1

Title:

Auctioning Government Securities: The Puzzle of Overpricing

Author:

Dr. Riccardo Pacini Department of Economics and Institutions, University of Rome “Tor Vergata”, Via Columbia 2,

00133 Roma, E-mail: riccardo.pacini@uniroma2.it.

Field:

Empirical Finance: Asset Pricing and Market Microstructure

Keywords:

Overpricing, Government Securities, Primary Dealers, Commodity

Bundling, Multi-Unit Auctions

2

Auctioning Government Securities: The Puzzle of Overpricing

RICCARDO PACINI ∗

Abstract. This paper explores if the overall institutional setups to place Government

securities adopted by eurozone countries preserve a regular functioning of the market,

principally looking at the primary market pricing performance with respect to the

secondary market. However, I find that Government securities are always overpriced.

The empirical analysis shows that this cannot be viewed as a success of the issuing

technique or as the result of the discretion enjoyed by some Treasuries in the stop-out

price setting procedure, but rather as a consequence of the Treasuries bundling the

securities auctioned with a number of commodities, such as the syndication rights. Since

the overpricing may harm the long-run competitiveness of Treasury auctions by driving

the smaller primary dealers out of the primary market, I suggest discontinuing the

bundling or exploring alternative auction mechanisms such as the clock auctions, and in

particular those which put up different items simultaneously.

JEL No.: G10 (General Financial Markets), D44 (Auctions), H63 (Debt; Debt Management)

Keywords: Overpricing, Government Securities, Primary Dealers, Commodity Bundling, Multi-

Unit Auctions.

∗ Department of Economics and Institutions, University of Rome “Tor Vergata”, Via Columbia 2, 00133 Roma. I would like to thank Davide Iacovoni for institutional information and for making many data available to me. I am grateful to Michele Bagella, Gustavo Piga and especially to Giancarlo Spagnolo for useful comments. A special thank to Lawrence M. Ausubel to whom I am particularly indebted for crucial insights on the bundling issue and its consequences. Finally, I would like to thank many public debt managers participating at the 15th OECD Global Forum on “Public Debt Management and Emerging Government Securities Markets” (12-13 December 2005) in Amsterdam for providing me with key institutional information and participants of seminars at the University of Rome “Tor Vergata”. I also wish to thank Rita Zajko. All remaining errors are mine.

3

1. Introduction

The debate about the effectiveness and efficiency of the auction mechanisms adopted by

sovereign issuers of Government securities is long-standing. Someone may question why are the

distinctions theoretical literature makes between different auction formats so important in Treasury

security issuances, if award prices may change from one format to another just for few basis points,

for auction peculiarities to play an important role. However, just to have an idea of the effect of a

saved basis point on a gross public debt of billions of € and to understand why every element of the

overall auction procedure is so important, consider the gross issuances in 2005 and the outstanding

debt at the end of the same year of the main EMU sovereign issuers (see Table 1 in Appendix I).

Moreover, the growing needs of funds push sovereign issuers to focus more and more on the

liquidity of issuances, trying to lower as much as possible their funding costs. Indeed, according to

a recent review of the latest developments in the management of Government debt in the euro area

(see de Haan and Wolswijk, 2005), the main objective of debt management agencies is to finance

the public debt at low costs with acceptable risks.1 In addition to cost and risk related objectives,

many debt managers have other goals such as duration, relative performance and broadening the

investor base, promoting the liquidity and infrastructure of the debt market, and consistency with

the aims of monetary policy. Hence, a well functioning Government debt market should result in

lower issuance cost and issuance risk. To this respect, the ways to produce cost savings may be

different, given that debt management decisions typically deal with the choice of instruments,

issuing techniques and institutional arrangements that minimise debt-servicing costs, given a certain

risk profile.

The objective of this paper is to raise the broad question of whether the current institutional

setup is optimally configured to make sovereign issuers match their objectives in the most efficient

way, principally looking at the pricing performance of Treasury auctions with respect to the current

secondary market prices. There are several studies which survey Government primary market

practices: Bartolini and Cottarelli (1997) document the auction formats adopted in 42 countries,

Arnone and Iden (2003) bring evidence on the diffusion of the primary dealership system

worldwide, Sareen (2004) makes a review of the overall placement system of a number of OECD

countries, Brenner, Galai and Sade (2007) investigate the issuer’s choice between the multiple price

and the uniform price auctions on a global scale, and Bagella, Coppola, Pacini and Piga (2007) 1 Low risk is generally defined as stable interest expenditure and would imply that a debt manager should issue primarily at the longer end of the yield curve. Such a strategy may, however, conflict with the low cost objective as the yield curve is normally upward sloping and so favours issuance at the short end from a cost perspective. A variety of approaches have been adopted to balance these potentially conflicting objectives, including cost-at-risk (CaR) analysis, the use of derivatives (e.g. interest-rate swaps), extending the investor base, fixed/floating ratios and duration targets.

4

along with the auction formats review all the institutional details to place Government securities

throughout the eurozone.

