Auctionettes: HMT’s Reply

Julian D. A. Wiseman

Abstract: an open Letter was sent to the Commercial Secretary to the Treasury, then Sir James Sassoon now Lord Sassoon, on the subject of the dangers of the current auction rules, and how they could be repaired. Two weeks later a reply was received, to which this is the response.

Publication history: sent to Christopher Goodspeed of the UK Treasury on 20^th June 2010, and also published at www.jdawiseman.com/papers/finmkts/20100620_auctionettes.html. Usual disclaimer and copyright terms apply.

Contents: HMT’s Reply; Introduction; JDAW’s Response; Stakeholder support; Auctionette issues; Supplementary gilt distribution methods; Post Auction Option Facility; The DMO’s incentive; Conclusion; Footnotes; Afterword.

Introduction

Whilst working as a gilt analyst for J. P. Morgan in the 1990s, the author realised that some of the risks faced by a dealer at an auction were unnecessary, and slightly lowered achieved sale prices. A solution was devised, which—with some subsequent small tweaks—has been advocated since. Until the start of the credit crisis the solution was of light importance: my guess is that it would reduce issue yields by something of the order of 2bp. If selling, say, £50bn a year, with a duration of 8 years, this increases sale prices over the whole year by a mere £80mn. Worth having, but it would never make a real difference to national solvency, and as it would be difficult to prove unambiguously that this increase had happened, it would never make a difference to a minister’s career.

But in late 2007 markets changed, and the other feature of the better auction mechanism—guaranteeing access to funds at a price close to prevailing market prices—became important. Hence it was recommended to the new government in An Open Letter to the Commercial Secretary to the Treasury: Trouble Coming, Easily Avoided.

The reply of H M Treasury to this letter appears on the right, and the author’s response follows.

JDAW’s Response

20^th June 2010

Dear Mr Goodspeed,

Thank you for your letter of 7^th June 2010 replying to my letter of 26^th May. It makes several important observations about the UK’s debt programme, and “issues” relating to my proposal. These require further comment.

Stakeholder support

Innovations also have to meet other objectives of a practical nature including … engendering stakeholder support.

◊ Though seemingly desirable, there is a serious problem with “stakeholder support”. Stakeholders’ interests do not always coincide with good public policy. In particular, high transaction costs for the DMO and for gilt investors can be excellent for the GEMMs: why would they object to a transfer of wealth from the taxpayer and pensioners to the investment banks? Hence GEMMs, substantial stakeholders, could hardly be expected to support innovations that reduce the advantage of those at the centre of the market relative to those at the periphery. So HMT should not require “stakeholder support”; HMT should listen to stakeholders’ arguments, and indeed others’ arguments, and judge them on their public-policy merits.

In particular, if a non-small GEMM credibly says that a proposed change would cause its withdrawal from the market, that would be an important argument. But if a GEMM argues that a reduction in profits—which of course could be a reduction in others’ transaction costs—might cause other GEMMs to withdraw, then that is not really an objection at all. The purpose of having many GEMMs is to reduce transaction costs for the DMO and for investors. Needlessly failing to lower transaction costs, in order to have more GEMMs, would be a perverse confusion of ends and means.

Auctionette issues

your proposal raises a number of issues in practice and in principle that require further consideration. These issues include issues of principle such as any implied underwriting by primary market dealers at each "auctionette", the role of price discovery and practical issues such as IT and systems modifications.

◊ Yes, there is de minimis underwriting at each auctionette. But the quantity is tiny: for a £100mn auctionette in units of £100k, with 15 GEMMs, each GEMM is at risk of buying £6.7mn at a yield marginally above the market price of 60 seconds previously. Compared to the risks faced in a real underwriting in the equity market, with the possible purchase of billions of pounds of equity at a price fixed several days before, this risk is negligible.

For example, imagine that, during an auction of a 50-year gilt, the yield jumps up +8bp. The GEMMs would have deemed bids at the next auctionette of +1bp, and at the following few auctionettes at +1bp above each previous. So the GEMMs would lose money, a total of about £5.9mn.^† But under the current system a +8bp move in yields, after bids are submitted but before positions could be sold, would cost the GEMMs £68mn, about 11.6 times more. And the current system is at greater risk of generating such a jump in yields, it being a very plausible consequence of the announcement of an uncovered or nearly-uncovered auction. So if there is ‘underwriting’, it is of much less risk than that embedded in the current system.

