Exit polls were controversial too in the Hugo Chavez recall referendum in Venezuela:
http://www.usnews.com/usnews/opinion/baroneweb/mb_040820.htmSome excerpts from the article:
...final exit poll results from Penn, Schoen & Berland Associates, an independent New York-based polling firm, show a major victory for the 'Yes' movement, defeating Chavez in the Venezuela presidential recall referendum." The poll showed 59 percent in favor of recalling Chavez, 41 percent against.
The next morning, Chavez was declared the winner by an almost exact opposite margin. "About 58 percent said 'no' to a recall, while 42 percent said 'yes,'" wrote the Washington Post.
...
Venezuela's President Hugo Chavez has been running an authoritarian regime. By various means he has taken control of the legislature, the courts, the armed services and the police. His thugs have been intimidating and even killing the regime's opponents. The literature on this is voluminous, but consider these reports from the Wall Street Journal: www.opinionjournal.com/extra/?id=110005494 and www.opinionjournal.com/wsj/?id=110005478. Chavez is an ally of Cuba's Fidel Castro and an enemy of the United States, and he has shown no commitment to democratic principles. He sought to block the referendum by extralegal means and, having failed at that, resorted to intimidation to win it. There is no reason to believe that he would stop at election fraud.
One weapon against such fraud is the exit poll. As Doug Schoen of Penn Schoen points out, his firm has conducted exit polls in Mexico and, just a few days ago, in the Dominican Republic, which produced results very close to the election results. His partner Mark Penn points out that the firm conducted two previous exit polls in Venezuela, both of which were on the mark. Warren Mitofsky's firm, Mitofsky International, has produced exit polls with similar results in Mexico and Russia. Mitofsky recalls that in 1994, Mexican President Carlos Salinas, seeking credibility with foreign investors for that year's Mexican elections, asked him for advice on what to do. Allow independent exit polls, Mitofsky advised, sponsored by the media, and allow the results to be announced soon after the voting. Mitofsky's exit poll results, announced soon after the polls closed, did in fact come close to the official results, as did another Mitofsky poll in 2000. More important, they provided independent confirmation of the fairness of the count.
...
In contrast, it would be far easier, given the touch-screen voting method and central tabulation used in Venezuela, for the central counting center to falsify the results. All you would have to do is program the computer to count every sixth "yes" vote as a "no." That would transform a 59-41 vote to 42-58. And the results would still show pro-Chavez areas voting for him and anti-Chavez areas going the other way—just by different margins.