RCE: Visualizing the Results, Part II

In the last post on the RCE, I presented some basic facts about the data you gathered to show that it was generally reliable. Your data was highly correlated with existing professional measures of democracy, and there were few cases of countries where the judgment of POLS209 was too different from that of other well-known datasets on democracy (including V-Dem and Freedom House).

In this post, I’ll show you a few comparisons between your data on democracy and two kinds of epidemic data: historical epidemic data gathered by the Centre for Research on the Epidemiology of Disasters (CRED), and current data on COVID-19 deaths and cases from the European Centers for Disease Control (nicely packaged by the sociologist Kieran Healy). In another post, I will give you some comparisons with data on policy interventions from the CoronaNet project. You will need to use this post in the "General Results" section of your essay - ask questions!

We can use these comparisons and correlations to get a sense of whether there is any likely association between how democratic a country is, and how well it responds to epidemic disease. But a few caveats are in order before we begin. First, data on epidemic disease mortality is not necessarily reliable. Its collection is sometimes difficult and subject to political pressures for falsification. For example, different countries apply different criteria for testing and for classifying deaths due to COVID19. In some countries (e.g., Russia and Mexico) there are indications of political pressure to understate the numbers of deaths from Coronavirus, and in other countries too few people are being tested to trust the numbers of reported cases. For more on some of these issues, take a look at the Our World in Data Coronavirus webpage.

Moreover, epidemic disease mortality is affected by many different factors, not all of which are political. Climate, timing of disease, age structure, geography, and many other things will interact with political and social factors to affect overall mortality. For a good review of how these factors seem to interact with COVID-19 mortality so far, take a look at this paper on Political and Social Correlates of Covid-19 Mortality, on which I draw here.

With these things out of the way, I will first look at the correlation between the data on mortality from various types of disasters in the EM-DAT dataset (International Disaster Database) and your measure of democracy. First, here’s the “raw” correlation between democracy and disaster mortality, across biological, climatological, geophysical, hydorological, and meteorological disasters:

This is not very promising! There is very little (if any) correlation between mortality from disasters of any kind, and how democratic the country is according to you.

There is a slightly larger correlation between disaster mortality and democracy (less mortality with more democracy) if we use the much more extensive measure of democracy from the V-Dem (Varieties of Democracy) dataset, though it’s still very small, and swamped by the wide variation in outcomes:

One obvious problem here is that richer countries should be more likely to mobilize more resources, and hence be more able to prevent mortality from disasters, regardless of the regime; and that larger countries are more exposed to natural disasters, so should experience greater total mortality overall. Since economic development (and other factors) and democracy may all affect disaster mortality, in theory what we want is to understand the correlation of democracy net of economic development, population size, and potentially other factors. An extremely simple regression model of total deaths against these factors suggests that democracy is not correlated with mortality for any type of disaster after we control for economic development and total population:

What this graph says is that economic development and total population are correlated in the expected way with mortality from disaster, and these correlations are robust to the other factors: larger countries experience greater disaster mortality (the coefficient, indicated by a dot with a black bar, is positive), and richer countries experience lower disaster mortality (the coefficient is negative). But it is not really possible to tell whether democracy is associated with lower democracy, whether or not we use the V-Dem measure or this class’s measure of democracy, as the error bars on the coefficient are too wide, and the measure of democracy from this class and that of V-dem point in different directions.

So it doesn’t look like democracy matters much for disaster mortality. Maybe it matters a little, and perhaps if we added more factors to our model and tortured the data enough we could find a correlation; but certainly not to the extent that it pops easily out of the data. Why not? Why aren’t the advantages of democracy in terms of accountability and information more obvious here?
Let’s repeat this exercise using current mortality data from the COVID19 pandemic. First, we look at the raw correlation between democracy and COVID 19 total mortality (as well as mortality per capita):

As we can see, there is no particular difference between democracies and non-democracies here; if anything, the more democratic countries seem to have done worse. Why might that be the case?
Finally, let’s do a bit of modeling using GDP per capita and total population:

Whichever measure of democracy we use, it seems that democracies have done slightly worse than non-democracies in dealing with COVID19 mortality (the coefficient is positive, so more democracy has meant on average more mortality), though it is possible that this effect would go away if we added further controls, as in this paper. What might be going on here? Why aren’t democracies doing better?