A sideways look at economics
Identified cases of COVID-19 are rising in Europe, including in countries that were among the hardest hit in the first wave of infections. Popular support for government handling of the outbreak is particularly low in France, Spain and the UK, perhaps unsurprisingly. These are all countries where the reproductive rate of the virus seemed to move decisively above one in late summer. Support has always been low in France and Spain, averaging 39% and 41% respectively in weekly YouGov polls.[1] By contrast, the Italian government appears to have retained the confidence of its electorate, with support averaging 66%, even though rates of mortality have been on a par with those seen in France and Spain. Support for the UK government’s handling of the crisis has dropped from a peak of 72% just after lockdown to 30% in the latest survey conducted on 18 September.
Our chart shows that low support for government handling of the crisis appears to go hand in hand with a high rate of infections. The correlation coefficient is -0.7. But, to fall back on the econometrician’s mantra: ‘correlation does not imply causality’. So, what’s going on here? Does a high rate of infections cause support for government to fall? Or does low support for government cause a high rate of infections? I don’t know, and this simple scatter plot certainly won’t tell me.
Econometric techniques can be used to identify, with a certain degree of confidence, whether relationships exist among a group of economic time series. We might, for example, use the technique of cointegration to find that, over time, nominal GDP and a measure of the money supply tend to move together. But it cannot tell us more than that. It cannot tell us whether increases in nominal GDP cause the money supply to rise, or whether increases in the money supply cause nominal GDP to rise. What to do?
The Nobel prize-winning econometrician Sir Clive Granger suggested a partial solution.[2] When time series X is found to be helpful in predicting time series Y, over and above information contained in the history of time series Y, then time series X is said to ‘Granger cause’ time series Y. Granger causality sometimes runs in both directions, with time series X helping to predict time series Y and vice versa. Of course, this solution is only a partial one, as it does not prove causality in the true, philosophical sense of the word. We might notice, for example, that high levels of expenditure on mince pies through December are useful in predicting the arrival of Christmas in the UK. But does expenditure on mince pies cause Christmas, in the true sense of the word? No, of course it doesn’t. With this caveat in mind, the concept of Granger causality, combined with a theory about why one particular time series might help to predict another one, is nevertheless a useful one.
Back to the matter in hand. By regressing the weekly change in a proxy for ‘R’, the reproductive rate of the virus causing COVID-19, on lags of itself, and on lags of the change in support for government, and by repeating that exercise in a separate regression where the change in support for government was the variable to be explained, we found evidence of Granger causality in both directions. Specifically, when a country’s ‘R’ rate rises, support for government tends to fall a week later. Moreover, when support for government falls, a country’s ‘R’ rate tends to rise two weeks later.
The message as I see it is that the relationship between the spread of COVID-19, and support for government, is a complex one, with each feeding off the other. A country’s ‘R’ rate is, at least in part, a function of human behaviours. If governments around the world want to influence behaviours in a way that brings ‘R’ down, they need to carry their public with them.
[1] Since 12 March YouGov has asked residents of 22 countries to describe how they perceive their own government’s handling of coronavirus. For some countries, the survey has been conducted weekly, for others less frequently. Our chart plots the proportion of respondents to the most recent survey who believe their government has handled the crisis either ‘very well’ or ‘somewhat well’. Only countries where the survey has been conducted in September are counted.
[2] It was also Sir Clive Granger, writing with Robert Engle, who first described the concept of cointegration, as a means of identifying when two series that both tend to rise over time, such as nominal GDP and the money supply, truly move together in the long run, or whether any apparent relationship is in fact a spurious one. Writing in 1987, they found that US nominal GDP and US M2 were cointegrated, but US wages and US prices were not.