Monday Program Note (on Tuesday this week)
May 14, 2024
This week I am in Atlanta to attend the American Association for Public Opinion Research (AAPOR) conference. The newsletter’s weekly schedule will change while I am here. I plan to write on whatever interesting things I learn at the conference about this year’s polling. If you do not already follow me on Notes, please do so. I may be posting a lot there this week with shorter reactions to what’s happening in Atlanta.
I hope to hear some opinions on why this weekend’s NY Times/Siena poll seems to make no sense. There is one thing that needs to be noted: the headline focused on the margin among registered voters. Biden does better with likely voters, and that was true in this poll as well. That wasn’t hidden from the reader, but as I have noted before, the problem is how media outfits frame their polling stories. It’s in the headlines and graphics they use. Those are what sticks with most observers because even among the politically-engaged, few people read the entire story and dig into the poll itself.
There are plenty of things that don’t seem right about this poll. How well Trump is doing with young voters and voters of color seems implausible. It’s not that he might not be doing better among them, but the change we see in this and other polls makes little sense. Why is that? The honest answer is: we don’t know - because we can’t check it against actual election results yet. It’s possible there is a historically significant demographic shift going on in the electorate. I don’t buy it, but it is possible. I think we are seeing polling overstating Trump’s support.
We are still six months out from the election, and polling on presidential races has not been that reliable this far out. There is also a phenomenon where incumbents poll worse early on as voters appear to register dissatisfaction with the president, even if voters end up for voting for him. We saw that in the last three elections in which an incumbent ran. Trump was losing by double-digits to Biden in the polling for a long time, but the actual margin for Biden was 4.5 points. Obama was losing at times to Romney in 2012 and in the run-up to the election was barely eking out a one point advantage in the polling. He won by 3.9 points. Bush was for long stretches of the campaign losing to Kerry as well, and the race appeared to be a toss-up going into Election Day. Bush won by 2.5 points.
If we are seeing a polling “discount” for the incumbent similar to 2004 and 2012, then Trump may be winning (although it’s more likely a toss-up). If we are seeing a polling discount similar to 2020, then Biden is winning.1 The appeal of this hypothesis is that it would help explain why all four Democratic Senate candidates in the battleground states Siena surveyed are outpolling Biden. It would make more sense for some or all of these candidates to be struggling if Biden were actually losing in those states.
There is something else about this poll that makes me think we might be getting a very skewed sample: interview methodology. According to its methodological note, Siena reached its respondents by phone, whether landline or cell. At this point, I am not sure type of phone matters anymore. I assume it would be ridiculous to survey only people with landlines, but it would also be cost-prohibitive since so few people even have one anymore. There was a time that the prevailing wisdom was that you needed to mix in cell phones to reach young voters. Well, those young voters are now my age, and today’s young voters are not using their phones like old-fashioned telephones. To be fair, neither am I. I don’t answer any call that is not from my contacts - and even then I often just get back to the person by text.
The Siena poll did get a adequate share of young voters in its raw sample; in fact, it appears it had to weight the results in favor of older voters. However, the issue here is whether that share actually represents young voters. The same is true for voters of color in its sample. I think we know less about whether voters of color are answering their phones or not, we have plenty of evidence that young voters are very hard to reach on their phones, especially from an unknown number. Hopefully, I will get hear some ideas from pollsters this week about what they think is happening.
In 2020, the incumbent (Trump) was underperforming in the polls by at least six points. He still lost convincingly. If the same discount is impacting the incumbent this year (Biden) and he is underperforming by six points, then that would mean he is beating Trump right now by about his margin of victory in 2020.