Polls Aren't Always Accurate. Here's Why They Were Right in New Hampshire

Feb 11, 2016

Sanders supporters at a New Hampshire Rally

The polls had it right when it came to New Hampshire’s presidential primary results—for the most part, anyway. With just a few exceptions, the polls predicted that Donald Trump would win on the Republican side, and Vermont Senator Bernie Sanders would beat former Secretary of State Hillary Clinton by a wide margin.

But, historically the polls in New Hampshire haven’t been this accurate. So, what accounts for this increased accuracy? For an answer to that question, we turn to Steve Koczela, President of the MassINC Polling Group. 

Q: Steve, the polls pretty much nailed it, and historically they aren’t so accurate. So, what accounts for this increased accuracy? 

Right, I mean polls in primaries and caucuses do tend to be much more difficult than polls in general elections, and nowhere more so than Iowa and New Hampshire. The polls there have an especially tough job to do, which is figure out first who’s going to vote, which is harder in primaries and caucuses, and then look at a race before really much else has happened. You know, this isn’t a race where we’ve got a lot of other votes that have been turned in, or we have a lot of results that kind of give us a sense of how things are going. In a way, it’s picking a really tough race and polling it essentially in a vacuum.

So, hearts predict primaries more so than in a general election. What about compared to New Hampshire’s primary in 2008. How does the accuracy of the polls this time around compare to 8 years ago?

2008 was one of the toughest years for pollsters in New Hampshire. On the Democrat side, right before the primary it suddenly looked like Barack Obama had caught up to and overtaken Hillary Clinton, but then when the votes came in, Hillary Clinton still won. So, it was really a tough year for pollsters all around there. And it led to questions, in combination with some other sort of big misses of the margin in New Hampshire, as to whether or not the polls in New Hampshire were going to be off again this year. But, this year on the Republican side, even with the big messy field, they were within one or two points of every single candidate with one exception, which was Marco Rubio. They really did a very good job overall of predicting the final percent of each candidate, which is I’d say quite a remarkable thing given the history.

Why exactly did these polls do a better job of predicting it? Was it just that there were more polls and a larger sample size yielded better results?

I think that’s probably a part of it, but the interesting thing is that there’s not really an obvious explanation. Because, in the past, yes there were fewer polls, and some would say that was a good thing, but all of the polls that there were in 2008 showed Barack Obama leading Hillary Clinton. It wasn’t just that there weren’t enough and that they were kind of scattered all over the place and made it hard to see what was going to happen—none of them really showed what ended up happening. Whereas, this time, none of the polls really hit it perfectly, but on average they really did hit it pretty much perfectly. So, I don’t know if there really is an obvious explanation. Truth be told, polling is actually a lot harder now than it even was in 2008.

Why? Why do you think that is?

Way more people have cell phones now than they did back then. Cell phones are far more expensive to call than land lines. So, lots of pollsters will try to get away with calling as few as they possibly can to still have a defensible sample. But, you know, a lot of people don’t really use their landline, even if they still have one. So, polling’s harder for that reason. On a cellphone, a lot of people just don’t answer it if they don’t recognize the number. Caller I.D. is now universal, whereas it wasn’t in 2008. So, all of this is combined to mean that the response rate, which roughly means how many calls do you have to make in order to get the number of calls that you need, has gone way down, whereas a few decades ago it might have been 35%, 40%, 50% of people would respond to a poll. Now it’s more like 8%, or 9%, or 10% might respond to a poll, even if you exert a lot of effort to try to get people to respond.

Compare this to other states. Is there a reason to believe that states like South Carolina or Nevada, which have their contests coming up, will have less-accurate polls than the polls here in New Hampshire?

The thing about South Carolina, and really all rest of the states beyond Iowa and New Hampshire, is there’s been hardly any polling there. So, similar to candidate visits, similar to ads and ad money, polls pretty much focus on Iowa and New Hampshire for the months leading up to the first two contests. You know, there’s been a couple polls in South Carolina, maybe one in the Nevada caucuses in the last few months, but it’s not like Iowa, New Hampshire, where for the last three or four weeks you could look almost every day and see what the state of the race was right at that minute, and if you didn’t like it wait another few minutes. 

South Carolina and Nevada are not at all like that, and the Super Tuesday states even more so. You know, there hasn’t been a poll in Massachusetts, which votes in just three weeks for months now. So, I think we don’t really know whether or not the hit in New Hampshire is going to be how things go for the rest of the cycle, and we don’t really know whether what we saw in New Hampshire in terms of the actual outcome is what we should expect in the rest of the states. Because, we just haven’t looked in on them in a while.

So, what polls do you think voters should be watching for in the coming months?

Over the next couple days, you will start to see polling in the contests that are coming up in the next two weeks or so, but really I’d start to pay attention to the polls on March 1st and March 15th states. March 15th states I think are the most important, because they’re when the winner take all states start in the Republican side. So, it’s starting to look like we may have a crowded field for a bit longer, which will keep the contest going for a bit longer. And it will also mean that any candidate who wins one of the big winner take all states on March 15th could still close the gap in a big hurry.