[thu2111]

They also look at South Korea.

However, that doesn't really matter. Sweden is a counter-example because lockdown theory doesn't posit any sort of Swedish exception, so the existence of this place alone disproves the models. Going further or using longer term data is good, but strictly speaking unnecessary.

Also, you're claiming that there was some sort of "big spike" in Swedish cases and that this would change the conclusions. It might do but if Sweden did anything since January it would have been tightening restrictions - around Christmas the international pressure got to their leaders and they started to get tougher, not laxer. So it would be hard to say that this would invalidate the conclusions. At any rate, lockdowns weren't justified on the basis of PCR positive results, they were justified on the grounds of deaths. Swedish death curves are the same as other countries that did hard lockdowns and there has been no big Swedish spike:

https://ourworldindata.org/explorers/coronavirus-data-explor...

This is important because for Sweden the models predicted predicted ~90,000 deaths from COVID alone, there were 98,000 deaths in total from all causes in 2020. That's clearly a huge miss.

Like the other anti-lockdown papers I've spot checked, I'm not impressed.

And of the papers arguing lockdowns work, they're all free of methodology errors in your view? Because that's definitely not what I've seen. The thing with lockdowns is, it's ultimately very simple. You don't actually need any complex analysis to see the truth. You can just look at the predictions and then go to ourworldindata for the reported curves. You should see huge sudden drops or surges when lockdowns and mask mandates are imposed or released, and it should be visible in all the countries that made changes. Additionally countries that did nothing should be have dramatically bigger curves than those that did lots. You can't see anything like that in the data.

[jjoonathan]

> lockdown theory doesn't posit any sort of Swedish exception

"Lockdown theory" doesn't posit an absence of lockdown variables. Perhaps the straw man you would like to imagine does, but it's alone in this regard. A "Swedish exception" doesn't disprove "lockdown theory" any more than a lottery winner disproves the idea that earning money is generally hard work.

> you're claiming that there was some sort of "big spike" in Swedish cases

    April 2020: 800 cases/day   <---- paper data ends
    November 2020: 3000 cases/day
    December 2020: 5000 cases/day
    January 2021: 7000 cases/day   <--- paper published
 

> around Christmas the international pressure got to their leaders and they started to get tougher, not laxer

    Feb 2021: 3000 cases/day

Looks like it worked.

> And of the papers arguing lockdowns work, they're all free of methodology errors in your view?

I've spot checked a couple and the quality of the analysis was considerably higher. They generally rested on much sturdier statistics, like estimating effect size of policy interventions on R, which don't rely on cherry picking pre-spike Swedish data.

> You should see huge sudden drops or surges when lockdowns and mask mandates are imposed or released

No, you would expect to see hard-fought dips in R after major interventions. Booms and crashes will act as a threshold on integrated R because such is the nature of exponential growth.

To be clear, these interventions frequently lead policy -- they can come from other countries, they can happen through policy of major institutions, ad campaigns, even distribution of news stories regarding other countries. Also, policy lags and suffers from inefficacy from people ignoring and even fight mandates, and this particular brand of inefficacy should be factored out for normative purposes. All of this dampens the effect and spreads it out in time. You would expect the dips in R to be messy but measurable given generous windows, and they are.

More than that, virus transmission isn't a cosmic mystery. We know exactly how it happens and the mechanism by which lockdowns impact its transmission is so straightforward that even medieval plague doctors were able to get it correct. The "lockdowns are ineffective" hypothesis needs some serious magical thinking even before you start cherry-picking data and ignoring confounding variables. It seems like an odd hill to die on.

[thu2111]

"Lockdown theory" doesn't posit an absence of lockdown variables

Yes, it does. What you're calling a strawman is the papers published around this time last year that motivated lockdowns globally. No model presented to justify these policies can explain why Sweden (and other places with similar outcomes) did not experience the large rate of deaths that were universally predicted in the absence of the harshest measures.

