This is our last dive into the mailbag before the election. But first I wanted to mention the news of the last 10 days.
This may be a “newsletter,” but we don’t usually write very much about what’s happening in the news. That’s because most of what happens from day to day doesn’t usually have a lasting effect on the campaign. The news cycle moves on, and so do voters. People have short memories. Remember when Donald J. Trump was convicted of a felony?
But with less than a week until the election, many less engaged voters are tuning in and making up their minds — right now. The news today can make a difference, and the recent news cycle has been driven by two items that haven’t been especially favorable to Mr. Trump.
First, John Kelly, his former chief of staff, said Mr. Trump fit the definition of a fascist, naturally turning the political conversation toward Mr. Trump’s conduct on Jan. 6 and the issue of preserving democracy, where Kamala Harris has the advantage.
Second, a comedian, in the context of a political rally for Mr. Trump at Madison Square Garden, referred to Puerto Rico as an “island of garbage,” sparking a backlash on social media and reminding many voters why they didn’t like Mr. Trump.
These events are drawing attention to some of Mr. Trump’s greatest liabilities at the most important part of the cycle. It’s hard to say whether it will make a difference, but the possibility ought to be taken seriously given the timing and the very close race.
Can you extrapolate meaning from early voting numbers?
This is a question I get often. In this case, a reader, Richard Robbins, asks:
Is there anything that might be learned from looking at early voting numbers, such as by comparing them to previous years?
I think it’s hard to glean too much from early voting. A vote cast on Election Day counts just as much as one cast early. It’s not easy to make a comparison to previous years either.
For one, the last presidential election was held during a pandemic. This led millions of voters to cast ballots by mail, and there’s every indication that mail voting will decline this time. In the states that had huge increases in pandemic-era mail voting — Pennsylvania, Michigan and Wisconsin, in particular — comparing today’s mail vote totals with those from four years ago is not a very good idea.
Second, the rules change from cycle to cycle. Nevada, for instance, now sends everyone a mail ballot, whereas it was almost all in-person early voting before the pandemic. There are some states where the voting in 2024 is fundamentally similar to, say, 2016 — Georgia, North Carolina, Arizona and longtime vote-by-mail states like Washington and Oregon — but there aren’t many.
Third, the campaigns change their emphasis. Democrats usually do best in early voting, but this wasn’t inevitable. It happened because the Obama campaign and subsequent Democrats promoted and emphasized early voting. This cycle, Republicans have made at least some effort to compete in early and mail voting, while I don’t hear nearly as many Democrats telling people to vote early (perhaps because of how Mr. Trump targeted such voting methods to try to subvert the result of the last election).
Fourth, most of the people who vote early are highly regular voters. When one party does better or worse in early voting, it mostly reflects which side’s most engaged voters are likeliest to decide to participate well ahead of the election. This can be indicative in a low-turnout election, but most highly engaged voters will turn out in a presidential election, one way or another.
All that said, I do think there are narrow opportunities to learn something about the level of turnout or the makeup of the electorate, especially in the states where early voting has been around for a long time, where the rules haven’t changed much, and where most people will vote early (like Georgia, North Carolina, Arizona, Oregon, Washington).
To this point, most of the evidence from those states suggests that the Black share of the electorate might drop and that the overall turnout might drop as well. Republicans are participating more vigorously than usual (even compared with prepandemic elections) in early voting. These are very tentative conclusions; the next week or Election Day voting could cancel it out.
Even if those limited inferences are correct, it still doesn’t tell you much about who will win the election. Even if you told me exactly who would vote in November, I’m still not sure I would be able to make a confident prediction unless the partisan turnout was clearly lopsided, which doesn’t usually happen in battleground states in a presidential contest.
How do we weight on gender?
The makeup of our polls is mostly determined by two things: the demographic makeup of registered voters (based on actual voter registration records), and the historical turnout of each demographic group.
Jon Staudt wants to know what that means for gender, specifically:
When you model your polls, what target do you shoot for as a percentage of men voting and as a percentage of women voting?
This varies a bit by state and race: Black voters, in particular, are disproportionately women, so women will make up a larger share of the electorate in states where Black voters also make up a larger share of the population.
Overall, we expect that women will make up 53 percent of the electorate nationwide. In Georgia, where Black voters might make up around 30 percent of the electorate, 55 percent of voters are expected to be women.
How bad can the polls be?
The polls have had some high-profile misfires in recent cycles, but it’s not really unusual over the longer course of history. Karen Raymond wants to know an example of when polls were truly terrible.
Let’s go way back in time when polls were exclusively by landline phone. Was there ever a poll that was wildly wrong, completely underestimating the winner, and why was that? What were the conditions and lessons?
1980 is a good one. The polls showed a close race between Jimmy Carter and Ronald Reagan, and Reagan ultimately won in a landslide.
What happened? Well, a few things.
First, in hindsight, Reagan did actually lead in the polls throughout the 1980 campaign. That’s not necessarily how it was understood at the time, but if there were polling averages, they would have shown a modest Reagan lead.
