This graph from Minnesota Public Radio is making the rounds on social media today, as it’s … provocative. There’s a dramatic effect here. The question is, what’s the cause?

Did the mass protests against racism and police brutality seed an outbreak? It’d be weird if they didn’t given the scale of them, but the difficulty in saying yes definitively is that Minnesota (and other states) also reopened for business as the protests raged. Which was more of a factor, commerce or the demonstrations? Is there any way to tell?

The first question to ask is whether there’s a spike in cases in Minnesota. We need to know that there’s an outbreak before we start trying to tease out what’s driving it. But if we look at the state’s data, we find that cases are stable, not rising:

There was a small one-day spike last week but the recent data doesn’t like much different from how it looked in early June. What about the positivity rate? Can that tell us anything? Well…

The positivity rate has doubled since about two weeks ago, a sign that there are more infections now in Minnesota. But the doubling was from a very low rate of 1.6 percent to a still-low rate of 3.2 percent now. To put that in perspective, the latest positivity rate in Arizona was 23.9 percent. If there’s an outbreak happening up north, it’s not a big one.

I haven’t seen any COVID data about protesters in Minnesota specifically over the past week but this was published on June 18. It may be outdated now, as more people have had time to demonstrate and potentially infect each other since then. And some people who were infected earlier this month may have only gotten around to being tested after June 18. But for what it’s worth, here was the picture 11 days ago:

Of the 3,200 people [who had been to protests in Minnesota] tested so far at the four popup sites across the metro, 1.8 percent have tested positive for Covid-19, says Ehresmann. HealthPartners, one of the largest health care providers in Minnesota, also reported to the state that it had tested about 8,500 people who indicated that attendance at a mass gathering was the reason they wanted a test. Among them, 0.99 percent tested positive. These numbers have been one of the few pleasant surprises since the outbreak began, says Ehresmann. “Right now, with the data available to us, it appears there was very little transmission at protest events,” she says. “We’re just absolutely relieved.”

In a handful of other US cities that have rolled out free testing for protest-goers, the first round of results look similarly encouraging. In Seattle, fewer than 1 percent of the 3,000 people tested after attending protests were positive for coronavirus, according to a statement put out by the city’s mayor last Friday. This week, Boston officials announced that 14 out of 1,288 people tested so far were positive for coronavirus, or 1.1 percent.

Maybe the numbers have changed by now. It’s worth noting that the positivity rate in Washington state has nearly doubled over the past two weeks, but in Massachusetts it’s lower now than it was then. New York, another site of mass protests, continues to see rock-bottom positivity. And here’s what the rate looks like in D.C., which has hosted plenty of protests:

If there’s a “protest effect” in coronavirus spikes, it’s not showing up consistently in the rates across these protest-heavy districts. It could be that there’s a “hidden” outbreak among protesters, whether because they’re less likely to get tested or because there’s a corresponding decline in cases among older people that’s offsetting an increase among younger adults to keep the overall number of cases statewide flat. That would make some sense: Older adults are more likely to self-isolate because of the greater risk posed to them by COVID-19 whereas younger adults are more likely to venture out and take risks by socializing.

But in that case it’s worth noting that the “young increase, old decrease” effect isn’t netting out in hot-spot states like Arizona, Florida, and Texas. Cases overall there are soaring because the number of infections among young adults is widely outpacing the reduction in cases among older people. If there’s a “protest effect” in Minnesota, with rampant infections among demonstrators, why aren’t we seeing an overall surge in cases a la Texas?

Could early reopening be causing the spikes instead? Here’s a fascinating possible correlation identified by an analyst at JP Morgan:

Analyst Jesse Edgerton analyzed data from 30 million Chase credit and debit cardholders and from Johns Hopkins University’s case tracker. He found that increased restaurant spending in a state predicted a rise in new infections there three weeks later.

He also said restaurant spending was the strongest predictor across all categories of card spending…

Edgerton said in-person restaurant spending was “particularly predictive.”

That’s exactly what we’d expect based on our (still-evolving) understanding of the virus. If it’s true that masks combined with completely ventilated outdoor spaces make transmission difficult and that indoor spaces where masks need to be removed make it comparatively easy, then we’d expect restaurants to drive outbreaks more than protests. Greg Abbott, the governor of Texas, seems to believe strongly that it’s bars that have driven the soaring outbreak there, going so far as to say recently, “If I could go back and redo anything, it probably would have been to slow down the opening of bars, now seeing in the aftermath of how quickly the coronavirus spread in the bar setting.” Bars seem like an even more obvious petri dish than restaurants since they’re not only indoors and offer a product that has to be consumed without masks but they’re more likely to attract a young, risk-taking clientele. (Gyms and nightclubs are in the same boat, unless people are working out with their mouths and noses covered.) It’s going to be a rough second half of 2020 for those industries as other governors absorb the hard lessons Texas and Florida are learning right now about them.

To bring this post full circle, Minnesota entered Phase II of its reopening on June 1, allowing outdoor-only service for bars and restaurants. It moved to Phase III, allowing indoor service with limited capacity, on June 10. How/whether that correlates with the surge among young adults in the graph up top is left to your own judgment. It’s possible, I guess, that Phase III caused an immediate burst of young people going to bars and getting infected, which began showing up a week later in the data. It’s also possible that reopening had little to do with it and that something else, like the protests, did. My guess is that, as each state relaxes restrictions, it sends a signal to the more adventurous segments of the population that it’s now safe(r) to socialize, and some of that socializing is destined to take place in high-risk environments (house parties, etc), if not necessarily in bars and restaurants specifically. Maybe Minnesota’s spike is coming in a week or two.

Here’s Dan Crenshaw on Fox today laying the blame for Texas’s outbreak squarely at the feet of the protesters. As I say, the governor doesn’t seem to agree and if there’s data supporting Crenshaw’s conclusion apart from the youthful trend among new cases, I’m unaware of it.