Monday, April 30, 2012

Floating-rate Treasury Debt?

Bloomberg has an intriguing April 29 article on Treasury Floaters According to Bloomberg, the Treasury may announce on May 2 that it will issue floaters. It quotes Cam Harvey, who testified that floaters as being a great idea in 1993, as disapproving.   Knowing Cam,  I suspect he had a more sophisticated view in mind.

Issuing floaters and converting a lot of debt to floating-rate debt is a great idea, if done right, even if the maturity structure of government debt should be much longer now. Let me explain.

Floating-rate debt is like an adjustable-rate mortgage. If you buy a $100 floating rate bond and short-term rates are 5%, you get $5. If next year short term rates rise to 10%, you get $10. And so forth.

The alternative to floating-rate debt is our current practice of rolling over short-term debt. The government issues one-year bills,  then next year issues new bills to pay off the old ones.  The big danger for our -- or any -- government is that markets refuse to roll over its debt. Greece didn't get in trouble because it couldn't borrow new money to pay its bills for a month; it got into trouble because it couldn't borrow new money to pay off the old money. The US has to roll over about half our $15 trillion debt every two years, so we face a similar danger. You may say, interest rates are low now, who needs to worry about that? I would answer, Greek interest rates were low in 2006 as well. You need fire insurance even if a fire seems remote. 

From a frictionless finance point of view, however, floating-rate debt is the same thing as rolling over short term debt.   If the one year rate is 5% in the first year and 10% in the second year, the payments from government to bondholders are exactly the same. Similarly, rather than get a floating-rate mortgage, you could get a one-year loan, then borrow again next year to pay off this year's loan and so on.

So, if you like floating-rate debt, you must have some market friction in mind. It certainly feels safer for bondholders to have long-term bonds with adjustable coupons rather than explicitly trying to sell new debt every year to pay off the old debt. It isn't, really, however. In the circumstance that markets refuse to lend new one-year debt, the short-term interest rate would rise arbitrarily high; people would all be trying to sell the long-term "floating rate" bonds, and you have exactly the same crisis.

So where does the feeling that floating-rate debt is safer come from? I suspect part of it is a little behavioral, and perhaps sensibly. Many holders of government bonds may act passively around an interest rate reset, where they would actively have to make the decision how many new bonds to buy after their old bonds retire.  It's easier to have a bank run if everybody has to go to the bank every day, than if most people sit at home and watch interest rates change.  On the other hand, most Treasury bills are held by large sophisticated institutions. Maybe the idea that we can get them not to pay attention by essentially making a roll over the default option is optimistic.

I think the real reason is a bit deeper. To really evaluate floating rate bonds, we need to know how the interest rate is set. Will it be an index, based on some other market? Or will it be some auction mechanism? Can rates rise arbitrarily high, or is there some cap on how far rates can rise? The devil is in the details here!  Rolling over one-year bonds is the ultimate auction mechanism. Both sides really know they're getting the market rate on bonds.

Another way to put the issue: will the market value of floating-rate debt always be exactly $100? True floating-rate debt has an interest rate set by auction every day, and the principal value is exactly $100. In that way it really is functionally the same as rolling over debt every day, but avoids a lot of needless churn.

I suspect the answer is no. As with auction-rate securities that fell apart in the financial crisis, I suspect there will be some cap on the floating rate, or that it will be set relative to some index. In a crisis, the market value will fall below $100, and bondholders take a hit. Now it is a combination of short-term debt and some sort of option which makes it much more fun for financial engineers.

That might actually be good for some purposes. I think Cam was objecting, as I do, to the short maturity structure of US Government debt in the present situation. Interest rates can only go one way, up. If the US massively lengthened the maturity structure of government debt, we would be insured against a rollover crisis, or the fiscal impact of rising interest rates. If interest rates rise to 5% now, that's going to add something like $500 billion to the annual deficit. If the US switched entirely to long-term debt -- fixed-coupon perpetuities -- then interest rates rising to 5% would cost us exactly zero. Bondholders would take the hit. Sure, it's a little more expensive on average -- the yield curve is upward sloping. That's the premium you pay for insurance. The premium seems mighty small.

So, in that context, short-term debt that turns into long-term debt in a financial crisis has some advantages. It doesn't stop the fiscal consequences of rising interest rates, but it includes an option for the government to make the debtholders take a hit during a potential rollover crisis or sudden spike in rates. For which the government will pay a premium. 

In sum, the issue of the maturity structure of government debt is different from the issue whether our short-term debt is rolled over or consists of floating rate debt. I suspect what Cam Harvey really said is that he likes a longer overall maturity structure but also likes conversion of the shorter debt to floaters. 

And, the key question to ask of Treasury floating-rate debt is, just how will that floating rate be set? Maybe the right answer is a spectrum: some truly floating rate debt (next subject) and some debt that resets less frequently and with an explicit cap on interest rate changes.


A second aspect of floating-rate debt intrigues me.

We already have floating-rate debt. It's called reserves. Reserves -- accounts banks hold at the Fed -- are nothing more than overnight, floating-rate government debt. They are curiously absent from the Congressional debt ceiling, but that's it.

Interest on reserves is a great innovation. Together with more and more widespread electronic transactions, it means we can live the Friedman rule. The economy can have all the money it wants without losing anything to foregone interest. All of the cash-management shenanigans of the past can be forgotten. One of the main justifications for massively levered banks goes out the window -- the idea that the economy needs a vast supply of bank deposits as very liquid assets.

The only problem is, you and I can't access interest-paying reserves, only banks can do so. And the interest rate is not a market rate, it is set by whatever the Fed feels like paying. And the Fed loudly announces that it will lower the rate on reserves below market rates when it feels like stimulating the economy. But smart cash managers will not want to suffer lower rates, so all the financial engineering starts up again.

A good supply of floating-rate Treasuries would be great in this situation. I mean really floating - rate, with an interest rate set each day by auction at whatever value makes the price exactly $100. And they should be electronically transferrable in arbitrary denominations.

These securities would be the ideal asset for money market funds to hold, and offer (at, I presume lower cost than the Treasury) all sorts of transactions services. Now you and I can have perfectly safe, interest-paying money.  And we don't have to put up with the "demand for liquid assets" story justifying huge bank leverage, very high levels of deposit insurance, and the other pathologies of our banking system.  

Will we get it? Let's see.