Did the Black Lives Matter protests fail to cause a spike in coronavirus cases?

While numerous mainstream media outlets claimed that the BLM protests did not increase COVID-19 infection rates based primarily on an NBER working paper (which had not been peer reviewed), subsequent research (which was both peer reviewed and published in a mainstream academic journal) came to the opposite conclusion.

According to a paper in Oxford's Journal of Public Health by Randall Valentine, Dawn Valentine, and Jimmie L Valentine:

In the eight cities analyzed, all had positive abnormal growth in infection rate. In six of the eight cities, infection rate growth was positive and significant...it was apparent that violations of Centers for Disease Control and Prevention (CDC)-recommended social distancing guidelines caused a significant increase in infection rates.

Further research might overturn these findings, but at the very least many media outlets were premature in making such a claim.

edit: I need to clarify: This is not a complete answer to the question. But it contains relevant information for reaching a conclusion.

---

JPMorgan has a new study (paywall) showing the strongest correlation to new cases is eating in restaurants.

The firm analyzed spending by 30 million Chase credit and debit cardholders and coronavirus case data from Johns Hopkins University, and found that spending patterns from a few weeks ago "have some power in predicting where the virus has spread since then," analyst Jesse Edgerton wrote Thursday. The study found that the "level of spending in restaurants three weeks ago was the strongest predictor of the rise in new virus cases over the subsequent three weeks," in line with the firm's recent studies using OpenTable data.

Some more discussion on twitter with this graph:

The conclusion of the study is that protests have led to increased social distancing elsewhere which offsets the growth in cases induced by the protests themselves.

Firstly, this is akin to saying that a serial killer on the loose isn't contributing to increased murder rates because other people, fearing the chance of running into the serial killer, are staying home and therefore the overall murder rate falls because there are less victims out and about. Does this mean we should encourage people to go around doing some serial-killing because hey, it reduces murder rates?

Secondly, the implicit conclusion of the study is that it puts the credit for reduced growth rates on the protestors, since the protests "caused" the social distancing. It seems more fair to state that the credit for the social distancing should go to the people who actually made the sacrifice of social distancing, and not the protestors.

Thirdly, the study justifies using only 10 to 21 days of data by pointing to a mean incubation period of about a week. This is irrelevant. What matters to judge whether there has been an increase in COVID rates based on these 21 days of data is the time between infection and diagnosis. This may well be much larger than 1 week, and indeed, in the case of protestors that are busy protesting, it may be larger still. Consider a protestor that gets infected on the 10th day following the first protest. Let's say it takes 7 days for this protestor to experience symptoms (if they even experience them at all). Let's say it takes another 7 days for this protestor to get tested and diagnosed with COVID (reasonable assumption: do you run to get tested immediately every time you sneeze?). Now 14 days have passed. This protestor's infection is not included in the data (14 + 10 > 21). Hence, even the overall conclusion of the article may be completely incorrect, since the range of data may bot be large enough.

Fourth, in relation to the previous point, the study ignores the fact that the reproductive number of COVID is larger for an infected protestor than an infected non-protestor, since the protestor gathers in large crowds, hence making it easier for the virus to spread. Thus, consider the protestor from above point who was not included in the data. This protestor will go back to protesting, and will spread the virus to far more people than the average infected would. These new infections are not included in the data either.

Summary: the article's conclusions do not paint the protestors in a good light even if the conclusions are true, and on top of that, the conclusions may not even be true, due to the limited data available. To bring this latter point fully home, the study has data up until June 20. It is now July, 03. Here's a graph of the daily cases in the US. Something tells me if they redid their study today, they'd get different conclusions .....

Fox News has a new article titled LA Mayor Garcetti admits 'connection' between coronavirus outbreak and protests, after downplaying link that quotes the LA mayor:

He claimed he'd consulted with Dr. Barbara Ferrer, LA County's director of public health, and determined the protests were in fact contributing to the spread of the virus.

[…]

"We do believe there is a connection, we don’t believe that everybody has been doing this safely and wherever you can, please stay at home.”

Covid rates have gone up in the USA, not down.

The number of cases/day in the USA has grown tremendously in June 2020, after undergoing a somewhat flat or slightly downward trend in April and May.

You wrote:

One news organization suggested that it may have slowed down the spread of the coronavirus

If this is correct, then the above graph would have shown even more growth in June if it weren't for the protests. This is hard for me to believe because the growth rate in June that this graph shows, is already unusually high.

In my opinion it is impossible to be 100% sure whether or not this enormous growth in June was caused by the protests or not. We have to accept that as mere mortals, complex fields like sociology, economics and epidemiology, do not lend us evidence that is so crystal clear such as the evidence that Newton's laws fit well with Keppler's observations. All we know is that the protests happened and then the # of cases rose a lot (but other things also happened such as more businesses opening up and good weather attracting people to beaches).

