prev next

Keith Naughton

Pollsters Were (Sort of) Right

by Keith Naughton

Did the 2016 Presidential election result mean the humiliation and death of polling? Not by a longshot. The fact is many of the polls were right — it's the people who got it wrong. The pro-Clinton mainstream media and the Washington Establishment echo chamber interpreted the data the way they wanted to, instead of objectively. If they really understood what their own polls were saying we would not be talking about this election as the upset of the century, but at most as a mild surprise.

Given the result (essentially a popular vote tie, 47.7% Clinton to 47.4% Trump), the perfect Monday poll would have showed Secretary Clinton at 46-47% and President-elect Trump at 43-45%. That's right. A truly accurate poll would have had Trump behind. Why that result and why did the media look so foolish yet again?

FACT #1: Margin of error is real, not just some cop-out.
FACT #2: The result was a classic challenger/incumbent result when the incumbent can't close the deal.
FACT #3: The media still have not the first clue about polling, statistics and voter behavior (in spite of the best efforts of The Daily Caller to educate them).

FACT #1: The margin of error is real. All polls have a margin of error. Since you are talking to just a sample of people rather than everyone, you cannot be absolutely sure the results can perfectly mirror what the entire population thinks. The number reported by the pollster is really just the most likely number in a range. That's not spin or hedging, that is basic statistics.

When a pollster says a candidate is polling 50% with a margin of error of plus or minus 4%, it means that there is a 95% chance the true result (if you asked everybody) is anywhere from 46% to 54%, but the most likely number is 50% — 49% or 51% is a little less likely, 48% or 52% even less and so on.

The fact is that 8 of the 11 polls released the Monday before Election Day had Clinton ahead by a difference that was within the margin of error (3-4%). Oddly enough, the USC/LA Times poll (now widely hailed as the "correct" poll) was not so accurate. USC/LA Times did show a 3-point difference, well within the margin of error (47/44), but it wrongly gave Trump an advantage in the popular vote and suggested that Clinton gained all the undecideds at the end — a highly implausible scenario.

To be fair, Clinton had been polling a consistent lead, although that lead had rarely moved outside the margin of error. However, Clinton's lead deceived the media, but should not have deceived experienced pollsters and political hands.

FACT #2: Trump over Clinton was classic incumbent vs. challenger. Any political pro will tell you that an incumbent is in trouble if they aren't polling at least 50% by Election Day. If you add voter anger to the mix, then no incumbent's lead is safe. (If you give Gary Johnson, Jill Stein, etc., the nuisance candidates, about 5%, then 47-48% is a winner)

Clearly Clinton was the de facto incumbent and Trump was the challenger. While there were points in the race where Clinton managed to climb above 48%, she had a tough time sustaining that number. People drifted to Clinton briefly and then drifted away — she could not close the deal. Meanwhile, there was a very important statistic the media resolutely refused to discuss: Right track/wrong track. On the eve of the election, Rasmussen reported 30% thought the country was on the right track — and 63% wrong track. The Economist (30/62) and NBC News (31/62) were similar.

When you combine a massive right track/wrong track deficit with an incumbent who can't stay at that critical winning percentage, you have a candidate in a lot of trouble. Given those numbers, Trump should have run away with the election. The last time the wrong track number was this high was 2008, and the "challenger" Obama won in a walk.

What the media instead focused on were the approval numbers for the candidates and President Obama. Those numbers gave Clinton a modest advantage. The last favorable/unfavorable numbers for Clinton were 43% favorable and 56% unfavorable (Economist). For Trump they were 39/60 and for the media's great hero, Obama, just 52/47 (Fox) and 50/49 (Economist). While the media obsessed over Trump's unpopularity and the (not really) popular Obama's backing of Clinton, they ignored the awful wrong track numbers.

The conclusion: At a time when over 60% of the public thinks the country is on the wrong track, an incumbent who was stuck under 48% lost to a challenger. No surprise there, except that Trump perhaps should have won by more.

FACT #3: The media is still in a fog on polling. First off, they don't even understand what a poll is. Polls predict nothing, they just tell you what the public thinks at a specific point in time. In other words, polls describe, people predict.

Polls are not magically produced overnight. They take time to conduct. The typical serious presidential poll takes 4 days to finish. So, all the polls released on the Monday before Election Day really show voter intent 2-5 days prior to voting (except for early voters).

And that leads to another issue: every election has a small group of holdout undecideds who don't make their pick until the last days — or even the last minute. As a result, the very late polls should show at least one, and usually both candidates with numbers at least a bit below their final total.

Not understanding this dynamic (or willfully ignoring it), the media saw Clinton ahead in the polls reported on Monday and excitedly guaranteed her victory. That's like declaring a baseball game won by the visitors before the 9th inning when the home team is behind by one run. The media called the election before it was over and clearly there were voters still deciding.

(There were some truly predictive models that had Trump winning, but they were generally ignored.)

So, what does all this mean? Given that the polls were telling us what the electorate was thinking the weekend before the election, that most undecideds tend to break for the challenger, and that the margin of error means you have a good chance of being off by a point or two no matter how good your polling is, a "correct" late poll would have had Clinton at 46-47% and Trump at 43-45%, with about 5% for the nuisance candidates and 5-6% undecided.

That puts ABC (47% Clinton; 43% Trump) and Rasmussen (45/43) in the winner's circle. Meanwhile, USC/LA Times (44% Clinton, 47% Trump) and Monmouth (50/44) get a trip to the doghouse.

Share   Share

Featured Columnists
Featured Audio Links