The central theses of the posts are that the IHME model is trash (it is), that the U.S. hasn’t implemented the social distancing seen in Italy, Spain, and Wuhan (it hasn’t), and that other models are predicting much worse outcomes (they are). Despite being right about the main points, these users are about to lose all their tendies because they have a false belief that because the IHME model is trash that the other models are much better. I have a background in consulting, which means I’m somewhat of an expert on selling broken models while pretending they work.
There are two central issues with all all of the epidemiological models
- They underestimate how effective moderate social distancing is
- They are not accurately accounting for how much further along Italy and Spain were when they implemented social distancing or how many more cases there are in these countries
First off, it shouldn’t be understated how terrible the IHME model is. It is not even an epidemiological model, it’s simple curve fitting based on deaths. Many people thought that the IHME model was underpredicting deaths even when it predicted 200k, but prior to the latest revision it’s consistently over-predicted deaths and hospital shortages in the U.S. and actually continues to do so now (by much less). The reason is that they fit Spain and Italy and…
Spain and Italy Have Lots of Corona
If you look at Covid19 Trackers, it shows the US as having 3x more cases than anybody else. We probably do have the most cases, or close to the most cases, but Spain and Italy are way closer than you might think. On March 30, ICL put out a report estimating the number of infections for each European Countries. They estimate on the March 28 that 15% of Spain was infected and 10% for Italy. That means there were over 6 million cases in both Italy and Spain at the end of last month. No wonder their hospital systems were past capacity. Italy and Spain basically only test old people, so even though the number of cases in the US and elsewhere are vastly underreported, it’s actually much worse in Italy and Spain. This also means that models based on reported data from Italy and Spain are just as bad as models based on China, although in this case they are building in the medical system being overwhelmed. This is because testing is unreliable, so everybody relies on deaths, and Italy and Spain have amplified deaths rates due to being overwhelmed third world countries that are just letting older people die. Things aren’t going to be this bad anywhere else in the world. They’ve literally managed to somehow be worse than Iran where the official advice was to oil up your buttholes.
Epidemiologists Can’t Model Social Distancing
Just like the IHME model simply fits a curve to deaths and calls it a day; other standard models aren’t much better. The vast majority of models simply assume that everybody has an equal chance of bumping into everybody else, which is why social distancing is so ineffective. The Chime Model out of UPenn has a single parameter for social contact and then just make up numbers for how much various things impact chance of contact. School Closings lower social contact by 4%, Closing all non-essential businesses by 8%. Do you really think you bump into new people 92% as much as before with non-essential businesses closed? The model used by the MN Department of Health also has a single parameter and decides that Social Distancing reduces contact by 50%, and shelter in place by 80%. The best model is the ICL Report, which has a contact matrix that says how often you bump into people of different age groups; but even then it’s just a matrix where you have an equal chance of bumping into anybody’s grandma. If you look at the methodology they simply say full paper coming soon, so we have no idea how they handle social distancing.
Now, I think that social distancing has been much more effective than epidemiologists have predicted. Why? Because you don’t bump into other people randomly. You mostly see the same clusters of people every day, and while mild social distancing might only reduce your contact with other people by 20%, it probably reduces your contact with strangers by 90%. If you can stop the spread across different groups of people you’ll dramatically reduce the spread.
If epidemiologists knew what they were talking about then they wouldn’t have flip-flopped on masks (April 1 vs April 4), said that travel bans are a bad idea, wildly fluctuated in their death estimates, or said they need 30,000 ventilators less than a week before saying NVM 8,000 is enough. Don’t bet your tendies on nerds drawing curves. That said, this virus isn’t a hoax, it’s just that it’s hard to predict things that you can’t rely on past data for. This is a novel virus with interventions implemented on a scale never seen before. You’re flying by the seat of your pants on this pandemic and everything is going to be a guess at best.
That said, we know U.S. style social distancing is working because we’re seeing the curve flatten in NY. I’m not sure if we’ll see a big drop in deaths or not. There’s a chance that big drop is from when the hospital system stops being overwhelmed, and if we weren’t then we won’t see one. But we are nearing the worst of it in the U.S., and that’s not prediction that’s data. Once the data becomes available, always trust it over predictions (except if it’s Chinese data).
Why We’re Still In Trouble
This blog post by an Economists at Chicago Booth basically sums up why the U.S. is still screwed. The U.S. got lucky that we’re not all dying in hallways like the Italians. We had no testing so we had no idea how far progressed this virus was. It turns out we implemented social distancing early enough that things aren’t going to get truly apocalyptic. That said, we’ve still destroyed our economy, while somehow considering 60-80k deaths from the first wave a success.
The same governmental incompetence behind a lack of testing? That’s still there. Good luck opening up. There’s no way we can implement the test-trace-isolate type programs in Singapore. Which just had to implement a full lockdown despite having those measures. Singapore is also currently hotter and more humid than most of the U.S. gets all summer, so don’t count on the virus disappearing by itself either.
When Fauci says things are back to normal by November, he’s straight up lying to you the same as Trump did for opening up by Easter. Everybody studying this virus knows that this isn’t just goingaway. Come November we’ll just be getting out of full lockdown #2. The market is pricing in things starting to go back to normal after we open up. Morningstar thinks Disney Parks will open in June. Bad news. After we open up, we’re simply back where we were in January. Most pandemics come in waves, the second wave of the Spanish flu was the most deadly. There are things that help, but it’s looking more and more like a miracle treatment isn’t coming; otherwise we would know by now.
The first wave of the pandemic and unemployment are fully priced in now. The next play is the failure to open up. I’m not sure if the drop will come when they admit we can’t go back to normal in mid-May or when we do go back to normal in mid-May and then see a spike in June and July. Nobody knows, which is why it’s still not priced in yet. Definitely, by September it will be clear the U.S. is in big trouble.
Pandemic priced in. Incompetence in reopening the economy probably not.
Disclaimer: This information is only for educational purposes. Do not make any investment decisions based on the information in this article. Do you own due diligence.