Monday, January 20, 2014
Does America's survival mean that it's a resilient country?
Author: Noah Smith
A friend asked me an interesting question yesterday: If a country survives a long time, does that mean we should expect its institutions to be resilient to future shocks? For example, American democracy has been around for over two centuries now, and has survived wars and depressions and ideological movements that have brought down the regimes of most other nations. Doesn't this mean that America has proven itself to be especially resilient?
The answer is: Maybe. It depends on what causes countries to topple.
Which is more reliable, an airplane fresh off the factory floor, or an airplane that has been in service for 30 years? Which is likely to be sturdier, a wall that has withstood 2 cannonballs, or a wall that has withstood 200? Which is likely to be more vulnerable to injury, a 20-year-old basketball player who has never been injured, or a 40-year-old player who has never been injured?
Let's think about this a little more concretely.
For example, suppose that countries get hit by "shocks" - wars, depressions, political crises. And suppose (for simplicity) that each country has a probability p_c of collapsing in response to any particular shock. p_c is a constant, but when a country begins, we have no idea what p_c is. We start with some prior distribution that represents our guess as to what p_c might be. As we see the country survive more and more shocks, we update our belief about the value of p_c. And, in keeping with my friend's hypothesis, the more shocks we observe, the smaller our guess of p_c will become. In other words, the longer the country survives, the more resilient we will expect it to be.
But that's not the only possible model for country survival. Here's an alternative: Suppose that a country has a set number of "hit points" - in other words, it will die after being hit with some number of shocks T_c. When the country begins, we don't know what T_c is, so again we start with some prior distribution. As we observe the country survive repeated shocks, our belief about T_c will change, and it will (weakly) grow - in other words, our guess about the total number of shocks the country can withstand will get larger and larger as we see the country survive more shocks. BUT, as more shocks occur, the country is also getting closer and closer to T_c - it's "hit points" are being depleted. Which effect is larger? It depends on the prior distribution. For example, if we start off believing that T_c is uniformly distributed between 1 and some number N, then our guess about T_c increases only at half the rate that the total number of shocks increases. Thus, the expected resilience (survival time) of the country goes down as the country withstands more shocks.
Why would a country be able to survive only a certain number of shocks? Well, a country might survive each shock by developing some sort of new institution - a regulation, a bureaucracy, a coalition of interest groups, etc. Those institutions might survive long after the shock has passed, and might make the country less efficient and less resilient to future shocks. This seems kind of like what Mancur Olson was thinking of. For example, America survived World War 2, but to fight the war we built up a huge military-industrial complex that might weaken our economy in the long run. We survived the Enron scandals, but we did so by implementing the Sarbanes-Oxley law, which may have severely hurt our capital markets in the long run.
Another even simpler model is "old age". Suppose that as countries age, the original problems that they were good at solving become less relevant, and new problems that they are not good at solving become more relevant. For example, the Qing Dynasty was great at subduing South China while simultaneously conquering Central Asian horse nomads. But when faced with the industrial revolution and the threat of seafaring empires, they collapsed. Again, "sticky institutions" mean that countries may have limited lifetimes. In math terms, take the "p_c" model, but assume that p_c is a decreasing function of time. If that function decreases faster than your belief about p_c(t) increases, then expected resilience goes down as a country survives longer.
There are a lot more, and more complicated models, but you get the idea.
(How about looking at data? What's the expected survival time of a country that has already survived 238 years? Unfortunately even this simple-seeming kind of exercise involves lots of assumptions, since different past countries might be more or less comparable to America - you don't want to lump Attila the Hun's empire in with the British monarchy. Also, there's the Industrial Revolution to contend with, which probably caused a structural change in the nature of the shocks that hit nations, and which came relatively recently in history.)
So to sum up: America's long lifetime might indicate that we're very resilient and durable, or it might indicate that we're due for a fall. Of course, I hope it's the former, but we really don't know. It all depends on what actually causes the rise, decline, and collapse of nations. And I don't think we know that yet.
Update: It has come to my attention that the argument that past longevity predicts future longevity was made by Nassim Taleb in his book Antifragile, and was referred to as the "Lindy Effect". I certainly hope Taleb did not attempt to use this principle to pick stocks, since this would constitute the Hot-Hand Fallacy.
Posted at 11:18 AM