A Deeper Dive into Increased Deaths by Race/Ethnicity and Geographic Patterns, U.S., 2020-2021
Top 12 locations in U.S. mortality data, come on down!
In my prior post, I looked at how the number of deaths increased by racial/ethnic group in the U.S. in 2020-2021, breaking out COVID but also looking at deaths overall.
I ended the post with a few maps, particularly with “the Acela corridor”, which reaches from DC to Boston. These maps came from the (threatened-to-soon-be-removed-because-it’s-ten-years-old) racial dot map, and it shows not only the very uneven population density in the U.S., but definitely shows a very uneven density of people distributed out by race/ethnicity.
In this post, I’m going to focus on top locations by population in the U.S., and four major racial/ethnic groups: White, Black, Hispanic, and Asian.
My point here is just to look at the patterns of mortality and population, and how they differ by different areas of the U.S. I’m not trying to look at any particular conclusion, but simply explore how deaths have increased differently over the period.
For this post, I’m going to look at 2015-2019 versus 2020-2021 (through 10/2/2021, about 1.75 years). I may look at the pattern of mortality (the four waves we’ve seen thus far) in a future post.
Have a map of the United States
Here’s the continental U.S.:
Even without labels, you can see where major cities are, major highways and rivers, where there must be large mountain ranges (and, in Florida, huge swamps) where nobody can live.
You can read the documentation of the racial dot map project (and get their little begging pop-up… it would be nice to have a new map, but this is really not high up on my donation priorities (what is? Thanks for asking! The Movember Foundation! I’ll be writing about that later.))
Some of what you see on the map (this kind of net you see in the very sparsely-settled midwestern/northwestern areas) are simply artifacts of how closely they can place people. Out where few people live, the census tracts are huge.
But on the East Coast, and along the West Coast, people are bunched up pretty well, and you can really see the major groups: White (blue dots), Hispanic (orange), Black (green), and Asian (red). They have brown for a catch-all group, but there are not many of those on the map. Indeed, in my own graphs, I’ve had to drop those categories, as I just don’t have enough deaths from those groups to have statistically reliable results other than at the national level.
My main observation has been: there has been a large geographic component to the effects of COVID mortality in the U.S., and a lot of the patterns we’re seeing have come out of that.
Caveating out the wazoo
And I have a very big wazoo. (TMI)
For the following, as with my prior post, I am using data from the CDC, and my underlying spreadsheet can be found here:
The file is 80 MB, btw, because I have all the CDC data in there, not only my particular aggregations. The underlying data have deaths by week, race/ethnicity, and state, and the data go back to 2014 (though I don’t use the 2014 data). So downloader beware.
For 2020-2021, I am using MMWR weeks 1-53 for 2020 and 1-39 for 2021 (which ended 10/2/2021). While the data do go to week 42, and I am using their “predicted” totals, I think weeks 40-42 are too undeveloped to use for analysis.
My definition of “excess deaths” is the number of deaths above comparable periods for 2015-2019 for the same jurisdiction. Now, there are many reasons this is not exactly right in defining excess deaths — it is essentially assuming the population is in a “steady state”, where the age distribution is the same, cause of death trends are the same, etc. That’s not true, of course. However, in general, the numbers tend to change very little over a 5-year period, which will probably be okay as a rough, order-of-magnitude analysis.
To have large enough populations, and to grab major geographical areas, I’m looking at the following jurisdictions:
U.S. (everybody!)
California
Texas
Florida
New York (this excludes NYC)
NYC
Pennsylvania
Illinois
Ohio
Georgia
Michigan
New Jersey
Virginia
[people who are checking this against biggest state lists, yes, I jumped over North Carolina, because NC data is crap. Yes, still.]
Percentage increase in total deaths by race/ethnicity
Last time, we saw a graph for the whole U.S.:
As noted above, I need to drop the smallest racial/ethnic groups as they usually don’t have enough deaths in the states we’re looking at.
Here is a table & sparklines showing the percentage increase in deaths by group:
Observations:
- Whites have the smallest percentage increase for each location, with California being the only place in the group with a less than 10% increase for white deaths
- Blacks do not have the highest percentage increase for any of these locations.
- In 7 of the locations, Hispanics had the highest percentage increase in deaths; in 5 of the states, Asians had the highest percentage increase. For the U.S. overall, Hispanics had the highest increase, as we saw earlier.
- Some of these increases in death counts for location were extremely large — compared against the national average of 21% increase for the period, we saw, for example, an 80% increase in deaths for Hispanics in Georgia and a 96% increase in deaths for Asians in Ohio.
