Do not jump to conclusions about lockdowns before all the data is in and analysed
Wolfgang Münchau
Sweden did not follow the same strict lockdown measures that much of Europe adopted © Jonathan Nackstrand/AFP/Getty
Only a fool would draw strong conclusions from sketchy data. The biggest fools this year were those who prematurely declared the spike in Swedish infections from April until June as evidence that the Swedish decision not to lock down their economy was wrong. I recall many armchair epidemiologists hyperventilating about Sweden’s obstinate refusal to follow the rest of the world.
Over the summer, Sweden took other steps to control the virus, including local lockdowns, and cases started to rise again in other parts of Europe. Now, Sweden’s new infection statistics look better than much of the EU. But we shouldn’t draw any conclusions yet. It was wrong two months ago to condemn the Swedish strategy based on that data, and it would be equally wrong to draw the opposite conclusion now.
It took many years for epidemiologists and biostatisticians to understand the infection rate and progression of the 2003 Sars outbreak. It will not be different this time.
Experts are most at risk of error when they go beyond their narrow field of expertise — and particularly when they venture into the world of statistics. In some cases, they get the maths wrong. But often they fail to see subtleties.
Years ago, when I was researching the asset bubble that later gave rise to the 2008 financial crisis, I studied value-at-risk data for banks. These statistics are the way bankers measure their risk exposure on a day-to-day basis. Back then, senior bank executives treated VAR like football scores, looking for winners and losers.
I found that the tiniest shifts in a measurement parameter had massive implications on the final result. The obvious conclusion is that you cannot reduce something as complex as a bank’s risk exposure to a single number. Today’s equivalent fallacy is the idea that you can compare the infection rate of one country with that of another and draw policy conclusions in real time.
It is a more profitable use of time to look behind the data. In Sweden, it is now clear that a major reason for the spike in infection rates in the early stages of the crisis was the failure to protect care homes. Protecting the elderly is where Germany, for example, did really well.
The infection rate in Sweden also showed strong geographical variation. Most of the Swedish cases were concentrated in two regions, including Stockholm. Meanwhile, the southern Swedish city Malmö is close to the Danish capital, Copenhagen, separated by the narrow Oresund Strait. Malmö’s rates look good by comparison with Copenhagen, even though the two operated under different lockdown regimes.
I don’t know why regional gaps were so strong, and my interlocutors in Sweden do not either.
If you want to make grand pronouncements about Swedish lockdown policies and infection rates, you should probably make an effort to understand this first.
Policy in times of Covid-19 amounts to decision-making under extreme uncertainty. The latest Swedish numbers do not prove or disprove anything. But before policymakers order something as extreme as another lockdown, they should have had incontrovertible statistical evidence, not just a bunch of numbers that feed their confirmation bias. As long as statistical doubt persists, we certainly do not want to do this twice.
A lockdown is an extreme policy measure and its consequences will not become apparent for some time. I have no doubt that it will end up increasing inequality. Unemployment and corporate insolvencies will rise once the support measures are withdrawn. Although stock market indices have fallen and recovered, these are just averages.
Behind them stand huge shifts of capital from old to new sectors. If people continue to work from home, this will boost residential and rural areas at the expense of city centres and shift resources from commercial to residential property.
I consider the lockdown reflex as currently the biggest threat to western capitalist democracies. The data at this point do not tell us what we need to know, but they inject useful uncertainty into the consensus that a lockdown is the only way to respond to a global pandemic.
To put it another way, next time we had better make sure that the data justify such actions beyond reasonable doubt and put policies in place to deal with the consequences. We did not do that the first time.
It is my hope that Sweden’s experiment will eventually provide us with enough data to make a valid cross-country comparison. Until then, we should keep watching closely.
0 comments:
Publicar un comentario