Comparing COVID mortality rates: What do governmental and behavioural factors tell us?

Comparing international death rates from COVID is difficult because of different counting methods. However, we need to make some progress here, and work out why some countries have done better (and worse) than others. We are in the first draft of history, and what follows is an attempt to begin to start a debate about how politics has functioned. There are more detailed methods below, but given that most people will want to see results, we’ll start with them.

All the calculations below are based on COVID deaths per million of population as of the morning of 23/5/2020 (UK time).

Countries that have done comparatively poorly in terms of COVID deaths

Of the countries included in the sample here, three patterns of causal factors appear important, with the first two being more important than the last one.

First, where countries have high levels of smoking and drinking, and low levels of government executive power, they are 0.84 (out of 1) consistent with a higher COVID death rate. 10 countries are included in this solution.

Second, where countries have high levels of smoking and drinking, and low levels of unitary government, they are 0.79 consistent with a higher COVID death rate. 8 countries are included in this solution.

Finally, where countries have high levels of social expenditure combined with high executive power and low of unitary, they are 0.85 consistent with a higher COVID death rate. However, this solution covers two countries only, and one of the countries actually achieves a low COVID death rate – it is an inconsistent case (Australia). This sounds a bit confusing (how can one county out of two have a consistency of 0.85?), but it is based on looking at patterns of these two factors across all the 25 countries in the sample, and so the 0.85 consistency occurs across them.

The overall solution is 0.8 consistent and covers 0.71 of the cases (out of 1).

The countries included in these three solutions, along with their political mapping (executive power and federal unity) is presented blow. Empty dots are other countries that are not included here:

In this we can also see three countries which have done relatively well in terms of COVID mortality – Australia, Austria and Japan. They appear because they have the same pattern of causal factors as countries which have done poorly, and so their success clearly depends on other elements. Australia, for example, may be related to strong border controls and early-lock-down, as well as relative geographic remoteness. We are never going to be able to account for something as complex as COVID with only four causal factors.

We can also see we are missing some countries which have done relatively poorly from the list above – including Great Britain. More on them below.

What do successful countries look like?

There are three solution terms for relatively successful countries.

The first combines high executive power with high unitary government, and covers 5 countries with a consistency of 0.69.

The second combines low social expenditure with low smoking and drinking and high executive power, and covers two countries only with a consistency of 0.84.

The third combines low social expenditure with low smoking and drinking and high unitary government. Again, it covers only two countries with a consistency of 0.81.

The overall solution has a consistency of 0.7 and covers 0.634 (out of 1) of the cases.

Again, we can produce a graph of the countries covered:

We can see there are fewer countries included in this solution than that for the high mortality graph – that is because the countries which have done well, in terms of the four causal factors included here, have less in common.

We can also see the appearance of ‘GBR’ at the top right of the graph – but also ‘CAN’ (Canada) at the bottom right – both of which have a pattern of causal factors which suggest they ought to have low rates of mortality – but empirically don’t. In other words, they are doing worse than we might expect.

What does all this mean?

This is an early analysis of an evolving situation, with relatively few causal factors. However, there are some conclusions we can draw.

First, the countries which appear to have the higher rates of COVID mortality also have higher rates of existing smoking and drinking. This goes against work done by Ben Goldacre and others and which is based on analysis of individual patients. This is an interesting tension – at a societal level, higher rates of smoking and drinking appear to be linked to higher rates of COVID death, but at individual levels, smoking appears to act in a small way as a protective factor. This is clearly a factor which is worth investigating further.

Second, lower levels of unitary government appear in two of the three solutions for higher rates of COVID mortality. This appears to point to the need for co-ordinated governmental action, and the weakness in some countries not to able to achieve this.

Third, countries with higher levels of unitary government appear to do well in terms of achieving low COVID mortality, with that factor appearing in two of the three solutions.

Fourth, countries with lower levels of smoking and drinking also appear to do better in terms of overall COVID mortality, again in combination with other factors, but with that factor appearing in two of the three solution terms.

Finally, we have Great Britain, which has governmental factors (high unitary government) associated with good COVID mortality, but has clearly done worse than it ought to have done. This is because of other factors which have offset this advantage. Although the GBR is graded as having high federal unity, it is perhaps open to question whether it has displayed this in its COVID response – with government struggling to achieve a coherent approach to testing, and to tracking and tracing those infected by the virus – especially in care homes. Perhaps this points to a lack of governmental capability in the face of a crisis, as well as to an implementation gap once the approach had been decided.

What did you do?

The above is the result of combining OECD data on smoking and drinking, along with Lijphart’s ‘Patterns of Democracy’ measures for unitary government and executive power, for 25 countries (which had data for both the behavioural and governmental factors), combined with data for COVID mortality per £1M people at 23/5/2020, using the data from https://www.realclearpolitics.com/coronavirus/ as a starting point, and checked against other sources.

Once this data had been compiled, it was analysed using Qualitative Comparative Analysis using the ‘QCA’ package in R by Adrian Dusa. Data was calibrated both graphically and through cluster analysis to establish crossover points, and then calibrated using the direct method. Necessary conditions were looked for, truth tables constructed looking for both consistency and PRI consistency, and sufficient conditions calculated. The intermediate solutions are presented above, but conservative and parsimonious solutions were also calculated, based on assumptions from OECD work that lower mortality rates from higher rates of social expenditure and lower rates of smoking and drinking. Even when these assumptions were relaxed (conservative and parsimonious solutions), they produced solutions very much in line with those presented above.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


%d bloggers like this: