Let's look again at Brad DeLong on fiscal policy:
The U.S. Treasury can borrow for thirty years, taking all CPI risk onto its own books, and pay only 1.83% per year in interest?
It's not just that a greater amount of government investment meets the benefit-cost test when the government can borrow at 1.83% in inflation-proof bonds for thirty years, a whole bunch of tax postponements do as well. And so do a whole bunch of expanded social welfare programs. And so do a whole bunch of government issues of debt which are then invested in risky private ventures.
…the cost of borrowing for the government has fallen–the market value today of future cash tax flow earmarked for debt repayment has gone way, way up–therefore we should dedicate more future cash flow to debt repayment by borrowing more. There is no "but even." Expansionary fiscal policy is a good idea,
To explain why I view it differently, let's start with a parable from the private sector. Managers in large corporations often are given project "hurdle rates" of twenty to thirty percent, not because their cost of capital is so high, but because of agency problems. If the manager were given a hurdle rate of eight percent, he or she might go crazy with new projects and the company would too easily meet its ruin. It then follows that changes in the actual cost of capital, within moderate ranges, don't so much influence private investment and in this second or third best sense shouldn't. If the hurdle rate is thirty percent, and the cost of capital falls from ten to seven percent for a business, that maybe looks like a big change but actually it's a change in a largely non-binding constraint.
All that is standard and it has long been stressed by Joseph Stiglitz, among others. Furthermore in regressions the cost of capital often fares poorly in predicting private sector investment and that is one reason why. I don't regard the above as "proven" but it does seem to be "likely."
Now let's turn back to the government. Pre-crash, it already could borrow at low real rates of interest. And there already were unexploited public sector projects with high potential rates of return. But we don't want our government to go crazy, spending money without limit. There are agency problems and a lot of the money ends up spent poorly. We instead wish to impose a hurdle rate on government projects. Let's say we trust our government more than shareholders trust private sector managers. That could mean a government project hurdle rate of, say, fifteen percent, which would be below the typical private sector hurdle rate. I'm being generous in that comparison of course; note that mistaken private sector projects get reversed more easily, through bankruptcy, than governmental mistakes do.
OK, there's a hurdle rate of fifteen percent and the cost of capital to government falls from three percent to one percent. How much does it matter? Should it matter? The hurdle rate — the most binding constraint — hasn't changed.
There are many ways of complicating and modifying this model, and I am not saying that the correct "elasticity of projects," with regard to the real interest rate should be zero. Still, I think this model explains why I am less impressed with the real interest rate changes than is Brad.
There is the separate — and very important issue — of covering for the worst-case scenarios. This is stressed, correctly, in analyses of global warming but it matters in fiscal policy too.
One could make the quite different argument that we today have a better administration than average and that therefore the hurdle rate should be lower and government spending should be higher. Now is not the time or place to evaluate such claims, but in terms of logic that to me is a stronger argument than citing the real interest rate effect. It focuses on the most binding constraint, namely the hurdle rate stemming from the agency problems. At the same time, it is an argument for higher spending with or without Keynesian factors or unemployed resources playing a major role. It's the common sense point that a more competent manager should undertake more tasks.