How Trump’s victory put under question all favorite concepts of digital world Articles in English / In the Air

The very next day after surprising results of US elections, medias have exploded in an urgent rush to find out why nobody saw this coming…

The reliability of some favorite concepts of the digital world, like Big Data & predictive analysis, were questioned – some of them even incriminated for letting Trump win.

 

So, why did almost everybody miss this bet?

The New York times, Reuters, NBC news and countless others…We all (well, almost) took the predictions of polling aggregation as a gospel.  And indeed, results looked so convincing, forecasting Hillary chances of winning of as high as 99 percent. What happened ?

It looks like all these big institutions have done a basic mistake : they didn’t choose the right data from the right source.

Polls were the main source of information. And this was wrong. In fact, polls have never been designed to be forecasts. They are simply one piece of data among many others. So, the real failure was to ignore or underestimate other sources.

When looking at this type of data, we should take into account that humans are extremely volatile – they can change their minds, they can decide not to share their opinions or they can flat-out lie. All that happen before you even get to some of the statistical issues that make polling inaccurate.

As we can see, both the data set and analysis approach were not adapted to the reality of these elections.

 

What is the right approach then?

Here is one of the brightest examples, that was also picked up by the media (although too late). Allan Lichtman, a political analyst who has correctly predicted the results of every presidential elections since 1984, relies on another methodology. His model, developed in 1981, uses historically-based system of 13 true-or-false questions looking at economic indicators, military failure and success, social unrest and third-party candidacies.

And here’s what he claims about polls usage:

“Polls are not predictions. They are snapshots and they are abused and misused as though they are prediction”

 

Should we call Big Data the biggest loser of these elections?

Last Tuesday was not a failure of data but a failure of analysis.

Obviously, one mistake doesn’t mean that the game is lost. On the contrary, it highlights that it is important to gather more data, crunch it and work on what it means. And most importantly, putting it into a broader analytical picture.

By the way, after correctly predicting the winner of this election, Allan Lichtman now says that we can expect the impeachment of Trump by Republican Congress.

Any bets?


Avatar de Ekaterina Pavlova
Ekaterina Pavlova

MSc in marketing from ESCP Europe business school, Ekaterina joined Equancy in 2014 while completing her next degree, an MBA in web management and digital marketing at INSEEC Supdepub. Ekaterina holds a solid experience of working with international accounts. Before arriving to Equancy, she was a part of Nissan EMEA account team at TBWA/G1. Here at Equancy, Ekaterina works on a range of strategic marketing and web analytics missions for such clients as Cartier, Michelin, Disneyland Paris and Sephora. She is also running the AB testing and Conversion Optimisation Stream for Michelin.

Leave a Reply

Your email address will not be published. Required fields are marked *