Joined: 10 May 2008
|Posted: Mon Jun 18, 2012 1:57 pm Post subject: Ultimate Smoking Gun- 300 state pres. exit polls (1988-2008)
|The Ultimate Smoking Gun: 300 state presidential exit polls (1988-2008)
I recently calculated the probability that 134 of 300 state presidential exit polls since 1988 would exceed the margin of error in favor of the Republican – and only 3 for the Democrat. Normally 15 would be expected to exceed the margin of error.
137 of 300 presidential state exit polls exceeded the margin of error (only 15 would normally be expected at the 95% confidence level).
P = 7.42E-80 =Poisson (137, .05*300, false)
P = 1 in 1 million trillion trillion trillion trillion trillion trillion
134 of 300 elections exceeded margin of error in favor of GOP (only 8 would be expected)
P= 5E-115 = Poisson (134, .025*300, false)
P= 1 in 2 million trillion trillion trillion trillion trillion trillion trillion trillion trillion
In the Exit Poll notes following the crosstabs, the pollsters indicate that there were 2547 exit poll respondents and the margin of error (MoE) was +/-4%. Presumably, this includes a 30% cluster factor. But the MoE seems too high, considering the number of respondents.
The theoretical MoE is given by the simple formula: MoE =.98/sqrt(n), where n is the number of respondents. For the recall: MoE = 2.0% = .98/sqrt(2547). It is 2.6% after adding the 30% cluster effect.
The National Exit Poll is always forced to match the recorded vote to within 0.50%. Yet the pollsters claim that the MoE is 4.0%. Why do the pollsters even bother to mention the MoE? It has no meaning since the exit poll is always adjusted to match the recorded vote anyway.
If we had unadjusted exit poll data, the margin of error would be applied to determine the interval where the vote share would fall 95% of the time. This is why unadjusted exit polls are necessary. The standard practice of forcing the exit poll to match the recorded vote implicitly assumes zero fraud, i.e. the recorded vote is identical to the True Vote. It never is.
The conventional wisdom is very conventional – and very misleading:
The NY Times Election site has the FINAL, adjusted exit poll crosstabs.