Covid surveys are one of many inputs into complex models of human behavior 📊 May 23, 2021
Partisan cues and random noise have mattered a lot more than the epidemiologists are probably happy with
This is my weekly data-driven newsletter on politics, polling, and democracy — and whatever else I’m thinking about this week. If you want more frequent posts on these subjects, please sign up for subscribers-only posts which come twice a week:
Covid surveys (and CDC advice) are one of many inputs into complex models of human behavior
Alternative title: Why I’m going to Popeyes and not inviting Anthony Fauci
I am sitting in a coffee shop for the first time in a year. This is one of many firsts I have had over the last week: first time inside a bar again, first time attending lunch reservations, and first time going to the gym unmasked. I feel slightly uneasy around other people but am overwhelmed with relief.
Polls show many people are having similar experiences right now. According to tracking from Morning Consult, 63% of adults say they currently feel comfortable dining indoors at restaurants. Masking among vaccinated individuals also continues to fall.
Epidemiologists are less sure, however, on what is safe to do. One survey of them from the New York Times found that only 30% felt comfortable eating indoors at a restaurant. Only 32% would eat at a friend's dinner party and only 45% would interact within 6 feet of someone else without a mask. The science is not conclusive, but the CDC has said vaccinated people can do these activities and not be worried about contracting or spreading the virus. Six percent said vaccinated adults should not socialize indoors, bucking the science, and 37% said people should not socialize with more than one other household of people at the same time — certainly more reserved than the official government guidance on the subject. The survey found 26% of epidemiologists would not go on a hike outdoors with friends, even though open-air transmission is [around 0.1% or lower — and that's for unvaccinated people!
A reader has found these puzzling, asking why we keep asking this overly-cautous group what they feel comfortable with:
Seriously?? It’s like asking a group of 828 people with severe OCD how we want to organize the pantry. And they completely downplay or overlook the main factor that any doctor or anyone with a sense of stats would tell you here: community spread and rate is the overwhelmingly important variable here. Don’t ask me “is it safe to go to a pool?” Ask me, “is it safe to go to this pool,” meaning, in this place, where the current Covid rate is X—in this neighborhood vs in this city in India right now? These are knowable things, and epidemiologists discredit themselves when they speak into a statistical vacuum with a bunch of generic warnings sound like total worry warts about everything. Of course there’s no zero risk other than total isolation but there are also ways to minimize risk to infinitesimally small, and they won’t speak to that; they take the easy way out and just say “don’t do it!” Driving a car any day of the week is more dangerous than a large majority of the things they are warning against here, when you get down to the level of the concrete in many places in the US right now.
Why aren’t they also asking doctors and statisticians and risk theorists and people who have something to say here, too? Epidemiologists are so myopic about this stuff because all they see is what they are trained to believe is going to happen, i.e. the outbreaks that they study.
This is a reasonable question! The reader lives in a small, outdoorsy town with a high vaccination rate. If he were translating the epidemiologists' advice, he should feel cautious (but not "worried") about taking his kid on a walk through the nearby forest with his neighbors. Twenty-six percent of epidemiologists wouldn't do that, so why should he?
Obviously, we don't know what the "true" risk is here. But one group has to be wrong; either the 63rd percentile person is closer to the truth and vaccinated individuals should feel comfortable dining indoors, or epidemiologists are closer and only 30% should.
This has got me thinking about bias and variance in sample surveys. Fans of election polling will know that any individual poll is likely to miss the mark from a variety of factors. But surveys can also be off _on average_, suffering from the same underlying biases in data collection. A national survey of epidemiologists, the reader points out, is going to be biased against activity regardless of (a) what community factors are in any given area and (b) what the emerging science is on activity. That's because they are primed to see the past year of uncertainty in what w know about covid, even if the _truth_ is not so uncertain. There is a lot we don't know about the disease, but we certainly know that nobody should feel uncomfortable going on an outdoors hike where the risk of transmission is ignorable.
The point, as I've been saying to a lot of people recently, is that advice from any official source is only one input into the complex models that produce human behavior. A survey of epidemiologists might make the signal from that input stronger, but it will not rule out the other factors. For more on this, see here and here.
Posts for subscribers
May 11: Yes, you can generally trust issue polling — but there’s a catch. A new study of polls on hundreds of state ballot initiatives reveals no bias towards liberals or conservatives, but rather a consistent underestimation of the status quo
May 20: The Immunity Treadmill: How covid-19 vaccine resistance could perpetually push herd immunity out of reach. Collective resistance depends on how many unvaccinated people have already been infected, how many of them have immunity, and where they live.
Plus the usual Sunday subscribers-only thread on the latest polls and the news.
Links to what I’m reading and writing
Please read this Economist piece from me on how violent racial justice protests last year (many of which instigated by the police) may have increased support for Donald Trump in some cities, such as Kenosha, Wisconsin.
Here is another Economist article on how attitudes toward race relations, racial justice and the police have changed since George Floyd was killed a year ago this week.
And a final Economist newsletter item on how American politics is becoming more divided by racial justice, rather than racial identity.
On the reading front, first see this Twitter thread about a humorous but serious error in Kahneman, Sibony and Sunstein’s new book NOISE, and then read this blog post from Andrew Gelman about its broader meaning. I still found the book insightful, if not terribly new. (Really, not new at all — review coming.)
Feedback / Share and subscribe!
That’s it for this week. Thanks so much for reading. If you have any feedback, please send it to me at this address — or respond directly to this email.
If you want more content, you can sign up for subscribers-only posts below. I’ll send you one or two extra articles each week, and you get access to a weekly gated thread for subscribers. As a reminder, I have cheaper subscriptions for students.
In the meantime, follow me on Twitter for related musings.