Reply #4: I'm going to post the following as a separate thread because...

[understandinglife]

...I do not want anyone to miss it. I am posting it here for your information and the information of those who read your post.

A very, very, very "orange" and happy New Year to you.

________________________________
To Those Awaiting A Preliminary Analysis of Exit Polling Data Released 12/31/04:

It's a lot of data to wade through but what has been released is more or less in line with what we had. What is curious and potentially very misleading is that the data from the 12:20 - 12:25 a.m. time of updates (which I was able to capture on election night, but which otherwise apparently would not exist anymore) is missing from this otherwise complete set. This is crucial, since it shows that (to take the full national sample), Kerry maintained his 2.6% lead (rounded to 3%) when 13,047 of the eventual "13,660," rather than the 11,027 from the last pre-"adjusted" batch released here (7:33 p.m.), had been counted.

From the data released it would look somewhat plausible that a late Bush swing (between 11,027 counted and 13,660 counted) could have accounted for the shift in the EPs from Kerry51/Bush48 to Bush51/Kerry48; but the missing sweep (the 12:20 a.m. timeframe) shows that this was not possible.

So my question is, where is that group of data; or did whoever released this stuff forget that I printed out the missing link? It appears that a partial set of data was released (missing a crucial piece) possibly in the hope of creating the impression that all was according to Hoyle. It is critical to examine this newly released data in conjunction with the screenshots which I possess and have distributed to certain recipients who (for obvious reasons) will not be named.

Among these screenshots, the national sample at 12:23 a.m. is public and can be referenced as Appendix A of the Simon/Baiman paper at <http://www.freepress.org/departments/display/19/2004/1054>. It reproduced a bit fuzzy so I'll recap: 12:23 a.m.; 13,047 respondents; Male(46%) 52%B/47%K, Female(54%) 45%B/54%K; Total 48.2%B/50.8%K.

It is also very much worth scrutinizing the breakdown by party ID for that 12:23 a.m. sample:

Dem 38% (B9%/K90%), Rep 35% (B92%/K7%), Ind 26% (B44%/K52%). Compare that to the "adjusted" sample from 1:24 p.m. Wednesday: Dem 37% (B11%/K89%), Rep 37% (B93%/K6%), Ind 26% (B48%/K49%). Remember: the # of respondents barely changed, so the changes are due almost entirely to "renormalization" which, if it is without justification, can better be called flat out fudging.

What we see is that the sample shifts from 38%Dem/35% Rep to 37%Dem/37%Rep (because of the huge effect of party ID on candidate preference, this shift in weighting is very much more powerful in altering the overall results than any reweighting by gender) and Independents lurch over to Bush by 7% (from B44%/K52% to B48%/K49%). Without the missing screenshot from 12:23, an argument might be made that the 2500 or so late exit poll respondents (after 7:33 p.m.) account for these shifts—anyone analyzing just the data released today would be excused for drawing such a conclusion. The missing 12:23 a.m. data shows that such a conclusion would be erroneous (it was data manipulation to match the "actual" vote counts, and not an increase in the size of the respondent group, which produced the pro-Bush shifts).

Edison/Mitofsky must present a legitimate reason for skewing their own polls to overrepresent Reublican and underrepresent Democratic voters (remember the controversy over some right-wing pre-election polls which did essentially the same thing?), while throwing a substantial share of the Independent vote from Kerry over to Bush. The only arguable reason for doing so is what has been dubbed the "reluctant Bush responder" hypothesis—the assumption being that (against all logic and observational evidence and against the evidence of the polls themselves) Republicans won the turnout battle in virtually every state across the country but that this implausible Republican advantage was masked by their innate comparative reluctance to participate in the exit polls. No EVIDENCE, however, has been advanced to support the reluctant Bush responder hypothesis other than the tautology that the vote count had to be right and therefore the exit polls must have been wrong and this is the only way to explain it.

We still await release of the missing late exit poll data and, of course, of the raw data which would permit independent analysis of the raw numerical facts of the case.

—Jonathan Simon

(source, private email forwarded to me)

Peace.