The two men that have correctly predicted most US elections
Almost everything now is tracked, measured and analysed. With less than a week to go all eyes are on the US election to see if Donald Trump can win again. No surprise that the amount of data being collected and analysed is huge.
Last time the polls and many pundits got it wrong, at least the part about Trump taking the electoral college win. The assumption that Hillary Clinton would win the popular vote was correct, but America actually decides who will be president based on the votes of a much smaller group designed by each state to make the final call.
In 2016, it was the votes by that smaller group that secured the White House for Trump not the popular vote.
In 2020, the question is will he be able to do it again. Besides the two men with long running models that have accurately predicted most of the previous outcomes - including 2016, film maker and ardent Democrat supporter Michael Moore realised that the polls were not reflecting the support for Trump that he was seeing in swing states that had been very negatively affected by globalisation and the previous financial crisis.
Those voters were angry and found the message of making America great again to be just the words they needed to hear to vote for someone with no political experience.
Those same voters may prove critical again, but this time the Democrats appear to have realised that they don’t just need the support in the polls, they need their supporters to go and vote.
The two models
One of their models will prove to be the one to predict even the most unpredictable election, a race between two of the oldest candidates in a country ravaged by a pandemic and from parties that are now polar opposites.
Being able to predict something as complex as this will see many wanting to try apply the principles to make better political, business, investment and even personal choices about many other complex subjects.
Here is what their models predict and how they get there.
The Primary Model - Trump wins
Helmut Norpoth is a professor of political science at Stony Brook University. In 1996 he first used his model based on the outcomes of elections going back to 1912 based on how a presidential candidate performed in the primaries. It draws its inspiration from the cycles first observed in sun spot activity. Norpoth explains how it relates in his TEDx talk below.
His model uses the amount of support a candidate had in their primary election to compare the relative chance of success in the actual election. His model correctly predicted the winner of the popular vote for each election from 1996. He did predict Al Gore would win in 2000 which eventually was won by George W Bush, but Gore did win the popular vote and the outcome of a very close recount decided by the courts allowed him to win the electoral college vote. It was the 4th time the two counts were not the same and considering the previous time it occurred was over a century earlier it was considered a major upset.
To the surprise of many not in the Trump camp, it would be true again in 2016 and allow Trump to beat Clinton in the electoral college vote even though Clinton had over a 2% lead in the popular vote (Gore had just a 0.5% lead)
Norporth’s model predicted Trump would win based on the Primary and so even as polls were still showing a victory for Clinton right up to the election date, the earlier prediction proved correct.
Norpoth has predicted Trump to have a 91% probability of winning in 2020.
The Keys to the White House - Biden wins
The model is based on factors that indicate the likelihood of an earthquake occurring in that when more than a certain number of factors can be detected the probability of an earthquake occurring is greater than it not occurring. In the same way when six or more of the 13 true/false statements about the incumbent presidency are false, it predicts the incumbent will lose or a political earthquake will occur.
In answering the 13 questions, 7 for this election are false which predicts an earthquake or a loss for Donald Trump and the Republican party.
#BusinessUnusual with @brucebusiness and @colincullis look at the profitable but risky work of those that predict the future and profile two men that correctly predicted the last US election and believe they know who will win in 2020. Details at 7pm. Who will win?— 702 (@Radio702) October 28, 2020
A word on predictions
While the impact of the outcome might seem remote for the average South African, the winner of the US election will affect us and the world.
Likewise predictions that relate to our country, our work and even our favourite sporting teams can move us to spend lots of time and energy trying to predict what will happen.
Getting it right can be very profitable or give your career a boost if not only bragging rights about having your ear to the ground.
In most predictions we use two sets of information. What we know and what we think. The work about what we know is analytics and the work about what we think will happen is statistics.
We need both but need to remember that they work differently. Analysis of what you know might help you understand how and even why it occurred. To predict the future requires a model based on what you know and statistical methods to extend the current results into the future.
Many of us might assume what happened recently will predict what will happen next, for some scenarios that is true. Your phone uses this model to show you the apps that are running on your phone, the first one displayed is the one you most recently used. You may not have thought about it, but odds are you do most often reuse the app you most recently closed even if it is not the most often used app.
But there are many models and there is a risk you will favour the one that fits what you think happened rather than one that reflects what did happen.
This is where the quality of your predictions will be determined by the quality of your analysis rather than the accuracy of the model you used.
In the next month only one of the two men with very different models and a good track record for success in predicting presidential elections will be proved to still be correct.
The actual outcome will give one more confidence to keep the model as it is and the other to reconsider if the factors that have proved to be so indicative in the past are still a good indicators in the future.
In the same way you should routinely check your data and your analysis to determine if your preferred model for predicting successful outcomes is as good as it can be. Don't believe in your model like some long-suffering fan hoping that this season finally your team will win.
One of the worst years this century did not stop some industries from having their best year yet.Read More
The headline relates to vaccine origins, this is about how close we are to a Covid-19 vaccine.Read More
500 million metric tons are produced a year, but environmental considerations are making alternatives look more attractiveRead More
At the moment, the IPO has been postponed, this is the story of how we got here.Read More
The future of agriculture may draw more from its past than the presentRead More
Over 1000 MW of renewable energy will be added to the grid in the next year, we need a way to store it.Read More
New investment platforms are making it easy and cheap to invest, what could go wrong?Read More
Most plastic was never intended to be recycled, but we have believed the story for decadesRead More
The cost and benefits are making home automation easy and affordableRead More
Almost a century after it began the last original stock is removed from the indexRead More