Reply #71: I couldn't agree more. [View All]

The first thing any scientist does with a hypothesis is to test it.

And having tested it, and found it supported, the next thing a scientist does is to try to disprove it. Possibly by turning to different data; possibly by running a different test on the same data.

Where I think misunderstanding has occurred is that people assume that those who investigated causes of error in the exit poll (including me) failed to consider the hypothesis that the discrepancy was due to fraud. It is certainly true that pollsters were likely to regard methodological factors as highly likely - after all, they know, from past analyses and from the nature of polling that a number methodological factors sometimes lead to bias, and indeed they knew in 2004, before a single result had come in, that some of their poll results were out of line with their pre-election estimates. So they suspected bias from the get-go. But that does not mean the alternative was not considered or tested.

And I can state, unequivocally, that I both considered it and tested it. One finding was presented in public by Mitofsky, and appears, with commentary, by my colleague Mark Lindeman, here:

http://inside.bard.edu/~lindeman/slides.html

A critical mind, as you say, is about refusing to take things for granted. In fact, it was because I refused to take the explanation that the discrepancy was due to bias in the poll for granted that I found myself in a position to test both that hypothesis and the hypothesis that it may have been, at least partially, due to fraud. The fact that the data suported the first, and not the second is not due to any failure to take the second seriously. It's due to the data itself.

Elizabeth Liddle