Thursday, April 18, 2013

Is empirical analysis really needed to debunk a foolish claim?

I am of course just an observer, not an active participant, in the debate concerning the Reinhart-Rogoff paper claiming that evil magic starts happening, in an oddly discontinuous fashion, when the debt-to-GDP ratio hits 90 percent.  It's dispiriting to read how sloppy and apparently unprofessional the paper was, and to think that it may have substantially worsened public policy outcomes around the world (although, more likely, it merely served to rationalize misguided austerity that would have been pursued anyway).  I certainly expected better from the authors.  (My fond memories of Rogoff date back to his participating in Shelby Lyman's Channel 13 broadcasts of the 1972 Fischer-Spassky championship chess match.)

But, while I certainly accept the need to take nothing for granted, and to pursue empirical inquiry wherever it leads - thus creating the theoretical possibility that the claimed results could have been confirmed, and if so, would have needed to be explained - the Reinhart-Rogoff hypothesis verges on being something that can't be true, as a logical matter.

Both public debt and GDP are artificial measures, as well as being snapshot measures.  OK, let's not worry too much about GDP, and about such conundra as Arthur Pigou's joke about the unpatriotic Frenchman who shrank the country's economy by marrying his housekeeper, thus taking her services (which she continued to render) outside of the official measure.  Public debt is also an artificial measure, as Laurence Kotlikoff has been arguing for years (although, in my view, he sometimes pushes the argument a bit too far).

For example, the U.S. has future Social Security obligations that have an expected present value and yet are not included in public debt, but that matter for the same reasons as public debt.  To wit, they might need to be paid out in the future.  Suppose we explicitly converted expected Social Security obligations into explicit public debt instruments with the same present value and payment schedule.  Then we'd have almost the same policies, but official public debt would be much higher.

Now, it's true that converting Social Security obligations into express debt would probably make them harder to renounce, meaning that the underlying policy wouldn't actually be wholly unchanged.  (This is the point that I believe Kotlikoff underappreciates.)  But the set of commitments that we actually had, and that mattered for the reasons that a debt overhang matters, would have changed far less than the official public debt measure.  So "90 percent" is just not an economically meaningful measure of anything.  It thus would be quite a surprise if significant economic consequences consistently followed around the world from changes in what is merely an artificial measure that has only a very rough relationship to the underlying set of concerns.

Then there is the snapshot point.  Reinhart and Rogoff have been rightly mocked for in effect assuming that the rise in U.S. public debt past the magical breaking point at the end of World War II led to an economic slowdown, when in fact they were observing the short-term macroeconomic consequences of rapid demobilization.  But in any rational economic world where the players have any foresight, how could a 90 percent U.S. debt-to-GDP ratio in 1945 have been the same as one in, say, 2025?  We had so vast a public debt in 1945 because we had been fighting a two-front World War that had just ended.  So it was perfectly obvious that the current U.S. fiscal policy path was about to change.  No matter how one assesses the long-term U.S. fiscal situation today, what with an aging population and concern about the growth rate of healthcare, it simply isn't comparable to that in 1945, even if we have, at a given moment, the same debt-to-GDP ratio. So it shouldn't behave the same.

I therefore rank the Reinhart-Rogoff hypothesis as not far above claims that, say, if the Redskins win their last home game in October, the Democrats are going to win the presidential election.  There simply can't be reasonable theoretical grounds for believing that it would be true.

2 comments:

blissex said...

«So "90 percent" is just not an economically meaningful measure of anything. It thus would be quite a surprise if significant economic consequences consistently followed around the world from changes in what is merely an artificial measure that has only a very rough relationship to the underlying set of concerns.»

While the paper by RR is suspicious, high levels of debt, whether public or private, do have a large impact on the economy and potential growth rates because of two very important factors:

* Interest payments.

* Debt repayments.

If they are significant they impact either the national distribution of income, or the balance of exports and imports, and in extreme but common cases can precipitate a financial crisis if they result in a financial structure with high sensitivity to shocks.

In general the specific credit and debit financial structure is important at high levels of debt, and the most illustrative book is the excellent "The volatility machine" bu Michael Pettis.

At low levels of debt even a poorly designed financial structure matters a lot less, and this is especially important if national politics tend to push for very pro-cyclical, very shock-sensitive public and private structures, which happens often in developing economies, but has happened on a colossal scale in credit-bubble countries like the USA, the UK, Australia, ...

Daniel Shaviro said...

Agreed that a country's fiscal balance, including current cash flow, is important. It's the idea of a discontinuous break point, based on an artificial measure of what I agree is a real and consequential underlying phenomenon, that is extremely unlikely have the significance R-R were ascribing to it.

Critics of R-R have generally agreed that, yes, rising debt levels are bad for growth. But that's no defense of their very naive specification of how a discontinuous break point at 90 percent supposedly might work.

BTW, just to make the point that I recognize the importance (on multiple grounds) of fiscal sustainability, see my books (e.g., on Amazon) "Do Deficits Matter?" and "Taxes, Spending, and the U.S. Government's March Towards Bankruptcy."