While we await further useful information about the state of the Justice Department’s investigation into *why* Donald Trump saw fit to abscond with top secret information, some other items of interests this week:
1)A new poll in the Florida Senate race shows Representative Val Demmings with a four-point lead over Senator Marco Rubio. In recent years, Florida is where Democratic hopes have gone to die. In 2018, for example, former governor Rick Scott won election to the Senate over three-term incumbent Bill Nelson, by just 10,000 votes out 8.5 million cast, less than two tenths of a percent. In the governor’s race that year, friend of the blog Ron DeSantis beat the highly touted Andrew Gillum by less than half a percentage point. Trump beat Hilary Clinton by less than two points in 2016, and Biden by around three points in 2020. Those outcomes all suggest that Florida is a competitive state. But it’s also one in which Republicans consistently win statewide office.
Demmings, a Black woman, is a former chief of the Orlando police. She was on the short list of Vice Presidential hopefuls for Biden in 2020 and her resume is impressive. But I mention this polling result not because I think it’s good news, but because it’s a useful illustration of how to think about and not think about polling. As a friend said to me today, if there is one thing we all ought to have learned from Nate Silver, it’s that one poll essentially means nothing. It’s only in the aggregate and over time that we can evenly plausibly glean useful predictions from polling about an upcoming election. In isolation, no single poll should really ever be newsworthy. Polls invite clicks, of course, including from yours truly. But single polls don’t tell us anything, really.
The aforementioned Florida poll has a notable feature. In Florida, something like 31% of adults over 25 have a four year college degree (the national rate is around 35%). One of the biggest issues in polling in recent years has been education weighting. Pollsters weight for race and gender, for example, to make sure their samples reasonably reflect the larger population they’re trying to understand. But after significant polling errors in 2016 that led to an undercounting of support for Trump, pollsters realized they had failed to adequately account for educational attainment among their respondents. This is a critical omission, since education level has become an increasingly significant dividing line between the two parties. Non-college whites, specifically, are increasingly voting Republican. So, if you don’t have enough of those folks in your polling sample, you’re likely to exaggerate support for the Democratic candidate.
The Florida poll above is from the University of Northern Florida. It’s a respected poll. But its sample in the Demmings-Rubio race this week raised some eyebrows. Why? Because 70% of its respondents identified themselves as having attended college. I haven’t been able to find exactly how they ask people about their college backgrounds, but it doesn’t look like they asked whether respondents had a four year degree. This is significant because while about a third of the over 25 population has completed a four year degree, a much higher proportion, closer to 60% have some college, meaning they took classes at a two-year or four-year institution.
What does this mean? If 70% of the sample had a four year degree, the poll would essentially be useless, so poorly would that reflect the underlying population from which the poll is trying to draw its inferences. If it’s some college, the poll still has an educational skew, though not as significant.
Regardless, this poll should be treated with skepticism. It’s one poll, as I said. Beyond that, even if the education profile of respondents in the poll doesn’t diverge wildly from the voting preferences of the underlying population the poll wants to assess, it appears that there is an education “bias” in the poll. I thought it might be helpful to share a bit of context about how to think about polling and what pollsters are trying to do to make their data as useful as possible. The UNF poll is a good way to illustrate some of the challenges to doing so.
2)It’s gotten very little attention, but President Biden appears to have dramatically reduced the American drone program. Ryan Cooper documented this development late last year, but it started pinging around Twitter this week when Cooper tweeted some updated data showing that the United States has sharply limited the number of strikes in Yemen, Syria, Iraq and Somalia in the past eighteen months. Last August, in the throes of the chaotic withdrawal from Afghanistan, the United States launched a horrific strike in Kabul that killed ten civilians, including seven children. American drone strikes have been killing civilians for decades, of course. What stood out about this particular strike was that it got far more attention than usual, eliciting strong criticism from some Democrats and advocacy groups. The Biden administration had clearly been rethinking drone policy anyway, and the Afghanistan bombing appears to have given that reappraisal a further push.
The drone program has been a standing temptation for American leaders to visit death and destruction on innocent people in parts of the world that, to be blunt, most Americans don’t give a shit about. So, from a political standpoint, there’s just been too little cost. For a generation, if a drone intended to take out a terrorist in addition, or instead, killed and maimed a family or two or three at a wedding party, a few people who pay attention to such things yelled about it. Otherwise, there was little to no domestic political downside.
According to the Costs of War Project at Brown University, the United States and its coalition allies killed an estimated 4000 or more civilians between 2006 and 2020. That included nearly a thousand killed in 2008 and 2009 alone, before the Obama administration tightened its rules of engagement. The result of those new rules was a decline to about 200 civilian deaths a year through the end of the Obama era. In 2017, Trump’s military command again significantly relaxed those rules of engagement and, as the Brown projects notes, the results were immediate, with over 1200 deaths combined just in 2018 and 2019 from drone strikes. Every intelligence assessment comes to the same conclusion - civilian deaths from drone strikes undermine the United States’ standing and serves as a powerful recruiting tool for the very forces we’re supposed to be suppressing. And drone strikes, however targeted they attempt to be, exact a human toll that, in effect, if not intent, imposes collective punishment on people who happen to live in proximity to those we’ve deemed legitimate targets.
The evidence so far, despite the massacre in Kabul last year, is that civilian casualties from drone strikes are far lower under Biden than they’ve been at any time in the past two decades. The drone program has been a moral disgrace and likely strategically counterproductive. If the Biden administration’s reticence to use drones persists, it will be a significant milestone. So far, it’s been an underreported one.
3)One final comment on Liz Cheney, who got walloped in her primary on Tuesday night, losing by nearly forty points. I’ve given her props before for what I consider to be the meaningful political sacrifice she made when she went after Trump. As she said Tuesday night, she won easily in 2020 and could have done so again had she maintained devotion to Trump. So, one more time, good for her. But she’s also long been part of the problem. As recently as 2019, Cheney and her dad hosted a million dollar fundraiser for Trump. Later that year, Democrats prepared to impeach Trump for the first time, because he secretly conditioned release of already approved funds to Ukraine on whether they would agree to dig up dirt on a potential political opponent. It was an egregious abuse of his office. Cheney responded by vowing to stand unified behind the President because:
we’re going to continue to stand up for the rule of law, and we are going to look forward to the day when we are back in the majority. And we certainly are not going to commit the kind of offenses to the Constitution and abuse of power that the Democrats have been engaged in now ever since they took over.
It’s laughable for Cheney to have contended that there could have been any compatibility between support for Trump’s egregious and serial lawlessness on the one hand and fealty to the Constitution on the other. This was all painfully obvious long before January 6, 2021.
Jonathan:
When I get home next week, I will send you an analysis of the 2016 presidential election exit polls. You will see that these polls were accurate in all states except for NC, PA, FL, WI, and OH. The polls were conducted the same way in every state.
Tom Henkel