Conversation

Taking nominal fatality estimates as true, the naive per capita fatality rates, without adjusting for time or reporting accuracy, are Spain: 0.031% Italy: 0.029% France: 0.016% Iran: 0.0049% USA: 0.0045% Germany: 0.0027% China: 0.00024 I just don’t believe China numbers
13
88
There’s an entire order of magnitude separating China and Germany. And another order of magnitude separating Germany and Spain. I don’t think this is explainable by political system. It’s underreporting + demographics + incidence of lifestyle diseases (diabetes etc)
2
16
Also possibly something like the rumored off-target BCG vaccine effect. Though California vs New York suggest that even short differences in lockdown might have order of magnitude effects, so China data are probably not off by an order of magnitude. Maybe 2x-3x off.
Replying to
Another possibility, if you believe the Chinese numbers, is that group quarantine has a huge effect, as opposed to self-isolation. In that case, you’d have to concede that the cost of liberal democratic civil liberties is a 10x-100x higher naive fatality rate.
3
13
Korea is an outlier due to much higher tech use for testing. So more tech can offset the disadvantage of more civil liberties.
7
9
People like to compare China and the west in very abstract, ideological terms, but really I think it boils down to the political feasibility of a *single* tactic: group quarantine. Everything else is pretty much legitimately available within all political systems.
1
9
It would be interesting to apply Dictator’s Handbook lens to this and determine actual extent of power. For example India is nominally a liberal democracy, but Modi probably has tactics available to him that democratically elected leaders generally don’t.
1
5
One of the difficult questions is: assuming there really is a “cost of democracy” holding tech-level constant, should you be personally willing to pay it? Can’t honestly answer that without normalizing for your relative safety as a function of class/wealth.
2
3
I’m totally willing to accept a 10x-100x higher fatality rate (on a low base like 0.1- 1%) for higher civil liberties but then I’m not most at risk from poverty, crowding etc. Healthcare workers at least knowingly sign up for a risky profession.
5
4
Hmm. Would you, in the wake of this, vote for a measure making forced group quarantine an emergency power, weakening habeas corpus in situations well short of martial law? I honestly don’t know.
1
1
If you could be administered a medical test against your will, by people who have entered your house without a specific warrant, and taken away to quarantine, would you be okay with that? I think it’s technically feasible in the US right now only under martial law right? 🤔
5
2
So... the case that per capita is meaningless because outbreaks are scale free. I’m not sure I buy this. You can’t just assume scale-free based on lack of correlation. A country is a logistics/travel connectivity cluster, not a meaningless fiction.
Quote Tweet
A quick chart for those who keep asking for per-capita adjustment: Here’s population vs total death toll one week after 10th death. No relationship. As I’ve been saying, population does not affect pace of spread. All per-capita figures do is make smaller countries look worse.
Show this thread
Image
2
I think the way to combine cluster analysis and country analysis is some notion like mean-free-path. The right unit of analysis is neither the “country” nor the “outbreak cluster” but something else. Populations are not structureless Petri dishes. They gave a mobility topology.
1
1
I’d buy the scale free argument if cities were large open spaces within walls where people all just stood still 3 feet apart and virus was jumping like a wildfire through a forest. No. We move, at varying speeds, through striations. The connectivity matters.
1
Per capita is likely a terrible way to control for country effects, but country effects must exist and show up if modeled right. Wuhan and New York are both important economic hubs with a lot of travel for example.
1
1
Maybe the right way to adjust is to Ignore country boundaries but segment graph of air travel and road travel connectivity bandwidth by a min-cut/max-flow logic, to define de facto “logistics countries”. This would make Seattle part of an Asian hub, NY part of an EU hub.
1
3
Add an MLM attribution logic. Wuhan gets a fraction of the case count of all 1-degree removed air linked cities, regardless of country they’re in. Recursively back from newest hubs. Many big cities are better connected to international trading partners than national hinterland.
2
Idea for a graph theory project: use the flight connectivity data and dates of travel restrictions to construct city-clusters that should be treated as “Covid19 nations”
1