One Conundrum of Governance

Gary Becker writes:

While average rates of growth do not appear to differ much between democracies and authoritarian regimes, the variability in performance does differ more among authoritarian governments.

While developing democracies may linger in the purgatory of mild or no growth, autocratic rulers tend to offer heavens or hells.  Here’s one comparison data graph by Almeida and Ferreira:

Hat to tip to Bryan Caplan, who also adds, “Interesting finding #2: When a dictator accidentally dies, growth rates persistently change. When a democratic leader accidentally dies, in contrast, they don’t.”

One thing that might cloud this issue is that catch up economic growth is often faster and of greater magnitude than that in the developed world. After innovative countries have slogged through the trial and error process of invention, autocrats have the luxury of copying what works. But once most countries approach the technological frontier as other sources of growth have been tapped out, I’m curious as to what the comparison between growth rates would be–will angel autocrats still attain superior rates? Singapore is making a run for it….but China may soon tell a different story.

Now if some kind of peaceful filtering process were at work (and the infernal hell of a failed autocracy weren’t so dreadful), I would see the variability of growth rates as a virtue. It would be evidence of a discovery process driven by rule set experimentation. Bottom rung countries would fail fast, fail safely, and fail often, quickly adopting the best practices and rule sets of the top rung batch. The decision making costs in autocracy are lower–one guy or a committee calls the shots–so that kind of process would work best in a market full of tyrants, especially if their customers could easily change government service providers, thereby avoiding onerous impositions. But, alas, the human costs of autocratic variability are very great, men aren’t angels, exit is expensive, and the top rule sets aren’t so easily emulated.