March 16, 2020
Why many of the Covid-19 statistics are misleading
We are no virologists, but successive financial crises have taught us how to think about genuine uncertainty. They taught us especially not to be over-reliant on easily-available statistics. When you are dealing with catastrophic risks whose probability is fundamentally incalculable, it is best to acknowledge the existence of genuine uncertainty - one that is not calculable. Economists will know the notion of Knightian uncertainty, a useful concept that distinguishes between risk and uncertainty.
When we heard about the impact of the virus in Wuhan in January, we argued that the EU should stop flights from China and close the Schengen border. We had no information about the spread of the virus, but thought it would have been a comparatively cheap insurance policy. When the cases in Italy started to rise, the EU missed another opportunity to shut the Schengen borders. This would have been an easily available lock-down measure because the infrastructure already exists.
On this point we think that some of the experts miscalculated. They estimated that border closures would have only bought a day. But that calculation ignores the importance of statistical outliers. The average rate of spread, as measured by the R-zero statistic, tells us nothing about super-spreaders. If only a single super-spreader could have been prevented from entering the EU, the situation in some of the hotspots in northern Italy would have been very different.
Over the last few days the EU has moved from complacency to panic. The partial border closures that are now being introduced are still useful in our view, but less so than they would have been two months ago. The experiences from Singapore, Korea and Japan are telling us that early lock-down measures work.
Statistics can mislead us in many ways. Another example are emergency beds in hospitals. Again, it is not the average that matters. Italy has more emergency beds than France. The problem is that the hot spots of Lombardy and Emilia-Romagna cannot cope. Even the much better-equipped German system would not have been able to cope with what happened in Bergamo. The average number of hospital beds per 100,000 people is meaningless. What counts is the number of beds in your area, in every area.
And finally, also be careful about worst-case scenarios like those that are making the rounds in the UK papers. An infection rate of 80% is extremely unlikely. We do not know the true mortality rate because the number of infected people outnumbers reported cases by a factor 10 or 20, possibly higher. The higher that number, the lower the mortality rate. So don’t combine the worst-case scenario for the spread of the virus with the worst-case scenario for the mortality rate.