Yesterday we tried to sort out the confusing and conflicting headlines we’ve been seeing about who’s really leading in the Democratic 2020 New Hampshire primary. This all centers around the concept of an outlier poll. Let’s say you do ten polls, and nine of them all say the same thing, while one of them says the opposite. That one poll is the outlier, because its results are way off from the rest of the results.
Statistically, this means that there’s a small chance that the one outlier poll is correct and all the other polls are wrong. There’s a far larger chance that the one outlier poll is wrong, and all the other polls are right. This all seems pretty obvious, right? And if you want the most meaningful look at what’s going on, you’ll take all these polls and average them together, either with a straight-up average like RealClearPolitics uses, or a weighted average based on each poll’s past accuracy like FiveThirtyEight uses.
The trouble comes when the media sees one outlier poll with a surprising result, and only reports that poll, for shock value. If you’re a TV pundit and you see nine new polls all saying the same thing, and one poll saying the opposite, and you report that one poll while pretending the other nine don’t exist, then you’re basically lying to your audience, right? Or maybe you’re just that clueless about how numbers work. Yet we see this from pundits all the time. So the public routinely hears the most suspicious and absurd poll numbers stated as if they were fact, and then when the results end up being something different, the public presumes the polls can’t be trusted. In reality, the polling averages usually tend to have been correct the entire time – but TV viewers never know it.
This brings us back to New Hampshire. As of yesterday there was one poll, from the Boston Globe, which showed that Pete Buttigieg had taken the lead by one point. But all the other new polls conducted in New Hampshire this week show Bernie Sanders still in the lead. Guess what’s happened now? Take a look at the latest numbers, screen capped from RealClearPolitics today:
You can see that the Boston Globe poll has been updated again, and it now shows Bernie being back in the lead. This isn’t surprising. Not only do outlier polls tend to be wrong, they often tend to quickly correct themselves. So now every poll says Bernie will win New Hampshire. Could they be wrong? Sure, but they’re probably correct. Any pundit who ran around yesterday touting the one poll showing Pete in the lead, while pretending the other polls didn’t exist, is looking pretty foolish right now. We see this happen on a national scale as well.
Over the past forty years, we’ve only had one election (in 2004) where the Democratic and Republican nominees both won Iowa and New Hampshire. The two states simply don’t have diverse populations representative of the Super Tuesday states. So it’s a mistake to presume that the winner(s) of Iowa and New Hampshire are the new frontrunners, or that they’re even in contention. Take a look at the latest results, again screen capped from RCP, in the national polling:
Here you can clearly see that – despite his poor finish in Iowa – Joe Biden is still the frontrunner for the nomination. The latest averages show that Biden is ahead of Bernie by five points nationally, and that Biden is ahead of Pete by twenty points nationally. Again, polls can change, and polls can be wrong to begin with. But if these numbers are correct and don’t change, Biden will be the nominee. If Biden’s national poll numbers finally drop from their months-long perch in the high twenties, or if someone surpasses him while his numbers remain the same, then he’ll be in trouble – and not until then.
Of course that doesn’t make for a ratings-friendly narrative, which is why the media is instead insisting that Biden is in deep trouble, and that Pete is on the verge of being the frontrunner. And because outlier polls always make for great ratings, some pundits are running with the one WSJ poll showing Bernie in the lead nationally, when every other new polls says it’s Biden. Bernie isn’t ahead nationally, just like Pete isn’t ahead in New Hampshire. Again, the polls aren’t the problem here. Stick to the averages and you’re fine. The problem is how the media reports, or more often misreports, the polls.
Bill Palmer is the publisher of the political news outlet Palmer Report