Actually, this is not so much a counterargument as a corollary. I will illustrate how an entirely different set of assumptions could explain the discrepancies between exit polls and vote tallies, while not changing a single thing about Febble's mathematical analysis. Words can often obfuscate what is really happening, and prevent rigorous understanding. That has happened in the case of Febble's analysis.
Febble's Fancy Function (and the surrounding analysis) is fundamentally not about whether a Bush voter was more reluctant to respond. It is fundamentally a model of the discrepancies between official vote tallies and exit polls (as functions of precinct or "partisanship" of precinct or whathaveyou). It would have been more illuminating had the analysis proceeded in reverse: begin with the math, and end with speculation about the nature (with words attached to phenomena).
Consider this assumption: Votes were fraudulently turned for Bush in a way analogous to the 56% and 50% Kerry and Bush voter exit poll response rates of Febble's work. See what I've done there? I've assumed the discrepancy was about a phenomenon regarding the vote tallies rather than one regarding the exit polls. It is easy to see that there is no mathematical difference.
Let's get more specific. Let's say that the Bushies had decided to turn 3 out of every 100 votes. A precinct that had 50 actual votes for each candidate, the tally (post-fraud) would be 53 for Bush and 47 for Kerry. Now, in Febble's analysis, you have a disproportionate number of Kerry voters responding to exit polls. In my argument here, you have a disproportionate number of Bush votes registering in the vote-counting machines (or in the Secretary of State offices or wherever). These two models of the discrepancies are completely identical. Exactly!
So what has Febble actually shown. Forget for a second about Kerry and Bush and an election and an exit poll and Republicans and Democrats. Think only of two different counts of an event, by different methods, each of which has some unknown error. What Febble has shown, and no one seems to understand, is this:
A systematic error can explain the discrepancies between the two counts of this event.
I wish to put particular stress on the "can". In USCountVotes language, the systematic error is "plausible". Now put Kerry, Bush, an election, an exit poll, Republicans, and Democrats back into the puzzle. One hypothesis might be that of Edison-Mitofsky, that Bush voters were less likely as a whole to respond to an exit poll (the reluctant Bush responder hypothesis, or "rBr").
Another mathematically identical hypothesis (which I'll call "futotr" for "fucking undemocratic thugs on the right") is that votes were systematically stolen for Bush.
Febble's work does not lend credibility to the rBr hypothesis, it lends plausibility; it lends an exactly equal plausibility to the futotr hypothesis, and, for that matter, any other hypothesis that involves a systematic error (or bias) in one (or even both) of the counts.
There is no evidence that the "reluctant Bush responder" hypothesis is more likely that the "fucking undemocratic thugs on the right" hypothesis. Febble's mathematical analysis applies identically to each of these hypotheses. Febble has not shown any of the following things:
- There is a problem with Within Precinct Error;
- Exit polling methodology is flawed (in a way not already known);
- There is a disproportionate response rate in exit polls; or
- Fraud is less likely than disproportionate response rates.
She
has shown that each of these might be true, but they have
not been demonstrated as fact. I think Febble's work is top-notch and highly valuable, but we should all take care that we do not suggest it is something more or other than it is.
The rest of this is pure speculation. And Occam's Razor. I only wish to consider the rBr and futotr hypotheses. Which is the simpler explanation?
rBr: A Bush voter is less likely than a Kerry voter to respond to exit polls.
futotr: The Republicans stole votes for Bush.
I know of no way to support or refute the rBr hypothesis, but it does not seem altogether unlikely that there is a correlation between responding to polls and voting for Bush. Forget about the swing voter who thinks Bush and Kerry are both no-good scoundrels without a lick of difference betwixt. There must be something very different between the minds of the average dedicated Bush and Kerry voters. Can that something also have an effect on willingness to respond to polls? We have to admit that there is a possibility, but I, for one, am skeptical that it could be a 6% effect.
What we do know is that Republicans had the motivation and the power to change votes. The fucking Presidency of the United States is the motivation (and, of course, Democrats have this motivation as well). The power is the Republican owned-and-operated black boxes we call voting machines. The only question is did they have the initiative and the utter disrespect for democracy to try it. I have little trouble believing that they probably did.
Update [2005-4-29 22:19:10 by R Lucian]: I am unperturbed by any of you who do not answer my poll, because I've nearly completed a simulation that will accurately model what option you would have chosen had you chosen one.
Update [2005-4-30 7:56:12 by R Lucian]: I just wanted to draw everyone's attention to Febble's comment here, as well as my response and her comment directly above that, also at the root level. Also, my apologies for causing so many heads to explode. I did not anticipate that effect occurring in greater than 50% of all who participated. Apparently friends don't let friends read this diary.