Read on Twitter
This thread by @NAChristakis, on the Swedish corona strategy, appears based on a simple causal fallacy. I know it sounds implausible, but let me explain. /1 https://twitter.com/NAChristakis/status/1262012815663149058
Christakis's central claim has a clear causal flavor: "[The] results are in. As expected, the price of this #COVID19 strategy is death." For sure, this is not the way scientists write when they wish to avoid causal (mis-)interpretations. /2 https://twitter.com/NAChristakis/status/1262012822768226305?s=20
What is the evidence for the causal hypothesis? Christakis cites a preprint, by Karin Modig and Marcus Ebeling, available here. /3 https://www.medrxiv.org/content/10.1101/2020.05.10.20096909v1
The paper explores excess mortality from COVID-19. It answers the question "How many more people died this year, relative to a normal year?," or in counterfactual terms, "How many more people died this year than what would have been the case if this had been a normal year?" /4
Notice that this question is very different from the one Christakis is asking, which is "How many more people died this year than what would have been the case if Sweden had chosen another policy?" /5
The answer provided by Modig and Ebeling is an answer to their question, obviously enough — not Christakis's. Their paper simply does not address the causal hypothesis, about how many people died as a result of the Swedish policy (if any). /6
I'm only aware of one paper that has tried to test, in a serious way, the causal hypothesis. It's "Do lockdowns work? A counterfactual for Sweden", available here (via @lilleberlin). /7 https://www.hhs.se/en/research/sse-corona-economic-research-network/research-reports/2020/do-lockdowns-work-a-counterfactual-for-sweden/
The paper uses synthetic control techniques to examine what would have happened in Sweden if the policy had been different. The paper, by Benjamin Born et al., finds that a lockdown would not have made a big difference in Sweden, although it might have worked elsewhere. /8
Inferring that A caused B simply because A preceded B is a fallacy known as post hoc ergo propter hoc, and it's one of the first fallacies you would teach in intro to logic and critical thinking classes. /9 https://plato.stanford.edu/entries/fallacies/#CorFal
Arguably, this is the very fallacy that Christakis is committing. The policy happened before the excess deaths, and he infers that the policy caused those deaths. I hope it's clear why this is a fallacy. /10
I don't know what the answer to Christakis's question is. (Neither do you.) But I do know that we will need to commit to thinking clearly and carefully to sort out these questions, since they are literally a matter of life and death. /11
