Chad Orzel asked for a regional look at the COVID-19 death rate in the United States. As it happens, this is a huge pain in the ass because the boffins at the COVID Tracking Project make their data accessible in a virtually unusable fashion. That’s why this is the one and only time you’ll ever see these charts (from me, anyway).

I would have preferred to put all five regions in one chart, but they were just too far apart. The Northeast peaked out at about 20 deaths per million while three of the other regions peaked out at only 3 deaths per million. Meanwhile, the Midwest was right in between:

The Northeast is declining after its mid-April peak, and the Midwest is declining too after its late-April peak.

The South and the Pacific Coast are both doing OK after peaking in early May. Alone among regions, the Mountain West has been steadily rising for over a month. This might not be the best time for them to eagerly start reopening their economies, but I don’t imagine they care what I think about it.

UPDATE: The bottom chart originally had the South and the Mountain West reversed. It’s now corrected, along with the text.

This chart shows the total number of people who died in 2018 in nine different developed countries:

Yikes! People are dying like flies in the United States. What’s going on? Obesity? Lung cancer? Deaths of despair?

Of course not. We have a lot of deaths because we have a large population. Here’s the death toll per million:

As it happens, the US is a little below average in deaths per capita. This is probably because our population is a little younger than the other countries, but in any case, the differences aren’t huge.

Nobody in their right mind would present the top chart as evidence of anything much. It’s obviously meaningless thanks to the large range of populations. If you want to study death rates, you need to look at deaths per capita. The same thing applies to deaths from COVID-19:

Our death total of 90,000 sounds bad, but it’s mostly because we’re a big country. If you look at COVID-19 deaths per capita, as you should, we’re about average among developed Western countries.

This might not always be true, of course. If other countries have the virus under control while we see a resurgence next month, then our death rate per capita could become one of the highest. But it isn’t yet.

This is a Sara orangetip, which looks kind of mothish to me but is apparently a butterfly. It’s got one of those Seurat-like patterns of black dots, which makes it look fuzzy and out-of-focus even when the image is sharp. Kittens and ducklings have this same superpower sometimes.

I mentioned the other day that all projections of coronavirus spread going forward are based on estimates of how well various social distancing restrictions work. For things like school closures, this is easy to measure: schools are either open or closed, and we know the dates of both. Other things are not so easy. You can open up restaurants, for example, but that only matters if people actually start coming back. Are they? Here’s some data from OpenTable:

Oklahoma is in a league by itself for some reason, but aside from that there’s about a dozen states where restaurant reservations have increased by 15-25 percentage points since May 1. Likewise, there are a dozen or so where there’s been essentially no increase at all. This real-life data is what we need in order to figure out how compliance with (in this case) restaurant closures correlates with COVID-19 cases a few weeks later.

Headlines can provide another sense of what’s happening now that social distancing restrictions are being relaxed:

Not so good! And I can report from my morning walk that I saw only one other person wearing a mask. This was outside on a non-crowded path, so I don’t think there’s any real harm, but two weeks ago more than half my fellow walkers were wearing masks. On the other hand, in my local supermarket mask-wearing is near universal—though most likely because the supermarket itself requires it.

Anyway, both data and anecdotes suggest to me that nothing surprising is going on. We know the dates that various state governments have lifted restrictions, but we have only weak evidence of how quickly people are returning to their old lifestyle. My best guess is that it takes a couple of weeks for lifestyle changes to get to the point where they’re causing more infections, and another couple of weeks for death rates to start increasing. That’s just an amateur guess, though.

As you probably know, a small number of people who have recovered from COVID-19 later test positive for the virus. The latest example of this was some sailors on the USS Theodore Roosevelt. Today, however, we got some good news on that front:

Scientists from the Korean Centers for Disease Control and Prevention studied 285 Covid-19 survivors who had tested positive for the coronavirus after their illness had apparently resolved, as indicated by a previous negative test result. The so-called re-positive patients weren’t found to have spread any lingering infection, and virus samples collected from them couldn’t be grown in culture, indicating the patients were shedding non-infectious or dead virus particles.

So once you recover, it’s safe to go out in public. What’s more, there’s little danger of relapse once your immune system has produced the antibodies necessary to kill the virus. Good news indeed.

Back when I first started charting the spread of coronavirus I decided not to use a logarithmic scale. I figured that log scales were fine for communicating with other professionals, but most laymen have no clue what a log scale is and have difficulty interpreting it even when they’re given an explanation.

