Sender: -2.6 (spamval) -- NONE
In-Reply-To: <45513735.4070603 &AElig umich.edu>
Message-ID: <Pine.LNX.4.64.0611072050430.9540 &AElig boston.eecs.umich.edu>
References: <Pine.LNX.4.64.0611072014550.9540 &AElig boston.eecs.umich.edu>  <45513735.4070603 &AElig umich.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed
Date: Tue, 7 Nov 2006 21:05:18 -0500 (EST)
To: Matt Rudary <mrudary &AElig umich.edu>
cc: improvetheworld &AElig umich.edu
From: Daniel Reeves <dreeves &AElig umich.edu>
Subject: Re: 40% chance that democrats take both house and senate
Status: O

But it's only been in recent elections that the bias has shown up 
(according to the GOP website -- I haven't researched this myself).

In any case, those are hypotheses to test, as you say.  The conspiracy 
theory is plausible enough that it's important we get a scientifically 
rigorous answer to this.  Reliable sources solicited!

And the conspiracy doesn't have to be vast and across the board -- if 
there's an exit poll bias then there should be a measurable exit poll skew 
factor that can be adjusted for to detect instances of fraud.

Actually, just checking if the exit poll skew is universal should reveal a 
lot.  As Matt notes, a vast across-the-board conspiracy is implausible.


--- \/   FROM Matt Rudary AT 06.11.07 20:47 (Today)   \/ ---

> What makes 1 more plausible than a vast, right-wing conspiracy that changes 
> vote counts across the board in every district without detection? Seriously?
>
> OK, here are a couple hypotheses to test:
>
> 1) Young people are more likely to vote Democratic and are also more likely 
> to answer exit polls than older people.
>
> 2) People who work 60+ hours a week are more likely to vote Republican and 
> are less likely to answer exit polls than people who work fewer than 60 hours 
> per week.
>
> 3) People without jobs are more likely to vote Democratic and are more likely 
> to answer exit polls than people with jobs.
>
> There are all sorts of reasons opt-out surveys are less accurate than 
> mandatory surveys. These hypotheses may have explanatory power.
>
> Matt
>
> Daniel Reeves wrote:
>> and 6% chance that republicans keep both.
>> 
>> ... according to the prediction market at tradesports.com.
>> 
>> 
>> Question: What do you all think of allegations that Republicans are 
>> cheating?
>> 
>> Here's some damning evidence from the GOP itself:
>>   http://www.gop.com/News/Read.aspx?ID=6718
>>
>>  From that page:
>>   "For reasons that remain unclear, Democratic voters are more likely than 
>> Republicans to agree to interview requests from pollsters."
>> 
>> (Just to spell it out:  If the exit polls favor democrats and the real 
>> polls favor republicans, there are two explanations:
>>   1. a democratic bias in exit polls.
>>   2. a republican bias in the real polls.
>> What makes 1 more plausible?  Seriously, how do Republicans answer that? 
>> This is scary!)
>> 
>

-- 
http://ai.eecs.umich.edu/people/dreeves  - -  search://"Daniel Reeves"

   A programmer started to cuss
   Because getting to sleep was a fuss
   As he lay there in bed
   Looping 'round in his head
   was: while(!asleep()) sheep++;
      -- David Jayne  (submitted by Dave Morris, 2004)