Women live longer than men on average, meaning they can vote more often on average. Credit…Arin Yoon for The New York Times
Watch out: Women outlive men
You may be partly joking, Mary, but this is probably a minor reason Democrats do a bit better among older voters than people might guess!
Why? American women, who tend to support Democrats, live almost six years longer on average than men. Women make up 55 percent of registered voters over age 65 — including 58 percent of those over age 80 — according to data from L2, a political data firm. In comparison, women are 52 percent of registered voters under 65.
I know all of this is a little morbid, but longevity strikes me as an underexplored dimension of electoral trends nowadays. We know higher life expectancy is correlated with socioeconomic status and tends to be higher in Democratic-leaning areas. Could this be a factor in why Democrats are performing better among older voters than usually thought? I think so.
Are you sure these people exist?
It would be a longer list than you might think. More than 20 million people with a college degree voted for Donald J. Trump in 2020. In our last New York Times/Siena College poll, 13 percent of likely Republican primary voters were self-identified moderates or liberals with a college degree.
There’s a bigger lesson here: A small percentage of a huge group can still yield a large number of people. To take another example: There are more Republicans in California than in any other state. There are more Republicans in Brooklyn than in Wyoming, the state where Mr. Trump fared best.
If not Biden, who?
Well, they’re not sure. In an open-ended survey question, no alternative candidate earns any meaningful amount of support from Democratic voters. And I don’t think that’s entirely unreasonable, given there aren’t any mainstream Democrats running against President Biden.
To me, the interesting question is whether many of these voters would wind up preferring Mr. Biden if alternatives like Gov. Gretchen Whitmer of Michigan or Gov. Gavin Newsom of California actually ran. It seems quite possible.
Just forget about national polling?
I’m sympathetic to the general sentiment here, Tim. Over the years, we’ve done many more Times/Siena polls in the battleground states than nationwide.
But I wouldn’t go so far as to say that national polling is misleading. The difference between the national vote and the battleground states isn’t that large — and might even be shrinking.
There are advantages to national polling as well. There are more historical questions for comparison. It’s much less expensive than battleground polling: It might take six state polls to get a decent picture. And there are plenty of cases — say, a Republican presidential primary or a battle for control of the House — where the national picture is much more relevant than the core battlegrounds.
Polling and nonbinary people
It’s worth adding some historical context. First, most pollsters typically asked whether someone was male or female — which is to say someone’s “sex,” not gender. That’s what the Census Bureau does as well, and pollsters generally find it advantageous to have their questions align with the census for comparison or even statistical adjustment. And as someone who loves historical data, I’m also always loath to lose a consistent measurement of something over time.
Second, you might be surprised to learn that many telephone pollsters haven’t actually been asking about the sex or gender of respondents. Instead, many have relied on the interviewer to record the respondent’s sex or gender based on voice. That might seem strange, but many respondents find it strange or even offensive to be asked if they’re a man or a woman.
Nonetheless, this is an area where survey research is evolving. In the last decade or so, many pollsters have started asking about gender. A smaller number of pollsters have offered respondents options beyond “male” or “female” or “man” or “woman,” though this is complicated in its own right. Respondents could identify in any number of ways, whether as transgender, non-cisgender, nonbinary, gender fluid, queer or something else. They could identify as a “man” or a “woman” to reflect a gender that does not align with the sex they were assigned at birth.
There’s another issue with adding small categories: measurement error. If one in every 300 respondents is trolling, or if one in 300 interviewers mistakenly clicks the wrong gender button, this mismeasured 0.3 percent of the sample will have no discernible effect on our results among men and women. But it could make up a huge share of the tiny number of transgender respondents.
In our most recent Times/Siena poll, “male” and “female” were the only explicitly listed options when we asked about gender. But if respondents said they identified in some other way, the interviewer would record it. In the end, we had three respondents who said they were transgender or nonbinary. This sample was too small for us to report. I’m not sure whether we — or anyone else — is handling this exactly right; I expect the industry to continue to experiment and evolve.
If you’re looking for a niche explanation for recent polling errors in Wisconsin, this is an interesting place to start.
Unlike with most states, pollsters have very little data on the partisanship of Wisconsin respondents, making it much harder to ensure an unbiased sample.
The absence of party registration is the best example, but the issue runs deeper. We don’t have data on whether our respondents participated in a partisan primary (like voting in a Republican presidential race). In most of the states without party registration, this primary participation data is a decent alternative.
The results by precinct aren’t very helpful, either. Outside of Madison and Milwaukee, very few voters live in overwhelmingly blue or red precincts. Even the most Republican counties in Wisconsin aren’t so Republican that we can be especially confident that an individual respondent will be a Trump supporter.
One of our major goals in recently collecting more data in Wisconsin was to improve our ability to estimate whether someone was a Democrat or a Republican, based on the relatively limited data at our disposal. I wouldn’t say we’ve found anything revolutionary: There’s just no substitute for knowing whether someone is registered as a Democrat or a Republican.
Don’t forget Arizona
Chris, I think we deserve a little bit of credit! In 2020, Arizona was one of our core six battleground states in Times/Siena polling. We surveyed it five times during the cycle, tied for the most of any state. We surveyed it in 2018 and 2022 as well, something that can be said only of Arizona and Nevada.
Heading into 2024, Arizona remains in the top tier. We’ll poll it just as much as Wisconsin, Michigan, Pennsylvania and Georgia.
Six kinds of Democrats?
You’ll most likely have to wait until 2028! In the meantime, consider reading our breakdown of Democratic voters from 2019. There were five types of Democrats in that analysis, based on data from the Hidden Tribes project: progressive activists, traditional liberals, passive liberals, moderates and the politically disengaged.
Another option to get you through until the next Democratic primary: Pew Research’s 2021 typology, which identified four Democratic-leaning groups: progressive left, establishment liberals, Democratic mainstays and the outsider left.