This is the very first time that I've heard of anyone asking this question on a science forum.
The real answer is pretty disturbing, and is best explained by the discussion here at DailyKos.com: http://www.dailykos.com/story/2012/10/2 ... for-Romney
Back when I was following the story there were two writers that understood the issues well, and they reported on several political sites and fora. They went by Michael Collins and TruthIsAll.
The discussion at DailyKos.com covers the problem rather well, although there were better articles at the time that covered this issue. Ron Paul supporters were pretty much all aware of the problem and tried to raise it as an issue during the primaries, to no avail.
My personal views are that it has only gotten worse since then.
For a short synopsis, they electronically switched votes from the leading candidate to Romney. "They" repeated this scheme at least 12 times, until all the leading candidates dropped out, one after the other.
It has been shown mathematically that the official certified vote (from the states' election boards) is the only data necessary to prove that this is so. The reason that the vote flipping can be shown by just looking at the published vote count is because they used an algorithm that increased the vote for Romney based on the number of votes in the precinct. Remember, this is based on the number of actual votes, not the size of the precinct. This means that neither precinct-to-precinct voter turnout, nor precinct size, nor geographical location can account for the anomaly in the data.
It was reported that an academic group tried to study the odds of the anomaly occurring organically but it outran their software's ability to calculate the odds. They estimated over one google-to-one odds.
After Bush v. Gore, Voter News Service (a scientific organization) was retired and an outfit called Edison/Mitofsky (a pseudo-science political organization) was used instead. Their main contribution to the way we do things now was to convince the public that weight factors should only be added to the exit polling after
the polls close and the vote is in.
Sort of defeats any scientific reasons for doing exit polling in the first place.
Think about it. Their methodology changes the exit polling figures in order to make them match the actual recorded vote. What's the point of doing that? There is no scientific reason. The only purpose that I can think of to do it that way is for the propaganda value.
Now, today, the exit polling is all done by the main stream media's National Election Pool (a propaganda organization):https://en.wikipedia.org/wiki/National_Election_Pool
Thank you for asking. I just had to get that out. It's about time the scientific community started asking questions.