A "Yes, but" for Brad DeLong and Paul Krugman

Sorry for the almost week-long blogless period. I am trying to restore my posting frequency, as the tragedy noted in earlier posts recedes, but I was on vacation with family and never got around to it.
A couple of interesting points I mean to comment on appeared in the press while I was away. For one, the following column from Paul Krugman, which I am linking to Brad DeLong as his site (unlike the Times) doesn't require registration.
Herewith Krugman:

Sometimes you really have to wonder. It should be obvious that the Social Security Administration's estimate of the growth of unfunded liabilities says nothing - nothing at all - about the cost of delaying a "fix", whatever that might mean. But it seems that even many economists - to say nothing of Joe Lieberman - don't get it.
So here's an example, to illustrate the point.
Suppose that an asteroid is bearing down on our planet. If nothing is done, it will strike in 2019, inflicting $20 trillion in losses. At a nominal interest rate of 5 percent, that's a present value of $10 trillion.
If we do nothing about the asteroid, by next year the present value of the future losses from the asteroid strike will be $10.5 trillion. So the "unfunded liability" from the asteroid strike rises by $500 billion a year.
Suppose that there is a way to fix the problem: we can send Bruce Willis into space to blow up the asteroid. So here's the question: if we wait a year to send Bruce Willis into space, does that cost $500 billion?
Of course not: it could cost either more or less. If waiting a year means that we've lost our last chance to stop the asteroid, it costs $10 trillion - the full present value of the avoidable losses the asteroid would inflict. On the other hand, if Bruce Willis can still blow up the asteroid next year (or any year before 2019), there is no cost at all to waiting. In fact, if waiting increases the Willis expedition's chances of success, there's a benefit to delay.
In other words, the $500 billion increase in the present value of the future costs from the asteroid says nothing about the costs of delaying action. All it says is that the future is getting closer.
The same is true for Social Security. The future is getting closer, so the unfunded liabilities of Social Security are rising in present value (though not as a percentage of GDP). This says nothing at all about the cost of delaying a "fix.' Those costs, if there are any, depend on the nature of the fix.
And it's hard to see any costs of delaying the Bush version of a fix. After all, the problem is that in the absence of changes in the system, at some future date Social Security may have to pay reduced benefits. The only thing the Bush plan does to help the system's finances is - guess what - reduce future benefits. Why does waiting a year to announce benefit cuts that won't happen for several decades have any cost?

DeLong then dutifully chimes in:

"Paul is, of course, right. There is no real economic cost associated with delay by itself: the $600 billion per year number is just a standpoint-of-valuation and choice-of-units effect. There is a real economic cost associated with delay only if delay robs you of the opportunity to undertake the most efficient and effective Plan A and forces you to adopt an inferior Plan B for fixing the problem instead. That's not the case here."

This is an interesting little chestnut for me, because I am generally a fan of infinite-horizon forecasting (notwithstanding its absurd misuse in connection with the Bush Social Security plan that does nothing to restore long-term solvency), and indeed I have argued that the proper economic-accrual measure of what we call the budget deficit would be the year's increase in the fiscal gap - $600 billion in this example. One point I plan to think about a bit more is that, important a principle though economic accrual of future liabilities is for various purposes, for other purposes, relating to the probability of future events, it may miss the mark. For example, although I like to say that the Bush prescription drug enactment cost $16.6 trillion (the estimated present value of the unfunded benefits it promises), obviously it would have been worse for Bush to waste $16.6 trillion that was spent today, as this would be 150% of a single year's GDP. The difference is that we might not (surely will not) actually spend the $16.6 trillion now projected, so in a sense the current policy because we know it can't and won't happen. (Which is not to say the words have no bad consequences given how they shape expectations and the political consequences of inertia.)
Anyway, Krugman and DeLong are right about the Bush plan because it doesn't actually cut benefits until later. But if we were talking about a rational response to the Social Security fiscal gap, as opposed to the Bush plan, the fact that the gap under present policy keeps growing each year actually does indicate that there is a cost associated with delay. This is the opportunity cost actually to shift policy sooner rather than later. Some of the fixes, such as those involving the better-off among the current elderly, do become impossible if we don't do them now.
Also, by delaying benefit cuts (as distinct from the announcement date for future benefit cuts, which only affects planning and perhaps the political probabiliies), we lose the ability to cut them less steeply when we do cut them. Similarly, suppose you had to cut your consumer spending over the rest of your life. The longer you wait to start the cuts, the steeper and presumably more painful the cuts will be when you get around to making them. The familiar argument in the public economics literature for "tax smoothing" is similar.
SO: Krugman and DeLong are right that a year's delay in adopting the Bush plan does not cost $600 billion. But they would be wrong if they argued that the accrual of the future liabilities does not indicate any grounds for responding (actually, not just by pre-announcement) sooner rather than later.
The question of how to think about the government's long-term fiscal sustainability problems is quite intricate and complex; more so than I had thought for a while even after thinking about it for the last 10 years.