Five years ago, demographer Richard Cincotta claimed in a piece for Foreign Policy that a country’s age structure is a powerful predictor of its prospects for attempting and sustaining liberal democracy. “A country’s chances for meaningful democracy increase,” he wrote, “as its population ages.” Applying that superficially simple hypothesis to the data at hand, he ventured a forecast:
The first (and perhaps most surprising) region that promises a shift to liberal democracy is a cluster along Africa’s Mediterranean coast: Morocco, Algeria, Tunisia, Libya, and Egypt, none of which has experienced democracy in the recent past. The other area is in South America: Ecuador, Colombia, and Venezuela, each of which attained liberal democracy demographically “early” but was unable to sustain it. Interpreting these forecasts conservatively, we can expect there will be one, maybe two, in each group that will become stable democracies by 2020.
I read that article when it was published, and I recall being irritated by it. At the time, I had been studying democratization for more than 15 years and was building statistical models to forecast transitions to and from democracy as part of my paying job. Seen through those goggles, Cincotta’s construct struck me as simplistic to the point of naiveté. Democratization is a hard theoretical problem. States have arrived at and departed from democracy by many different pathways, so how could what amounts to a one-variable model possibly have anything useful to say about it?
Revisiting Cincotta’s work in 2014, I like it a lot more for a couple of reasons. First, I like the work better now because I have come to see it as an elegant representation of a larger idea. As Cincotta argues in that Foreign Policy article and another piece he published around the same time, demographic structure is one component of a much broader and more complex syndrome in which demography is both effect and cause. Changes in fertility rates, and through them age structure, are strongly shaped by other social changes like education and urbanization, which are correlated with, but hardly determined by, increases in national wealth.
Of course, that syndrome is what we conventionally call “development,” and the pattern Cincotta observes has a strong affinity with modernization theory. Cincotta’s innovation was to move the focus away from wealth, which has turned out to be unreliable as a driver and thus as a proxy for development in a larger sense, to demographic structure, which is arguably a more sensitive indicator of it. As I see it now, what we now call development is part of a “state shift” occurring in human society at the global level that drives and is reinforced by long-term trends in democratization and violent conflict. As in any complex system, though, the visible consequences of that state shift aren’t evenly distributed.
In this sense, Cincotta’s argument is similar to one I often find myself making about the value of using infant mortality rates instead of GDP per capita as a powerful summary measure in models of a country’s susceptibility to insurgency and civil war. The idea isn’t that dead children motivate people to attack their governments, although that may be one part of the story. Instead, the idea is that infant mortality usefully summarizes a number of other things that are all related to conflict risk. Among those things are the national wealth we can observe directly (if imperfectly) with GDP, but also the distribution of that wealth and the state’s will and ability to deliver basic social services to its citizens. Seen through this lens, higher-than-average infant mortality helps us identify states suffering from a broader syndrome that renders them especially susceptible to violent conflict.
Second, I have also come to appreciate more what Cincotta was and is doing because I respect his willingness to apply his model to generate and publish probabilistic forecasts in real time. In professional and practical terms, that’s not always easy for scholars to do, but doing it long enough to generate a real track record can yield valuable scientific dividends.
In this case, it doesn’t hurt that the predictions Cincotta made six years ago are looking pretty good right now, especially in contrast to the conventional wisdom of the late 2000s on the prospects for democratization in North Africa. None of the five states he lists there yet qualifies as a liberal democracy on his terms, a “free” designation from Freedom House). Still, it’s only 2014, one of them (Tunisia) has moved considerably in that direction, and two others (Egypt and Libya) have seen seemingly frozen political regimes crumble and substantial attempts at democratization ensue. Meanwhile, the long-dominant paradigm in comparative democratization would have left us watching for splits among ruling elites that really only happened in those places as their regimes collapsed, and many area experts were telling us in 2008 to expect more of the same in North Africa as far as the mind could see. Not bad for a “one-variable model.”
Dart-Throwing Chimp 
Artem Kaznatcheev
/ April 14, 2014To build on my previous comment a little, I wanted to play with this post:
I think this concern, is a great example of what I meant when I was writing that political science “is still searching for the terms and relevant variables.” If you are used to folk variables not being good predictors then it is reasonable to expect that no simple one-variable theory exists. However, there is also no reason to believe that the variables that feel most ‘natural’ to us should necessarily be meaningful. Instead, work like Cincotta’s points us to a potentially synthetic seeming variable (although ratios of young-to-old in the population still seems relatively natural) that we should consider as a central word in a new vocabulary. Similar to how Newton wanted to relate Force to acceleration and so made up an arbitrary synthetic concept called mass to be their ratio.
You write that the young-to-old ratio, like the infant mortality ratio, aggregates a lot of other relevant variables. Here, you are using a folk theory and its associated causality to suggest that these very predictive variables are derivative from some more ‘fundamental’ folk-theory variables. However, what happens if we try flipping our language? And instead say that “young-to-old ratio is important in the prediction of a lot of other natural seeming variables” and just for the sake of exploration quiet those voices in our heads that cry fowl based on what seems like natural chains of causality. Can we build a better theory from this angle?