With a recent publication in Nature that reported lockdowns have no effect on covid-cases or covid-deaths, there are now over 30 studies that fail to find any covid-reducing benefits of lockdowns. Worse, across countries and time, more severe lockdowns are just leading to more deaths as they enforce unhealthy behaviour and reduce the capacity of the health system to look after people.
Yet, many think they can clearly discern the benefits of lockdowns in the available data. And not just for lockdowns: also for masks, which have been aptly described as garden fences supposed to stop mosquitoes. An instructive method to challenge these beliefs is to play the game “spot the policy effect” wherein one assembles graphs of how claimed cases or deaths developed over time in a set of countries and then asks the onlookers to predict when lockdowns or compulsory masks (or other restrictions) came into effect, if at all. Take the following graph of a recent Financial Times article:
This picture shows the number of claimed covid-deaths per 100 thousand inhabitants per day for two European countries. Guess which one of these countries instigated ‘tough’ lockdowns at what moment, and try additionally to guess when masks or other restrictions might have come in? It is basically impossible to get it right unless you already know the answer.
If you think lockdowns or masks work, you’d think they were introduced a few weeks before some clear decline or change in the trend of the numbers of covid-deaths, right? And you’d think the place with less deaths in the same continent did tougher lockdowns, right? Well, you’d then guess wrong: the blue line is for the UK which instigated its first lockdown at the end of March 2020 after the trend had already switched, instigated mask policies in the summer and then a more or less continuous new lockdown from early November. That second lockdown was instigated just before the first peak of the second wave, but 2 months before a spectacular third peak that made the UK the country in Europe with the highest proportion of covid-deaths among all large European countries.
The other country (in red) is Sweden which had no lockdowns, but interestingly enough has had the most stringent average restrictions among all Scandinavian countries and has had more covid-deaths than any other Scandinavian country. So whilst compared to the UK, Sweden shows a similar disease pattern without copying the UK restrictions, among Scandinavia Sweden is the ‘bad boy’ because it was more restrictive than others and has ‘thus’ had more covid deaths. So whilst I too initially thought it possible lockdowns could have some covid-reducing effect, the data forced me to reject that idea and to start wondering how lockdowns were causing covid-deaths.
Yet, it is a very prevalent human tendency to look at the same graphs and discern some ‘proof’ that lockdowns and masks work. You can’t make up the strange reasoning lockdown-believers latch onto to find what they expect to see, namely that restrictions surely ‘must’ have an effect. They invariably want to see more graphs in more countries. Even if you tell them that the 10 countries in Europe with the highest proportions of covid-deaths have all had variations of ‘tough’ lockdowns with big covid-deaths waves coming months into those lockdowns, they simply dismiss it out of hand. They insist there is some version of lockdowns that surely did work. They know it for certain. Just like the UK government advisers insist lockdowns in the UK were useful.
As it so happens, I did research into this “seeing evidence where none exists” phenomenon in 2008 in Australia. Together with Juan Baron, now at the World Bank, I ran lab experiments wherein we tried to elicit cult-like behaviour among regular university students. The published paper was thus called “The Cult of Theoi”. Let me explain the gist of what we found.
We invited groups of 20 students into a lab (over 500 students in total) where those 20 students were then set behind a computer in which they had to follow instructions for 20 rounds. Each round consisted of a first phase in which students could earn ‘units’ with some simple task. Let us call those units ‘apples’ for ease of exposition. One of the ways these apples were earned was via simple mathematical sums (addition/multiplication), but also via competitive games with the other students. It turned out to be unimportant how the apples were earned for the second phase of each round, which was the phase we were interested in.
The second phase of each round was all about the price of apples earned in the first round. The students were told that it was unknown how the price of apples came about, but they were given the option of sacrificing a proportion of the apples earned in the first phase to “Theoi, the market maker”. Deciding on how much they would sacrifice was the only thing they needed to decide in the second phase. After their sacrifice, they would receive an apple price that was unique to them (each student got a different price in each round). Students were indeed paid what their remaining apples were worth, so whatever they sacrificed was truly lost to them.
The reality was that the apple price was set by a random number generator, as if determined by a throw of the dice. So the apple price differed round to round and student by student in an entirely random manner, with no relation at all to any sacrifices by any students. The research question was how much students would sacrifice and whether they would learn over time that their sacrifices were futile, just like (arguably) with lockdowns and masks today.
The main result is in the graph below that shows per round what the average proportion sacrificed was. You see that in the first round students on average sacrificed 40% to Theoi, and even after 20 rounds still sacrificed over 25%, even though there was no benefit ever of doing so. You do see some aggregate learning, but limited. A more close-up analysis revealed that something like a third of students had figured out sacrifices were useless and stopped donating anything substantial to Theoi, whilst two-third kept on sacrificing 40% to 50% to Theoi every round right up till the end. They never learned.
As if we were trying to make the experiments even more relevant for today, we did a version of this experiment wherein we gave the students all the information on the previous apple prices and sacrifices of the other students in the experiment, which thus gave them 20 times more information than just their own choices and outcomes. It was as if we were giving them information on 20 countries rather than just their own.
What would you expect? Faster convergence on the truth? The next figure shows the average sacrificed under this version of the experiments.
As you can see, there was an effect, but not a huge one. If anything, the level of sacrifices in the first few round is even higher with more information, but the drop is bigger than before and after round 15 the average sacrifice is just below 20%, which was significantly less than in the main experiments. So there is some learning with more information, but not that much.
What we could see in the experimental lab was that the more serious students would take a lot of time peering at the history of apple prices and sacrifices of others in their experiment, usually ending up deciding on a big sacrifice. So their long gazing at the data gave them some reason to believe the sacrifices were having benefits, even though that was entirely false. They were seeing things that weren’t there.
This is how one should also view the deductions of many of the lockdown and mask believers: because they expect these things to work, they stare at the data long enough to find something. Just like the students in my 2008 experiments. They ignore the headline evidence that is normally used to decide whether something is likely to work: information across time, regions, and countries. They simply insist ‘it must be so’. The underlying logic is that of the sacrifice: pain must surely come with a gain. Well, sometimes pain is just pain without gain.
The futility of lockdowns, masks, curfews, and many other restrictions was the accepted scientific consensus from before February 2020. At best, such measures were thought to delay matters with huge costs and no long-run benefits. It was a consensus codified in textbooks and to-do blue prints in the UK, the Netherland, the US, and also Australia. In a bout of panic that consensus got ignored in March 2020 as fear and the ‘something must be done’ logic overwhelmed the science of decades. A draconian medical experiment got enforced upon Western populations without even attempting to ascertain the likely costs of doing so, something that is a clear crime under the Nuremberg code on medical experiments that underlies public health laws in most Western countries. Now, slowly, the scientific consensus is returning to what it was before. We are finding out empirically that there is not even a delay benefit from lockdowns or masks (which has surprised me because I did expect lockdowns to at least have a delay effect). The folly and horror of the experiments is thus becoming clearer and clearer, though there are still millions who peer into the covid data soup and see the benefits of lockdowns they expect to see. For months, I expected to find some delay-benefit too, but it just aint so.