Obviously the main issue in the debate concerning the best way to place Government securities

is the auction format, in particular contrasting the multiple price auction (also pay-your-bid or

discriminatory auction) with the uniform price auction. The multiple price auction is the traditional

format used for Government securities worldwide, and it was also used in the U.S. for all types of

securities until the early 1990s. However, in 1993, the United States started to run uniform price

auctions for 2- and 5-year Government notes as an experiment, following an old proposal of Milton

Friedman made in 1959 and adopted in the Joint Report on the Government Securities Market

issued in 1992 by the Department of the Treasury, the Securities and Exchange Commission, and

the Board of Governors of the Federal Reserve System.2 This change was prompted by violations

by Salomon Brothers in 1991.3 Afterwards in 1998, the United States switched entirely to the

uniform price auction for all issues. Indeed, to some extent auction theory supports the uniform

price format, since it is argued that it is less prone to the winner’s curse and to the bidders

collusion.4 Ausubel and Cramton (1998) show that the auction revenues associated with the best

scenario which may occur in the uniform price auction are higher than those possible in any

scenarios adopting the multiple price auction. Nevertheless, Back and Zender (1993) and Ausubel

and Cramton (2002) show that the uniform price auction is often subject to inefficiencies which

lead to poor revenue-raising performances. Hence, the revenue ranking of the uniform price auction

and the multiple price auction is ambiguous and determining the better pricing rule is necessarily an

empirical question.5 This is reflected in the varied practices worldwide, as Sade, Schnitzlein and

Zender (2004) observed:

“Many countries use auctions to sell their Treasury securities. While the institutional

details of the auctions differ in a variety of subtle ways across countries, most use a variant

of either a uniform or a discriminatory pricing rule. In recent years there are examples of

countries switching from the discriminatory to the uniform-price format (e.g., the United

2 Besides Friedman, other economists too have argued that U.S. Treasury securities should be sold by uniform price auction (e.g. see Miller, 1991, and Chari and Weber, 1992). 3 The amount of a security any single firm can purchase was limited by U.S. Treasury to 35% of amount issued. By submitting unauthorized bids on behalf of its customers, Salomon repeatedly violated this restriction. Indeed, in the May 1991 auction for 2-year notes, for instance, Salomon admitted to having cornered 94% of the amount issued. Salomon then took long positions on the when-issued (forward) market and squeezed short-sellers. 4 See Umlauf (1993) on the effects of the switch operated by the Mexican Treasury from a disriminatory to a uniform price format on collusion between the major banks participating at the Mexican auctions. 5 Castellanos and Oviedo (2004) list 7 changes between discriminatory and uniform-price since 1990, whereas Brenner, Galai and Sade (2007) find in about 50% of a 48 countries sample that countries employed in the past a different selling mechanism than the one they currently use.

5

States) and examples of countries making the reverse change (e.g., France), indicating that

the choice between these mechanisms remains an unsettled issue.”

However, since very often the devil is in the details, I argue that restricting the attention just at

the auction formats may not be sufficient to understand the functioning and the outcome of the

placing procedures. Then in the empirical analysis I take into account both some critical features of

the overall auction procedure and factors exogenous to the auction mechanism per se, such as the

practice of granting the best auction participants the right to syndicate specific issuances, the

participation in profitable debt management operations or the option to buy Government securities

at the auction price the day after the same auction. Indeed, whereas the former may condition the

auction outcome, since even a trivial change in the auction design can have a dramatic impact on

prices,6 as Daripa (2001) and Kremer and Nyborg (2004a) respectively argued in the case of

uniform price auction:

“.......The ranking also shows that uniform-price auctions are susceptible to institutional

change. Discriminatory auctions, on the other hand, are very resilient to institutional

changes. The revenue from a discriminatory is the same across institutions (with or without

a minimum quantity, fixed-supply or random-supply). Thus discriminatory auctions are

probably a ‘safer’ option for the Treasury in situations where markets are affected by a

large set of institutional features and the effect on uniform-price auctions is therefore

uncertain.......”;

“.......the micro design of a uniform price auction can dramatically improve its

performance.......”;

on the other hand, the latter may bias the pricing of auctioned Government securities bearing on the

incentives of auction participants with respect to their bidding behavior. In this regard, it is

common practice worldwide for Treasuries to avail themselves of a primary dealership system to be

sure of the placement of all the securities (see Sareen, 2006). In such a system, issuers often restrict

access to the primary dealers by imposing participation and activity obligations on them on the

primary and secondary markets against rents conditional on their compliance with the same

6 To this respect, besides the switch to the uniform-price auction for 2-year and 5-year notes, the Joint Report on the Government Securities Market (January 1992) by the U.S. Dept. of Treasury, SEC, and the Federal Reserve Board, proposed a stronger enforcement of auction rules and to broaden participation in the auctions for all securities. For further institutional details on US auction techniques see http://www.treasurydirect.gov/instit/auctfund/work/work.htm.

6

obligations, which is assessed according to a special ranking compiled each year by Treasuries. A

major weight in the ranking is given to the quantity of securities won in auction. This gives rise to a

sort of commodity bundling between the securities being auctioned and the rewards attached to

them. Then primary dealers actually bid for a bundle consisting of a Government security and an

enhanced probability of winning for example profitable syndication and/or debt management

operations rights.7

Therefore, I assess the overall institutional setup to place eurozone Government securities

looking at the auction pricing performances and the relating determinants: first I measure the

presence of misalignement between primary and secondary market prices along the auction day and

then I try to shed light on the causes of the mis-pricing between the two markets, testing some

implications coming from auction theory and controlling for the main features of the overall

institutional setup. Following the empirical results, I further focus on the latter making a tentative

evaluation of one of the main rents granted by the issuer, i.e. the syndication rights, to see whether

the auction prices bias can be explained in monetary terms by such conditional rents.

The remainder of the paper is organized as follows. Section 2 reviews the auction formats

adopted in the eurozone, discusses some of the main institutional arrangements to place

Government securities and illustrates the operation of the primary dealership system. Section 3

measures the auction pricing performance with respect to current secondary market prices. Section

4 carries on the empirical estimation to identify the determinants of the primary and secondary

markets mis-pricing. Section 5 makes a tentavive evaluation of one of the most profitable rents

granted to primary dealers and Section 6 further elaborates on the role of the commodity bundling

within the economics of Treasury auctions. Section 7 summarizes all the findings and discusses

some avenues for further improving the auctioning of Government securities.

