“Not every thing that counts can be counted”
The American election is, at the time of writing less that a week away. Joe Biden has, according to the polls, an unassailable lead. The book makers have him at unbackable odds and in the absence of divine intervention nothing can save Donald Trump from the most one-sided election since Reagan -Mondale.
Yet even among Democratic faithful there is a slight reservation, a paranoid caution. We were in a similar situation in 2016, and the unthinkable happened. Beyond that something seems discomforting in regards to the size of Joe Bidens lead- it is enormous. In almost all states, in many by double digit he leads.
Yet when we look at other data it sems strangely incongruous. Trump is holding enormous rallies with intense public support, while Biden struggles to fill a parking lot with cars. This we may attribute to a different strategy in regards to Coronavirus- that seems plausible.
Yet the polls are unanimous- a decisive majority of persons interviewed report that they will vote for Biden. So are the polls correct, or is something else going one.
A salient reminder from Albert Einstein reminds us that just because we can measure something does not mean it is worthwhile or even useful.
Computer programmers have a shorter version GIGO- garbage in garbage out. In short, if your methodology is wrong, it does not matter how much data you collect, you will not get any closer to a true value.
If the methodology of an Election Poll is wrong, it does not matter how often you conduct the polls- you will just accumulate misleading data. In this case, a huge data set all indicating Trump will lose. It does not however provide any increasing evidence.
So could there be a flaw in the polling system. It should be noted that polling is not an altruistic activity- it is an industry, a profit industry, a huge profit industry. Polling relies on people being honest when asked about their future intentions. There are an array of issues that have been identified with polling, including selection biases, leading questions, so called shy voters and social desirability bias. Pollsters have attempted as much as possible to adjust for these factors in an attempt to avoid the errors of 2016.
There is however one other important bias that is hard, if not impossible to adjust for in polling. It is a bias related to virtue signalling. Human beings have a tendency to genuinely misrepresent their own behaviours. In research people consistently under report their substance use, drug use and promiscuity, while over reporting their work hours and house work. Human beings see themselves as more virtuous than they really arm. This bias becomes even more pronounced when they try to predict their own future behaviour. At university most of us are guilty of under predicting the amount of alcohol we will drink at some future even, while significantly over predicting how much study we will do. As a whole we will tend to predict our future behaviour to be more virtuous than it really is.
In marketing we know that while a large percentage of the consuming public will verbalise a preference for free range, organic and dolphin friendly foods, only a minority of this same group will actually buy those same products. As people we usually fall short of meeting our high moral values when it comes time to making real decisions. In such circumstances practical issues, such as price, convenience and brand appeal is often more powerful that our lofty ideals.
In the current US election, we see a metaphor for our virtue bias. Most of us can agree that Trump is not seen as virtuous as his opponent. He is seen as a womaniser, misogynist, tax evader and unscrupulous business man. Our virtue judgments make many people verbalise a plan to not vote for him. After all, to vote for him would be to reward and unscrupulous person, and this goes against our values. It is reasonable, therefore, for voters, when polled to predict they will vote for Biden based on their virtuous side. Many of those people may genuinely believe that that is how they will behave.
However when it is time to actually vote, a different cognitive process tends to occur. Who the voter believes is the more virtuous candidate be comes a secondary consideration to a more pressing issue- which candidate does the voter feel will benefit them the most. Humans are buy nature virtuous in their words, but pragmatic and self centred in their actions.
So if humans cannot be trusted to predict their future actions, which in truth they can’t is there a better way to predict voting behaviour than just asking them who they plan to vote for. There are a number of research techniques which tend to better predict behaviours, than self reporting.
Firstly, rather than asking someone to predict their own behaviour, asking someone who knows them to predict it is often more accurate. Spouses for instance tend to more accurately predict alcohol consumption than self reporting from the person themselves. In polling this is the equivalent of asking people who they think their neighbour will vote for. In the current polling, despite the edge to Biden, a majority of persons polled predict their neighbours will vote for Trump and even that Trump will win the election. This is an interesting contradiction of the polling data, that even in polls with a clear majority for Biden, the same voters still predict a Trump victory.
Another way to predict future behaviours is to measure associated measurable behaviours. In studies of weight loss participants for instance self reporting is very inaccurate. Weight losers consistently under report how much they eat, even when examiners put video cameras in their refrigerator and kitchen. It turns out that how often you visit take away restaurants is a better predictor of weight loss, than how much you claim you have eaten when going to those same restaurants.
In the election, there are quantifiable behaviours we can measure which can predict voting patterns. Behaviour on social media and the internet is a good marker of future behaviours. In this case Trump has a huge advantage in social media followers. On twitter alone, Trump has 87 million followers against Bidens 6.1 million and similar ratios can be found on all social media platforms. Some researchers have analysed social media comments and posts, suggesting that there is a significant over representation of pro Trump opinions.
In live rallies we seem an even more striking dissonance between attendances for Trump vs Biden. Pollsters have referred to this measure as enthusiasm.
So if the polls are incorrect and Trump actually already holds a majority, why have the Democrats lost support. They are going against on of the most disliked presidents in US history. The Democrats have also been very supportive of black live matter movements, defund police and stronger coronavirus restrictions. The polling clearly indicates that these are important issues to the voters. If the hypothesis of virtue signalling in polls is correct, then even though people may report their support for such policies, at the ballot box they will not follow through with voting intentions.
My suspicion is that in an attempt to appease the polls, the Democrats have embarked on a policy suicide. My suspicion is that to a majority of American voters, Defund the police is ridiculous policy with very little widespread support. Similarly, no matter what the polls indicate. American, like most of the Western World are well and truly sick of the coronavirus lockdown and any party promoting a more stringent lock down than Trump will not attract vote. In regard to the American riots, I would suspect that mainstream voters are unlikely to resonate with the position the Democrats have taken.
I suspect that the Democrats have been led astray by the polls, even to the point of making Biden their candidate. Biden was chosen, because in polling he does well- he is after all likeable and kindly and does not alienate any specific group with any strong opinion. He is perfectly suited to an artificial virtue signalling bias in the polls. It is a shame, because the democrats had a deep pool of smart and talented people who could have presented a real alternative to the voting public.
When it comes time to vote however the average voter is left with the unenviable decision of weather to elect a successful sociopathic racist misogynistic business man, or a likeable and pleasant dementing politician. American are between a rock and a hard place.
Disclaimer: The author notes that he has no political affiliations and is a member of neither party. He has no specific preference for either candidate. The author has placed some bets on the election and would prefer to win both for financial reasons but mostly for bragging rights with his friends.