I’m aware that people make claims of this sort every election, and to be clear I’m not claiming that he will exceed his poll numbers. Only that these two arguments seem reasonable to me.
And the important thing is that, as noted by made by 538.com and possibly others, overall Trump did not beat his poll numbers in primaries (he was over in some but under in others). So any argument that he will do so in the general election needs to be one that did not apply in primaries.
[ol]
[li]The “shy voter” argument. This has been discussed a lot, and refers to the notion that people are embarrassed to admit to pollsters that they’re planning to vote for Trump, and this deflates his numbers in polls. This has been discussed a lot, and Trump’s performance versus polls in primaries is generally brought up in the context of refuting this argument. But the counter – and this is courtesy of my (17 yo) son the other day – is that conservatives who despised Trump had other options in the primaries, and Trump’s votes would tend to be those who actually held him in high regard. In the general election there would be a lot more people who hold Trump in contempt but intend to vote for him anyway. So you could get more people who don’t want to admit that they will vote for him, but will.[/li][li]The “likely voter” issue. As the general election gets closer and closer, poll focus more and more on trying to assess who actually will vote. And one big part of the equation in making this assessment is whether a person has a history of voting in prior elections. I would suggest that this may break down in elections involving Trump. Much of his support is thought to come from people who are angry at the system as a whole, and think all politicians are worthless, other than Trump the anti-politician. So you could have a lot of people who never voted in prior elections, which featured two politicians, but will come out to vote for Trump in their first opportunity to vote for an anti-politician. It’s hard to imagine how the likely voter models used by pollsters might account for this, since Trump is an unprecedented phenomenon. And it’s worth noting in this regard that the LA Time polls have consistently been stronger for Trump than the consensus, and one reason for this (according to them) is that they do not eliminate people based on their past history of non-voting.[/li][/ol]
Again, this is not any sort of prediction, and if I had to bet my house I would go with the consensus of the polls. But it’s a bit wild-card, IMHO.
Trump is certainly inspiring people who have never voted before to come out and vote this time. But it works both ways: There are probably a lot of people who are coming to the polls just to vote for him, but there are also a lot of people who are coming to the polls just to vote against him.
That’s probably true, but I’m thinking that’s not the type of people who never vote. People really opposed to Trump would be people who are not that violently opposed to the status quo and would thus sometimes vote.
Point being that while it’s true that Trump would raise turnout for Clinton, that increased turnout would tend to be reflected in the “likely voter” polls more than the increased Trump turnout would be.
As noted, the LA Times people think this is one of the differences between their stronger Trump numbers and other polls, which suggests - since they can see the actual numbers - that people who think they will vote but haven’t done so in the past tend to be Trump supporters.
There’s the “shy Hillary” voter too - people, especially women, in households where the dominant personality supports Trump.
Item #1 I can attest to personally - canvassing Saturday, I had no fewer than two women do the following when I asked them if they’re voting for HRC/DT/JS/GS/UND:
Cast quick glance back towards interior of the house, turns to me and whispers “He’s voting for Trump. I’m voting for Hillary.” One woman even added “… and he thinks I’m voting for Trump.”
Frank Luntz, Republican Pollster: “Only 33% of husbands say their wife is voting for Hillary Clinton, but 45% of wives say they’re voting for Hillary Clinton.”
As regards #2, Nate Silver has discussed it a few times in articles. He points out that the “angry disillusioned first time voter” theory implies a larger than normal number of new registrations, which has not been observed.
I was going to make that point too - at no time in the past 15 months have we seen an upswing in Republican-lead voter registrations on any scale approaching what the Democrats have done. “Shy voters” need to be “registered voters” and that ship has effectively sailed. The electorate, for all intents and purposes, is already baked in - people wanting to join the game at this late date are shut out.
The reason he might do worse than his poll numbers is Hillary’s state of the art get out the vote efforts vs his non existent one. This one is an actual thing that exists as opposed to the two made up ones in his favor.
The likely voter theory is a poor one. The polls of registered voters favor Clinton. So an overly strict likely voter screen is far more likely to benefit Trump than Clinton. Moreover, early voting gives us a lot of insight into who is turning out, and on the whole Clinton seems to be the one outperforming 2012. Finally, there are many different kinds of likely voter screens. One of the most common is simply to ask people how likely they are to vote. They are not showing big internal divergence on these methods.
