Complex patterns in data and looking beyond what we might be disposed to see

Interesting just to stare at data sometimes to try and see what patterns might lurk. In this instance, I am looking at Worldometer's data on Covid cases in the US and by state.

I am focused on the death rate per million as a broad measure of performance. If you look at the Deaths oer 1 million population, the average for the nation is 405 per million. The fact is that Covid-19 hit different states at different times so this is still currently a lag factor which we cannot properly take into account (our knowledge is too constrained.) Eventually, over 9-18 months, these numbers will likely converge to some degree. However, there will remain marked differences which will reflect underlying populations factors (age and morbidity prevalence) as well as government policy effectiveness.

For example, New York will likely remain at the high end of deaths per million into the future simply because of their ill-considered policy of parking Covid-19 patients in assisted living facilities.

But that is down the road. Where are we now and what might the numbers tell us. Or, more properly, what might the numbers suggest.

State and July 8, 2020 cumulative deaths per 1 million residents

New Jersey - 1,728
New York - 1,660
Connecticut - 1,217
Massachusetts - 1,192
Rhode Island - 915
District Of Columbia - 795
Louisiana - 715
Michigan - 626
Illinois - 574
Maryland - 540
Pennsylvania - 534
Delaware - 528
USA Total - 405
Indiana - 404
Mississippi - 389
Colorado - 295
New Hampshire - 282
Georgia - 273
Minnesota - 268
Arizona - 265
Ohio - 255
New Mexico - 248
Iowa - 232
Virginia - 220
Alabama - 211
Washington - 182
Florida - 179
Missouri - 178
Nevada - 178
California - 166
South Carolina - 164
Nebraska - 146
North Carolina - 139
Wisconsin - 138
Kentucky - 135
South Dakota - 111
North Dakota - 110
Oklahoma - 102
Arkansas - 100
Kansas - 99
Texas - 97
Tennessee - 97
Vermont - 90
Maine - 82
Utah - 61
Idaho - 53
West Virginia - 53
Oregon - 52
Wyoming - 35
Alaska - 23
Montana - 22
Hawaii - 13

There are 12 states/territories above the average of 405/million deaths and 39 below. For those 12 above the national average, they are way above. Together, their average deaths per million is more than double that of the national average with 919 deaths/million.

For those below the national average, their collective average is 158 deaths per million a mere 17% that of the twelve worst. This suggests a very high concentration of the tragedy in a small number of states. It also likely reflects, at least to some degree, the lag time between early exposed states and later exposed stats.

What else is noticeable.

It is too easy to note that of the twelve bad performers, ten of them are pretty solidly Democratic and the remaining two have Democratic Governors. Is there difference in outcomes due to differences in either state policies or differences in state competence? Plausibly yes, and almost certainly yes in the case of some particular cases. Still, I think some of the numbers will even out over 18 months once the initial onset factor is taken into account.

While we are taking the political view, it is worth noting that the three state Governors who are routinely excoriated in the mainstream media for their cavalier, callous and ineffective governance, especially in terms of reopening their states all three have declining number of deaths per day and together the three have a death rate per million of 183, less than half the national average, and only 20% that of the twelve worst performers and only 11% that of the mainstream media's darling governor Cuomo of New York.

While politics and state policy will, I think, eventually be shown to have some clear consequence in the eventual measure death rates per million, I see some other patterns that I suspect are probably more significant.

I have lived, worked, or traveled extensively in 23 of the states and have visited another fifteen or so. Combining that with just general knowledge, I would point out some other differences.

That the good performers are more rural is another easy pattern, but I am not sure just how explanatory that is. What we really want to know is how urbanized a state might be. Georgia is the largest geographic state east of the Mississippi, and it does have a low death rate. But it has a high urbanization rate of 75%, just under the US average of 80%.

California is another huge state but it is 95% urbanized.

Are the bad performers more urbanized? The twelve are 87% urbanized. All the rest are 72%. So urbanization is certainly part of it.

But I suspect that there is one layer further of refinement and that is by urban density. Not just how urbanized the state is but how dense that urbanization is.

The average density of the cities in the twelve worst states is 7,100 people per square mile. The average urban density of the in the better performing states is 3,580 people per square mile.

But the definition of City is pretty constrained. What we really want to know is the density of big cities. Philadelphia MSA, not just Philadelphia City proper.

Using US Census Data for 2010, we can get some additional insight. Counting the 283 cities with a diameter of ten miles, we can get to the following numbers.

The average density of those cities in the twelve worst states is 9,266 people per square mile at one mile, 3,173 at five miles, and 2,002 at ten miles. For the better performing states, it is 5,450 people per square mile at one mile, 2,388 at five miles, and 1,643 at ten miles.

So the worst performing cities are 50% more dense in their core and 25% more dense in their suburbs than the better performing states.

That's still not quite getting at the pattern I think I see. Other than Louisiana, the other eleven states have large mass transit systems. But that can't be the distinguishing factor because plenty of the well performing states such as California, Georgia, Texas, also have large mass transit systems.

What I think is see is that large dense cities with many high rises, large mass transit systems which are heavily used are the ones with the highest deaths.

There are some exceptions to the pattern, but it seems that states with high urbanization rates and where those urban conglomerations are high rise/high density cities with heavy dependence on mass transit are the ones most disposed to high death rates.

Georgia is relatively highly urbanized and it has an extensive mass transit system but the city density is low (few high rises) and the mass transit system is lightly used. And the state has low death rates.

At least that's what the data seems to indicate for now.