As we approach the Republican and Democratic conventions, coverage of polling results has become more and more frenetic. Data on Trump’s major deficiencies among African Americans and Latinos is juxtaposed with his proximity to Clinton in a number of major national polls. It is endlessly exciting and hopelessly confusing. Honestly, though, you should probably ignore everything you see right now. In fact, you should probably ignore every poll until at least August.
The problem is that polls from May, June, and July have historically been off the mark. If you think about it, this makes a lot of sense. Often, one or both parties are still caught up in the internecine primary struggles, and it is rare for either party to have seriously been going after the other for any meaningful period of time at this point in the race. This certainly describes the current election: Clinton is still being dragged down by Sanders, and Trump “hasn’t even started on Hillary yet.” Therefore, there is too little data on where people are going to fall. We simply don’t know how many Sanders supporters will back Clinton, and we don’t know how many Republicans will defect to the Democrats. Many independents and moderates in both parties are still weighing their options, and the vast majority could still go either way. There is even a chance that a significant minority of Sanders supporters end up backing Trump. The point is that there is a lot of speculation, but there is also a dearth of meaningful evidence. Polls right now are fairly useless because they don’t capture the important questions. The interesting questions regarding how Sanders supporters and independents will vote have yet to be answered.
Indeed, the polls this year are even more confusing than usual, as the conservative American Enterprise Institute has pointed out. In short, the polls are shifting too much for them to possibly be accurate. To quote Norm Ornstein, “It went from a 13-point Clinton lead on May 4 to a tie just five days later. Six days after that, on May 15, Mrs. Clinton had a six-point edge. But an NBC/Survey Monkey poll showed a bare three-point margin for Mrs. Clinton nationally over Mr. Trump.” This makes absolutely no sense. There simply cannot be this much volatility in voters’ opinions. Obviously, voters’ views are fluid and constantly changing, but there is no way that people change their opinions by such large margins in such a short period of time. What’s even more puzzling is that there is no clear trend line in any of this data. It’s not as if there was a massive shift in voter support due to some major scandal breaking. Instead, the support just swings up and down over and over. Clearly, there are some methodological problems, then.
Ornstein has broken down a number of the polls, demonstrating their flaws. For example, the NBC poll seems to overestimate Trump’s support among Latinos by around 10%-12%, the Quinnipiac poll grossly oversamples whites, and the Reuters/Ipsos polls have been way too volatile to be believable. This is not to argue that every poll is wrong, although there is emerging scholarship suggesting that polling accuracy has seemed to decrease over the past few years. It is to say, however, that we should be very dubious about current polling data. Many polls have problems with sample size or demographics, and many of the key groups who will determine the outcome of the election (moderate Republicans, independents, and Bernie Sanders supporters) have yet to make up their minds. This election has been harder to predict that usual; both Trump and Sanders outperformed most analysts expectations. Thus, weaknesses in polling data during normal election cycles are only going to be magnified this time around.
The point is not to ignore everything you see this summer. Obviously, after the conventions are over and people start to make up their minds, the data will become more revealing. The point is that you should not be too excited or upset about any one poll. Often they are misleading, and even if they are quite accurate, they don’t reveal mid-term or long-term trends. Yes, Trump is leading Clinton in some polls, that does not mean he is winning the election. Yes, Clinton is demolishing Trump in other polls, that does not mean the election is going to be a blowout. As Ornstein explains, “When polling aficionados see results that seem surprising or unusual, the first instinct is to look under the hood at things like demographic and partisan distributions. When cable news hosts and talking heads see these kinds of results, they exult, report and analyze ad nauseam. Caveats or cautions are rarely included.” My advice is to not be like the talking heads on TV. They don’t know the exact methodology of the polls, and neither do you. Wait for more representative polls to be produced and aggregate trend lines to emerge. Only then will we have some idea of how things will turn out come November.