Naturally, then, I’m thrilled to pass along a study from the LSE School of Public Policy that confirms I was right. They showed people the following two charts:

I think they put their finger on the scale by starting the y-axis of the log chart at 0.1, but I don’t suppose many people actually noticed that. In any case, here are the results:

We find that the group who read the information on a logarithmic scale has a much lower level of comprehension of the graph: only 40.66% of them could respond correctly to a basic question about the graph (whether there were more deaths in one week or another), contrasted to 83.79% of respondents on the linear scale. Moreover, people in the logarithmic group also proved to be worse at making predictions on the evolution of the pandemic: they predicted, on average, 71,250 deaths for a week after the experiment was taken, whereas the linear group predicted 63,429.

The first finding here is peculiar: since both charts feature lines that are steadily increasing, why would anyone have trouble saying if there were more deaths in one week vs. another? The later week will always have more. That’s odd.

The second finding, however, is the key weakness of a log chart: people have a hard time interpreting the scale. In the log chart, the final dot looks like it’s at around 60-70,000 deaths or so. It’s not, of course, because the the distance between 10K and 100K increases on a log scale just like the rest of the chart. But most people simply don’t understand that. The linear chart, by contrast, shows clearly that the final dot is at 40K and the death rate is growing quickly.

So how does this affect attitudes toward the COVID-19 pandemic?

First, we find that despite predicting a higher number of deaths, people who were shown the logarithmic scale chart declare to be less worried about the health crisis caused by the coronavirus.

Divergences, however, don’t stop there. The scale of the graph they see affects people’s responses concerning their policy preferences and stated behaviours. Ceteris paribus, respondents who see the information on the linear scale graphs support less strongly the policy of keeping non-essential businesses closed than those who look at the logarithmic one — although they also favour reopening them later. At the same time, those who see the linear graph are more willing to support a hypothetical state-level tax aimed at providing citizens with masks.

It’s not surprising that people shown the log chart are less worried about the virus. The log trendline simply doesn’t look like it’s growing that fast. It’s peculiar, though, that folks shown the linear chart are less supportive of keeping non-essential businesses closed. That’s hard to explain. In any case, I guess I’d like to see a complete breakout of the policy stuff so that it’s clearer how strong the effects are.

Bottom line: Ditch the log charts if you’re writing for a non-professional audience. They leave a seriously misleading impression.

Here’s the coronavirus death toll through May 18. Switzerland is down to 0.7 new deaths per million. Germany is down to 0.6. The United States recorded 785 new deaths, its lowest weekday number since the end of March.

The raw data from Johns Hopkins is here. The Public Health Agency of Sweden is here.

I’ve been asking for a while about the effect of specific COVID-19 countermeasures, and a new study in Health Affairs finally delivers. This is the first study of this type that I’ve seen, and it should naturally be taken as tentative until we see what other teams come up with. But with that said, here are the results:

The error bars in this chart are large, but the point estimates suggest that school closures and bans on large gatherings have no effect on reducing the spread of the virus. In both cases the effect is statistically insignificant, and in the case of school closures the effect is most likely to increase the spread of the virus.

Conversely, closing restaurants and issuing shelter-in-place orders both had statistically significant effects and both slowed the spread of the virus considerably.

There are, of course, several things that the study didn’t test. The most important is probably mask wearing. “Future work,” the authors say, “should also examine the impacts of other social distancing policies such as closing public parks and beaches, the requirement to wear masks in public, restrictions on visitors in nursing homes, state announcements of first cases or fatalities, and federal government actions such as prohibiting international travel.”

I’ll caution once again not to take these results as definitive yet, but further research along these lines is critical. If follow-up studies confirm the results on school closures, for example, it means we can send kids back to school in September. That would be a huge benefit for everyone.

In these plague-infested days we could all use a reminder of the vast expanse of the natural world where six-foot distancing is no problem. And what better place than the Grand Canyon? These four pictures are all panoramic shots, each comprising eight or nine pieces stitched together in Photoshop. From top to bottom they were taken from: Desert View Watchtower; Lipan Point; Moran Point; and Maricopa Point.

Here is the result of a poll taken a few days ago by YouGov for CBS News:

Roughly speaking, it looks like about 70 percent of Americans wear masks routinely when they leave the house. That’s not bad, and a number that high normalizes the behavior enough that it’s likely to increase over time.

This was on my mind because I was thinking about whether we’re headed for a big increase in COVID-19 cases thanks to the relaxation of social distancing restrictions. I think this is pretty likely, but if it doesn’t happen it will most likely be due to widespread mask wearing. Even as we ease off of other restrictions, mask wearing is the one thing that everyone still agrees about.

I hope this is something that’s currently under intense study, both here and in other countries. We need a better handle on which specific measures are most effective at stopping the spread of the virus, and it would certainly be useful to know just how effective mask wearing is even if you do nothing else.

Now just imagine how close to 100 percent we could get if our president and all our governors got on board…