Main index | Financial Markets index | About author |
Julian D. A. Wiseman
Abstract: Sovereign governments should issue long-dated debt, as doing so stabilises the macro economy, provides information to the issuer about demand for assets, and provides benchmarks for private-sector issuers. The situation is more complicated for EMU-zone governments.
Contents: Introduction, Non-governments, Governments, other advantages of long-dated borrowing, Disadvantages, Surpluses and buy-backs, What are governments actually doing?, What should governments do?, Footnotes.
Publication history: only here. Usual disclaimer and copyright terms apply.
Governments borrow money, typically by issuing fixed-coupon bonds, which can be of any maturity from a few days to several decades. At which maturity is it optimal for governments to borrow?
One objective for governments is to minimise the cost of borrowing over the long term. This requires an auction calendar (to signal credibly that governments are not using their inside information to ‘time’ the market); an efficient auction mechanism (about which see a earlier paper dated December 1997); and avoiding complex tax features so as to attract the widest possible audience. By and large, most governments do most of this well, and the rest satisfactorily.
But at which maturity should a government borrow? Let us begin by considering the position of poor governments, and of corporates.
For some governments, choices are very much constrained. For example, a poor country with a history of political instability might be charged a punitive price by the international financial markets for long-term borrowing. Short-term debt, of maturity no longer than a year or so, may be acceptable to lenders, because they can judge the likelihood of repayment and of devaluation over such a horizon. But longer-term debt, with a maturity over (say) five years, may be seen as too risky, and hence lenders would demand a very high interest rate. While these constraints are very real for many countries with a history of default, they are not directly relevant for most rich-world governments, including the US, the UK, Japan and most EU members.
So instead let us start by considering the borrowing strategy of a typical company, the income of which rises and falls with the state of the economy. If the economy slows, and the company’s income slows, the company would like its fixed expenditure to fall. For such an entity the interest-rate component of the debt should be short-term, so that when the economy weakens and the central bank lowers its interest rate, outgoings soon start to fall. The company might issue short-dated fixed-coupon debt, or floating-rate debt (of any maturity), or longer-dated fixed-coupon debt that is then swapped back to floating.
An example of such a company, with income rising and falling with the state of the economy, is the French 5th Republic. France, like a private-sector company, cannot print money. And the income of the French state is mostly derived from taxing economic activity, which falls if the economy is weak, a problem exacerbated by a simultaneous rise in government spending. A very reasonable strategy for the French state would therefore be to issue long-dated fixed-coupon debt, and then to swap this back to floating. Indeed, an announcement in February 2001 by the French Tresor stated that it would continue to issue fixed-coupon securities, but would receive fixed on the swaps to shorten the average maturity of its debt.
But the United States Treasury can instruct that money be printed, and so is not constrained by its own creditworthiness. In particular, during a period of sustained deflation, the US government may well wish to push money into the economy, and in extremis the US government may wish to print the money to do so. Generally, printing is money is frowned upon because it leads to inflation, but if the inflation (or disdeflation) is desired, then printing money can be helpful. A policy of issuing long-dated fixed-coupon debt would, if deflation set in, push money into the economy. This would not impair the government’s credit, because however far tax revenues fell, the money could be printed. Of course, issuing long-dated fixed-coupon debt does not guarantee that the government would be printing money; because if inflation were to take off, the government would instead have a windfall reduction in the real size of its payments.
How effective at macro-economic stabilisation would be this strategy of issuing long-dated bonds? Imagine that a government has issued at par $25bn of a 50-year 5% semi-annual bond, and that 4 years later deflation has set in and the bond yields only 3%. The price of the bond would have increased from 100 to 149.72, and the purchasers of this bond would be showing a profit of $12.4bn, which could be realised immediately, either by selling the bond or by using it as collateral in the central bank’s monetary policy operations. Whichever, the profit goes some way towards repairing banks’ balance sheets and thus ending the deflation. But a long-dated-borrowing policy does not give money to the rentier class: if the yield had increased to 8%, the bond’s price would have fallen and the financial sector would have a loss of $9.1bn. So deflation would cause a wealth transfer to the financial sector, but inflation would cause a transfer from the financial sector, and all this would be done without the moral hazard of subsidising incompetent banks.