Differences in magnitude – some numbers are small
Now, let us go back to the disclaimers. In some of these cases, the numbers involved aren’t huge.
Let’s look at that 96% increase in deaths for Asians in Ohio.
Well, over 2015-2019, they averaged about 500 Asians per year dying in Ohio. That’s not really a very large number. Contrast that against about 4,400/year dying in New York City alone and 26,200/year dying in California.
That’s before COVID.
The period I’m looking at is about 1.75 years, so I was comparing the total of 1,600 Asian deaths in Ohio seen so far in 1/1/2020-10/2/2021 against the expected 800 deaths for the period of interest.
So that’s an increase of 800 deaths — 100% increase (I’m rounding for simplicity).
On the other hand, NYC saw a 61% increase in Asian deaths, for a total of 12,500 deaths (all causes).
These are very different amounts, with different repercussions.
Distribution of deaths among races
First, given the various states, let’s look at how the total deaths were distributed among the races/ethnicities.
Pre-pandemic:
Post-pandemic:
No, don’t squint too hard. There’s little difference between the two graphs. Here’s a table so you can see the percentage point difference:
The darker cells are where a higher percentage of that state’s deaths went to that group compared to 2015-2019; the lighter cells (and negative numbers) are where it went down — pretty much the White category for all the states.
We can see that the biggest disparity comes from California.
Let me tell you the underlying numbers – in 2015-2019, 19% of deaths were Hispanics. In 2020-2021, 25% of all deaths were Hispanics — in California.
Let’s contrast that to Texas. In 2015-2019, 21% of deaths were Hispanics. In 2020-2021, 26% of all deaths were.
Those are actually pretty substantial differences in distribution by race, primarily for Hispanics across the board.
It gets much more stark if we solely look at COVID deaths:
And a similar table as before:
Let’s point out California again — 47% of deaths were Hispanics, making for a 27-percentage point disparity between the “baseline” mortality distribution and COVID.
Similarly, for Texas, 43% of deaths were Hispanics, with a 22-percentage point difference against baseline.
When I graphed the table with the percentage increase of deaths for each group, we saw deaths were up across the board, most for Hispanics, then for Asians, and then for Blacks. Whites were the least hit.
When we look at these distributions, to see the relative risk between these groups, again we see Hispanics were more heavily hit compared to other groups. And it’s all over the map, with California and Texas looking the worst, in relative terms, and then there’s FL, NYC, NJ, and Illinois all having 10-percentage point disparities or higher.
Road trip! (virtually)
To round out this post, let’s just take a trip around the U.S., in map form.
Let’s start out with the good ole Acela corridor again:
I didn’t actually extend the picture all the way to Boston, stopping at New Haven. I really think of it as I-95, having driven it many times from NY or CT down to NC (or even Florida). I used to take Amtrak from NYC to Raleigh, NC as well for many years – it’s the same route, essentially. I am extremely familiar with what it looks like along the way. In particular, I know what a really densely-populated area looks like versus… not so much. I used to take Amtrak from NYC to Raleigh, NC as well for many years – it’s the same route, essentially.
Of course, NYC is the most densely populated area in the U.S. along this pipeline, and it spreads tendrils out into Long Island, into CT and NJ, and has all sorts around. The categories “Hispanic” and “Asian” (and even “Black”) are hilariously broad given how extremely specific the groups can be by block in NYC, even now. If I remember correctly, Queens is the county with the highest percentage of foreign-born residents.
Along this corridor, you have the multicolored splotches around the very large cities, and the more southern you go, the more green dots (i.e., Black people) you see. There are lots of red dots (Asians) clustering around DC and NYC, especially in New Jersey. But not many clusters outside of that. Orange dots (Hispanics) are noticeable in the Bronx and in various places, but not that easy to see… not compared to my next area of the map:
TEXAS! (and New Mexico and a few other places)
If we couldn’t see orange dots before, we can certainly see them now. Huge clusters around major Texas cities of Houston, Dallas, and San Antonio — but also large groupings on the Mexican border at key spots, like near Juarez.
Something to note is the changing of dominant colors — obviously, blue is generally everywhere in the east (white people), but gets very sparse out west (everybody gets sparse out west, except at key points). I want you to see in this particular map that green dots (Blacks), increase in density as you go east, especially past the Mississippi River.
But the point is to notice clustering.
Southeast U.S.
Here we have some of the southeast, plus Florida. Note the big green swathe across Mississippi, Alabama, South Carolina, and Georgia. It continues through North Carolina and into Virginia. Obviously, this pattern is the result of history, like most of the settlements we see. Lots of people stay near where they’re from. Some, like me, go far afield, but I’m not normal. And I didn’t even really go that far — I’m still along the East Coast.
