A reader recommends Mark Pennington talking about his book Robust Political Economy. The talk is very well organized and concise, and I recommend it (I didn't stick with it through the Q&A). He says that one has to keep in mind two facts:
1. Humans have limited knowledge (he calls it "limited rationality," which I am not sure is the best choice of terms here).
2. Incentives matter.
What follows from those facts are two important conclusions.
1. Societies benefit from continual experimentation. Since no one knows enough to design a perfect system, more experimentation is better.
2. Exits works better than voice. The incentive for consumers to make good choices in the market and for the market to respond to those choices is pretty much always going to be better than the incentive for voters to make good choices and for politicians to respond to those choices.
He then proceeds to demolish a number of rationales for government intervention, based on that straightforward logic.