Why Polls it is essential to ignore all polls

Owing to the absence of a proper methodology these poll turning out to be as reliable as projection models for Covid19.

 

It was 29th November 2015, election results day in Bihar, one of the most populous states in India. Almost every poll, including exit polls, had predicted a landslide victory for the challenger the National Democratic Alliance, i.e. NDA.

For weeks, news channels had panels of ‘experts’ vehemently expounding the reasons for these poll results. The incumbent state government was blamed for cozying up with activists instead of governing, and was also branded as arrogant and incompetent. The leaders of the NDA were lauded for their prodigious campaigning skills, their impeccable ground game, and their strategy.

When the vote count commenced, the early trends had the NDA leading. The ‘experts’ replicated their previous proclamations; they lambasted the incumbent while they lavished blandishments on the challenger. They even began advising the incumbent on remedial measures after their humiliating defeat. As the pompous pontifications continued the tide began to gradually turn. The NDA lead began to wane, in an hour then the current government had a slim lead. The lead grew bigger and bigger, by mid-day, it was clear that they were re-elected.

Some news personnel offered perfunctory apologies, others blamed the people for lying to the pollsters, and some refrained from talking about the polls and proceeded as if nothing had happened.

This isn’t the first time and it most certainly won’t be the last time that polls got it so appallingly wrong. The big question is why do these polling agencies that consistently committed calamitous mistakes continue to be treated with the utmost reverence and reverence by news organizations?

We focus on the polling process. The first step is the actual sampling. Most agencies have demographic-based sampling on parameters such as age, race, gender, religion, income groups, region, political affiliation, etc. In some cases, it is a combination of parameters. The next step is the actual interview. The resulting information is compiled and extrapolated. All of the above-mentioned stages are replete with elements of randomness that present challenges and huge risks.

First, it is almost impossible to define the optimum size and decide the proportion of diversity for any given sample such that it is an appropriate reflection of the entire voter base. The sample size and variety are probably based on the time, budget, and instinct of the polling agency.

Secondly, the composition of the sample into demographic groups is deeply flawed. Just because individuals are from the same income group, age group,  religious group, and race group does not mean that their life experiences and outlook on life are identical. There are several cases of deep ideological and political differences among the same age group of people within a family. This means voting choice will never identical.

Polling five thousand African American men in their early thirties with college degrees cannot possibly capture the voting choices of five hundred thousand African American men in their early thirties with a college degree who run small businesses.

The conducting of the interview has its own inherent shortcomings. To question strangers about their voting the choice may either yield evasive answers or out of sheer politeness, an answer that they think the interviewer wants to hear though in some situations the truth will present itself. For exit polls, the politically charged atmosphere outside the polling booths presents another challenge.

Also, to understand any individual and his voting choices, the interviewer will have to spend a considerable amount of time with the interviewee, perhaps ask specific questions related to key issues, observe their lives, their social and financial situation. Most pollsters do not have the time and budget for that sort of focus.

Not conducting an interview mano a mano, perhaps doing it over the phone or online or a paper form presents its own risks. In addition to the fear that the opinions may be leaked, people may just falsify and there is no way to verify.

Thus the core of the problem is the lack of statistical guidelines, methodological protocols, interviewing techniques, and transparency. The fact that we have millions voting across deeply diverse states exacerbates the already existing challenges. This is probably why we can see starkly contradicting results from various polls for the same election.

Now for the ‘experts.’ These are usually adept wordsmiths and have covered politics for so long that conjuring up a narrative is almost second nature. You want ten strong reasons why any given political party will certainly be back in power, they can give it to you. You want ten strong reasons why the same political party will receive an emphatic drubbing at the polls, they can give it to you, too.

Even if they do interact with voters, they will have identical challenges that pollsters do. In the end, the opinion of the ‘expert’ maybe backed by years of experience, education, and research, but their opinion may be no different from anybody else with some basic understanding of politics. The expert opinions are mere conjecture after looking at the polls.

If polls were to be believed, Brexit was never supposed to occur and Donald J. Trump was had no chance of becoming President of The United States

In what seems like a repetition of 2016, the polls show Biden beating Trump comfortably. 

There are many reasons not to trust these polls. First, we are three months away from the elections and there is a lot to occur in between. Second, the fast-growing cancel culture, where an actress was targeted merely because her boyfriend was thought to be Trump supporter or the President of Goya Foods faces backlash for praising Trump. It is difficult to expect individual to be truthful when the consequences could be dire?

The general narrative, in addition to usual baseless allegations of Russian collusion, sexism, racism, homophobia and other varieties of bigotry, is that the people disapprove of Trump’s handling of Covid-19 and this will translate into them voting against Trump.

Even if one were to accept this premise, the same individuals may support Trump because he is tough on immigration, strong on national security and still believes in the greatness of his country while the other side is busying attempting to dismantle the nation.  They have seen Trump build a strong economy and is sure Trump will rebuild it while Biden would return them back to the Obama days of economic turmoil.

Like a broken clock that is right twice a day, there will always be a group of pollsters who get it right and earn overnight credibility. This temporarily earned credibility will come crashing down after they commit some monumental but inevitable errors during the next elections.

Despite all the uncertainties and risks, polls are always broadcasted with great fanfare. This is no surprise for 24-7 new channels desperately seeking to fill news space. Another could be to psychologically influence them into thinking that the results are caste in stone and that their vote is inconsequential.

In the end the only polls that matter is the actual votes in the ballot box.

An elderly friend of mine known for his humor, would pose this question to his doctor whenever he felt the treatment or tests prescribed seemed gratuitous: “Is this for me or for you?”

As consumers of news, we should also be asking this question news organisations who dedicate reams of news space to polls that are almost always wrong.