The curious case of Donald Trump: Is polling a useless predictor of outcome?

By Jamari Mohtar | Nov 10, 2016

The reliability of political polling to predict the outcome of an election is put into question when despite and in spite of most polls predicting Hillary Clinton as the favourite to win, albeit in a close fight because all polls are within their margin of error, Donald Trump against all odds clinched the trophy of the presidency.

Before we come to the conclusion that polls are a useless predictor of outcome, let’s hear some quotable quotes on statistics:

“There are lies, damned lies and statistics.” Mark Twain

 “It is the mark of a truly intelligent person to be moved by statistics.”  George Bernard Shaw

 “Smoking is one of the leading causes of statistics.”  Fletcher Knebel

I like most the quote by Fletcher Knebel because it hinted at arriving at something with no concrete substance as the aim of statistics, and to wit, we are all indeed “smoked” by the polls that said Hillary Clinton has a 90% chance of winning the presidency on the eve of Election Day.

Ignoring historical precedent at one’s own peril

 There are more than one ways to predict the outcome of presidential election other than polls.

A few hours before the results of some exit polls were announced on Election Day, I told friends through one of my WhatsApp groups that Hillary Clinton might not be elected as President, if we go by historical precedent.

Since term limit was imposed in 1947 – curbing presidential term to no more than two terms (eight years) – there has never been an instance where a Democratic presidential nominee won an election after eight years of incumbency by a Democratic president.

That is why Clinton lost after eight years of a Democrat Obama; Al Gore too (2000 election) after eight years of a Democrat Bill Clinton; and finally Hubert Humphrey (1968) after eight years of Democrats John Kennedy and Lyndon Johnson.

Whereas Donald Trump has a greater chance to win because there is one instance of history in which a Republican nominee won the election after eight years of incumbency of a Republican presidency. Who was he? None other than the one term President HW Bush who won in the 1988 election after eight years of Republican Ronald Reagan.

I am an ardent fan of history (and of ‘isteri’ too) although I’m aware that students pooh-pooh the study of history, one reason being it does not make you fabulously wealthy as compared to the study of law or medicine although I have come across poor lawyers and poor medical doctors. But as the outcome of the recent US presidential election, the Brexit vote and in fact most significant global events – even significant event at the personal level – have shown, one ignores history at one’s own peril.

This peril of ignoring history is famously encapsulated in the adage that history has a tendency to repeat itself. Even the natural phenomena of life have a habit of repeating themselves such as the repetition in the observable change in the days following the nights, of being healthy followed by being sick, of birth and death, and the boom and bust of the economic/business cycles.

Nonetheless, I’m not that naïve to believe that historical precedent is the only thing that matters. My view of history is as follows:

History seldom moves in a linear fashion. And that is why we don’t see new changes or new things everyday. Instead it moves in gradual non-linear twists and turns, giving us glimpses of an approaching historically repeating event in the making, where we feel things on the surface are the same as of yesteryears, yet with some qualitative differences in their essence. Once we get to feel this sensation of the same yet different, there will be many more non-linear twists and turns for years, before the full force of the repetition occurs.

At times history does not repeat itself at all but propels forward with a quantum leap as if in a three dimensional setting that demolishes every known assumptions with the onset of new inventions and discoveries or simply paradigm shift, heralding the emergence of a brave new world instead of the repeated old world. This then becomes a new normal and ultimately a status quo normal when it keeps repeating time and again before another quantum leap occurs.

Statistical disinformation or the fallibility of statistics

Now, let’s come to the crunch. Are political polls really useless as a prediction of outcome? One cannot blame those who say they are when the example of the Brexit vote is still fresh in our mind. Despite the narrowing of margin in polls as voting drew near in June, the majority of the polls were still predicting the Remain in EU would win, albeit with a small margin.

And last year in Askenazi Israel, despite exit polls had forecast a dead heat, Bibi Netanyahu’s right-wing Likud Party still won a surprise victory over its main rival, the centre-left Zionist Union.

But if you understand statistics in the context of the probability theory, you’ll be humbled enough to know that a poll which said that Hillary Clinton has a 90% chance of winning the election does not mean it’s a sure 100% win and that the 10% chance of Clinton losing is something that can take place in the realm of reality. In this sense, there is really no big deal in blaming polls for the different outcome than what was expected, as long as the different outcome is not a regular feature of the US election polls.

