Traditional Polls Are Dead: Long Live Online OSINT Surveys
The polls that consistently forecast Hillary Clinton’s victory against political outsider Donald Trump in the Nov. 8 US presidential election crashed resoundingly when the dust began to clear.
The pollsters, notably ABC News/Washington Post, Fox News, or CBS News/New York Times, and the media which ran these surveys missed the trends leading up to the final outcome, including the Republican candidate’s last-minute raid on traditional Democratic battleground states that clinched his win.
Was flawed methodology the cause of this run of misjudgments? Or were the figures manipulated?
Traditional pollsters work by framing the questions and canvassing individuals drawn from a sample crosscut of mixed demographic, social and other factors – but only through landline phones, which cuts a large voting segment out of the model sample.
The questions, the sample and the time of day for canvassing are determined by the pollster. The table of data collected is analyzed mathematically, but the data itself and the sample model may vary from poll to poll.
Poling by this method failed to pick up on the trends that led to the election’s final outcome – and not for the first time.
DEBKA Weekly’s analysts recall that the pollsters who put all their prestige behind Binyamin Netanyahu losing the last election had to eat their hats. They also failed to forecast the British public’s majority vote to quit the European Union, Brexit, and the end of David Cameron’s career.
As a long list of pundits confessed after the election shock, “We were wrong” about the Trump victory,” the regular pollsters were forced to admit their reliability was in tatters and their long dominance of the media as the trendsetters in politics, business and other fields was at an end.
Where they came unstuck, two different kinds of survey using different methods succeeded. Their findings never reached the front pages or screens of US mainstream media – or interrupted their unswerving support for a Clinton triumph. This raises another set of questions about media credibility.
And so, whereas NBC News misreported that Clinton held a six-point lead on Election Eve, the USC/Los Angeles Times Daybreaker poll, which had put Trump ahead of his rival for months of the campaign, placed him three points ahead on the morning of the vote.
And Sanjiv Rai’s MoglA forecast victory for the Republican candidate and gauged him as 25 percent more popular than Barack Obama in 2008.
The LA Times got it right by basing its results on a sample that was double or more the size canvassed by their competitors, and reaching a wider spectrum of voters by interviewing people on mobile phones as well as landlines.
The Indian scientist, who founded MogiA in 2004, uses artificial intelligence online for analyzing and evaluating common phrases and words harvested on search engines; the frequency and nature of content appearing in chains of posts and tweets on all social media; as well as comments on articles appearing on news sites, and the level of Internet users’ engagement in the subject of his survey.
Forecasts based on OSINT (Open source intelligence) are becoming more and more accurate, prevalent and broader-based, as the numbers of Internet and social media users proliferate by leaps and bounds (up to app. 85 percent in the US) and reach into almost every corner of the electorate.
As experience accumulates in the definition of algorithms, which are the basis of online opinion surveys, traditional polling will be relegated to the dustbin.