2. Auction formats, institutional features and primary dealership systems

This Section aims at exploring the overall auction environment, with a particular emphasis on the

main relevant institutional details to place Government securities: from the auction format to the

stop-out price setting procedure, from the syndicated issuances to the practice of ranking primary

dealers, from the access to auctions to the obligations imposed upon primary dealers. Special

7 See Grimm, Pacini, Spagnolo and Zanza (2006) for a comprehensive analysis on the economic principles behind the practice of bundling, though in the context of procurement.

7

attention is given to the operation of the current primary dealership system, through a detailed

analysis of primary dealers’ requirements, duties and rewards.

2.1 Multi unit auction formats for Government securities and the main institutional arrangements

Eurozone sovereign issuers adopt about the same issuance procedures to place Treasury bills

and bonds.8 Government securities are issued either through auction, tap procedure or syndication.

As Table 1 shows, their usage does not vary significantly by country, the prevalent scheme being

the use of auction for both T-Bills and T-Bonds and syndication for specific T-Bonds issues. Few

Member States (Finland and The Netherlands) adopt more than one of the above techniques

depending on the type of security. Relative to the main institutional structures of Government debt

management, issuance is generally either outsourced to a debt management agency or remains

within the Ministry of Finance (in some countries performed by the Central Bank). From now

onwards I refer either to a debt management agency or a Ministry of Finance as “Treasury”.

In all eurozone countries auction mechanisms are used, usually through the multiple price

format (in ten countries). In particular six countries run the multiple price auction for both T-Bills

and T-Bonds. Three countries (Finland, Italy and The Netherlands) instead use the uniform price

method for some of their issuances (T-Bonds in Finland and Italy, T-Bills in The Netherlands) and

one, Spain, adopts a hybrid format for both T-Bills and T-Bonds. Finland and The Netherlands

issue a relevant part of the debt by tap: the former to issue T-Bills, while the latter for T-Bonds.

However, the operation is not performed in the same way in each country.9 In most Member States

syndication remains a common practice or an open option especially for new benchmark bonds,

long term bonds and index linked lines.

8 The securities placed through specific programs (e.g. Global Bond Program, Medium Term Note Program, Commercial Paper Program, etc.) and retail instruments (e.g. they are often specific instruments issued through dedicated placement channels, available to the public at large), which typically account for a small part of the total debt, are not included in the following analysis. 9 Finnish Treasury Bills are currently issued under the T-Bills programme, via what is called the ‘daily window’. Whenever there is a need to issue T-Bills, the State Treasury notifies Dealers of the reference price and quantity for sale. The reference price is the price Dealers are allowed to buy at. In The Netherlands, the Dutch State Treasury Agency (DTSA) issues T-Bonds (DSL) through an innovative procedure, called “tap auction”, which borrows elements both from a standard tap and a standard auction. This technique consists of the seller announcing the price and the buyers indicating the desired amount. Every second Tuesday of the month, the DSTA announces, at 10 o’clock, the initial ask price at which primary dealers may take up the issue, without communicating the quantity on sale. During the tap-auction, the DSTA adjusts the price upward or downward if capital market developments in general, or demand for the bond in particular, so require. The DSTA places two to three billion euro per month via this procedure. Only Primary dealers are entitled to participate to the tap-auction. The tap-auction usually lasts 5 to 25 minutes, depending on the interplay between DSTA and the Primary dealers. No statistics are released during the tap procedure. According to the DSTA “…Tap auctions have two advantages compared with the issuance technique commonly used in many countries. No time is lost between subscription and allocation, thus eliminating the auction risk. The winner's curse phenomenon, whereby the market falls if a high volume is allocated and vice versa, is also less likely to occur…” (see p. 51 of the Annexes of the DSTA’s annual report “Dutch Government Securities 2005”).

8

Syndication Auction Tap Auction Tap AuctionReopening after the auction

AUS Austrian Federal Financing Agency

Option subject to investor demand Multiple - Multiple - PDs PDs No

BEL Belgian Debt Agency To launch new lines Multiple - Multiple - PDs and Recognized Dealers PDs Yes

FIN State Treasury, Finance Division To launch new lines Uniform - - X PDs PDs Yes

FRA Agence France TrésorOption to launch new long

term and index linked lines

Multiple - Multiple - Unrestricted PDs Yes

GERFederal Ministry of Finance, German Finance Agency

Option to enter a new market segment or to use a

new instrument*Multiple X Multiple - Auction Bidding

Group - Yes

GRE Public Debt Management Agency

To launch new long term lines Multiple - Multiple - PDs PDs No

IRL National Treasury Management Agency No** Multiple - - X PDs PDs Yes

ITAMinistry of Economy

and Finance, Department of Treasury

To launch new long term and index linked lines Uniform - Multiple - Unrestricted PDs No

LUX Trésorerie de l’Etat Yes Multiple - - - Unrestricted - No

NTL Dutch State Treasury Agency No DDA*** X Uniform - PDs and Single Market

Specialists - Yes

POR Portuguese Government Debt Agency

Option to launch new long term lines Multiple - Multiple - PDs and Other auction

participants (bonds) PDs No

SPA Treasury Department Option to launch new long term lines Hybrid - Hybrid - Unrestricted PDs Yes

* The syndication procedure was adopted the first time on March 2006 to issue the first index linked bond.

** The NTMA employed a syndication procedure for the first time to launch the 4.5% Treasury Bond 2018 during 2007.

*** According to the Dutch State Treasury Agency the DDA (Dutch Direct Auction) is an alternative issuance method to launch new benchmarks, which borrows elements both from a standard syndication and a standard auction, since it employs a book-building process via Primary Dealers, to obtain greater direct engagement during the issuance with the international investment community, and adheres to the principles of the uniform price auction. The responsible for the book-building process is the DTSA itself and three Primary Dealers are appointed to assist the DSTA in the formulation, preparation and execution of the whole procedure (see p. 3 in "Dutch Direct Auction: DDA Rules", October 2004, Dutch State Treasury Agency).

