Speaking to reporters on Wednesday, FBI director James Comey, lacking any evidence beyond the anecdotal, said that he believed a “viral video effect...could well be at the heart” of a spike in violent crimes in more than 40 cities during the first quarter of 2016.

Comey first raised this idea in October, when he said the “Ferguson effect” had sent a “chill wind” through the law enforcement community. He also admitted that he had “no evidence” for this theory. And, a month later, U.S. Attorney General Loretta Lynch said during a congressional oversight hearing that there was “no data” to support Comey’s theory.

“There’s a perception that police are less likely to do the marginal additional policing that suppresses crime—the getting out of your car at 2 in the morning and saying to a group of guys, ‘Hey, what are you doing here?’” he said Wednesday. From the New York Times:

Asked about his past views on the “Ferguson effect” as a possible explanation, Mr. Comey said he rejected that particular term, but added that he continued to hear from police officials in private conversations that “lots and lots of police officers” are pulling back from aggressive confrontations with the public because of viral videos.

He said that the phenomenon “could well be an important factor in this.”

Comey is actually making two claims here: (1) that police are more hesitant to engage in aggressive confrontations with citizens and (2) that this has contributed to a wider spike in violent crime. That there has been such a spike is not in dispute; what is in dispute is the spike’s cause.

It’s not quite fair to say that Comey doesn’t have any data to back up his first claim: He does, it’s just anecdotal. There is no reason to believe Comey is lying about speaking to police officers who feel this “chill wind,” and there is no believe that those police officers are themselves lying. The question is whether that is a systemic phenomenon, and whether it has anything to do with the concurrent rise in violent crime. For that, Comey has no statistical evidence.

So if cops are becoming more hesitant to beat up surrendering suspects, and it doesn’t necessarily have any impact on overall crime rates, what’s the problem?