Similar to the western democracies analysis, we compare COVID-19 case growth rates across the United States. While the analysis shows that the severity of the pandemic is mitigated by social distancing measures, there are some key clusters of outliers that reflect that other factors are in play.
We previously looked at the impact of social distancing for the aggregate United States. In this post, we break that analysis down on a state by state basis (plus the District of Columbia). We also restrict our analysis to education and stay-at-home orders, as the non-essential business orders seem to be subsumed into the stay-at-home orders. We did not analyze cross-state travel quarantines at this time, as the number of people affected is likely many fewer than those covered by the education and stay-at-home orders.
As with the country-level analysis, we must first account for differences in population and when the pandemic reached each state. This involves the following steps:
To account for the different sizes of the states, we divide total cases in each state to the state's population. This is expressed in Cases Per Million People.
To account for the pandemic reaching different states at different times, we shift the starting point for each state to when the state crossed a threshold of 10 per million people. (Note that this is the same threshold we used before, but saying 10 per million takes fewer syllables than 1 per 100,000.)
As with the country as a whole, the case numbers in each state grow exponentially. Accordingly, we look at the 5 day growth rates for evidence that we are bending the curve. (As a reminder, the 5 day growth rate is ratio of the number of cases on a given day to the number of cases 5 days earlier. This smooths out the noise in the daily reporting numbers. It also matches the incubation period of the virus, which is the shortest time period over which we would expect to see changes in social distancing policy manifest themselves.)
Putting all 50 states and DC on the same graph is a mess, but here it is:
To visualize the effects of social distancing measures, we plot the correlation of the number of days that a state waited to impose a stay-at-home or school closure order versus the growth rate 4 weeks after hitting the 10 cases per million people threshold. For education orders:
As noted in the previous post, the aggregate case growth rate in the United States appeared to peak and turn over ("bend") when pervasive education orders were employed across most of the United States, and this graph confirms that correlation at the state level. Note that while almost every state waited to impose stay-at-home and non-essential business orders, education orders were applied much earlier, and in some cases, schools were closed before the state reached our (arbitrary) 10 case per million people threshold. Those states certainly benefited from that early action.
It also bears repeating just how important school closures were in terms of starting to control the virus, particularly because the benefit is not just in protecting the students themselves, but in keeping the students from exchanging the virus between their families. For this reason it is difficult to conceive a plan for re-opening the country in the next couple of months that includes letting students go back into the classroom.
For stay-at-home orders, the correlation of growth rate and lag in when the order was implemented looks like this:
This is a more complicated picture than the education orders. This graph shows four distinct groups of states:
The cloud of states in the lower left, which comprises a majority of the states, confirms the behavior that we observed at the country level, namely that states that lock down sooner reduce the growth rate of the virus, even 4 weeks later.
The seven states in the lower right waited a long time to implement social distancing orders and in some cases are still waiting to do so. (Note that if a state has not yet implemented a statewide stay-at-home order, we used today's date in order to include them in the analysis.) Despite the delay, the case growth in these states is comparable to states that acted early. A possible explanation, and one that these states themselves posit, is that these states have low population densities and are naturally self-distancing. We will explore the urban/rural factor in much more detail in a later post.
There four northeastern states (and the District of Columbia) that have high growth rates (doubling every week or less). The reasons for these high growth rates vary. Massachusetts may be suffering from the consequences of "social compression" on St. Patrick's Day (similar to what happened to Louisiana at Mardi Gras), despite taking action to cancel the parade in Boston and other cities. The other states in the Northeast (Rhode Island, Connecticut, and Delaware) are likely suffering a knock-on effect from Massachusetts and New York. The District of Columbia has a dense population with a less-affluent core.
And then there is South Dakota. While Governor Kristi Noem (pronounced gnome, but really more of a troll) has the right looks and assertive mendacity for her next job as Fox News commentator, it hasn't turned out as well for her constituents, as South Dakota is second only to Rhode Island in the current growth rate of cases.
Because of the compounding nature of exponential growth, a little hesitation can end up being a big difference. To see just how important it is to act early, compare the trajectories of California and New York. California issued it's stay-at-home order only a couple of days after reaching the 10 per million threshold, while New York waited 11 days to issue its stay-at-home order. (This was compounded by waiting too long to close schools.) Certainly there could be other factors that need to be considered, but on the face of it, waiting another 9 days resulted in over 10 times the number of cases per capita as California (as of this date).
Commenti