There’s No Hint of a Negative Association Between Stringency Index and Excess Mortality in Europe

I previously reported on Youyang Gu’s analysis of Covid death rate and average stringency index across U.S. States. As you may recall, Gu found essentially zero association between the two variables: there was no evidence that states with longer and more stringent lockdowns had fewer Covid deaths.

One weakness of Gu’s analysis is that he used the official Covid death rate as a measure of mortality. This is problematic for two reasons. First, different states may count Covid deaths in slightly different ways. And second, the official Covid death rate doesn’t account for differences in the age-distribution across states.

Since older people are much more likely to die of Covid, states with more old people will tend to have more Covid deaths. For example, more than 20% of Floridians are over 65, compared to only 12% of Alaskans. Hence you’d expect more Floridians to die of Covid, all else being equal.

Incidentally, Gu did find an association between average stringency index and the unemployment rate. (States with longer and more stringent lockdowns had higher unemployment.) This suggests that average stringency index at least partly captures the extent to which different states curtailed economic activity.

As I noted before, the ONS recently published estimates of age-adjusted excess mortality for most of the countries in Europe, covering the entire period from January 2020 to June 2021. And this measure doesn’t suffer from the two problems outlined above.  

I therefore decided to check whether there’s a negative association between average stringency index and age-adjusted excess mortality. Other commentators have produced similar plots before, but I haven’t seen one based on the latest estimates from the ONS.

Results are shown in the figure below. The left-hand chart corresponds to the ONS’s earlier estimates, covering the period up to 18 December. The right-hand chart corresponds to the latest estimates. (The reason I included the left-hand chart is that the one on the right is somewhat affected by the vaccine rollout.)

In any case, neither chart shows any hint of a negative association between average stringency index and age-adjusted excess mortality. States that had the longest and most stringent lockdowns do not have lowest mortality. In fact, both the correlations are positive.

This doesn’t mean lockdown had no effect on the epidemic’s trajectory in any country. But it does suggest that any effect it did have was swamped by other factors (e.g., geography, population density, household structure).

The countries clustered at the bottom of the right-hand chart are the Nordics, Cyprus and Malta. As I noted last time, these are all geographically peripheral countries that used border controls to contain the virus. The only one that made extensive use of lockdowns was Cyprus (see lower right-hand corner). And this appears to have paid off.

However, larger countries that made similar use of lockdowns (see centre of chart) have had much higher mortality. I take this evidence that lockdowns may work in countries like Cyprus or New Zealand when combined with border controls. But that they don’t seem to do much in large, dense, highly connected countries like Britain.