To me, things like air traffic control and other public goods are a no-brainer. Monetary policy, not so much. Sure, unemployment can be alleviated, but the problem is that nobody really understands the long term and unseen effects of interest rate and money supply manipulation. We are messing with a chaotic system. It might become even more unstable as a result of our intervention. Also, there is a lack of consensus about an "optimal" rate of unemployment and so on.
While there definitely is some dispute about optimal rates of unemployment (a few percentage points of dispute at most, however), there is certainly not dispute about unemployment being a cause of lost output, which is inherently detrimental to growth.
While I won't say that we are able to perfectly model the effects of monetary policy, or that central banks and policy makers get it right frequently (or even occasionally), I will say that systems with random variables can be dealt with mathematically to great success.
Stochastic processes, and, more specifically,
Ito calculus, are very effective means by which an entity could measure and determine probabilistic outcomes of policy action and respond dynamically as they play out. So, while it certainly isn't an simple or easily determinable question, there still exists the potential for answers outside of "we just don't know what will happen." Whether or not we can accurately gather recent data on the those variables, however, is another story entirely, so there is a case to be made regarding incomplete and inaccurate information influencing policy decision and causing detrimental feedback cycles. I believe this case is relatively strong, too.
I don't know whether democratic government is the best Leviathan. DACs seem like a promising and less error-prone alternative. Anyhow, a good government should not be a decision maker but merely an executioner of consensus. Thus it is incapable of making strategic errors in the first place.
I like DACs as well, and I'm also not entirely convinced that representative democracy is the most efficient system either. However, I do disagree that consensus decisions are error-free. There are a lot of arguments even in basic economic game theory that suggest unrestricted group consensus can, in some cases, produce less desirable results than restricted group consensus (i.e. the classic example where a system's nash equilibrium leaves both parties at a zero gain instead a higher collusive gain because of some larger profit providing a motive to cheat). There are a good number of relatively popular strong empirical arguments that have been presented over the years demonstrating this. Ultimately, it sort of boils down to what degree a government's actions are based in political vs economic philosophy.
Good commentary though, Timo. Out of curiosity, what do you think of the model and assertions in the OP?