What explains the variation in COVID-19 death rates across European countries? One factor that has been mentioned since early on in the pandemic is population density.
As a BBC article from May of 2020 states: “We know that the density of population is important for the spreading of coronavirus.” The article compares the U.K. and Italy, noting that “the U.K. is considerably more densely populated”.
And indeed, it seems plausible that the virus would spread more easily in areas where people are crammed tightly together than in areas they’re spread further apart. For example, aerosols might linger in the elevator in a multi-storey building, whereas they’re unlikely to travel between one semi-detached house and the next.
This raises the question of how to measure population density at the level of whole countries. The simplest measure is the just the number of people divided by the total land area.
However, this measure doesn’t take account of urbanisation. An ostensibly sparse country could have vast swathes of land where nobody lives. So while the average population density would be low, the majority of people might still be crammed tightly together in cities.
An alternative measure, population-weighted density, is provided by the E.U.’s Urban Data Platform. To calculate this measure for a particular country, the country is first divided into ‘parcels’ of 1km2. The population density within each parcel is computed. Then the weighted average is taken, with weights equal to the population of each parcel.
For example, suppose a country comprises ten parcels, nine of which are completely empty and one of which has a population of 100 people. The average density would be only 10/km2, but the weighted density would be 100/km2.
Using average population density, the densest country in Europe is the Netherlands. But using the UDP’s measure of weighted density, the densest country is actually Spain. (And contrary to the aforementioned BBC article, Italy is slightly denser than the U.K.) While this might seem counter-intuitive, it is consistent with other evidence.
The map below (which was created by the geographer Duncan Smith) shows the distribution of people across Europe at a fine level of detail. Light green and turquoise shades reflect low and moderate population densities; blue shades reflect high population densities.
In Spain, most of the country is unoccupied, but there are a few highly dense areas corresponding to Madrid and other major cities. In Germany, by contrast, there are a much large number of moderately dense areas. As a matter of fact, four of Europe’s five densest cities are in Spain.
As the statistician Alasdair Rae notes: “Spain contains within it more than 505,000 1km squares. But only 13% of them are lived in… So even though the settlement pattern appears sparse, people are actually quite tightly packed together.” (By way of comparison, 60% of Germany’s 357,000 1km squares are lived in.)
I checked to see whether the UDP’s measure of weighted density is associated with COVID-19 death rates across European countries. Two measures of COVID-19 death rates were used: COVID-19 deaths per million up to August 10th, 2021; and age-adjusted excess mortality up to week 51 of 2020. Results are shown below:
There is no relationship on the left-hand chart. And while there is a moderate positive relationship on the right-hand chart, it is mostly driven by Spain (which has both high population density and high age-adjusted excess mortality). When Spain was removed from the analysis, the relationship became substantially weaker.
Earlier in the pandemic, Philippe Lemoine looked at whether weighted density was associated with COVID-19 death rates across European countries, but found “no clear relationship”.
Note that these analyses are highly simplified (and should not be taken as anything definitive). It is plausible that in a multivariate analysis based on a larger sample of countries, clear evidence for an independent effect of population density would emerge.
However, if the true effect of population density were large, you’d expect to find it even in a relatively simple analysis. This suggests that, while population density probably does matter, it’s not the most important factor affecting COVID-19 death rates in Europe.