Second, undecided voters appeared to break for Reagan. The final polls showed him picking up ground after the final debate — in which he delivered his famous “are you better off than you were four years ago” line. Studies that recontacted polling respondents found signs that voters had shifted toward Reagan. The Carter campaign’s polling also showed a collapse over the final stretch.
Finally, at least some polls found that failing to account for household size was a contributor to the polling error. Most households had only one landline, but some households had one person while others might have a full family. Single-person households were likelier to be unmarried and Democratic-leaning, so a poll that didn’t account for household size would have had too many unmarried lower-income voters, who leaned Democratic.
The Amish
As someone who routinely drives along the Susquehanna River through Pennsylvania — which means seeing the Amish on horse and buggy — I do appreciate this question:
Hello! I live in Pennsylvania (Pittsburgh area). How do you factor in the significant Amish voting bloc when calculating polls in Pennsylvania, Ohio, etc.?
Are they Trump supporters? Are they appalled by his personal behavior, and therefore likely to not vote for president, while pulling the handle for down-ballot Republicans? How do you factor in this population? Thanks! — Emily Viehland
Needless to say, this is a tough-to-reach population. Realistically, they do not take polls.
Can the polls account for them, anyway? It depends on the extent that the Amish population is very different from voters with the same characteristics — like their party, the partisanship of their precinct or educational attainment.
Imagine two groups of unreachable voters, each hypothetically 100 percent for Mr. Trump: one group registered as Republican and living in precincts that voted for Mr. Trump; the other living in swing precincts and not registering with a party.
In the former case, the polls might be fine: Yes, we might not reach this particular group of Republicans, but the poll is weighted by party and consequently we will basically just speak to other Republicans in their place. In the latter case, the polls have a problem: We will miss a super-Republican group, and basically replace them with a bunch of swing voters.
The Amish vote in Pennsylvania is probably close to the safer first case. They appear to be overwhelmingly registered as Republicans and live in overwhelmingly Republican precincts. There’s still a risk that they’re systematically more Republican than the others who might replace them in a poll, but they’re a small enough share of the population that this minor difference is probably not material to the state as a whole.
The Republican-leaning Cuban vote in Florida, by the way, is a version of the second problem for pollsters. If you don’t get enough hard-to-reach older Spanish-speaking Cubans in Miami-Dade, you might replace them with English-speaking Hispanic voters from Orlando and Tampa — who outright support Democrats.
How does recall vote really work?
We’ve written a lot about recall vote in polling, and in doing so we’ve glossed over one important technical question: Do pollsters ever try to account for changes in the makeup of the electorate, or who actually voted (as opposed to saying they voted) in the last election? James Devitt asks:
I may have overlooked this, but how does recall-vote weighting figure in those who didn’t vote in the last election? (Mostly voters under 18 in 2020 who are now of voting age.) I assume this is done in some fashion?
Most public pollsters who weight on recall vote simply don’t account for changes in the electorate. They take a poll of, say, Florida. They look up the result of the election in Florida in 2020 (Trump +3.4), then weight the people who say they voted for Biden or Trump in 2020 to match that result. That’s it.
There are some pollsters who weight on recall vote and who try to do more. They might weight only the people with a validated record of voting in a state (which would essentially allow the partisan makeup of any potential new voters to “float”). They might even try to model how new voters who were eligible to vote would have voted in 2020, and then adjust their targets to account for changes in the makeup of the electorate. That said, this is not what’s happening among public pollsters.
Split ticket voting?
You might think knowing the turnout in advance is enough to know who will win, but that’s absolutely not the case. In Arizona, for instance, the same R +8 midterm electorate voted for Republicans for U.S. House, narrowly voted for a Democrat for governor, and voted for a Democrat for U.S. Senate by five points.
John Hendel has a very narrow question about a specific kind of ticket splitting:
Some people have said voters consciously split their ticket to try to assure a different party holds the executive branch while the other controls the legislative branch. Personally, I don’t think this happens — people tend to vote a party line or perhaps pick and choose individuals but not with that underlying intent of balancing power. Do you find this happens?
It does! There’s evidence that a small but potentially significant group of voters support divided government and cast ballots with the specific intent of balancing the party in power.
Possible reasons for Harris’s solid polling in Northern swing states
We’ve written that Ms. Harris is holding her own in Michigan, Wisconsin and Pennsylvania despite Trump gains in noncompetitive areas, potentially reducing Republicans’ Electoral College edge. Gary Hirsch asks:
“Could this be the result of a focused campaign in these states with heavy in-state appearances and overwhelming ads and door-knocking? Do campaigns make a difference?”
It certainly could be. I’ll also note that Ms. Harris is still not especially well known nationwide, relatively speaking, and her campaign has spent a lot on positive advertisements in the battlegrounds. Maybe it’s made a difference.
One other factor I’ve wondered about: whether Mr. Trump’s campaign to overturn the election left a deeper mark on Republican-leaning Biden voters in these states. It’s their vote that he didn’t respect, after all.
The post What We Can (and Cannot) Learn From the Early Votes So Far appeared first on New York Times.