A Tale of Two Cities

March of this year coronavirus exploded in New York, especially New York City (NYC), plagued by nearly 2,600 deaths (over 300 per million population) by month’s end. Worse yet, new deaths were spreading rapidly. The disease was slower to arrive in Chicago, with only 61 total deaths by the end of March: a mere 12 per million population, less than 1/25th the number seen in NYC.


Not only did Chicago have far fewer total deaths, they were seeing fewer new deaths each day:

Clearly Chicago, although victim to coronavirus like all of us, was far less stricken than NYC. It’s no surprise that New York state put serious lockdown measures in place — the toughest in the nation — while Illinois implemented social distancing but with far less severity (it has been ranked 23rd out of 50 states on the scale of lockdown strictness). One wonders, how are they doing?

Both strategies worked; both slowed the rapidly rising epidemic. But NYC’s stricter approach worked better. A lot better.

Chicago has managed to slow the rise of the epidemic, but not stop it. The death rate is still rising, albeit slowly, but shows signs of accelerating recently. Already well over 10 deaths/day/million population, it’s a severe strain on the health care system. Loosening even more their lockdown measures, will accelerate the curve, bending it upward and raising the death rate even further. It seems to me, a crisis will be unavoidable.

NYC, however, has shown us how to recover after being blasted by this epidemic. Not only has the rate of new deaths dropped as rapidly as could realistically be expected, NYC has shown the fortitude and courage to stick with their winning strategy. If they can hold the course, the death rate may soon fall below 10 deaths/day/million population, not what anyone would call “good” but for NYC, some relief to the nation’s most overworked health care system.

The above graph incluces solid lines to show “smoothed” values, for better clarity; it helps remove the confounding influence of the large fluctuations. An easier way is just to plot 7-day running means:

My conclusions: 1. Social distancing works, but when done poorly the death rate can still rise, albeit more slowly. 2. Lockdown works better, much better, and can actually drive the death rate down. 3. When the rate keeps rising, it will exceed tolerance levels — sooner or later, probably sooner.


Thanks to all who have kindly donated to the blog. To help support this blog, please visit the donation link below.


This blog is made possible by readers like you; join others by donating at My Wee Dragon.


19 responses to “A Tale of Two Cities

  1. The thing is, once lockdowns are eased, the virus will start to spread again. In the end, perhaps the same proportion of the population will catch the virus in NYC and Chicago. No one knows. Perhaps we’ll see the same number of deaths. No one knows. If Chicago keeps the level of serious infection to a level that the health system can handle, then theirs might prove to be the best strategy but measures have to be put in place early, not when the disease is already rampant. It may turn out that Chicago does better, overall, than NYC, which may have had overwhelmed hospitals in the earlier phase. This is all just speculation. We’ll only see what really has worked when the virus is under control or just reaches a steady state.

    • You’re right, Mike. All along the way there have been surprises: just this morning, I was surprised to see that Brazil had posted over 15k new cases and (with a caseload already over 200k) a 7.5% daily increase. Not good!

      They are clearly destined for a top 3 place on the global leaderboard–and, given that their population is over 100 million greater than Russia’s and that their current increases are considerably greater as well, probably #2 spot behind the US within two weeks (ish). That wouldn’t necessarily have been predicted even a couple of weeks ago; while they were clearly on the rise, I did not expect, and probably few others did either, that from that point their growth would be near-exponential. (It’s been more often the case that the curve continues to flatten with growing cumulative cases.)

      But that’s just one currently notable example. Covid ‘denialati’ that I’ve seen are concerned to minimize the significance of the virus, presumably because it is economically and politically unpalatable for them. And there are two strategies that are being tried out: the first is to argue that the damage is only ‘measly,’ to use willie’s word. One instantiation of that actually misuses the concept of “mortality rate” to do so, by redefining it as the proportion of deaths to overall population rather than to infected persons. So, for example, they take the 268 deaths per million the US has experienced so far, and restate that in percentage terms: .0268% They then call that “mortality rate” (a step apparently too egregious even for willie.)

      “But wait! There’s more!” They then invert that to arrive at the very reassuring overall population survival rate: 99.9732%! See ma, nothing to worry about!

