How it might be done right, and why it would matter
I've been thinking about fiscal policy lately, in case you can't tell. In particular, one of the ideas I've been bouncing around is a variant of President Obama's proposal to create a national "infrastructure bank," a part of his stalled "American Jobs Act" legislation, which would finance public-works and transportation projects in conjunction with private investment.
I don't really like the infrastructure bank plan as it stands--but I see some remarkable potential in the idea. I would want it to have official independence from the political process, as does the Federal Reserve, to avoid pet projects from making their way into investment. I would want to require it to seek maximum profits*, which would be reinvested, and to study every proposed project with rigorous cost-benefit analysis* methods imported from the best of the private sector. I would even consider requiring it to be a publicly-traded company, as are many of China's state-owned corporations. (I use an * because I would have the bank include in the profit estimates a measure of broadly-distributed gains to social welfare.)
"Socialism," you say. "Government ownership of the means of production." No. These projects have massive positive externalities--that's why I write above about the social welfare gains. Not every social-welfare-enhancing project (think "publicly profitable") would be privately profitable. (A lighthouse. A bridge. A tunnel. Improvements of the electricity or communications grids.) Thus the private sector underfunds them, and our government does a poor job of funding the most valuable infrastructure, too.
And we know that, one way or another, some projects will be funded. How would you rather that be done? By a Congress which allocates for gross political favor? Or by something which is run with a bit more maturity? (I would package the proposal of an infrastructure bank with a binding ban on earmarks. That'd be a policy I think in a rational world both sides of the political aisle could embrace.)
I would further seek to maximize the gains from this infrastructure bank idea: let it absorb the whole idea of fiscal stimulus. When the 2009 recession came around, members of Congress saw the spending opportunity as if they heard the music coming from the ice cream truck. So extend the ban on earmarks to prevent Congress from spending on particular areas of infrastructure they favor--Obama's quixotic high-speed rail plans come to mind--perhaps softening that ban to permit it if a supermajority votes to approve. Otherwise, limit Congress' hand in infrastructure to the funding level for the infrastructure bank. When Congress wants a fiscal stimulus in the form of infrastructure, let it provide a boost to the bank's capital and do no more.
Furthermore, let the states, regional consortiums, counties, and cities recommend projects to the bank. The bank's job would be to sort through these projects with the cost-benefit analysis, in addition to generating their own. These recommended projects would have to gain support through the relevant deliberative body, such as a state legislature, a city council, or a county assembly. By making these decisions at a more local level, the planners become stakeholders, and thus avoiding the classic Milton Friedman incentives problem of someone spending someone else's money on someone else that is inherent to the federal government.