Source: Sovereign issuers' official websites.

Member States

Government Debt

Management

TABLE 1 - Issuance Procedures in the EMU

T-Bonds T-Bills Access Discretion in the stop-out price

setting procedure

Therefore, three are the auction formats employed to issue Government securities in these

countries: the multiple price auction, the uniform price auction and the “Spanish” auction (the name

obviously refers to the fact that Spain is the one using it). In all three formats, bidders submit

sealed bids comprising price-quantity pairs in advance of a deadline. After the deadline, the

auctioneer unseals the bids and aggregates them, determining the clearing price (stop-out price) at

which demand equals supply. Hence, securities are awarded in the order of descending price until

supply is exhausted: all bids submitted at prices higher than the stop-out price are allotted for their

full amount, whereas bids submitted at the stop-out price may be rationed in the case of excess

demand. So, while each bidder wins the quantity demanded at the stop-out price in all the formats,

the three auction formats differ in the payment rule. In the multiple price auction, bidders pay what

they bid. In the uniform price auction, all winning bidders pay the same price, namely the stop-out

9

price, which is the lowest accepted bid price. In the “Spanish” auction bids made at the minimum

price are accepted at the same price; bids falling between the minimum and rounded-up weighted

average price also pay the price actually bid; and bids higher than the rounded-up weighted average

price pay the rounded-up weighted average price. Hence, the Spanish auction combines elements of

the uniform price auction with elements of the multiple price auction.10

Despite the rich analysis and the numerous results on auction formats obtained by auction

theory, there are other important dimensions of the entire auction procedure which may have a

critical role in determining the auction outcome. Hence, besides the auction format per se, I review

two critical features of the actual auction procedures (see Table 1), in particular:

- Access: in most “small” Member States, the access to auctions is typically restricted to primary

dealers and in three cases (Belgium, The Netherlands and Portugal) also to some other

“recognized” dealers. In the largest countries of the euro-zone (France, Germany, Italy and

Spain) the access is unrestricted but for Germany, where auctions are open only to those banks

belonging to the so-called Bund Issues Auction Group.11 It does look therefore that the pressing

need of small and medium countries to obtain a liquid issuance program requires them to grant

more benefits to primary dealers, in this case in the form of a more protected environment for

primary dealership. Indeed, in a repeated auctions framework with a group of informed bidders

(in this case the primary dealers), Daripa (2001) shows that allowing free entry to uninformed

outsiders (in this case the non-primary dealers) raises the prices paid in auctions by the

informed bidders, which makes it unprofitable for the uninformed outsiders entering the

auctions.

- Stop-out price setting discretion: during the auction, many Member States explicitly or

implicitly enjoy the right to choose the lowest accepted bid price with a great degree of

discretion, thus playing an active role in the auction, determining the revenues and the

borrowing costs deriving from the auction and possibly influencing bidders’ strategies, which

would reflect the uncertainty created by the seller’s discretion.12 In two Member States this

option is explicitly provided for, that is in Germany and The Netherlands, whereas in other

Member States (such as Finland, Belgium, France, Spain and Ireland) the right to choose the

stop-out price after observing the bids is implied in other provisions concerning the issuance

amount finally allotted. In all these countries there are no specific written down rules that drive

10 See Álvarez, Mazón and Cerdá (2003) for a first theoretical comparison between the Spanish auction and the two standard auction formats and Abbink, Brandts and Pezanis-Christou (2006) for some specific laboratory results on the revenue-raising abilities of the three auction formats. 11 See the next Paragraph. 12 For example, Back and Zender (2001) prove that such a right operates as a deterrent for some of bidders’ speculative behaviors which bring about the inefficiencies to which the uniform price auction is subject (for further insights see also Lengwiler, 1999, LiCalzi and Pavan, 2005, and McAdams, 2006).

10

the discretionary behavior, leaving them great room for manoeuvre.13 On the other hand, the

remaining Member States, which do not provide for any type of intervention in the stop-out

price setting procedure, waive an important tool to condition the auction outcome.14

2.2 The primary dealership system

To the face of monetary union, the Government security market is fragmented into twelve

domestic markets, some of which are quite small in scale. Moreover, the growing needs of funds in

the European market make room for an active competition among euro-area issuers, which are

focused more and more on the liquidity of issuances, trying to lower as much as possible their

funding costs. Such a competition forces convergence in some key-areas of debt management.15

However the most common practice in eurozone is to combine auctions with a primary dealership

system.16 This system was introduced to improve the price discovery process, by requiring primary

dealers to actively participate in auctions, to continuously quote two-ways in the secondary market,

to foster foreign demand and to reach more non-domestic investors by including foreign firms

among them. This is particularly true for small sovereign issuers, otherwise marginal to larger

markets. On the other hand, sovereign issuers should create some incentives to gain participation in

their domestic markets, that is to provide for some rewards conditional on that. Sareen (2006)

shows that the main reason for the adoption of a primary dealership system indeed relies on primary

dealers granting the full placement of the amount of securities put in auction (this is the main

obligation), although in return for profitable privileges with respect to non-primary dealers, giving

raise to a sort of commodity bundling.17 Indeed, this is the case for the eurozone countries too.