◊ Next comes “issues of principle such as … the role of price discovery”. There is price discovery. Bids are submitted to each auctionette, the top so-many million are accepted, the cutoff announced, all pay the cutoff price. Very rarely will the clearing price be the minimum price of the auctionette, and even it were, the following auctionette’s minimum price would be a DV01 lower, so normal price discovery would quickly resume.

Indeed, in many ways auctionettes facilitate price discovery. Some years ago, at an auction of five-year gilts, I wanted to buy versus selling ten-years. If I had bid, should I have sold the ten-years immediately (not knowing how much five-year I was about to buy); or after the results had been announced (that announcement changing the price of the ten-year gilt)? Neither was satisfactory, so the trade was entered after the auction. But with the auction split into forty auctionettes, I could have bid for some five-years at the first auctionette, then sold futures, and repeated. This would have been better price discovery.

◊ It might be thought that “IT and systems modifications” would be an “issue” as large IT projects do not have a happy record. It might be thought that this could introduce substantial implementation risk. But there are multiple, if not many, recognised exchanges that would love this business, and that already have the systems architecture to make it happen. The DMO’s consultation on this subject should invite exchanges to submit proposals, requiring that they should be accompanied by a working demonstration of the software. There is IT risk, but it need not be the Treasury’s risk.

Supplementary gilt distribution methods

the supplementary gilt distribution methods were introduced in Budget 2009 and have improved the Government's ability to smoothly access the investor base for long-dated and index-linked gilts and supported the gilt auction programme. Syndication does so partly by adopting an iterative 'book building' process which has some similarities to the sequential approach to gilt sales as outlined in your letter.

◊ What is the purpose of syndication? Let us hypothesise that a small country with infrequently traded debt, say, Paraguay, wishes to issue a £500mn bond (whether or not swapped into another currency). But there is no regular market in Paraguay’s bonds, excepting small pieces (≈$5mn) that trade infrequently with a non-small bid-ask spread. Further, most potential investors have no recent experience of Paraguay as an issuer: what drives the spread, what are the matters of concern? So an auction may well come at a price that is very risk-averse, i.e., cheap. From the viewpoint of the issuer this would not be good. This is, in some sense, a collective action problem, solved by issuing via syndicate.

• The investors collectively determine a fair price (or a fair spread), and all deal at that collectively determined price, knowing that all have done so.
• The Lead Managers also give their imprimatur to the transacted price, implicitly or explicitly stating that it is fair. For an investor buying a small piece, for whom it would not be economically efficient to research the product, this can provide some comfort.

But this hardly applies to a gilt, in which there is an active market, there also being an active market in a comparable gilt of nearby maturity. It cannot advantage the UK to describe—albeit implicitly—its own debt in its own currency as being partway-to-Paraguay.

So why is the UK issuing by partway-to-Paraguay syndicate? (To be fair, France also uses syndicate; Germany restricts it to foreign-currency issues; the US prefers auctions.) Of course, syndicate is used because the auction mechanism cannot guarantee finding a clearing price. But this should be seen as a reason to choose a better auction mechanism, not as an excuse not to do so.

However, it is not in the interest of the GEMMs to make this argument. So far they have been paid £72mn in fees (a number about which the DMO has been rather coy, nowhere stating it in terms sufficiently clear for it to have been noticed by the Treasury Select Committee). That sum of money, paid in fees for a series of zero-risk transactions, doubtless buys much agreement from the GEMMs, but that doesn’t make it the best deal for the UK taxpayer.

There are two possible good reasons for the UK to issue by syndicate.

• First, if there were to be issuance of something genuinely new. This might be—as a hypothetical rather than a recommendation—a GDP-linked gilt. For the first opening of the first such gilt, there would be no comparable: the fair price would be genuinely unknown. For this, syndicate might well be appropriate.
• Second, if the issuance were to a very different group of investors, as would usually be true for an issue denominated in foreign currency. And, rightly, previous non-£ issues by HMT and by the BoE have been sold via syndicate.

But neither of these apply to a conventional or a linker, even if twice as long as the previous longest.

Post Auction Option Facility

Moreover, the new Post Auction Option Facility (PAOF) encourages competitive bidding at auctions as the amount available to each participant under the PAOF is directly linked to the amount purchased at auction.

