Business booms for SA firm which predicted the result

NEW YORK: A handful of small public opinion polling companies that bucked consensus and accurately called the US presidential election for Republican Donald Trump are reporting being flooded with calls from investors and clients seeking their services.

Most pollsters wrongly forecast Democrat Hillary Clinton as leading Trump ahead of Tuesday’s election in the latest fiasco to hit the $20 billion (R284bn) public opinion research industry, only months after it failed to predict the British vote to leave the EU in a June referendum. Among the few that got it right was a new industry player using a different method, South African firm BrandsEye, which analyses social media posts.

With offices in Cape Town and Johannesburg, BrandsEye took an entirely different approach from traditional polling. The dataminer pays people around the world to sift through social media for relevant posts, a process known as crowdsourcing, and then uses a computer algorithm to rate consumer sentiment about products or politicians. Its method pointed to a Trump victory. It also correctly called Britain’s Brexit vote.

“My phone has not stopped,” chief executive and part owner JP Kloppers said .

Kloppers said venture capitalists have been calling about investing in the firm and polling companies have inquired about its methods. Kloppers said the company was open to licensing its technology.

Existing clients include ride-sharing firm Uber, banks, telecommunications firms and some government agencies, he said. He declined to identify the governments.

“Elections come every couple of years, but there are companies and governments that need to understand every day and every week what’s driving customer satisfaction and dissatisfaction,” Kloppers said. New methods, if successful, are of high interest to polling companies that are having difficulty reaching cellphone users or surveyweary Americans.

“The answer is ongoing research,” said Krista Jenkins, director of the PublicMind survey research centre at Fairleigh Dickinson University in New Jersey. “It’s basically digging in as an industry and comparing the findings from your traditional telephone polls to data that is collected online.”

Traditional pollsters base their results on questions posed to randomly selected people, often in interviews conducted live over the telephone.

Among other errors pollsters made ahead of the US election, they almost universally miscalculated how turn-out would be distributed among demographic groups.

High-profile London-based polling agency YouGov PLC’s stock dropped 4 percent on Wednesday after its new method of surveying people online failed to accurately predict the US presidential election’s outcome.

Analysts were split on the reason behind the drop and YouGov stock partly recovered on Thursday, closing up 2 percent in London trading.

The Reuters/Ipsos States of the Nation project gave Clinton a 90 percent probability of winning the 270 electoral votes needed to secure the election.

On its website, one of the possible scenarios was how Trump could win the election by adjusting turn-out levels. – Reuters