Richard Serlin has a great post on the optimal size of government (HT Brad DeLong). He talks a lot about the sort of pervasive "market failures" that aren't immediately obvious - he mentions asymmetric information specifically. I've talked here before about what I call "temporal autarky", which distorts any sort of intertemporal calculation. The point isn't to drop the market, of course. The market is still by far the best resource allocator available. But there are major underinvestments that we'd expect to characterize the market, which means that even in a market economy a minimalist state is completely unviable.
Serlin also makes some great points about data insufficiencies that I think Mattheus and Jonathan could appreciate after recent discussions of the adequacy of GDP. I still think our output measures are fine and important - as long as they're used for what they are designed for - treatment as a "national income and product account". But Serlin, quoting Paul Romer, notes that there are a lot of other important things to look at that are a lot harder to measure. Romer sounds a lot like Deirdre McCloskey in the sections that Serlin has quoted. If any of you are familiar with McCloskey's work on statistical significance you'll know exactly what I'm talking about.