      One important tacit dimension of that is also the second strategy I mentioned, namely, that of assuming that the present moment ‘is all there is.’ Trump has done this consistently–“It’s only 5 cases! It’s only 15 cases!” The examples above commit the same sin. And again, it’s familiar from climate denial: “CO2 is only 0.04%! Sea level rise is only 3 millimeters a year!” Flies in amber, bereft of past or, especially, future. But the risk is that failing to consider the future may deprive one of actually having a future at all.

      • Thanks Doc you hit on something i had been contemplating for a while.
        “Flies in amber” has struck me as a cognitive defect a lot in denial use.
        I am familiar with such thinking from the climate change debate but can not recall much direct examination of the phenomenon. I guess it could be considered a subset of motivated reasoning when you only focus on the present and ignore the clearly apparent impact of the progression of time on the data.
        Early on in this epidemic I hammered the concept of exponential growth on a forum I comment on and manged to get though to one or two the significance of what we were seeing . Some are still in denial and insist on such time challenged stupidity as directly compering total yearly flue deaths to two months of deaths from Covid.
        If any one has a link to a good every man examination of this phenomenon I would like to read it .

      • I estimate that Brazil must have already 7 Millions infections by now if you consider the number of deaths and the younger population. If less the IFR would be even higher than my estimate which would be bad.

      • Yes, Doc, I’ve seen that method of downplaying the severity by talking about percentage of deaths in the population, rather than percentage of deaths of those infected. Death numbers alone don’t really capture the full seriousness, either, because of potentially permanent damage to a multitude of organs and because the disease seems to drag on for months in some patients.

        But, back to the death rate, an interesting tracking is how the percentage of deaths relative to total cases and the percentage of deaths relative to resolved cases (deaths plus recovered) move towards each other over time. All recorded cases will eventually resolve in either death or recovery (of sorts), so the two ratios will get closer and closer together. The latter number is scary early on (50%, 60%) but gradually moves towards the former and seems to resolve to somewhere between 2% and 6%. In the US, deaths are currently 21% of closed cases and about 6% of all known cases.

      • Thanks, Griff. It’s quite the phenomenon, really.

      • Diaminedave

        Another thing with developed countries is that a lot of deaths are playing out unseen by the general population within nursing homes and hospitals. The psychological impact, in my uneducated opinion, is that it allows a kind of group deniability. Seeing is believing.
        Unlike generations from 50 years or more ago where that impact would have been more within the family home.
        This is probably not true across the entire household income range but maybe enough of those within groups that have a larger voice disproportionally.

  2. Good one as always Tamino. Keep hammerin’. I hope policy makers are reading this. I’m sure Trump isn’t because apparently he doesn’t read anything. Too bad we can’t get you a spot on Fox News.

  3. Hi, Tamino. Thanks for your work, but please clarify your conclusions in comparing NYC to Chicago, because as someone who moved from midtown Manhattan to downtown Chicago some decades ago, and still has family in NYC (and family working in Chicago hospitals), this post has me wondering.

    How do you compare the strictness of lockdown in the City of New York with the same in the City of Chicago. Where’d you get that rating from and is it “illinois” or “chicago”?

    Illinois put a mandatory shelter-at-home rule into effect the day before New York did and has barely started relaxing things, and Chicago (where the first US covid transmission was reported) isn’t trying to be less strict.

    I’ve left my building here in Chicago four times in the last five and a half weeks, to walk to the grocery store three buildings away. I see the lakefront trail, a harbor and parks from my home, and they are bizarrely empty (though like in all northern-tier cities, pressure is building). The office building high-rises (and street-level restaurants) in view are pretty much shut down, though some restaurants are making a go of delivery or curb-side pickup. Some hotels have population and income from the city paying to isolate low-level infected, or medical workers, or homeless. We dismantled most of the emergency hospital in the convention center, and the fraction of ICU beds or ventilators in use for covid patients is holding steady or declining.

    So what is the basis for your conclusions 1, 2, 3 – especially predicting a crisis here? You compare two very large, northern-tier cities that implemented lockdown at the same time and stuck with it, but it’s not even clear yet if the lockdown might have been a bit late in either case.