Not peer reviewed

The researchers from the University of Denver currently have their paper titled Black lives matter protests, social distancing and COVID-19 posted online as an NBER Working Paper. This means the paper HAS NOT gone through peer review and carries the same scientific weight as a blog.

To reiterate, while this work may be valid, it has not been cross-examined and the media SHOULD NOT have picked it up and sensationalized it!

The authors conjecture that the number of COVID-19 cases may have gone down due to people wishing to avoid the protests

For example, other individuals who did not wish to participate in the protests, perhaps due to fear of violence from police clashes or general unrest, may have chosen to avoid public spaces while protests were underway. This could have an offsetting effect, increasing social distancing behavior in other parts of the population. The net effect, on both social distancing and on the spread of COVID-19 is thus an empirical question, and the focus of this study.

Of concern to myself is that the protests are ongoing to which the authors point out

One concern regarding the lack of any strong effects for COVID-19 case growth is that the post-protest sample period might not be sufficiently long enough as of yet to detect a resurgence or increase in the infection rates. While this is a possibility, we also note that our sample includes at least 21 days of data following the early protests that took place in 154 cities (during the first five days following George Floyd’s death), at least 18 days of data following protests in 242 cities, and at least 16 days of data for 257 cities that experienced protests (during the first week following George Floyd’s death).

The author use a 21 day lag-time as their window of safety for evaluating if a city sees a rise in cases. They offer literature sources using this criteria. This does not make it valid, but, it appears to be the norm - insomuch as COVID-19 has norms.

In the body of the results/discussion the authors say that their findings show, with 95% confidence, that mass protest cities did not see a rise in COVID-19. However, in their conclusion they say

Likewise, while it is possible that the protests caused an increase in the spread of COVID-19 among those who attended the protests, we demonstrate that the protests had little effect on the spread of COVID19 for the entire population of the counties with protests during the more than three weeks following protest onset. In most cases, the estimated longer-run effect (post-21 days) was negative, though not statistically distinguishable from zero

So I am a bit confused. Are the 21 day results passing or not passing the hypothesis test that there is no rise in COVID-19 due to mass protests. They cannot conclude that COVID-19 decreased from the protests if the confidence interval contains 0. In fact, if it even contains 0.0000001, they cannot conclude that it didn't cause a rise in covid-19.

Edit: I previously made it sound like the paper was loaded with unsubstantiated jargon, and this is not accurate, as was pointed out in the comments. My opinion is that there are several areas that need tightening up.

Conclusions

Overall I will say that the work appears for the most part well written and of an academic level. This manuscript has potential but I believe the peer review process is required before I take it seriously. Good reviewers will make them substantiate all discussions involving statistics and will hopefully make their conclusions exactly match their discussion. Also, reviewers will go over their statistical methods to ensure they are acceptable. It is unfortunately the case that given a desired outcome, there is a way of evaluating the data to get that outcome. Reviewers will ensure this has not been done. I am not qualified to grade their metrics and choices.

Finally, the peer review process should take time. It is fine for the authors to publish a working manuscript, I do it myself, but it is absolutely not okay for the media to propagate it before it has been peer reviewed! If this is the case scientists can just start publishing things on blogs.

At this point there isn't much we can say about this but to point to the study itself. I'm a non-expert and I haven't read the entire study, but it seems to me that the researchers are careful to explain their methodology and not to overstate the conclusiveness of their findings. I quote from the conclusion of their paper.

This study is the first to empirically examine the linkage between the 2020 Black Lives Matter protests and the spread of COVID-19, which has been a point of concern among public health officials and the media (Bacon 2020; Goldberg 2020; Harmon and Rojas 2020). While it is almost certain that the protests caused a decrease in social distancing behavior among protest attendees, we demonstrate that effect of the protests on the social distancing behavior of the entire population residing in counties with large urban protests was positive. Likewise, while it is possible that the protests caused an increase in the spread of COVID-19 among those who attended the protests, we demonstrate that the protests had little effect on the spread of COVID-19 for the entire population of the counties with protests during the more than three weeks following protest onset. In most cases, the estimated longer-run effect (post-21 days) was negative, though not statistically distinguishable from zero.

The main caveat I would place on this is that it is a working paper, not a peer-reviewed publication. There may be room to argue that the media has overstated the evidence by reporting based on a single working paper. But unless and until there are experts who reject the methodology of this study or raise contradictory evidence, the claim is as well-supported as we can reasonably expect at this time.