Political Exit Polling

The Pew Research Center for the People & the Press released a polling summary concerning the 1996 presidential race and a U.S. Information Agency survey conducted in Bosnia-Herzegovina aimed at determining any long-term difficulties in nation-building in that region. The article can be found at: http://people-press.org/commentary/print.php3?AnalysisID=27

The article discusses that the exit polls provided far better predictions of the actual primary election results than did the “pundits” who were making guesses at the American public’s opinion of the various candidates. The polling information was based upon exit polling and in some instances telephone polling results from pre-election day surveys. A chart of the various polls found at the end of the paper show that individual polls could be off by as much as 60% from the actual election results, which can cause the reader to question the validity of the pollster’s methods of obtaining their data. Still, the polls tended to pick the first place winner 4 out of 5 times, while the “pundits” were correct only 15 out of 29 times.

The model used in exit polling is straight forward and is a simple statistical analysis. I believe there could be a number of reasons for the large margins of error encountered in these polls. First, could be that the samples were not randomly selected. Second, the individuals polled could have provided false information to the pollster. A third possibility for error is the sample size was not large enough to provide accurate information. In the case of pre-election polling predictions is the possibility that some of the individuals polled changed their minds after the poll was conducted and voted contrary to the way they informed the pollster that they were going to vote.

It has been my experience that most polls as they are presented on one of the major/national television networks are normally associated with a margin of error of plus or minus 3 to 5 points. The major network polls are also far more accurate in their predictions than those associated with any of the regional newspaper services. This is probably due to the size of the sample group, although it has been suggested in the past that some papers skew there results to favor a particular candidate that the paper is backing (biased results).