All eurozone countries, except Germany where auctions are open to all those institutions which

belong to the so-called Bund Issues Auction Group (see the end of this paragraph), developed a

formal system of primary dealership to ensure the placement of their debt in their primary markets

13 See Rocholl (2005) and Keloharju, Nyborg, and Rydqvist (2005) on the main drivers of discretion in this phase, in particular in German and Finnish Treasury auctions, respctively. 14 See Back and Zender (2001), McAdams (2006) and Lengwiler (1999) for a discussion on the benefits of enjoying some freedom to determine the amount to issue once received all the bids, which basically implies the right to discretionally intervene in the stop-out setting procedure. 15 For example, efficient links have been established between EU Member States settlement systems, market conventions have been harmonised, and the offer has focused on standard “plain vanilla” bonds which are easy to trade. 16 See McConnachie (1996), Gray (1997), Arnone and Iden (2003) and Sareen (2004) for previous reviews on the subject. In particular Arnone and Iden (2003) document the popularity of the primary dealership system versus the open participation system worldwide. 17 More specifically, Sareen (2006) argues that the primary dealer model is one channel to resolve the agency problem faced by an issuer in the primary market for Government securities, namely the presence of too few buyers who are willing to subject themselves to participation obligations of the issuer, raising the likelihood of failed primary issuances.

11

and to enhance the liquidity of the related secondary markets.18 With a view to the eurozone

Treasuries’ goals regarding allocation, transparency and liquidity of Government securities, it may

be desirable to be assisted by primary dealers in the purchase, issuance, trading, promotion and

support of Government securities. It is in the best interest of the Treasury that Government

securities be allocated adequately to end-investors, that market liquidity and transparency in the

secondary market be maximised, and that their status of first class debtors in the capital market be

maintained and strengthened. This system normally concerns both T-Bonds and T-Bills markets.

Typically, eurozone Treasuries make use of the services of primary dealers in auctioning,

promoting and distributing Government securities. Therefore, primary dealers are banks which

have chosen to dedicate part of their staff to placing and trading Government securities. In 2005 the

number of primary dealers in the eurozone ranged from eight (Finland) to twenty-five (Austria) and

tipically includes large banking networks, credit institutions, capital market specialists and

securities trading houses. A one-year contract is generally concluded between each primary dealer

and the Member State,19 involving on the one hand a composite system of requirements and

obligations, and on the other a series of privileges and rewards.

This Paragraph is aimed at reviewing the actual advantages and duties of being a primary dealer

operating in EMU Member States Government bonds and bills markets and at assessing their degree

of harmonisation. These rules reflect the belief that effectively dealing and placing Government

securities requires not only regular participation in auctions but also an active presence in the

secondary market and a willingness to play an advisory role with the State issuer. The main

characteristics of the primary dealership systems adopted in most Member States, in particular the

selection criteria, the rewards and the duties of primary dealers, are the followings:

- Requirements to apply for becoming a primary dealer: generally to obtain the status of “primary

dealer”, a number of market activities and structural requirements must be met and maintained.

The most widespread requirement refers to the existence of sufficient and qualified human

resources, whose presence is deemed by the issuer vital in establishing and maintaining a high

level of confidence in the execution of the task. Most Member States require from primary

dealers a clear capacity in placing securities and a continuous effective participation in the

secondary market. Furthermore, primary dealers are often selected on the basis of their 18 Whereas someone may argue that the Bund Issues Auction Group is a milder form of primary dealership, it is the same German Finance Agency which emphasizes that: “The German Finance Agency has a unique involvement in the country’s government bond market, helping to support the liquidity of the secondary market in the absence of primary dealers. Germany is the only large sovereign issuer of euro-denominated bonds that appoints no primary dealers. In all other large government bond markets, a selected group of international banks is committed to supporting the efficient launch and distribution of new bond issues at a primary level as well as a liquid market for the securities at the secondary level.…” (see the Euromoney issue of April 2007 on the Special Publications section of the German Finance Agency official website, www.deutsche-finanzagentur.de). 19 Usually such agreements are confidential.

12

financial strength, as well as their capability for the physical and financial settlement of

Government securities and in some cases also on a minimum market share maintained during a

specific period in the primary market.

- Primary dealers’ obligations: once a financial operator has become a primary dealer, there is a

number of duties to be fulfilled both on the primary and secondary market, which mainly relate

to participating in auctions, placing Government securities and maintaining a liquid secondary

market. More specifically:

• In primary markets, primary dealers must participate to the issuances, often by bidding on

each line of any auction, in the majority of cases compatibly with an average annual

minimum participation obligation in auctions. Other obligations may relate to the allocation

of sufficient resources to support the issuer in its debt management operations and to avoid

any distortion in auction prices;

• In secondary markets, primary dealers face a number of obligations aimed at maintaining an

active participation and enhancing the overall liquidity, by quoting firm prices or rates for at

least a minimum amount of securities and within a maximum spread, displaying indicative

prices on screens and achieving a minimum market share. Some Member States require

them also to ensure the liquidity of the related repo market;

• More generally, in most cases they are also requested:

o to advise and assist the issuer on matters related to issuance policy and debt

management, as well as on questions of a more general nature pertaining to the workings

of the market, and to keep the issuer informed on market developments;

o to promote Government securities by adequate analysis, research and publications;

o to report regularly the Treasury on their activity both on primary and secondary markets.

- Primary dealers’ privileges: as counterparts to the above obligations, primary dealers enjoy a

number of privileges, principally in the primary market. The main privileges offered to them

are similar across Member States: they have in most countries an exclusive right to make

competitive bids at the auction and non-competitive bids after the auction, to participate to debt

buy-backs and exchange operations and to strip and reconstitute Government securities. In

many Member States, they also benefit from an exclusive or privileged access to syndicated

issues and to the repo market. Moreover, the condition of Primary dealer will be taken into

account when the Treasury chooses counterparties for other debt management operations, such

as swaps, foreign currency issues or issues by systems other than auctions. Beside the

privileges on the primary market and the exclusive right to carry the title of Primary dealer, they

often have the right:

13

• To receive all relevant information about issuance policy and other public market operations

and participate in meetings with the issuer;

• To be privileged counterparties of the issuer in its overall debt management activity;

• To be compensated with commissions.20

The activity of primary dealers is constantly monitored in most Member States, according to

predetermined criteria. Indeed, most Member States evaluate and rank primary dealers looking at

all of the components of their obligations: their participation in auctions (or more generally primary

issuances), their presence in the secondary market and the qualitative aspects of the relationship

(information, advice, quality of research and promotion of Government securities). Typically, the

assessment of their activity is made by assigning different weightings to all such components.

