Astrologers Got It Wrong Last Time, Pollsters Get It Wrong This Time

By Niranjan Rambukwella –August 14, 2015

Polls often measure perception rather than behavior. This is one reason they often fail spectacularly in predicting electoral outcomes.

Election surveys at best imperfectly measure differences in turnout between voter segments. First, because some amateur polls don’t factor in turnout at all. Second, because even professional polls can only measure voters’ stated intention. Voters often say that they plan on voting but then don’t. There are two reason for this. First, they are ashamed of telling pollsters that they are not voting. Two, they are usually over optimistic of their commitment to voting.

Turning to our current situation. In this election poor UPFA base turnout and high UNP base turnout means that all polls, which primarily measure voter perception rather than behavior, are biased in the UPFA’s favour.

For example, imagine the Colombo District has 100 voters. Suppose I conduct a survey and find that 50 voters support the UNP and 50 voters support the UPFA. This does not mean that the UNP and UPFA will each get fifty votes. Suppose 80 percent of UNP voters actually vote in the election, while only 50 percent of SLFP voters do. That means that 40 votes will be cast for the UNP and only 25 will be cast for the UPFA.Read More