it is intended to prevent voter fraud but insiders like PA House Majority leader privately stated something else:

The Colbert ReportMon - Thurs 11:30pm / 10:30c
Blood in the Water - Mike Turzai's Voter ID Remarks
Colbert Report Full EpisodesPolitical Humor & Satire BlogVideo Archive

Opponents of the law demanded to see one case of voter fraud and stated that many of the state's poor and minorities would be disenfranchised because of inability to obtain an ID. 

Another example of the need for exactness is Mitt Romney's claim that the 47% of the US who were dependent on the US Government for assistance.  Many of his opponents, once they found out about his comments, were quick to point out who the 47% were: mostly the elderly and the working poor. 

I was also asked another question at the Goo Goo gathering about electronic voter machines that have no paper trail.   How can they be checked for accuracy.  He said he had a sample size of 20 machines to test for malicious software out of about 5,000 in Allegheny County.  He wanted to know if that sample size would be able to detect anything.  I computed the margin of error for the sample for the percent of machines found to be defective.  With a sample of 20 that would mean a margin of error of +/- 22.4% with 95% confidence.  This means that if 50% of machines were found to be defective in the sample (10 out of 20), the actual population proportion would be between 72.4% and 27.6%.  This is a wide margin but we could be confident that the population proportion was different from zero.  He said that these tests have been done before but they have always come out to be 100% not defective.

Another method of checking the machines I told him would be to compare the machine vote totals to the exit poll data or even hold a mock election.  The 2006 CNN exit poll accurately predicted that Bob Casey would defeat Rick Santorum with 59% of the vote (margin of error +/- 2%).  Granted one or two defective counting machines would not have an overwhelming effect on the statewide totals with millions voting.  If precinct level exit poll percentages could be compared to corresponding vote totals, that could provide a better indication on the reliability of the machines.  If that is not feasible then holding mock elections with no secret ballot could be the next best thing with the 20 machines.

I have a similar problem looking at my blog traffic statistics which I will discuss in depth on my second year anniversary post.  My post 100th post is listed as the top post for the month on the built in stat counter for blogger but hardly registers for Google Analytics and Stat Counter and I have a similar post on the PUSH blog with inflated statistics.  That is why independent verification is important.

  **Related Posts**

Making Sense of the Pat Toomey-Joe Sestak Senate Race

Progressive Soul Searching in Wisconsin, PA, & other Red Gov. States

Deep South Primaries