US Elections: Testing predictive diversity and the wisdom of crowds

The Trump-Biden election is coming to a close in a few days. Being Canadian I am less concerned about who will win this, and obviously not taking sides. I am more interested in the accuracy of predictions – especially the ones that are crowdsourcing the results and what these show as of today. It will be interesting to see whether these crowd-based data was either right or wrong and how accurate if they were right. Nearing election day, nearly 50 million voters have already voted – either through early voting or mail-in voting, or other mechanisms.

Crowdsourcing is based on a large data set. For this to work, you can’t make predictions based on single data points. The data analysis also needs to account for the voting in the swing states and influence in-between state voters in states physically near each other. Crowd-sourced data is called predictive diversity and different from cognitive diversity which utilizes human cognition in large groups to do work. Predictive diversity has a spectrum of predictions and the extreme ends even out because of the large data set and this is often more accurate than one-off polls, crystal gazers, or even a collection of experts from the industry – no one person, source or small groups of persons are going to get this right. The chances for the crowd’s wisdom in being right is that much higher. Sometimes, the crowds get it wrong too or the crowds do bad things and this is called the “madness of crowds”. That is a topic for another day.

One great analysis source I like is the one prediction that is done by the Economist. According to it, it looks like including the margin of errors in the polls Biden should be winning this one. And the win should not be even considering the swing states. See the slideshow below for details about it. Here are a few snaps shots of their predictions using a poll of poll data.

https://projects.economist.com/us-2020-forecast/president

https://www.realclearpolitics.com/elections/betting_odds/2020_president/#!

https://www.realclearpolitics.com/epolls/latest_polls/

https://projects.fivethirtyeight.com/2020-election-forecast/

Sources of data

Another view of the polls could be through Betting Odds. Here again, I don’t believe in a single source betting data. I use the Real Politics betting odds site which aggregates various betting odd data into a single view. Here too Biden is leading 64% to 35% – but I suspect most of these betting is outside the US and also may not be by people who actually are able to vote.

Well, time not too far away will tell whether the crowds were right (as represented above) – for me, I am betting that there is going to be tumultuous market post-election before the votes are finalized, and therefore am sitting on some part cash in my investments portfolio awaiting to make some killing on the stock market. And I might be wrong too… look for my follow up article after the election is over and the results are announced.