It’s easy to think back on the past two months and assume that Biden’s polling began to take off as a new wave of coronavirus erupted across the south and California. It’s easy because it’s largely true. Data nerds have pointed out that voters’ views of the pandemic correlate closely with their preferences for president; the worse the country’s doing with COVID at a given moment, the longer Trump’s odds at reelection seem to be. That’s especially true, I think, because Trump was so outspoken about prioritizing reopening over containing the virus, which placed him at odds with most Americans. According to this new WSJ story, he and his advisors have finally come around to that view themselves, which explains their pivot this week. The coronavirus briefings are back; Trump is all-in on masks; the Jacksonville convention is canceled lest it endanger anyone’s health.

Safety first is the new message. The polling trends left them no choice.

But the “as goes coronavirus, so goes Trump” theory is a little too cute. It explains a lot but it doesn’t explain everything. In particular, it doesn’t explain why Trump’s polling began to tank in early June even though the surge in cases in new hot-spot states didn’t begin until late June. Look closely at the dates in this graph to see when cases nationally began to tick upward…

…and compare it to the dates in this graph to see when the polling gap between Biden and Trump began to widen:

America briefly seemed to be on the right track with COVID-19 even as the president’s numbers began to decline. The southern “second wave” may have cemented Biden’s lead at an eight- or nine-point margin, but it clearly didn’t cause it.

So what caused it?

Vanity Fair has data from a Democratic firm suggesting that the George Floyd protests and Trump’s response to them are what drove the change. In particular, says author Peter Hamby, the fact that the protests calmed down after a violent beginning and turned into a mostly peaceful series of demonstrations was crucial in winning over public opinion:

Avalanche found resounding support for the protests not just among Biden supporters, but among persuadable voters and even soft Trump supporters. The hardcore Vote Trump respondents were against the protests, with 56% opposing them. But among the softer Lean Trump set, an eye-opening 59% said the protesters were “completely right” or “somewhat right”—probably not what the president had in mind when he commandeered Lafayette Square. And 72% of Americans with Mixed Feelings about the presidential race—precious undecided voters—said the protesters were right too. “There’s not a lot of issues where you get even a strong majority of Americans on the same page,” said Michiah Prull, the CEO of Avalanche. “It speaks to that historic moment, and it speaks to a degree of national alignment on something that’s honestly pretty rare these days.”

But just as remarkable were the shifts among those persuadables in the 10 days between June 1 and June 11, a window that opened with burning cities and Trump’s march to St. John’s Church, but concluded with mostly peaceful demonstrations nationwide. During that period Avalanche found that support for the protests grew 10 points among Mixed Feelings voters, 14 points among Lean Biden voters, and a head-spinning 25 points among Lean Trump voters. “I had never in my research career seen public opinion shift on the scale in this time frame,” Prull said. “When we look at this from electoral context, when you see a 25-point swing in Lean Trump supporters from disapproving of the protests to at least somewhat agreeing with them, that’s just a scale of public opinion shift you don’t see in this line of work very often.”

The reasons persuadables moved from opposing to supporting the protests, Prull said, can mostly be attributed to the demonstrations growing and becoming largely peaceful by their second week, with human stories of everyday police brutality saturating the media environment.

The president has a habit of believing that his own views are more widely shared than they really are (“silent majority!”) but this example of it may have been uniquely costly if even “Lean Trumpers” were becoming more sympathetic to the protests as he was threatening to crack down. The Times claimed a few weeks ago that the demonstrations had become the largest movement in U.S. history based on the sheer number of people who say they participated in them. Numerous polls taken over the last month show widespread support for Black Lives Matter. Maybe Biden was destined to benefit from all that no matter what, even if Trump had bitten his tongue and said nothing. But threatening to send U.S. troops to shoot looters after the looting had largely ended did him no favors.

Note in particular the shift in attitudes between June 1 and June 11 described in the excerpt. Prull thinks that’s a function of the protests turning more peaceful but BLM may also have benefited from a public backlash to some of the heavy-handed tactics used against protesters during that period. There were countless videos circulating online during those 10 days of cops getting rough with demonstrators. The most notorious example was the clip from Buffalo in which an old man was pushed by police, cracked his head on the sidewalk, and was left there to bleed initially. (Trump was on the wrong side of that incident too.) But there were other clips shared on Twitter, Facebook, and Instagram, and ultimately mainstream news. A movement aimed at curbing police brutality won’t lose followers at a moment when a number of cops are caught on camera behaving brutally. To the contrary. And because Trump is the “law and order” candidate who almost always sides with police, it may be that voters took it out on him.

But something else happened during that 10-day period. The spectacle in Lafayette Park, where protesters were pushed out so that Trump could have his Bible-waving photo op in front of the church across the street, took place on the early evening of June 1. It drew a ton of coverage, both the next day and for weeks afterward as the media tried to figure out who had given the order to cops to advance, what sort of non-lethal munitions had been used on demonstrators to clear a path for Trump, why Mark Esper and the Joint Chiefs chair were there, and so on. It was two days later, in response to the sight of Esper and Mark Milley strolling through the park with Trump, that James Mattis blasted the president for it in a short essay. Other retired brass like Mike Mullen chimed in with condemnations of their own. More importantly, the public seemed to find it distasteful as well. David Shor is a Democratic data guy who worked for a firm called Civis until recently. According to Civis’s data, said Shor to New York magazine, the Lafayette Park operation was the turning point of the campaign so far:

[O]rder is a winning issue for conservatives here and everywhere around the world. The basic political argument since the French Revolution has been the left saying, “Let’s make things more fair,” and the right saying, “If we do that, it will lead to chaos and threaten your family.”

But when you have nonviolent protests that goad security forces into using excessive force against unarmed people — preferably while people are watching — then order gets discredited, and people experience this visceral sense of unfairness. And you can change public opinion. And if you look at the [George Floyd] protests, there was some violence in the first two or three days. But then that largely subsided, and was followed by very high-profile incidents of the state using violence against innocent people.

And, you know, the real inflection point in our polling was the Lafayette Park incident, when Trump used tear gas on innocent people. That’s when support for Biden shot up and it’s been pretty steady since then.

If Shor’s name sounds familiar, it’s because he became a celebrated victim of “cancel culture” not long ago. That’s why he’s no longer with Civis. He was fired when leftists and some co-workers took offense to a point he had made on Twitter in late May, a few days before Lafayette Park. His point was that historical data shows … Democrats tend to suffer electorally when protests turn violent and to benefit electorally when they’re peaceful, which is exactly what seems to have happened with Biden and the BLM protests. Shor’s extremely woke cohort may have resented him for uttering a discouraging word about rioting but, irony of ironies, the protesters’ turn away from violence in the days after Shor’s tweet may end up getting Joe Biden elected.

The question now: What does this mean for the ongoing standoff in Portland between camouflaged federal agents and the hordes facing off with them outside the federal courthouse? On the one hand, the feds have taken heavy criticism for how they’re operating, dressed up like soldiers and having made at least one dubious arrest. On the other hand, some “protesters” clearly are violent, attempting to set fire to the courthouse and pinning the feds inside their own security perimeter. Under those circumstances, maybe it’s to be expected that Biden’s national lead hasn’t moved much since the start of the month. He was ahead by 8.8 points on July 3 and he’s ahead by 8.7 points today. The Portland vandals are harder to sympathize with than the BLM protesters so Biden gets no electoral benefit from what’s happening there. But Trump gets no benefit either, presumably thanks to the weird paramilitary trappings of the operation and possibly the sense that he’s making the situation worse with such a heavy-handed response. We’ll see if the numbers change as it plays out.