Reader Mailbag: Bots and Other Business
Voters in College Park, Md., in November. There are still some things to review about the midterm elections. Credit…T.J. Kirkpatrick for The New York Times
We’ve gotten a lot of great questions and feedback over the last few months, and I think we’re inclined to make a mailbag a regular feature, especially during the off-season. If you have a question, send it to [email protected]
Bots in polling
Let’s start with a question on an issue that I’d already been mulling:
This is a serious question. A Pew Research study a few years ago used a few simple techniques to find that 4 percent to 7 percent of respondents to nonprobability surveys were “bogus,” to use their term. These “bogus” respondents took the survey multiple times, took the poll too quickly, stated that they lived outside the United States, or offered nonsensical answers in open-ended questioning.
Anecdotally, I’ve heard more complaints about these kinds of issues from pollsters. I routinely hear about pollsters who toss as many as half of their respondents for data quality issues. It is reasonable to assume that the better-known pollsters take these challenges seriously, but we don’t know much about their practices, let alone whether they work.
Could this be part of Donald J. Trump’s strength in online panel polls? It’s hard to prove, but one emerging pattern has caught my attention: Mr. Trump has fared quite poorly in two online nonprobability polls linked to voter registration files, which require the panelist to provide personal information that matches a real registered voter.
Or put differently: Mr. Trump has done quite poorly in the nonprobability polls where there’s an extra layer of confidence that the respondents are human. It’s the pattern we would expect if bogus respondents were a factor.
Still, I wouldn’t get too excited about this emerging pattern if you’re a Ron DeSantis fan. It’s just two polls, after all. And Mr. Trump’s poll numbers have gone up over the last few weeks, including in telephone polling. So perhaps these voter-file-matched polls are really the outliers. As we’ve written before, voter-file-matched polls like these can be biased in other ways.
I’ve asked a few smart and well-positioned pollsters to look into the “bogus” respondent question in their own data. If you’re a similarly well-positioned pollster interested in digging in, you know where to find me.
The tilt of independent voters in the midterms
We’ve written that Republicans enjoyed a pretty meaningful turnout advantage in the midterm elections, but a few of you thought we overlooked one important group in this analysis: independents.
Democrats did quite well among independent voters, and that’s partly why the party held up despite a Republican turnout advantage.
Nationwide, self-identified independent voters backed Democrats by two percentage points, according to the network exit polls. It doesn’t sound so impressive, but it is by the standards of recent midterms, when the president’s party has tended to lose independent voters by double digits.
And in the high-profile statewide contests — think Senate races in Pennsylvania and Arizona — the Democratic candidates fared even better among independent voters than that, allowing Democrats to win marquee races by unusually wide margins.
Black turnout before Obama
On a related midterm note, we observed that the Black share of the electorate was the lowest it has been since Barack Obama won the presidency in 2008. That left an important question unaddressed:
It depends a bit on how you measure it.
By the simplest measure — the proportion of Black adult citizens who vote — Black turnout actually remains slightly higher than it was before Mr. Obama’s historic campaign in 2008.
The catch, however, is that turnout is higher across the board than it was before 2008. As a result, the racial gap in turnout in 2020 — that is, the difference between the share of eligible Black voters who turn out, compared with the share of eligible white voters who do — returned almost exactly to what it looked like in the 2004 presidential election. Based on what we’ve seen so far, the racial turnout gap in 2022 will probably look like 2006 or 2002.
There’s one last twist to consider: the Black share of the electorate. Oddly enough, the Black share of the electorate has usually remained above pre-Obama levels, even though the gap between white and Black turnout is basically the same as it was back then. That’s because white voters, who turn out at higher rates than Black voters, have dipped as a share of eligible voters, while Hispanic and Asian voters, who turn out at lower rates than Black voters, have increased their share of eligible voters.
Put it all together and the Black share of the electorate in 2020 was about halfway between where it was in 2004 and 2012.
Alternatives to Biden
We haven’t done much to touch on the possibility of a Democratic primary, but that hasn’t stopped many of you from asking about it:
Whoever wins the Democratic nomination will at least stand a chance, and probably more than a chance if Mr. Trump is the nominee. Whether there’s a Democrat who would do as well as or better than President Biden is a slightly different question, and it’s really quite hard to say. I hope you don’t ask me that next time!
Believe it or not, the most frequent question we received was about this sentence I wrote:
Here’s one (nice) version of the question it raised:
The answer is actually quite mundane. The Times uses people’s full names when they’re mentioned for the first time, but they typically receive courtesy titles, like Mr. and Ms., when mentioned again. In this case, Hillary Clinton had already been mentioned, but the male presidential candidates were being mentioned for the first time. Consequently, Mrs. Clinton, who prefers Mrs. to Ms., received her courtesy title while the male candidates got their full names.
I see how it ended up looking a little weird in this case, so we adjusted the sentence after publication.