Come election season, a lot of noise is generated because of opinion polls conducted by various media houses. Political parties that lead in these polls start peddling the results as gospel truth, whereas the ones trailing behind take no time to declare them as bogus. This happens in every election (Assembly or General), and the only thing that changes is the stance taken by each political party.
This time around the noise was shriller than usual, partly due to its amplification on social media and also because of the sting operation (named Operation Prime Minister) that implicated many opinion poll agencies for fixing their results. The public sentiment on opinion polls varies from considering it as complete junk to having more than warranted faith in their accuracy. Many of those who believe the former, perceive them to be unscientific, as they are based on the opinions of a very small percentage of voters. However, it should be noted that opinion polls are scientific and if conducted in the right manner can provide a good insight into the minds of potential voters.
A reasonable sample size (20-25,000) and proper adherence to the sampling methodology can provide a decent estimate of vote shares across various regions (or states) in the country. However, the conversion of this vote share into number of seats is a tricky business and hence the predictions are rarely accurate. In order to get a fairly accurate estimate of seat share, the agencies would need a sample size of at least a couple of thousand from each constituency (more than a million respondents for a general election). This would not just be a very costly exercise, but would also require a very large trained workforce (which most agencies don’t have).
Once the sample size has been decided, the next important task is creating an unbiased questionnaire. The same question may elicit a different response from the respondent, depending on how it is framed or the kind of response it requires (whether it is an open-ended, binary choice or a multiple-choice question). Opinion polls use random sampling to minimise bias.However, random sampling has limitations of its own; it measures the opinion of the entire population. With random sampling there is no way to segregate the potential voters who would actually go out to vote from the ones who won’t. It has been observed that potential voters belonging to upper castes/classes have a lower proclivity to vote in comparison to potential voters belonging to lower castes/classes. Therefore, while the polls would show the preference of the entire population (that is eligible to vote) the final results might witness a slightly higher influence of the population group that had a greater level of participation. This is quite evident from the fact that the Bahujan Samaj Party (BSP) has consistently performed much better than what most opinion polls have projected over the years.
The biggest limitation of opinion polls is that they are mostly conducted before the list of candidates is released. The responses recorded in these polls are based on a voter’s feelings towards a political party or a central leader; they do not account for the candidates who would be contesting from their constituency. While loyalties towards a political party or likeability of a central leader are important factors, the candidate contesting can often be the most important factor while casting a vote. This can be observed in the latest opinion polls for the general elections. According to most polls, the Aam Aadmi Party (AAP) is expected to get an abysmally low national vote share. While it could be true that people outside Delhi (and some metros) do not strongly associate themselves with AAP, if we consider their list of candidates, it cannot be denied that many of them are capable of impacting the local outcome on their own.
In light of these caveats and the inability of a lay person to understand them, there have been some demands to ban opinion polls altogether. Although it is true that the margin of error in extrapolating the results of a sample to the entire population and then using this vote share to estimate the number of seats is quite high. An average viewer/reader, with little understanding of the process, may get misled with these polls. However, this does not justify an outright ban. At best, a case could be made for their regulation by the Election Commission.
Opinion polls act as a catalyst in initiating political discussion even in the most non-political gatherings. The numbers churned out by these polls enable those people to take part in political discussions, who otherwise do not understand the qualitative aspect of politics and the jargon that surrounds it. The problem arises when media houses start pushing these opinion polls as hard news instead of acknowledging what they really are – an infotainment program. Media houses should make a greater effort in educating their audience about these opinion polls, their process and accompanying qualification, so that they are not misled in any way. A little self-regulation by media houses is all that is required to calm all the voices asking for a ban, and save the Election Commission from an additional burden of screening opinion polls.