The last time the polls were dead wrong was in 1948 when Harry Truman was predicted to have lost the election, with one newspaper having circulated an early edition the day after election, which showed Thomas Dewey to be the winner as its page one lead story. The editors had had a hell of time in withdrawing that early edition.

Hence, the utility of polls as a predictive tool lies not so much in its accurate predictive power of the outcome ALL THE TIME, but rather a prediction that is dead accurate MOST OF THE TIME – giving credence to the notion of the working in the real world of the principle of an exception to the rule.

We can’t even predict with dead accuracy what’s going to happen to us in the next few hours, and yet we don’t want to eat humble pie in accepting that our prediction – the prediction of mere mortals – might go wrong when it comes to election polls. That is indeed arrogance of the highest order!

Of course there is nothing wrong in doing a sort of post mortem to get the answer on why and where did the polls go wrong, especially after the humble pie has been eaten. The least that this will result is in the lessons learnt to ensure that there will be a less frequent occurrence of the principle of exception to the rule, which it is meant to be for otherwise we would be living in a world of chaos. And then He who is in Heaven will smile approvingly at our action to learn from past mistakes and to minimize the exceptions!

And that is why I’m deeply moved when an intelligent and scholarly man says, “It is the mark of a truly intelligent person to be moved by statistics.” (Ahem, ahem…)

A ‘new’ normal?

 So what went wrong? I’m in no position to tell what went wrong scientifically though I took statistics at the undergraduate level but didn’t do well in that subject (actually within a certain margin of error, I did well in the exam, hehe). Based on news reports in the US, it is not so much statistical error that is at fault but systematic error.

Statistical error has to do with the methodology the pollsters used which will lead to among others, the questions of the representativeness of the sample (sampling error), the sample size so chosen (size error) and the degree of freedom assumed which will impact the level of confidence in prediction.

Experts have all been unanimous that the statistical errors were all within the threshold of acceptability statistically. Remember that statistics is not a science that is about 100% accuracy all the time and if you perceived it as such and refused to accept the existence of acceptable statistical errors, you (the layman) are exhibiting arrogance of the highest order.

So it is the systematic error that is in question which in layman term can be phrased this way: “Yes, the sampling method was right, the sample size was right but what are the questions that you asked the voters? Is it leading questions such that the result of the poll is what you (the pollster) want to hear rather than asking objective questions that beget objective answers?

The systematic error could also be explained in the way the final consumers of the poll (not the pollsters themselves but the media and Clinton’s campaign staff who commissioned the pollsters) spin the pollsters’ analysis in accordance with their own agenda of supporting Hillary at all cost whether consciously or not.

In this regard, Trump actually made sense when he alleged during campaigning that the election was rigged but he was far off the mark when he said that these people (the pollsters) were interviewing each other rather than random voters.

Perhaps he gave this stupid reason out of desperation because the analysis of his own pollsters had shown him that he had a good chance of winning in the battleground swing states.

But instead of seeing all these in term of polling errors, I’m of the view that the 2016 US election is a watershed election because it sees the emergence of a new normal as exemplified in Trump getting away unscathed for:

  • Not showing his tax returns;
  • Speaking outrageously against women, the Blacks, Latinos, Muslims, China and Mexico, etc;
  • Mimicking the gestures of the handicaps and his opponents; and
  • ‘Brawling’ with his fellow Republicans including Speaker Ryan

Seeing the above as a new normal also implied that perhaps the Muslims and others too should ultimately judge him based on the policies that he will finally implement as a President, rather than based on his speeches during the heat of the moment when campaigning.

The fact is for about three months after winning the election he is not the President of the USA, Obama is. President elect Trump will be just as lame-duck as the real President during these three months until his inauguration in late January, and due to this, it does not make sense to be emotional about him during this period.

So how do we predict the outcome of a Trump presidency under this new normal scenario? Is there any historical precedent? There is, actually.

When Nikita Khrushchev succeeded Josef Stalin as the Soviet leader in 1954, his outrageous behavior at the UN Assembly in 1960 by repeated banging of his shoe in protest at a speech by the Philippine delegate, Lorenzo Sumulong, had made him a Soviet leader with a relatively brief reign as compared to his predecessors who ruled until their deaths.

So in light of a new normal and a historical precedent, the relevant question to ask about Trump in relation to predicting the outcome of his presidency is not so much whether he will be a one term president; rather the question is will he serve the full duration of his first term?

Only time will tell whether the latter outcome will see the light of day! So far since winning the election, Trump’s actions and sayings are presidential.

Advertisements