The governors of Florida and California have taken somewhat different approaches to managing the COVID outbreaks in their states. California’s Governor Newsom was the first to order a statewide shutdown. Florida’s DeSantis was among the last. Also, Newsom’s order was considerably more restrictive than DeSantis’s. And, much to the chagrin of the media, DeSantis has moved more quickly to reopen his state than has Newsom to reopen his. As a result, the media has largely lionized Newsom while trashing DeSantis. (e.g., these stories in Politico and The Guardian.)1
So, I thought it would be interesting to compare the actual results in the two states.2 What I found was that the experiences have been very similar and, if anything, the situation in Florida seems to be improving slightly more rapidly than in California.
Let’s start with the cases and fatalities per 100,000 persons. California is currently at 159 cases and 6.5 fatalities, while Florida is at 174 and 7.4, respectively.
Florida has tested a slightly higher percentage of its population than has California (2.23% vs. 2.17%).
When you look at the new cases and fatalities over time, it appears that Florida has done a little better at bending its curve than California.
The similarity in the results is perhaps even more surprising given the frequent travel between Florida and New York, home to the worst outbreak in the nation, and that Florida’s median age is six years higher than California (42 v. 36). Both factors should have made Florida more susceptible to a wider outbreak and a higher mortality rate. But, at least so far, they have not.
So, what are we to make of these similar results from two supposedly different responses?
First, I do not think the approaches were as dissimilar as the media would have us believe. Newsom ordered the statewide stay-at-home order on March 19, but DeSantis was only 11 days behind on April 1 and he had closed bars and night clubs and restricted restaurants to 50% capacity earlier on March 17. Much of the perception of a difference has more to do with the politicization of COVID by the media rather than actual substantive differences in their actions.
Second, and more importantly, the results call into question to what degree we can judge the efficacy of containment strategies like the stay-at-home orders. As the curves have begun to flatten, proponents of the various containment strategies have been quick to credit them with the improvement. But coincidence does not prove causation.
We know from history and common sense that if a disease is contagious, there will be some benefit from keeping people separated. But there is little scientific or statistical evidence to demonstrate the degree of the efficacy of any particular containment strategy. The emerging picture from the data is that it appears many of the restrictions put in place may have had limited impacts and that other factors such as population density, climate and mass transit usage may be at least as, if not more, important. (See this CNN story on the difference between Florida and California compared to the Northeast.) As former FDA commissioner Dr. Scott Gottlieb recently said, “While mitigation didn’t fail, I think it’s fair to say that it didn’t work as well as we expected.” 3
In the panic that ensued from the outbreak in March, public officials, attempting to err on the side of caution, rushed pell-mell to enact very broad efforts to contain the virus with little regard for the costs or collateral damage they would cause. Since we have now gotten past the irrational fear that the outbreak is going to grow exponentially forever (or, at least, most of us have), perhaps we can now start trying to sort out a more nuanced approach to help slow its spread while doing the least possible damage to people’s lives and livelihoods.
___________________________
1 To fully disclose my personal bias, I really don’t care for either one of these governors. Both represent to me the extreme partisanship to which I am opposed.
2 Unless otherwise noted, all of the data for this comparison comes from the COVID Tracking Project.
3 There has been a particularly robust debate within the Israeli medical and academic community about the wisdom of that country’s lockdown, click here for story.
The Florida stats were bullshit.
https://www.miamiherald.com/news/coronavirus/article242552796.html
I thought that Houston’s reliance on individual automobiles was actually a benefit in this case.
Even if you take the statistics Bill King graphs at face value, they don’t tell the story he presents. I would like to post the graphs, but can’t paste them through the interface. CA has 1.8x more people than FL. If you graph per-capita deaths per day, the two time series are very close. The big problem with Bill King’s graphs are the way he smooths over the extreme day to day swings. I think he has used the Excel exponential trend line, which has nothing to do with data like this. Using an appropriate smoothing function, like 7-day moving average, shows FL’s trend is consistently worse than CA’s since 1 April, and getting dramatically worse since 1 May.
Bill, thanks for writing.
Certainly we all understand that if you doing gross number comparisons between jurisdictions they need to be adjusted for population which is what I did with the bar graphs at the beginning of the post.
I do not agree that we should disregard new case data. I don’t know any tracking or modeling group that is doing that. And I think the comparison between Florida and California is fair because we know that both have tested just over 2% of their populations. Also, the fatality data a significant problem in that there is frequently a very significant delay between the actual date of death and the day it is reported. Here in Houston, the City yesterday reported a death that occurred on March 25! It is also “lumpier” because death certificates sometime come in “bunches” from local health authorities.
With regard to the line graphs I included, the purpose was not show the relative size between the two states but rather the course of outbreaks in each. Adjusting those numbers for population will move where the plot appears on the y-axis, but the not the shape of the curve.
Also, I did not use the exponential trendline, I use the polynomial (order 3). Polynomial trendlines are specifically frequently used to smooth datasets with a lot of variation. They are unused by numerous tracking and modeling groups.
However, I agree that moving averages can be used for a similar effect. I prefer polynomial because one atypical datapoint in the averaging period can have a disproportionate effect. For example, in your chart, Florida reported 113 fatalities on May 5, which is much higher than the other datapoints in that 7-day averaging period and is largely responsible for the spike in your trendline. If you extend the averaging period for longer, that spike starts to diminish. And once that datapoint moves out of the 7-day average, the trendline will turn back down.
But even accepting your interpretation of the data I do not think counters the principal points of the post which were (1) what Florida and California did was not that much different (2) the course of the disease in those two states have been roughly similar and (3) there are factors other than policy responses that will affect the course of the epidemic which are significant and perhaps much more significant than policy responses.
Finally, I would just like to say that I do not appreciate your acerbic, condescending and accusatory tone. I have noticed it your “style” on your social media posts as well. We are all trying to figure this out with poor and incomplete data and information. I think having respectful discussions about the very complex and in many ways unknowable course of this epidemic can be useful for all of us. You are obviously a very intelligent person that feels passionately about this issue. But if you continue with this tone, you are going to end talking to no one other than yourself.