When did election polling and actual reported counts begin to diverge?

[randomelement]

Sorry if this is a dumb question, just wondering.

During all those years of watching election returns on TV, I know the results were based on polling and, more often than not, those results were proven accurate. Exactly when did polling and announced results begin to diverge to the point where polling isn't considered "accurate" enough anymore? I have to think this is a recent occurrence ...

Thanks in advance

[marmar]

1. Since Diebold machines were created...

to steal elections. The Max Cleland election in Georgia (I forget what year, '98 maybe) was the first I'd heard of where the polling numbers and the election results were vastly different, and Diebold machines were used there.

[keopeli]

2. It was on November 2000

The sElection that saw the overthrow of the US government. The media's polling service VNS, was right on the mark with regard to the election, but Bush* manipulated the vote and declared that "old" polling services had become unreliable. In fact, the opposite was true. The polling service was correct and the election process itself had become unreliable.

In a court ordered coup d'etat, the Supreme Court declared Bush* the winner of the sElection without any factual evidence (such as vote tallies) to back them up.

We've been living in a fascist oligarchy ever since.


Welcome to DU! Chin up - we may just get our democracy back some day!

[madeline_con]

3. "... we may just get our democracy back some day!"

It will have to be wrested from the hands of fiends!!!

FIGHT!!! :woohoo:

[Poet Lariat]

4. Exit polling has been used around the world for many years

as the check and balance "most accurate" way of determining fair elections.

On "selection night" 2004, my wife and I watched with total disbelief as some of the news stations changed or dropped their exit poll numbers (that showed Kerry ahead) to reflect the vote totals coming in from precincts and states around the country.

To answer your question...Exit polling IS accurate. It's the election vote totals that are not.

[philb]

5. Exit polls showed Gore won Florida; and Media analysis later confirmed

that in a fair election with proper vote counts Gore would have won Florida by over 40,000. It wasn't even close.

Likewise for some other important U.S. Senate and Govenor races in other states like Georgia, Missouri, etc.

The 2004 elections were the most well documented farce in history of U.S. elections, as the EIRS system had a major effort to document the problems ; documenting major touch screen switching in about half the states; major systematic dirty tricks to reduce minority vote in most states; manipulation of machine availability and broken machines in minority precincts, other manipulation of registrations, absentees, provisional balloting,etc. in most states
as well as other fraud documented in many counties of states like Ohio and Florida and New Mexico(which all appear to have been swung)

http://www.flcv.com/ohiosum.html
http://www.flcv.com/ussumall.html
http://www.flcv.com/summary.html

But the non-presidential elections of 2002 also had similar problems

[morningglory]

6. When they decided to steal the 2000 election for Sh%thead. n/t

[applegrove]

7. I think it started in places in Texas - you know.. wherever Rove goes

the exit polls are off.

[OnTheOtherHand]

8. exit poll results have often diverged from official returns

but we don't know exactly how often because there is no routine practice of releasing projections based just on the exit polls. In any case, don't assume that the exit polls "beg(a)n to diverge"; as far as I know, there is no evidence that they have become more or less accurate. (TruthIsAll argued a few times that presidential exit polls had gotten more accurate from 1988 through 2000, which is plausible but based on unreliable data.)

Dunno how long you have been watching election returns, but exit polls were barely used in projections before 1980 -- which was the year NBC called the presidential election for Reagan around 5 in the afternoon Eastern time. But that projection didn't require great accuracy!

We know that one of the 1988 exit polls showed a very tight race between Dukakis and Bush, while others didn't; we know that the 1992 projections were pretty inaccurate. What happened in 2000 actually was more a problem with the incorporation of "quick counts" than with the exit polls; the Florida result from exit polls alone would have been, correctly, "too close to call." (Exit polls aren't designed to predict elections that are decided by a fraction of a percentage point.)

[Febble]

9. One way we can estimate

the degree to which they have diverged before is to look at the "within precinct error" rates for each state, which were given in the E-M evaluation issued in January.

http://www.exit-poll.net/election-night/EvaluationJan192005.pdf

This is probably the best measure, as it does not include adjustment to vote returns, nor any error due to unrepresentative selection of precincts, nor weighting errors. "Error" at precinct level (what Freeman calls Precinct Level Discrepancy) can only by caused by bias in the poll or fraud.

I plotted the mean PLD here for the five elections given, using a measure that I believe gives a more accurate estimate than the WPE itself, and in which a positive value indicates "redshift" (it represents the ratio between the rate at which votes for the Democrat were represented in the poll, and the rate at which votes for the Republican were represented in the poll).



What is clear is that the precinct level discrepancy in all five years was significantly in the "redshift" direction - i.e. the vote was "redder" than the poll. This was particularly marked in 1992, but as Clinton "won" both the poll and the count, it probably didn't attract much notice.

(On edit: vertical axis represents mean "redshift" across states; error bars represent 95% confidence limits of the mean).