[EM] advocacy by means of exit polls

Abd ul-Rahman Lomax abd at lomaxdesign.com
Wed Aug 30 20:19:17 PDT 2006


At 02:01 PM 8/29/2006, Michael Poole wrote:
>To paraphrase: Since you are convinced that there was official bias in
>favor of Republicans during that election, I should stop trying to
>convince you that well-structured exit polls are comparable to
>well-structure telephone polls.

This is frustrating, I'm sure. What was asserted was not that there 
was official bias, but that the possibility of that bias existed and 
could not be discounted. If this is true, then the exit poll results 
can't be faulted on the basis of deviation from official results, as 
Mr. Poole did do.

As to the comparison between exit polls and telephone polls, I'm not 
sure that sufficient research has been done to come to a conclusion. 
Telephone polls, I'd think, would be less expensive to run, and they 
could then, wtih extra care, compensate for the deficiency in 
correctly measuring the desired population, i.e., actual voters. Exit 
polls do have that advantage, but they suffer from other disadvantages....

But the issue, really, was independent analysis of election results 
using different election methods. For this purpose, it is not crucial 
what polling method is used; I'd expect that telephone polls, 
however, would be more satisfactory, because they could be more complex.

>You have almost convinced me to give up but first please consider
>this: The US is not the only place that exit polls have shown serious
>errors.

Once again, while it is to be expected that there is the possibility 
of error in any poll; after all, pollsters do make mistakes, people 
lie, etc., deviation from official results is not a proof of polling 
error. Further, analysis of the raw poll data could uncover a pattern 
of official fraud. But that was not really the point here. The point 
was to understand how election methods could shift election results, 
and why and how. We actually don't know a great deal about this. 
Analysis of so-called "election failure" has mostly attempted to 
presume or infer how people would have voted had they had, for 
example, the right to overvote. We don't really know. Actually asking 
people is quite a good idea.

Of course, Warren Smith has actually done this. But his poll was 
quite small, pretty much good only, in my opinion, for suggesting 
possible directions for further research. But far better than 
nothing, which is what most of us have as far as actual experimental 
data is concerned.

>   Emotion-driven appeals to authority are not convincing and do
>not address the substantive accuracy of exit polls.  My original mail
>was not trying to make a general point about official poll accuracy,
>but about exit polls.

And the whole line of argument was off the point. The original post 
and subject line could have read "polls" instead of "exit polls," 
though exit polls implies a closeness to the actual voting; but that 
can be done with phone polls as well.

("Did you vote today?")





More information about the Election-Methods mailing list