Polling in Elections

Legon Voting No poll can be a substitute for actual results, what matters is what happens on election day.

Wed, 30 Nov 2016 Source: Arthur Kennedy

In recent times, the use of polling to predict election results has lost some credibility.

The polls got it wrong in Brexit. Then, to top that, it got it wrong in the election of Donald Trump. To be fair, amidst the general error of most pollsters, a few, including the Los Angeles Times got it right in the US elections.

The Ghanaian punditocracy is all atwitter about polls leading up to the December elections. There have been polls by Political science departments, by Think Tanks and then, by Ben Ephson. And there have been prophecies, many prophecies. As you know, in Africa, unfortunately, prophecies are ranked ahead of polls by many, including candidates.

I learned about the art of interpreting polls at the feet of Professor Larry Gibson, who helped elect Ellen Johnson Sirleaf and Ravelomana as Presidents of Liberia and Madagascar respectively, in addition to his roles in the elections of Clinton and Obama.

When he looked at the first set of polls in 2008, nearly half of respondents from the Volta region were uncommitted. After poring over the figures for some days, he announced that most of the uncommitted were, in fact, committed but undeclared. He examined their positions on other issues and their demographics and predicted whom they would vote for.

Then, he added the summation that made him a legend in my eyes. "The NPP" he said, "will win the first round but not get 50%. As for the second round, it is too close to call but if you put a gun to my head, I will give it to Mills by a whisker". He was dead on, unfortunately.

Polls are predictions, based on certain assumptions. It assumes that the turnout will be similar to the demographic of those polled. If it turns out that turnout in Ashanti, Greater Accra or Volta is 10% higher or lower, it can make a poll inaccurate. Keep in mind that turnout in 2012 was79 % nationally and 85% in Ashanti. A poll also assumes that the respondents are being honest.

If they are lying to the pollster, the poll will be wrong. The third variable is the biases and integrity of the pollster. Democratic polls tend to favour Democrats while Republican polls favour Republicans. In Ghana, in particular, people and parties pay for polls and they are done, more for propaganda than for prediction.

Thus, the reputation and track record of a pollster matters. For instance, Synovate, Afro barometer and Ben Ephson are generally credible and must be taken seriously.

The final variable is the timing of the poll. A poll 2 weeks from an election may be different from one 2 days from an election. Voters perception change during a campaign and campaigns ignore this at their peril.

To those celebrating or panicking because of particular polls, remember, no poll can be a substitute for actual results. What matters is what happens on election day.

Indeed, even exit polling on voting day has been known to be wrong. After early exit polling in 2004, John Kerry believed he was President-elect till the results were tallied.

On election day, a party that is behind by a few points can win with a superior get-out-the vote operation.

May Ghana win on election day.

Columnist: Arthur Kennedy