Accordingly, each year the league table of the most active primary dealers in Government securities

is published. A good ranking is often needed to have a chance for a place in the syndicated

issuances and to access other profitable debt management operations.

Conversely, in Germany there is not a formal primary dealership system (see note 35). Still, the

access to auctions is not unrestricted, since they are open only to those institutions which belong to

the so-called Bund Issues Auction Group. Applications for membership of the Bund Issues Auction

Group may be submitted to the German Finance Agency at any time. Applicants have no right of

admission to the Bund Issues Auction Group. The only requirement for a member of the Bund

Issues Auction Group is the capability to take over 0,05% of the annual auction volume, along with

accounts at a branch of Deutsche Bundesbank and Clearstream Banking, and an operating unit in a

Member State of the European Union. The Bund Issues Auction Group members are then expected

to submit successful bids for at least 0,05% (unrounded) of the total issue amounts allotted,

weighted by duration, at auctions in one calendar year. The respective weights to be applied are

published by press release and notified to all the bidders. Accordingly, the German Finance

Agency simply ranks the Bund Issues Auction Group members by their total issue amounts

awarded, weighted with respect to the different maturities and the related interest rate risks. A

ranking list of the Bund Issues Auction Group members by size of their shares in the weighted issue

volume allotted, without quoting percentages, is published annually in a press release. Those

member institutions which fail to reach the required minimum share of the total amount allotted will 20 In Finland, commissions to the Primary dealers on primary market shall be agreed upon separately. The commission structure will be defined by the State Treasury and it will be valid until amended or terminated by the State Treasury. This fee structure is confidential. In The Netherlands, Primary dealers are entitled to compensation for their achievements. In Italy, such commissions are paid out by the Bank of Italy, which operates on behalf of the Italian Treasury in the settlement and delivery process, in return for the bidders’ commitment to resell the awarded securities to their customers at the stop-out price (weighted average price) in the uniform price auctions (multiple price auctions), without any additional raises.

14

be excluded from the Bund Issues Auction Group. Should this be the case, it will be possible for

them to rejoin the Group at a later date. Indeed, the number of members of the Bund Issues Auction

Group has decreased from 72 in 1998 to 38 in 2005 (in some cases because of consolidation among

the bidding institutions). There are no explicit rewards for being a member of the Bund Issues

Auction Group. However, a good ranking is taken into account when assigning a place in a

syndicate, even if until 2006 the German Finance Agency never held a syndicated issuance.21 Also,

additional business (interest rate swaps, etc.) are done only with members of the Bund Issues

Auction Group.

3. The auction pricing performance

In this Section, I examine the performance of most of the auctions held by the euro-zone

Treasuries in 2004, checking for the presence of possible mis-pricing for the securities put in

auction with respect to secondary market prices. The comparison with the secondary market prices,

if liquid and efficient, is certainly the most appropriate way to evaluate the auction performance.22

To this aim, I collected the auction weighted price for multiple price auctions, “Spanish” auctions

and tap auctions, and the auction stop-out price for uniform price auctions, for a total of 178

auctions grouped as reported in Table 2.

The 178 auctions are most of all reopenings of existing lines and cover about 88% of all the

Treasury auctions held in 2004 in the eurozone for the Government securities considered (see the

related column in Table 2).23 Each auction is identified by the bond’s ISIN code and a text

abbreviation of the bond’s type. These two variables make it possible to link these data with the

secondary market from the first release of the MTS Time Series database, which is a new source of

high frequency and daily data for a large number of European sovereign bond markets. In

particular, I collected the BidPrice1 series, i.e. the best bid quotes available at each moment, from

the Tick-by-tick Data BestProposals files for the 178 auction days around the bidding deadline.24 I

choose the bid leg of the bid-ask pairs following the assumption that bidders follow the buy and sell

21 In the Euromoney issue of June 2006 the German Finance Agency states that: “…the syndicate members appointed to distribute the Federal Government’s inaugural linker were selected on the basis of the consistency of their participation in the Bund’s nominal auctions together with their product and market expertise and the quality of their advice…” (see the Special Publications section on the German Finance Agency official website, www.deutsche-finanzagentur.de). 22 In this case, secondary market prices are considered as the “true” value of the securities. The European Treasury bond wholesale secondary market (EuroMTS) is competitive and efficient, with a large number of participants. Accordingly, it is assumed that MTS prices accurately reflect realized common values of issued Treasury debt. 23 The reopenings held in June and some of the initial auctions, for which secondary market data were not available before the auction deadline, were excluded from the calculation. 24 See Dufour and Skinner (2004) for an exhaustive overview of the MTS Time Series database.

15

strategy, as argued by Bikchandani and Huang (1993) and documented by Scalia (1997).25

Accordingly, three are the differences calculated: the first one is between the auction price and the

average of bid quotes during the 30 minutes between the 35th minute and 5th minute before the

bidding deadline (mis_(-35 to -5)), the second one between the auction price and the average of bid

quotes during the five minutes before the bidding deadline (mis_(-5 to 0)), and the last one between

the auction price and the average of bid quotes during the 30 minutes between the bidding deadline

and the 30th minute after it (mis_(0 to 30)).26

TABLE 2 - The auctions sample (year 2004)

Number Amount issued (mln of €) Number Amount issued

(mln of €)

3/5/10/15/30-year BTP, 7-year CCT and 2-year CTZ

7/10/12/15/30-year OAT and 2/5-year BTAN

3/5-year BONOS and 10/15/30-year OBLIGACIONES10/30-year BUND, 5-year BOBL

and 2-year SHATZE

Eurozone 178 411,473 203 507,551

Note: Finland and Luxembourg did not run any auction in 2004. Italy includes one auction for an index linked 10-year BTP, whereas France includes 19 auctions for index linked OAT with different maturities. Overall 25 auctions are missed (out of 203). This is due to missing data on the MTS Time Series database for the month of June or to a later start of negotiations with respect to the auction deadline especially on the days of the first auctions for a new security.