The effectiveness of such a program would, in part, depend on who owned the long bonds. The ideal would be for the bonds to be owned by those financial entities that also take on the banking risks in the economy. However, banks tend not to own many bonds longer than 5 years or so. However, bank-type risks are increasingly being securitised, and being held by longer-term investors, including pension funds and insurance companies. These investors do tend to own long-dated bonds. This ownership pattern means that the macro-economic stabilisation effect would be present, but wouldn’t be as strong as if the banks owned the bonds.
How large would such a borrowing program need to be? Imagine that a government has debt of 30% of annual GDP, half in the form of 5% bonds with maturities from 20 to 50 years, and half in the form of short-term debt. Then a fall in yields from 5% to 3% would be a wealth transfer to the private sector of about 6.3% of GDP, and a rise to 8% would tax the private sector by about 5.2%. Given that banking crises cost 15% to 20% of annual GDP*, issuance of long-dated debt could repair a substantial proportion of a problem without moral hazard.
Issuance of long-dated debt by a perfect-credit issuer has other advantages (other than stabilisation of the macro economy).
It provides information to the issuer. If there is great demand for (say) 50-year paper, there is currently no price to show that demand. But if the government has issued a security with about 50 years to maturity, that demand could be seen.
And that demand could also be seen by private-sector borrowers. In general, it is expensive for private-sector borrowers to issue much beyond the longest government bond, as the absence of perfect-credit interest-rate pricing increases the risk premium charged by investors. But if there were government bonds maturing in (say) 2040, 2050, 2060 and 2075, these assets could act as benchmarks to weaker-credit bonds maturing on or between these dates.
So, there are three advantages to issuing long-dated governments bonds: stabilisation of the macro economy; ongoing price discovery; and the provision of benchmarks to facilitate private-sector issuance. But are there any disadvantages?
There is a possible price disadvantage to long-dated issuance. If the market is charging a great premium for longer-dated borrowing, that premium would reflect a lack of end-investor demand. That lack of end-investor demand isn’t all bad: it increases the proportion of the issuance that would be owned by the banks, and hence increases the stabilisation of the macro economy. But governments’ desire for cost-efficient borrowing suggests that there’s no need to be stubborn: if the 50-year yields +0.50% more than the 30-year, don’t try a 60-year bond. However, it is unlikely that either the UK or the US yield curves would be anything like as steep.
Likewise, if inflation is high and the political authorities wish to signal their determination to lower it, shorter-dated borrowing might be more appropriate. This would have applied to many countries in the 1970s, but rich-world inflation is low and stable, and bond yields reflect this.
It might be thought that issuing long-dated debt would give the government an incentive to create inflation, or at least might create the appearance of such an incentive. However, inflation targeting is now performed by central banks not by finance ministries: perhaps this policy does require that central banks remain independent.
The last possible disadvantage of long-dated debt is its price variability. A government that is running a consistent surplus may wish to buy back some of its debt. If the debt were short-dated, this could be done simply by allowing the existing bonds to mature. But if the debt were long-dated, its price might have risen considerably, and the debt buybacks could be expensive.
But there is a simple solution to this apparent conundrum: don’t buy back long-dated debt. Instead a government running a surplus should leaves its cash pile on short-term deposit. This deposit might be at the central bank, in which case the central bank would lend the money to the markets as part of its monetary policy operations (again, see an earlier paper on this subject). Alternatively, the government could deposit the money with the private-sector financial system, accepting in return top-quality collateral.