In the southeast, there is a very large Black population, as well as White, and a lot of it is rural. There’s only one city of note in the area (outside of Florida), and that’s Atlanta. You get one of those multicolor splotches around the city, and while it has the usual White/Black mix you see throughout the south, you get all the major categories represented.
Going to Florida, which isn’t really “the south”, you see some interesting habitation patterns. No, the hurricanes didn’t sweep everybody to the edges of the state. It’s just that most of the state is swamp. Sure, they call them “wildlife preserves”, but it’s more you may as well try to make the most of the situation, which is that the state is a giant sinkhole.
(yeah, I’m biased.)
People are really, really crowded on the coasts, there’s one city smack in the middle (that’s Orlando), but most of the cities are on the coasts. Not seeing much in the way of red dots (Asians), but definitely plenty of Hispanics, Whites, and Blacks in the dense clusters on those coasts, and of course, the multicolored splotch around Orlando.
Let’s visit just one more place, and yes, it’s still on the coast.
California
Now you can really see some interesting features in this landscape from the patterns of where people live.
California has challenging geography. Like Florida, it has a lot of places people really can’t live, though it’s not swampy. So you see people on the coast (as per usual), but not all the coast is congenial. Then you have the string of towns in the Central Valley.
Indeed, when you get out west, the human settlement definitely looks a lot more like the shape of the land and scant water than what you see in the east (where water tends to be overly plentiful). In the east, our mountains have been worn down, and we’ve had plenty of time to fill in some of the swamps. I mean, we did build Orlando, didn’t we?
The demographics of California are very different from all the other places we’ve seen: more Hispanic, and more Asian… but the Asians are still clustered at the cities on the coasts, and the Hispanics are in the Central Valley as well. White people dominate in the north. There aren’t many Blacks in the state, definitely not compared to the national average or even someplace like NYC.
What’s the point?
My main point is just to see what’s there.
Too often, when people have been exploring these data sets, the person doing the exploring has a definite agenda in mind, and it is so obvious that one never gets to just see the data.
Early on, when all the data was a mess, it was understandable that people were grabbing at whatever metrics they could, just trying to make sense, but that sort of thing dissolved pretty rapidly.
Now, I’ve been disgusted that after 1.75 years of a pandemic, there are still people doing the idiotic: “This state which has a governor I hate was tops for COVID deaths for this extremely specific three-week window that I swear I just picked for its super-awesome meaningfulness, and no, I will not go back and revisit when the state with the governor I like is tops for a different three-week window.” I’m not even bothering to link to these news “stories”. Someone shared another one with me today, and it got me angry again.
If I have any hypothesis of my own, it’s that the policies and politics of specific states may not have had as much impact on the COVID and excess death outcomes so much as how people in those states live. Are they having to take mass transit everywhere (spreading NYC and its area)? Are a bunch of people crowded into small spaces, whether for living or working? Are they inside together a lot? And what time of year?
Both Texas and California have had similarly bad outcomes for Hispanics in their states (60% increase in deaths, disproportionate share of deaths), and their politics and policies are very different. They both have very large Hispanic populations compared to the national average, and they’ve got border towns with Mexico. There are some geographic differences between the two, but there may be aspects of how people actually live in those states, especially the Hispanic residents, that would explain the especially bad result for them there.
I am corrected
Side bar: one of my fellow pedants corrected me from my prior post, when I was talking about people going inside, and he wrote [just a copy/paste]:
It is not that it is cold is it that the climate makes you want to be indoors in a location that for energy efficiency recirulates a lot of air. The south was hot many started spending their time in AC. Now it is getting more temperate there and people are out hunting and doing stuff outside. A long the Rockies it cold enough for indoor activities and activities that require indoor recovery.
So I said I would make sure to include this.
He reminded of some other patterns, but I may do one more post on this, because my comparison was the whole 1.75 years, and I didn’t look at smaller chunks of time, because there definitely were different times the waves went through different states, especially in the south.
Facts before conclusions
My biggest lesson here is that people really need to get some concrete facts into their brains a bit more before they try making it all fit into some sort of hypothesis.
My own theory regarding geography & seasonal weather patterns affecting the different groups is just one of many hypotheses, and of course, it’s likely there are multiple factors interacting.
Unfortunately, too many people have decided on their conclusion first, and have worked backward at harvesting the data that supports it.
That is a very unlikely way to reach the truth.