With the 2016 election nearly upon us, the political world is once again awash with presidential polls. Many of them show very different things, with some showing Hillary Clinton ahead and others Donald Trump. They can't all be right. How can you tell which polls to trust?
Fortunately, there are data for this. And they clearly show that the IBD/TIPP Tracking Poll, over the past three election cycles, has been the most accurate poll out there.
It's no accident. Polling is both a science and an art, and our pollster, Ramsey, N.J.-based TechnoMetrica, excels at both.
The 2012 election is a case in point.
With less than two weeks to go, a number of polls showed a significant lead for Republican challenger Mitt Romney. Gallup, for instance, had Romney up by six points with just a week to go. Meanwhile, other polls showed an "enthusiasm gap", with GOP supporters far more motivated to vote than Democrats.
It looked like it could be a blowout — for the Republicans. But our own 2012 polling showed something different, with incumbent Barack Obama having the edge.
How did we stack up in the end? By our calculations, based on final results of 11 national polls that have been active over the past three presidential elections, IBD/TIPP was among the most accurate polls in 2012. Indeed, polling analyst Nate Silver of the New York Times' blog FiveThirtyEight, who ranked 23 presidential polling organizations in 2012 using a different method than the one we used, described IBD/TIPP as "the most accurate" tracking poll for the year.
Over the long term, our own tally shows that for the 2004, 2008 and 2012 elections IBD/TIPP's average prediction for the final presidential vote margin was the most accurate, with a difference of a mere 0.9% between our predictions and the final outcomes for all three years.
As the chart with this suggests, this was no fluke.
In 2008, we got Obama's final margin of victory exactly right, predicting a 7.2 percentage point victory for President Obama over Sen. John McCain.
As for 2004, with just days to go, some polls still had John Kerry beating George W. Bush by as many as two percentage points. IBD/TIPP's prediction: A 2.1 percentage point win for Bush. The final difference: a 2.5-percentage-point win for Bush.
How does IBD/TIPP do it? It starts with a random sample of registered voters and likely voters, who are identified based on a series of carefully crafted questions. There are typically between 850 and 1,000 respondents, giving a margin of error for both groups of plus-or-minus 3.4 percentage points, at a 95% confidence level. What that means is that, statistically speaking, we're 95% sure that our polls will be within 3.4 percentage points of the actual result.
TechnoMetrica collects its information via telephone, calling both landline and cellphones. Roughly 65% of the respondents come from cellphones, and 35% from landlines. All the calls are made live — there are no "robocalls."
To fine tune the results, proprietary data are used to accurately adjust for race, gender, region and political party affiliation of the nation as a whole. This assures a more accurate end result than simply relying on raw poll responses.
It's a matter of record that the IBD/TIPP Tracking Poll, which we will publish this year starting on October 20 and running through to November 7, has been the nation's best over the past three election cycles — something we're quite proud of. As we've said before, our polling partner, Raghavan Mayur, and his able staff at TechnoMetrica, deserve credit for their title as "America's Most Accurate Pollster." They've earned it.