The truth problem for opinion polls

The outcome of the US election has demonstrated that public opinion polls are increasingly becoming unreliable.

The polls were off again. Donald Trump did better than expected in many states so there was no "blue wave." However, Minnesota did as the polls predicted. The polls predicted Joe Biden to win by eight percent and the final count was some 7+ percent.

Those who had faith in the polls thought it was going to be a walkover for Biden. However, the Democrats did not buy the lead they had in the polls. They had been through a similar situation in 2016 and as they say “once bitten, twice shy.”

They therefore rolled their sleeves and pushed their supporters to vote early. They won with a thin margin, way below what the polls had projected.

Politicians like to dismiss opinion polls, yet, they often want a measure to guide them on strategies they want to employ. The importance of polls in any democracy cannot be underestimated. Polls help us to gauge how the population generally feels or thinks about specific issues that impact society.

In a diverse population, polls are important in getting to understand the views of others with the hope of forging mutual appreciation. To some extent, polls moderate media by enabling people to speak their opinion instead of letting others to interpret what is in their minds.

Indeed, an accurate prediction of elections can lead to making serious decisions by some individuals who may find policies of the winning party unpalatable. Therefore, the reliability of polls can enable the people to make informed decisions about their future instead of waiting until it is too late to react to the changes.

In most cases, polls do not explain why people believe in the polls or why they change their minds. More often than not, it is the work of the social scientist to interpret the responses. In the recent past, however, the polls have tended not to be accurate. It makes it harder for the scientists to interpret the outcome.

Several factors can be used to explain why polls don’t project accurately. These include poor sampling of the respondents, bias on the part of the researcher, small samples, poor questions and failure by the respondents to tell the truth. Most of these weaknesses can be corrected but the last one is difficult to deal with.

A situation where a respondent is not willing to tell the truth is referred to as the tendency to “choose responses they believe are more socially desirable or acceptable rather than choosing responses that are reflective of their true thoughts or feelings.” The concept is known as social desirability bias. It is a bias that Kenyans are particularly fond of. We generally vote along ethnic lines but you would never know it from our political discussions in public, as participants seem to be more rational in judging the candidates based on their issues. What is often not explicable is that when the vote is cast, rifts along ethnic lines emerge.

Public opinion poll respondents in Kenya as well as in America therefore would rather publicly associate with a socially acceptable norm but in private ethnic affiliations or emotions outweigh acceptable social norms. The reliability glitch is also happening basically because the survey method does not pick a shift in people’s attitudes.

Several studies show that people across the world have begun to have a negative opinion about polls and pollsters. Attacks on polling by politicians does not help as most of their followers believe what they see as fait accompli.

A recent poll on the performance of governors by Infotrack attracted as much condemnation as it did praise. The poll was necessary for people to gauge the performance of their leaders but instead the pollsters were unnecessarily being condemned.

There are also other factors like digitalisation that are changing traditional polling and survey methods. The implications of many of these new developments for survey data quality are yet to be fully understood.