Sovereign issuer

Government securities considered

Auction deadline (CET)

Auction format

Italy 53 146,750 58 160,750 h 11:00 uniform

Auctions considered Auctions held

France 44 89,837 51 116,025 h 11:00 multiple

Spain 24 29,327 25 30,750 h 10:30 hybrid

Germany 16 89,608 22 135,724 h 11:00 multiple

Belgium 10 11,843 10 11,843 h 12:00 multiple5/10/13/30-year OLO

The Netherlands 9 20,359 10 22,109 3/5/10-year DSL h 11:30 tap

Austria 8 9,100 8 9,100 10/15-year BUNDESANLEIHEN h 11:00 multiple

Greece 6 8,800 8 12,100 3/5/10-year BOND h 11:00 multiple

Ireland 4 2,550 6 5,050 15-year BOND h 10:15 multiple

Portugal 4 3,300 5 4,100 4/10-year OT h 11:30 multiple

25 See Table 6 page 24, Scalia (1997). 26 To make the Italian Treasury auction prices homogeneous with those on the secondary market, I subtract the fee returned by the Italian Treasury to bidders per each security awarded from the stop-out prices, which is equal to 20 cents of euro for the 2-year CTZ and 3-year BTP auctions, to 30 cents of euro for the 5-year BTP and 7-year CCT auctions and 40 cents of euro for all the other auctions. This fee is paid out by the Bank of Italy, which operates on behalf of the Italian Treasury in the settlement and delivery process, in return for the bidders’ commitment to resell the awarded securities to their customers at the stop-out price, without any additional rises.

16

Table 3 reports the average values of the three measures of mis-pricing for each country (in euro per

100 € of face value) and some other related descriptive statistics. The results show a strong

evidence of overpricing in all the three specifications (around 97% of all the auctions), with

statistically significance well below to 1% p-value for the t test. Moreover, this is true for all kind

of auction formats. This evidence is contrary to what found by all the previous studies on the

pricing performance of auction formats worldwide. Usually, the empirical literature has compared

the revenue-raising abilities of uniform-price and discriminatory auctions by looking at the level of

mispricing relative to benchmarks such as contemporaneous when-issued yields or secondary

market prices. In several primary markets worldwide, however, there has been growing evidence of

underpricing, smaller in uniform Treasury auctions than in discriminatory Treasury auctions (see

Umlauf, 1993, Nyborg and Sundaresan, 1996, Malvey and Archibald, 1998, and Goldreich, 2007).27

Only two studies employing more recent data, i.e. from 2000 onwards, are closer to my findings:

Rocholl (2005) finds that the average of primary market prices are higher than secondary market

prices, even if statistically not significant, showing that the seller in German Treasury auctions sets

the auction price equal to the market price on average; Elsinger and Zulehner (2007) show that

Austrian Treasury auctions are de facto overpriced. Moreover, the results in Table 3 appear even

more puzzling if compared to the theoretical works on multi unit auctions revenue raising abilities,

e.g. Back and Zender (1993) or Ausubel and Cramton (1998), which are all worried about

underpricing, and to those studies which try to come up with the issue of underpricing, proposing a

series of remedies.28 However, Nyborg and Sundaresan (1996) show that the liquidity on auction

days may vary substantially and that implications about the level of mispricing therefore have to be

considered with caution. To this aim, I measured the overpricing throughout the day, namely from

8:30 to 15:00 (in euros per 100 € of face value). Figure 1 shows that on average the overpricing

keeps stable on the auction day.

I refer to the following Sections for a discussion on the causes of such results, first looking at

the traditional arguments of auction theory and the associated empirical research and then exploring

alternative explanations related to the overall placement procedure. Next, I derive the possible

implications for the Treasuries and some remedies for a fair pricing of Government securities, in the

light of the findings of previous Sections on the primary market institutional arrangements.

27 See Table 1 page 10 in Scalia (1997) or Table X page 1896 in Keloharju, Nyborg and Rydqvist (2005) for further comparisons with other previous studies. 28 See Back and Zender (2001), McAdams (2006), Li Calzi and Pavan (2005), Kremer and Nyborg (2004a and b) and Lengwiler (1999).

17

TABLE 3 - Three measures of mispricing in euros per 100 € of face value

Issuer mis-pricing average weighted by duration min max st. dev. n. obs n. cases

overpricing %

mis_(-35 to -5) 0.070 0.0179 0.006 0.261 0.056 53 100%Italy mis_(-5 to 0) 0.076 0.0193 0.010 0.234 0.057 53 53 100%

mis_(0 to 30) 0.070 0.0184 0.009 0.237 0.052 53 100%mis_(-35 to -5) 0.083 0.0131 -0.071 0.442 0.094 39 89%

France mis_(-5 to 0) 0.087 0.0138 -0.105 0.407 0.095 44 40 91%mis_(0 to 30) 0.085 0.0136 -0.068 0.387 0.091 41 93%

mis_(-35 to -5) 0.085 0.0134 -0.076 0.215 0.057 23 96%Spain mis_(-5 to 0) 0.088 0.0136 -0.074 0.154 0.047 24 23 96%

mis_(0 to 30) 0.110 0.0156 -0.064 0.269 0.071 23 96%mis_(-35 to -5) 0.061 0.0134 0.024 0.146 0.033 16 100%