    From numbers I see, New York’s 8.4 million population declined by 0.27% and Chicago’s 2.7 million declined by 0.06% (so far, for both). New York had a terrible spike, Chicago has had a slow burn at much lower per-capita levels. And according to the data here, the daily death rate in Chicago seems to have declined – the 7-day average is back to where it had been April 21-22.
    https://data.cityofchicago.org/Health-Human-Services/COVID-19-Daily-Cases-and-Deaths/naz8-j4nc/data

    And the chart of Chicago cases and deaths here also show some slowing
    https://data.cityofchicago.org/Health-Human-Services/Daily-COVID-19-Cases-and-Deaths/kxzd-kd6a

    So please provide details for you conclusions to this downtowner who has adopted Chicago. And stay well – it will be important to figure out the results of the natural experiments as different states relax at different rates – kind of like comparing Sweden vs. its culturally/demographically/geographically similar neighbors.

    • Whachamacallit

      Hi B fagan, fellow Chicagoan here who lives in practical downtown. I will say that while the city has closed many areas, I’ve noticed that a lot of people still take walks and go to parks. Yeah, the Lakeshore is closed, but people still take the sidewalks and stuff. While I’m not entirely certain, I’m pretty sure New York doesn’t even allow people to go onto the parks.

      On the other hand, we are better than most other states, but that’s a pretty low bar.

  4. Stephen F. Blau

    A friend if a friend posted this yesterday. Curious if you’ve had the chance to analyze excess deaths?

    • The “It’s no worse than the flu” meme is a total crock, unless it refers to the infamous so-called Spanish flu outbreak of 1918-19. Annual mortality from flu never reached 50k annual deaths from 1976-2007:

      https://www.cdc.gov/mmwr/preview/mmwrhtml/mm5933a1.htm

      And more recently, it has ranged from ~12k (2011-12) to ~61k (2017-18).

      https://www.cdc.gov/flu/about/burden/index.html

      CO19 has already busted that mark by nearly 40%, and isn’t anywhere near done. So even *without* looking at the excess mortality, those making the claim were ‘just wrong.’

      And they could easily have known that. Perhaps some did, which would make them liars rather than bullshitters.

      • Which made me curious: what was the US toll of the 1918-19 outbreak?

        Says here:

        The number of deaths was estimated to be at least 50 million worldwide with about 675,000 occurring in the United States.

        The US population then was about 104 million, so the 675k deaths would equate to ~2 million today.

      • At some point, I’m sure we’ll see estimates of Covid-19 deaths and I suspect they will be much higher than the direct numbers recorded (as happens for estimated flu deaths). No idea, yet, if they’ll be higher than the Spanish Flu.

      • This does not even approach the Spanish (actually, Wichita, KS) Flu.

      • “This does not even approach the Spanish (actually, Wichita, KS) Flu.”

        Yet. But sanity-checking numbers, we’ve lost ~100k Americans in 3 months. The “Wichita flu” outbreak basically lasted a year (spring, 1918-spring, 1919). So, very naively, if we were to lose 100k people every quarter, we’s see 1.2-1.5 million Americans lost to this epidemic. That would be greater than the “Wichita flu” toll in absolute numbers, but considerably less on a per capita basis.

        So maybe–and I sure hope this isn’t how it plays out!–“approaches” is not out of the question, depending on one’s notion of what that verb means.

        Of course, all that is prior to factoring in cryptic Covid mortality, as Mike alludes to. Include that factor, and I reluctantly have to say that an event surpassing the relative magnitude of the 1918-19 event can’t yet be excluded.

      • Ack! Somehow threw in an egregious factor of 3 there, no idea how.

        Please disregard the previous comment…

  5. Tom Passin

    Where is the city data to be gotten? The GitHub site for the John-Hopkins US data only has counties, and they are not the same:

    https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_time_series

    [Response: Notice I said “the biggest counties listed in the data available from Johns Hopkins University.” That includes New York City NY, Dallas County TX, Miami-Dade FL, Los Angeles CA, and King county WA (home to Seattle).]