This solution appears to have a problem: if short-term interest rates fall, the government would be paying the old (higher) long-term interest rate on its borrowing, but receiving the new (lower) rate on its deposits. But this is intentional. If deflation were to set in, this policy would transfer of wealth out of the government coffers. And if inflation were to rise, the government would profit at the expense of the private sector. So, in the example above, the government would allow the short-term debt (totalling 15% of annual GDP) to mature, but would leave the long-term debt untouched, instead putting its growing cash pile on short-term deposit.
France rightly recognises that it isn’t a sovereign government, and is protecting its credit rating by shortening the effective maturity of its borrowing. However, the EMU-zone economy as a whole would benefit from an agreement amongst the member governments to borrow long, but in the absence of such an agreement, France is declining to take onto itself the burden of stabilising the macro economy. In a sense, this is a variant of the lender-of-last-resort problem: there is no perfect-credit entity able to take action in a crisis.
The UK has been lengthening its debt profile, but this seems to be in response to demand from pension funds, partly driven by regulation (the MFR and FRS17). An insight into the Treasury’s reasoning can be found the recent consultation document (local copy), on the subject of “whether to adopt a new design for new issues of index-linked gilts”. This document asks, amongst other things, whether a new design of index-linked gilt should have a floor at par, but remarks in ¶24 that “notably, a deflation floor would reduce the value to HM Treasury of having index-linked gilts in its debt portfolio by reducing their deficit-smoothing properties in some circumstances”. Very interesting: a deflation floor would only have an effect after a long period of zero or negative inflation, and it is under exactly these circumstances that the government should care less about “deficit-smoothing properties” than about printing money to push into the macro economy.
The UK can at least argue that it might in the future join EMU and so cease to be a sovereign government (in the sense required here). But the US? The Under Secretary of the Treasury for Domestic Finance, Peter R. Fisher, in remarks at the November 2001 quarterly refunding announced the cancellation of the 30-year bond program! Indeed, this announcement was made just as uncertainty about the US economy was at its highest for many years. The argument above says that it would have been sensible to replace the 30-year with a 50-year rather than a 10-year.
The French government is behaving optimally for itself: funding long, swapped back to floating. The EMU-zone government might wish to agree amongst themselves to fund long and not to swap, but such an agreement could be easily dodged by swapping via government-owned entities.
If the UK is not to join EMU, it should fund long to benefit from the macro-economic stabilisation. If the UK is to join EMU, it should fund long to benefit from the low long-gilt yields (substantially below those of the best EMU-zone names despite higher short-term interest rates). Yet the most recent new gilt was of a 5% 2025, some seven years shorter than the longest conventional gilt, the 4.25% 2032. Why the shorter maturity? Perhaps because the market makers (the GEMMs) like an easy life, and at the DMO’s quarterly consultations they always ask that a new long be no longer than the current longest long. Their protestations should be ignored. (This author, when at these meetings, has always recommended that a new long be as long as possible.) The next long should be a 2045 gilt, then a 2060, and if that yields less than +0.05% more than the 2045, the next should be a 2075 or 2080. Note that a par 5% 58-year has a Macaulay duration 469 days longer than that of a 43-year, and that a 73-year is longer than that by 223 days.
The US used to have so much funding to do that it could produce a new long bond several times a year. That is no longer true, and so the US can only produce an occasional new long bond. One should be made approximately every year, of a maturity specified several months in advance (to avoid fears that the Treasury is timing the market). The maturity could rotate around a selection: perhaps 30s, 35s, 40s, 50s and 60s? As with the UK, size and liquidity should be improved with reopenings, which may require a change to the Original-Issue-Discount regulations. This is very far from the recently announced current practice.
Julian D. A. Wiseman, January 2002
* Estimate taken from Costs of banking system instability: some empirical evidence by Glenn Hoggarth, Ricardo Reis and Victoria Saporta, presented at the Conference on Banks and Systemic Risk at the Bank of England on 23 May 2001.
Main index | Top | About author |