One size does not fit all

Well, kudos to the NYT for attempting some fact-based reporting. From What Is the Real Coronavirus Toll in Each State? by Josh Katz, Denise Lu and Margot Sanger-Katz. This is not an analysis seeking to establish the real death toll given all the ambiguities in the current reporting. Rather, this is an effort to establish an alternate definition of the likely deaths. I have advocated, given the problems with the Covid-19 deaths data reporting, that we should be looking at the all causes deaths numbers and that is what the NYT has done. It doesn't matter how you are defining a Covid-19 death.

The all causes death approach allows you to compare the total deaths by state against what the normal volume of deaths for this same month usually are. If you usually have 100 deaths in a state in a month and that has been the number for the past three months, then it suggests either that your state has not yet experienced Covid-19 or that it has and the consequence is negligible. If you usually have 100 deaths a month and you are now experiencing 150, then that suggests (but does not prove) that Covid-19 is indeed a problem.

The NYT has collected all causes death data for forty states and two other entities (NYC and Washington, D.C.) for 42 of 52 reporting governance entities. They have then grouped them by the degree to which the all causes death volume is up over 10% above the corresponding year last year (8), those states where it is between 0-10% above last year (16), and those states where it is the same as last year (18).

As the coronavirus pandemic cuts through the country, it is leaving behind large numbers of deaths that surpass those of recent history. A New York Times analysis of state data from the Centers for Disease Control and Prevention shows just how many lives are being lost in the pandemic in each place — as the virus kills some people directly, and other lives are lost to an overwhelmed health care system and fears about using it.

Our analysis examines deaths from all causes, beginning in mid-March when the virus took hold in the country and examines every state with reliable data. The death count so far is not uniform around the nation. Some places have seen staggering death tolls, while others have seen smaller aberrations from historic patterns. In some states, the number of deaths so far looks roughly in line with those in a typical year, suggesting that the virus and its effects throughout medicine and society have not yet had a major impact on survival.

New York City, long the epicenter of the U.S. outbreak, has experienced the most extreme increase in deaths, which surged to six times the usual number. Altogether, since mid-March deaths there are 23,000 higher than normal.

There are a lot of interesting tidbits which are in the numbers but not really pointed out or emphasized in the new report.

First is that the NYT estimate of Covid-19 deaths is about 34,000 compared to the Johns Hopkins estimate of about 70,000 deaths so far. We won't know for a long time what the truth is but this supports my supposition that we are currently doing some overestimating by factors. I have been crudely estimating that the Johns Hopkins numbers overestimate reality by a factor of two and this analysis is consistent with the eyeball estimate.

What is most striking is the political division. I am not trying to make a political point, I am simply puzzled by the existence of the divide.

Among the eight 10% excess and above states/entities, at least seven are deep blue: New York City, New Jersey, Illinois, Massachusetts, Maryland, Colorado and Vermont. The eighth state is correspondingly bright red.

What to make of that? Colorado and Vermont are inconsistent with a density explanation. Regardless of density, perhaps percent urbanization might be the explainer but then Wyoming has a very low urbanization percentage. All but Wyoming have extensive mass transit systems. I don't have an answer but the blue-red imbalance is striking.

Among the sixteen states/entities affected but not excessively (0-10% increase), eight are deep red and eight are different shades of blue.

Among the eighteen states where there has been no change increase in excess deaths, twelve are deep red and four are deep blue.

In aggregate, there is no hard and fast rule. There are deep blue states among the unaffected and there is a single deep red among the most affected. But overall, it appears that the deeper blue the state the more likely they are to suffer inordinately from excess deaths during this pandemic.

Are the low to medium impact states simply lagging behind the high impact states? Perhaps, but there is plenty of reports suggesting that most state are past their initial peaks.

What else might be driving this divide? I don't know. I can conjure a number of outer limits plausible explanations but none of them strike me as especially compelling.

However, it does seem to keep density, percent urban, mass transit, and prior population base health index as plausible drivers of the apparent division of outcomes.

It also highlights why extreme one-size-fits-all solutions are such a bad idea. It also suggests the continued need for humility in the face of our incomprehension. Bryan Kemp's orders for opening up Georgia look a lot more explicable when examined through this data than the mainstream media has been willing to acknowledge. It correspondingly makes Colorados near identical opening up much less explicable.