I feel reasonably certain that there is something wrong with my logic here, but I'm hoping someone can explain to me what is wrong.
A government should spend on something if the economic impact (call them the economic returns) of that spending will eventually exceed the cost borne by the government. This government cost is a blend of its interest rate, the economic activity displaced by government spending (crowding out), and the economic activity displaced by current and future taxation to pay off the cost of the project. In mega-recessions like this one, interest rates are near zero because of deflation fears, and the economic activity displaced by government spending is very low because there's so much extra capacity in the economy that government spending doesn't drive prices up much. Thus, project cost is reduced to the level of economic activity displaced from the private sector by taxation.
What this implies is that the government spends more in mega-recessions like this one. It does not imply that the government should will spending up (a la the Keynesian solution), and it does not imply that the government should cut back and do nothing (the tea party solution), it implies that the government should continue the exact same process of project evaluation but the hurdle rate for acceptance is lower, so more projects get taken on. Thus, government spending should be countercyclical.
Is this absurdly reductionist? It's how corporations are supposed to evaluate their projects, in theory - cost of equity capital (analogous to taxpayer cost), cost of debt capital (interest rates) and opportunity cost of projects, both in management time and attention and capital raising limits (analogous to crowding out).
Obviously this doesn't address nasty problems like wage stickiness and some of the other things that Keynesian economics tries to analyze (and, at least it seems to me, is dubious in its application). However, while the Federal Reserve needs to do those kind of analyses for determining interest rates and optimal monetary policy, the federal government as a whole should be able to take these costs (interest rates, crowding out and taxation), treat them as exogenous, estimate them and use that as a decision rule, right? What's wrong with this picture? I understand that politically it's hard to estimate in an unbiased fashion, because each party will find ways to make their pet projects pass the decision rule, but on a theoretical level, isn't this how it could work best?
(Before anyone starts yelling at me for inconsistency on deficits, there's a big difference between temporary spending now and permanent spending increases. We can't have permanent spending increases because of the debt. We can, however, run a deficit now, during the recession, as long as we ensure that it ends when the recession is done. We also can't have spending that is destabilizing from a taxation and business uncertainty perspective, but contributes no economic stimulus. Healthcare reform, I'm looking at you, and that's just scratching the surface on your flaws...)