Justin Katz is bad with data. Katz has been bad with data for a long time. During the Great Recession (are we calling that the Great Recession of ’08 now? The First Great Recession? The Little Great Recession?) Katz used to chart the states by what their current percentage of their pre-recession peak employment was. Katz charts would have a Y-axis that was percent peak employment and an X-axis of the states in alphabetical order.
He would then proclaim Rhode Island “an outlier”... because it wasn’t anywhere close to Pennsylvania or South Carolina.
Someone who cared about finding insight would do this differently. You might just do this in a table, so all the data was easy to read. You might do this with a bar chart, ordered from most to least recovered, so you could show RI as 49th out of 50. Either of those would’ve made Katz’s point better, but it wouldn’t allow him to say “outlier” so he needed to chart things this way. At the time, Michigan, Indiana, Alabama all had Republican governments, which wouldn’t reinforce his point that RI Democrats were uniquely bad.
Similarly, The Center for Freedom and Prosperity used to display their “Freedom Index” scores in a spiral. If you’re unfamiliar with the Freedom Index, it was the Freedom Center’s way of arbitrarily assigning scores to bills that it liked and disliked, and then ranking all the members of the General Assembly. Sometimes, no one got a positive “freedom” score.
It was obviously ridiculous, made more so by Katz’s spiral chart. Why are they spiraling? Because it looks interesting. Is this chart readable? Hell no. Even by its own internal logic, it makes little sense. If you’ll notice, there are separate Democratic and Republican spirals, despite a few Democrats ranking more freedom-y than some Republicans.
Again, a table would’ve been a better way to display this “data.” I think RI Rank is pretty arbitrary, but you have to admit that its tables are at least readable.
Mostly, I avoid the Freedom Center and their Ocean State Current blog where Katz writes, mostly for my own peace of mind. But then, in Ian Donnis’ May 8 TGIF column, he linked to something that piqued my interest:
A different finding? What could it be? If you follow the link, you discover that Katz is upset that people are focusing on how COVID-19 is disproportionately affecting Hispanic residents of Rhode Island, and he claims that Central Falls (which leads the state in cases per capita) is actually doing better than say, Burrillville, where by his estimate it’s much worse.
If you’re like me you’re wondering how this could possibly be, here’s what Katz writes:
…if we recalculated to rank our cities and towns by cases per person per square mile, Central Falls drops to 32nd in the state. Providence becomes number 1 in the state by that method, but it is followed by Burrillville, North Kingstown, Smithfield, and Exeter. That isn’t exactly a racial profile. And it isn’t a close thing. Burrillville has 25 cases per person per square mile to Central Falls’s three. Given its density, Central Falls is doing very well.
“Cases per person per square mile” is a brow furrowing statement. Is that cases per capita divided by square mile? Is that cases divided by population divided by square mile? Those tend to yield extremely small numbers. I couldn’t figure out how Katz arrived at 25 for Burrillville which has 73 cases and 55 square miles (when only counting land), versus Central Falls, which has 503 cases and is 1.2 square miles.
If you were to extrapolate the case count from Katz’s figures, this would mean Burrillville has 4,015 cases and Central Falls somewhere greater than 3. For context, at time of writing, Providence, which has the largest raw number of cases, has just 3,096 cases. Katz’s math would mean about a fourth of Burrillville is infected with COVID-19.
How did he get there? Katz links to the Wikipedia list of RI municipalities, which has a population per square mile stat (it’s based on the 2010 Census population divided by square land mileage). Lo and behold, when I just divided the number of cases in a town by Wikipedia’s population per square mile I got… .25 cases for Burrillville and .03 for Central Falls. When you times this by 100 (making it kind of a per capita rate) you get Katz’s 25 cases per person per square mile… per capita. You can check my math here.
This isn’t a just “different finding” in Donnis’ phrasing. It’s just bad (and deceptive) math. You don’t divide the raw number of cases in a town by its population density, because population density is an estimate of a single square mile, and the disease is not contained to a single square mile of a particular town.
With such a calculation, Katz doesn’t really demonstrate anything beyond that when a denominator is larger than the numerator, the result is a small number; and the smaller the denominator, the larger the result. This is about as arbitrary as the Freedom Index.
Now, Katz’s bad math is in service to larger point is that we shouldn’t focus on Central Falls and Providence have large Hispanic populations that are disproportionately being exposed to COVID-19. Katz claims infection rates are really about population density.
Except, not particularly. Certainly, looking at the data, the municipalities with the highest per capita infection rates are Central Falls and Providence (also the 1 & 2 most densely populated cities), but if you look at municipalities with the lowest per capita infections, they’re all over the place. 8th most densely populated Newport ranks 35th for per capita infections. Little Compton has twice the per capita infections of Portsmouth, despite being less a quarter as densely populated.
I’m no epidemiologist, but from what I’ve read, a large part of what causes COVID-19 spread is people infecting others in their living or work areas. When people are unable to isolate, they infect others. So it’s no surprise we’re seeing a lot of infection in places like nursing homes, and in areas where people are unable to socially distance because they are “essential” workers and/or live in large households.
A large part of the reason why we’re seeing high infection rates is because even as hotels sit empty, there’s only one place to self-isolate in the whole state. Worse, even if there were, for many it would be impractical because if you stop working to self-isolate, you’ll lose pay and be ineligible for unemployment insurance. Fred Ordoñez isn’t exaggerating when he says that we’re forcing people to their deaths.
Katz also makes note of how white people constitute 74% of the population, but 79% of fatalities in RI, while Hispanics are 14% of fatalities whiles being nearly 13% of the population. He links to the Statistical Atlas’s Race and Ethnicity page for RI. But what Katz omits, and is crucial, is that while he does have the racial and ethnic composition for both RI and its COVID-19 fatalities correct, that includes Rhode Islanders of all ages. Two thirds of RI’s COVID-19 fatalities have occurred in people 40 and up. When you look at the age cohorts above the 35-45 range, well 79% to 93.5% are white. Without the exact breakdowns for age cohorts in RI and COVID-19 fatalities, it’s very likely that within their age cohorts, Hispanics represent a disproportionately large rate of the fatalities.
These are the kinds of deceptive calculations the Freedom Center treats us to. But then, what do you expect a few days after another of their writers suggested that COVID-19 killing 6% of the people it infects wasn’t so bad?
The sad thing is, despite publishing bad math takes and “so what if your family members die?” takes, the Freedom Center still is taken seriously by the press. Thus, you get a day of discussion about non-issues like the Shepard Fairey poster “controversy” that occurred last week. There are dozens of organizations in RI that aren‘t acting in bad faith, but the Freedom Center is the one that gets the ink.