Maturity Diversification
In an earlier post I linked to a provocative proposal by Andy Harless in which he argues that the Treasury should be shortening the maturity structure of government debt. His reasoning is roughly as follows:
At current interest rates, money and bills are virtually identical assets: holders of bills are requiring no compensation for the additional liquidity or safety that money would provide. This makes conventional monetary policy (exchanging cash for bills) ineffective. On the other hand, an increase in the issue of bills can have expansionary effects, putting the Treasury effectively in charge of monetary policy.
In addition to its expansionary effects, a shift to shorter maturities on government debt should lower the expected value of the costs of debt service, since there is a liquidity premium to be paid on longer term bonds. However, it would also increase the vulnerability of the Treasury to unexpected increases in short term rates (expected increases are already implicit in the yield curve).
Maintaining long maturities to insure against this risk would be hedging against good news, assuming that an unexpected increase in short term rates would signal a more rapid recovery than is currently forecast. In this case (unexpectedly) higher rates would be accompanied by (unexpectedly) greater federal revenues and lower benefit payments, so the financial position of the Treasury need not be worsened despite the greater costs of debt service.
That's his argument, if I understand it correctly. The proposal is similar in some respects to one made recently by Joe Gagnon, in which he argues that the Fed should be buying substantial amounts of long term debt. Both proposals would result in roughly the same mix of short and long term securities in the hands of the public, and would lower long term interest rates. But there are two important differences. First, as Gagnon notes, "it would be better for the Fed to do it because they have the staff and expertise for gauging how much to do and when to stop." Second, there would be greater maturity diversification in Treasury issues, raising the expected value of debt service costs but reducing the vulnerability to unexpected fluctuations in interest rates.
How important is the issue of maturity diversification? A commenter (identified only as JKH) on Harless' blog thinks that it would be irresponsible to ignore it:
As far as the Treasury is concerned, it’s just acting according to prudent interest rate risk management considerations in locking in some interest cost on such a massive prospective debt load. It’s just a matter of judgement on how to diversify maturities, given the “risk” that the Fed may want to start tightening some time. Ignoring the issue of maturity mix is irresponsible. From there it’s judgement on the right mix.
This is fair enough as far as it goes, but what are the principles on the basis of which such judgment ought to be exercised? The trade-off here is between the expected costs of debt service and the risk of facing a situation in which costs are much greater than forecast. The basic problem was expressed very succinctly by Richard Roll (1971) as follows:
For example, consider a government agency borrowing for a specific long term project at currently high rate levels. It might be able to reduce the total expected interest payments (and expected taxes) by financing the project partly with short-term bonds rather than entirely with bonds whose term-to-maturity matches the project's life. On the other hand, even though the agency expects lower rates in the future, it would not feel secure in funding the entire project with short-term bonds that would require a later refinancing. It would prefer to pay the higher expected rate on a portfolio of long- and short-term bonds rather than accept the risk that rates will go higher contrary to expectations.
Given some specification of Treasury preferences and expectations about the future course of interest rates, this is a fairly standard optimization problem. But it is not obvious (at least to me) how the desired maturity structure should vary with changes in uncertainty about future rates. Other things equal, greater uncertainty should lengthen maturities. However, greater uncertainty will also steepen the yield curve and raise the expected costs of long-term (relative to short-term) financing, and this effect should reduce desired maturities. In any case, these effects should be possible to identify, given some specification of Treasury objectives.
Much more difficult is the question of what those objectives ought to be. How much weight should by placed on risk as opposed to the expected costs of debt service? And should the desired maturity structure of government debt be sensitive to macroeconomic conditions?
I don't have answers to these questions but (as I said in a comment on Mark Thoma's page) it is important to resist the temptation to view the problem solely from a corporate finance perspective. There will be a significant shortfall in government revenues over outlays for many years to come, so short-term Treasury issues will have to be rolled over repeatedly for an indefinite period, or converted to long-term debt at some point. If this were a corporation, such financial practices would be madness (for the company as well as its creditors). The company would be engaged in massive maturity transformation, and be highly vulnerable to changes in short term interest rates and credit availability. It would be unable to meet even interest obligations without borrowing - which is Hyman Minsky's definition of Ponzi finance.
But the Treasury is not a private corporation, and it faces very different constraints and objectives. Matching maturities to expected net revenues is simply not an option, nor should it necessarily be a goal. This much is straightforward. Much less clear is the set of principles that ought to guide the decisions that ultimately determine the maturity structure of government debt.
---
Update (12/28). In an email to me (posted with permission), Joe Gagnon adds:
Since the government includes the Treasury and the central bank, we could have one objective function and find the optimal policy from first principles. But I suspect the same result would obtain if the Treasury acted as first mover and followed a reasonable, modestly risk-averse, cost-minimizing strategy for maturity of issuance given the Congressionally set taxes and spending. Then the central bank would take Treasury behavior as given when it decides on the optimal maturity structure from the overall social welfare viewpoint, taking into consideration unemployment and inflation as well as risks to government finance and likely future behavior of Congress.
My own sense (buttressed by Table 6 in my paper) is that the risks to government finance from shifting into short maturities is very small compared to the benefits, and much less costly than outright fiscal-spending stimulus.
Also:
The point about the table is that even in the inflation scare scenario, when future short rates rise quite high for a while, the debt burden is lower with monetary stimulus (via maturity transformation) than without it. Of course, one can construct even more extreme scenarios in which that is not true (or in which the debt burden is reduced by high inflation) but I would argue those scenarios are highly unlikely given the state of the economy and the current membership of the Fed’s policy committee.