A few months ago, I complained about an article that asked what effect independence had on Ireland without addressing how the country would have done under continued Home Rule. Doug Muir stepped up to the plate with a simple counterfactual: imagine that the Irish civil war depressed GDP about 20% and that the country would have grown at U.K. rates as part of the U.K.
Is there another way? Well, yes there is! For a research project, I’ve been learning about a technique called “synthetic controls.” Normally we can’t run an experiment in which we take a country just like Ireland but keep it inside the U.K. while comparing it to the one outside. We can look for countries that look like Ireland, but countries rarely have a convenient twin sitting around for comparison. So what to do?
Well, one is to do the best we can with the comparison’s available. This will often take you fairly far.
Another is to compare the country in question with itself before the event you’re interested in. This can also take you fairly far, but of course raises all sorts of problems. For example, it assumes that past performance predicts future results, which is not always true. It also assumes that the changes you’re interested in (in the Irish case, income levels) are not caused by the same underlying forces that caused the event you’re interested in (in the Irish case, independence).
The final one is to use synthetic controls. When used for cross-country estimates, a synthetic control takes a bunch of countries that look more-or-less like the country that you’re interested in and calculated a weighted average of those countries that looks as close as possible like the country you’re interested in. You can then see how that mythical control country performed after a treatment compared to the real country. For example, if you wanted to see the effect of Irish independence, you could take a weighted average of British regions that looked very close to Ireland in 1914, and then compare real Ireland to the performance of that weighted average of British regions. In that case, the experiment would be with remaining fully in the U.K. rather than Home Rule, but it could still tell you quite a bit. (Hey! Anyone want to write that article with me?)
It turns out that someone has run that study for Communist Cuba! In 2013, Felipe Ribeiro, Guilherme Stein, and Thomas Kang did just that. They created a synthetic Cuba that was 55% Mexico, 15% Honduras and 30% El Salvador. These were not the countries that I would have imagined, but they were what provided the best fit with the real existing Cuba between 1929 and 1959. They then compared the GDP per capita of synthetic Cuba with real existing Cuba and found this:
By 1970, the GDP per capital of real Cuba is half that of synthetic Cuba. (The right axis is logged.)
What are the problems with this analysis? Well, three. First, while they got a good match on GDP per capita and the terms of trade, they do not come as close on two other key indicators. Synthetic Cuba had a lot more people in agriculture than real Cuba (68% versus 41%) and much higher illiteracy (65% versus 35%). Second, Mexico in particular was subject to radically different trade shocks than non-Communist Cuba would have been, whereas the analysis assumes that all post-1959 external shocks are identical. Finally, it is not clear that a non-Communist Cuba would have followed the same mix of extreme protectionism and hostility-to-FDI that Mexico followed between 1964 and 1974. (Plus, once the Cubans begin receiving large-scale Soviet subsidies around 1968, then the whole analysis goes out the window.)
That said, most of the problems cited above point to a real non-Communist Cuba outperforming synthetic Cuba. They are not fatal. And the authors admit that their study measures only counterfactual GDP per capita; educational attainment grew faster in Cuba than anywhere else. (Although Costa Rica matched Cuba in terms of reducing illiteracy.)
In short, they find about what you would expect: the Revolution made Cuba poorer by a large amount. What that does not measure, however, is the well-being of poorer Cubans as opposed to the poorer residents of synthetic Cuba. That experiment remains for another day.
P.S. I almost called this post a “tribute,” but Fidel was at the end of the day a bloody dictator and doesn’t deserve one. So I called it a “retrospective” instead.
Synthetic controls are neat. You know that I kind of hate poorly motivated regression analyses, but the weighting in synthetic controls is much more intuitive while still being numerically rigorous.
Posted by: Carlos | November 28, 2016 at 05:43 PM
https://abitmoredetail.wordpress.com/2016/11/28/brief-note-on-the-immiseration-of-cuba-under-communism/
Posted by: Randy McDonald | November 28, 2016 at 11:45 PM