Electricity might be important, but not uniformly soPosted: September 28, 2014
Sinclair Davidson has been playing around with World Bank Development Indicators data on electricity consumption, access and fossil fuel consumption, as well as the Human Development Index, to have a look at the impact of having access to and consumption of electricity and fossil fuels on development.
His point is that electricity—access and consumption—and, by implication, fossil fuels, is important for development. I don’t necessarily disagree, but I think his scatterplots are maybe a little bit crude, so I thought I’d run some non-parametric fits through the data to look a bit closer at the relationship between electricity metrics and development.
It’s also worth pointing out the usual caveats. Simple bivariate analysis can not only be incomplete—in this case, neither Sinclair nor I are trying to disentangle the ‘true’ impact of electricity measures on development—but also misleading. For now, bear in mind the problem of reverse causality—higher HDI scores might lead people to consume more electricity—and omitted variable bias—there might be something causing both HDI scores and electricity consumption or access. Here, since we know what goes into HDI, it’s worth pointing out that one metric used to create is is Gross National Income. A higher GNI leads to a higher HDI, but it also gives governments higher potential tax revenues that they might spend on expanding electricity access.
The other major problem is missing data. Many countries do not report electricity data every year, or at all. This propensity not to collect or report data may itself be linked with HDI and bias our estimates one way or another. If a country reported at all from 2009, I used the most recent figure in my graphs, and excluded countries that haven’t reported since then. I then ran local regressions (loess) to see the relationship between measures of electricity access and consumption and the HDI index.
Looking at per capita consumption and HDI, I find roughly the same scatterplot as the BREE report Davidson reported on; however, as the loess smooth makes clear, this relationship, very strong for low levels of electricity consumption, becomes much more muted as per capita consumption increases.
When it comes to access to electricity, there does seem to be a strong relationship that appears to become stronger as access to electricity is expanded.
Subject to the caveats above, there seems to be a fairly strong relationship between access to electricity and HDI, although this doesn’t necessarily mean of course that higher electricity access causes HDI to increase, although it wouldn’t surprise me if it did.
Part of the noisiness of the data might obscure other interesting relationships among different subsets of the data.
For example, let’s restrict the dataset to countries at a medium or low level of development (HDI<0.7). We get a somewhat more unambiguous linear relationship:
But consider what happens when we narrow our dataset to very highly developed countries (HDI>0.8) we get something of a different story:
To me, the story this suggests is a fairly banal one: the electricity and fossil fuel needs of developed countries aren’t as important for as they are for developing countries. It also reflects, by the way, prior development strategies and experience: it doesn’t necessarily mean that there’s no way for a poor country to become a developed one using renewable sources, particularly since the cost of these (especially solar) is declining rapidly.
Anyway, since Professor Davidson’s argument was a fairly tentative and vague one I’m not attempting a rebuttal, but I think looking more formally at the relationship between electricity usage and consumption with nonparametric fits might give a better, and slightly more nuanced, view of the link between energy and development.