Venezuelan Referendum Highlights Problems With Polling and Elections

Exit Poll Seriously Flawed; Pre-Referendum Polls Predicted Results

For Immediate Release: August 19, 2004

Contact: Debi Kar, 202-387-5080 

A CEPR issue brief shows that the results of an exit poll by Penn, Shoen, Berland and Associates, which is at the center of an opposition challenge to the results of the referendum, must have had serious methodological flaws.

The issue brief also shows that the results of the referendum were easily predictable -- and predicted -- by pre-election polling.

As Venezuelan opposition groups continue to dispute President Hugo Chavez' landslide victory in last Sunday's referendum, they have relied heavily on an exit poll by Penn, Shoen, Berland and Associates. This poll of 20,000 voters, released Sunday evening at 8 p.m., showed President Chavez losing by a margin of 59-41 percent -- almost the exact opposite of the official results, which were certified by international observers from the Carter Center and the Organization of American States.

CEPR's analysis shows that, given the certified result, the odds of finding a random sample of voters as Penn, Shoen, and Berland reported were less than 1 in 10 to the 490th power. This is less than the odds of winning the lottery every week consecutively for an entire year. The methodology of this poll must therefore be called into question.

CEPR's analysis further shows that the result -- that Chavez would win -- was easily predictable from pre-election polling. Many press accounts had reported the election as "too close to call," but the probability of getting a random sample of 2000 respondents that would fall within the margin of error -- i.e. too close to call -- were less than one in 50 million. And in fact the most recent (Aug. 4-8) poll by Evans/McDonough Company (EMC) and Varianzas Opinion of a nationwide sample of 2000 accurately predicted the results of the referendum.