Looks like it worked.

No. The period I'm talking about is December time. Thus if you insist on a link between lockdowns and cases existing then it'd actually prove the opposite, as the number of cases went drastically up as their policies started to change, not down.

I'm not arguing that though because when you look at the data, the fact that Sweden tightened restrictions and "cases" surged tells us nothing. In other places the opposite happened. And in still others they tightened restrictions, or made them laxer, and nothing happened. Most recent example: Texas. For a causal effect to exist, there should either always be a causal effect, or there must be some theory that explains why sometimes it doesn't exist. Neither is available.

I've spot checked a couple and the quality of the analysis was considerably higher. They generally rested on much sturdier statistics, like estimating effect size of policy interventions on R, which don't rely on cherry picking pre-spike Swedish data.

Which ones did you read? I've had the opposite experience. The papers claiming lockdowns worked all had absurd problems, often excluding Sweden or including it but cancelling it using arbitrary and huge Sweden-specific variables, because of course they must. If they included this country their models could not achieve fit, as counter-examples disprove the theory the model is trying to represent. An example of this problem is the paper by Flaxman et al: it was based on circular logic, hid the fact that their Bayesian priors assigned Sweden a 1 in 2000 chance of existing (i.e. was wrong), and the model couldn't explain US data.

BTW you're arguing here that including places like Sweden, South Korea, Texas, Florida, South Dakota etc in the data analysis is "cherry picking". This is an argument I've seen a few times lately. It's a logical fallacy; having control cases in a data analysis is not cherry picking, it's the opposite, it's including all the data. Once again, lockdown theory is very simple. No epidemiological group predicted last year that "lockdowns are critical to avoid mass death everywhere except Sweden and a few other places". How could they? There's no scientific theory that makes these countries special, leaving the conclusion that they aren't special and forcing us to accept the null hypothesis. Because obviously, for lockdowns to work there must be a clear difference in outcomes depending on how harsh the restrictions were. That requires all the data to be analysed, and when such analysis is done it shows no reliable correlation.

It's actually the studies that exclude the inconvenient data points that are the ones cherry picking.

Also, policy lags and suffers from inefficacy from people ignoring and even fight mandates, and this particular brand of inefficacy should be factored out for normative purposes

Another common argument made recently, e.g. Fauci is trying this now. In reality compliance has been measured to be exceptionally high. Obviously, compliance with business and offices closures is total. There have been no reports of any businesses refusing to comply because the tiny number that made a political stand just got big fines or shut down. And as for masks, compliance is usually measured as being very high e.g. >90% in the USA:

https://www.marquette.edu/news-center/2020/marquette-researc...

You would expect the dips in R to be messy but measurable given generous windows, and they are.

No, the models are quite clear about this. How many have you examined, I wonder? Effects are expected to be visible in test results within days (a serial interval) and deaths within a month, and were predicted to be massive, like 75% different. No such outcomes are consistently visible in the data, although they should be because, again, the theories underlying these mandates are quite simple and don't have exceptions that render them useless even when compliance is high.

We know exactly how it happens and the mechanism by which lockdowns impact its transmission is so straightforward that even medieval plague doctors were able to get it correct

Yet lockdowns don't impact the transmission, as proven by counterfactuals. And this is indeed very easy to explain. Most transmission occurs in poorly ventilated places where people are exposed to sick people for long periods, places like: homes. Care homes. Hospitals. Lockdowns do not affect these places and cannot, by their very nature. Places that contact tracing studies show very little transmission even in places without lockdowns: businesses, shops, offices. Why - because when people feel sick they stay home and don't go to those places. It's exactly what you'd expect to see and attempting to make it to be otherwise has required a lot of very poorly supported new medical theories to be invented on the spot, like massive scale asymptomatic transmission. Something that large scale studies have not actually found any evidence for (e.g. the Wuhan city-wide PCR testing programme couldn't find evidence of asymptomatic transmission).