Shy voters are slightly more plausible. But there too it’s difficult to explain the early vote numbers if there are enough shy voters to matter in important states. And, as with LV screens, if there were a big shy voter contingent you would expect different polling methods to yield divergent results (i.e., lying to a live person vs. pushing a number for a robot). But the results have not been divergent enough to matter.
Trump may outperform his polls, but that will be because the polls haven’t captured late movement in the race. We can be pretty confident that the polls are not 6 points wrong because of sneaky voters or some methodological error.
I don’t know if that follows. It’s absurdedly easy to register to vote these days, and I think most people register when they become eligible because it’s exciting at that time, but become disillusioned later. Registrations don’t generally lapse, but participation in the process does.
In addition to the above, I think Democrats always have an advantage in registered voters because younger voters tend to be Democrats. (Also, because in many larger cities, the only meaningful elections are the Democratic primaries.) The Republican’s need to rely on people becoming more conservative as they age, but many of these people don’t bother to change their registrations.
I’ve not looked into this but I strongly suspect that most of these new registrations are just kids becoming eligible to vote for the first time, and are not really indicative of anything more than younger people tending to be Democrats, which is accounted for separately.
Depends on how you screen for likely voters. If you use only expressed interest, then it favors Trump. If you also use voting history, then it may understate Trump’s support.
If you read the link to the LA Times methodology (in post #3), they give this as one of the suggested for their poll’s divergence, and it’s certainly been divergent enough to matter.
Correct, but “it may understate Trump’s support” as compared to other likely voter screens. So we can actually test that hypothesis today using the data. Does Trump underperform in voter screens that examine voter history versus ones that are purely subjective? No, he does not. The converse is true.
As for the registration issue, how many (swing) states have same-day registration? Trump voters going to the polls and registering then voting would not be seen in the statistics up to now.
It exists in swing states that don’t have enough black people to threaten Republicans with the kind of surprise scenario you’re describing–i.e., OH, NH, CO, IA.
There’s really not any good reason to think there’s a big well of unregistered Trump supporters. Polls of all eligible voters are generally more liberal than polls of registered voters. Registration rates are lower among Democratic-leaning demographics than Republican-leaning ones. And even if there were some marginal propensity for Trump voters to register late, you’d still expect to see the beginning of that trend in early registrations and early voting. It’s hard to make that case in more than a couple of states.
The real place where Trump could have pulled enough unexpected votes to swing this was non-college whites who are registered. Their participation rates are historically quite low, so if he really goosed them then that could throw off the demographic models in the media polls. But there’s just no evidence that is happening, and there would be, since they are RVs.
IIRC the candidate that has slightly outperformed the polling in the last 4 elections going back to 2000 is what I would call the “safer” candidate (i.e. either the incumbent or the one who is a better known commodity).
In 2000, Al Gore outperformed the polling data and actually won the popular vote. In 2004, George W Bush outperformed polling to win that race. John McCain, though still slammed at the ballot box, slightly outperformed polling in 2008. And Obama outperformed polling in 2012.
The one thing that could confound the trend is that Trump supporters are fired up in ways I’ve never seen. They are going to march to the polls and bring their hunting buddies and bartenders with them. But still, history suggests that if anyone outperforms polls it’s probably going to be Hillary Clinton. Might be by a few tenths of a percent but that could be enough in a tight race.
His most “fired up” supporters are also the ones most likely to swallow his claims that the election is rigged for Clinton. Which can’t be helpful for turn out: if the election is rigged, why bother showing up?
I think the key question is whether Trump can mobilize a few percent of white working class disillusioned voters who don’t normally vote and and don’t answer polls and who register late in states where that is possible. Doesn’t seem implausible to me but hard to say how big an effect it will have.
You can also tell plausible stories about shy Hillary female voters and the impact of her superior ground game.
Overall it does seem to me that this election is “different” in a number of ways and this increases the uncertainty around polling. This uncertainty cuts both ways and increases the probability of both a narrow Trump win and a Hillary landslide but obviously the two are asymmetric in that the size of the win matters much less than who the winner is.
I think my subjective probability of a Trump win is around 15%, a bit above Upshot’s 10% and below 538’s 24%.