◊ In the mid-1990s the author was part of a conversation with the Belgian authorities about the relative merits of issuing by syndicate (the fees being desired by my employer) and some form of post-auction options (also desired by my employer). The Belgian authorities were quite frank about the problem: if issuing by syndicate, fees must be paid, and those fees must appear in a cost column somewhere. But if giving away options, the accountants don’t notice that the OLOs are sold only at a below-market price. Please could somebody at HMT show me where, in any accounts anywhere, the UK’s equivalent cost is shown? And if it isn’t, can the authorities see the risk that this could be perceived as a dodge of accounting rules?

However, the UK is correct to issue debt via options. More debt would be issued into a rising market, and less into a falling market; and market participants holding options causes market-stabilising delta-hedging. But this requires options with multiple days to expiry, and these options should be sold rather than given away, as described in the author’s reply to the DMO’s consultation of December 2008.

But in the PAOF bidding by investors causes options to be given to GEMMs: hence this is a tax paid by investors to GEMMs. The GEMMs should like this more than the authorities.

The DMO’s incentive

Total Derivatives is a subscription-only newsletter for the fixed-income derivatives market, and the following comes from its newsletter of 28^th January 2010.

As one trader said “there’s no point for a DMO chief to try and tell people to buy gilts at very low yields and then trigger a failed auction. [Stheeman] needs to help massage yields higher, which is what he is doing, to a level where offshore investors will buy them. …”

Of course, the Treasury should deny that the DMO chief was trying to “massage yields higher” in order to prevent a failed auction (it is not the only possible interpretation of Robert Stheeman’s interview in the FT of 28^th January 2010). Further, the Treasury should deny even that the DMO has an incentive to massage yields higher. But it is very bad that market participants believe that the DMO has such an incentive.

Indeed, if that is what the DMO chief was doing, and one holds constant the refusal to fix the auction mechanism, it wasn’t a foolish tactic. An uncovered auction, the UK being seen to fail to fund itself, could be disastrously expensive. So it might have been sensible for Robert Stheeman to attempt to cheapen gilts by, say 5bp, by warning of “uncertainty in the bond markets, which could lead to rising government yields”, that we could “move into a different pricing environment”, and even “that gilt yields are close to historic lows”. Slightly cheaper gilts might have attracted more bids, thus avoiding a failed auction.

But even if this was a good tactic, fixing the auction mechanism would be a better strategy.

Conclusion

None of the issues raised by HMT’s reply really argues against the proposal, and one—syndication—strengthens the arguments in favour.

Please move this forward, before rather than after the current system contributes to a ratings downgrade.

Footnotes

† This assumes an auction of a 4% 50-year gilt, £4bn split into 40 auctionettes each of £100mn. The gilt starts with a market price of par and so with a DV01 of 21.55 pence, this therefore also being the maximum change in the minimum price of one auctionette relative to the previous. We assume that, for whatever reasons, the market yield then jumps to 4.08%, equivalent to the market price falling to £98.2995. Presumably the clearing prices of the next seven auctionettes would be their minimum prices: £99.7845, £99.5690, £99.3535, £99.1380, £98.9225, £98.7070, £98.4915, after which the clearing price would be the new market price of £98.2995. Losses therefore total £5,869,765, which if split amongst fifteen GEMMs would be ≈£391k each. This is not a trivial sum of money, but as a rarely incurred loss, effectively irrelevant to GEMMs’ reasoning about auctions.

Afterword

22^nd June 2010: from an interview with Total Derivatives of Robert Stheeman:

This week, independent analyst Julian Wiseman published details of communications between himself and the Treasury regarding his suggestion of replacing regular gilt auctions with a series of auctionettes – a plan he says would be safer and cheaper than the current process (see GBP Bonds: Time to change gilt auctions?).

The DMO’s Stheeman said today that any such suggestions are appreciated He also cautioned that any auction mechanism “must work on the practical side as well as from a theoretical perspective. We must know that it does not cause any practical difficulties for us or the GEMMs and must help, rather than complicate, the auction process.”

Is this an acceptance that it works in theory? Great. So what test can be done to show that it works in practice, short of doing it?

Meanwhile, the DMO should not omit to say that, in practice, the current rules do actually produce failed auctions. Rephrased: the current rules don’t work; the new rules at least “work … from a theoretical perspective”. Should we conclude that the DMO is waiting until another country does it? Let us hope that the delay isn’t too expensive.