Germany mis_(-5 to 0) 0.066 0.0134 0.029 0.195 0.042 16 16 100%mis_(0 to 30) 0.050 0.0115 0.008 0.138 0.030 16 100%

mis_(-35 to -5) 0.116 0.0149 0.038 0.241 0.051 10 100%Belgium mis_(-5 to 0) 0.140 0.0177 0.114 0.190 0.021 10 10 100%

mis_(0 to 30) 0.154 0.0184 0.104 0.261 0.047 10 100%mis_(-35 to -5) 0.051 0.0140 0.019 0.102 0.029 9 100%

The Netherlands mis_(-5 to 0) 0.045 0.0130 -0.013 0.103 0.039 9 8 89%mis_(0 to 30) 0.045 0.0133 -0.021 0.095 0.040 8 89%

mis_(-35 to -5) 0.172 0.0228 0.132 0.217 0.026 8 100%Austria mis_(-5 to 0) 0.176 0.0232 0.149 0.211 0.022 8 8 100%

mis_(0 to 30) 0.185 0.0245 0.163 0.230 0.022 8 100%mis_(-35 to -5) 0.222 0.0550 0.163 0.289 0.054 6 100%

Greece mis_(-5 to 0) 0.227 0.0561 0.160 0.312 0.060 6 6 100%mis_(0 to 30) 0.213 0.0522 0.136 0.326 0.064 6 100%

mis_(-35 to -5) 0.084 0.0075 0.047 0.110 0.025 4 100%Ireland mis_(-5 to 0) 0.087 0.0078 0.051 0.127 0.030 4 4 100%

mis_(0 to 30) 0.095 0.0086 0.065 0.126 0.024 4 100%mis_(-35 to -5) 0.155 0.0323 0.140 0.173 0.013 4 100%

Portugal mis_(-5 to 0) 0.168 0.0343 0.157 0.183 0.010 4 4 100%mis_(0 to 30) 0.159 0.0311 0.123 0.192 0.027 4 100%

Issuer mis-pricing average average min max t-statistic (p-value) n. obs n. cases

overpricing %

mis_(-35 to -5) 0.088 0.0169 -0.076 0.442 0.000 172 96.6%Eurozone mis_(-5 to 0) 0.093 0.0177 -0.105 0.407 0.000 178 172 96.6%

mis_(0 to 30) 0.093 0.0175 -0.068 0.387 0.000 173 97.2%

18

FIGURE 1

Overpricing every half an hour from 8:30 to 15:00 on the auction day

0.00

0.05

0.10

0.15

0.20

0.25

€

Austria 0.228 0.209 0.187 0.181 0.172 0.185 0.189 0.180 0.168 0.168 0.187 0.175 0.164

Belgium 0.132 0.130 0.132 0.132 0.112 0.107 0.122 0.154 0.197 0.203 0.206 0.174 0.225

France 0.062 0.051 0.054 0.067 0.084 0.085 0.073 0.071 0.073 0.065 0.069 0.079 0.138

Germany 0.071 0.076 0.080 0.064 0.062 0.050 0.048 0.039 0.042 0.045 0.053 0.059 0.059

Greece 0.231 0.220 0.213 0.223 0.222 0.213 0.206 0.198 0.190 0.192 0.190 0.189 0.198

Ireland 0.077 0.053 0.079 0.081 0.112 0.125 0.120 0.113 0.131 0.111 0.119 0.157 0.234

Italy 0.059 0.063 0.068 0.065 0.071 0.070 0.069 0.065 0.072 0.078 0.077 0.073 0.087

The Netherlands 0.049 0.054 0.048 0.050 0.049 0.050 0.045 0.031 0.025 0.022 0.013 0.015 0.037

Portugal 0.167 0.153 0.163 0.151 0.167 0.156 0.159 0.155 0.147 0.139 0.148 0.159 0.162

Spain 0.070 0.069 0.077 0.086 0.110 0.117 0.105 0.116 0.111 0.115 0.126 0.122 0.179

8h30m 9h00m

9h00m 9h30m

9h30m 10h00m

10h00m 10h30m

10h30m 11h00m

11h00m 11h30m

11h30m 12h00m

12h00m 12h30m

12h30m 13h00m

13h00m 13h30m

13h30m 14h00m

14h00m 14h30m

14h30m 15h00m

4. A puzzle for the conventional auction theory

The assessment of current issuance techniques to a great extent depends on the way we read the

actual evidence of the phenomenon of overpricing in the eurozone primary markets. Indeed, the

presence of overpricing is not consistent with the standard homogeneous multi-unit auction theory.

In general it is shown that when bidders have a pure common value and possess private information

(assuming they receive strictly-affiliated signals of that value), the issuer should expect revenues

less than the expected value of the securities being auctioned, since bidders can extract

informational rents from their private information (see Ausubel and Cramton, 1998). However, it

may still be useful to test for the main implications of auction theory to check if the existence of any

19

underlying factor, besides biasing the auction pricing upward, also disrupts such typical relations.

The latter typically relate to the winner’s curse, the degree of competition and the auction format, as

argued among others by Goldreich (2007):

“Overall, the results show that underpricing depends on the auction mechanism, the extent

of competition in the auction and uncertainty in the auction. These are exactly the

economic forces that theory would suggest should affect auction prices.”29

To this aim, along with the auction theory suggestions I review the possible underlying factors,

though all closely interrelated, among which the measurement method and the presence of market

manipulation, the distorting incentives in the evaluation criteria for primary dealers, and the

discretion enjoyed by some Treasuries in setting the auction stop-out price.

a) The measurement issue and market manipulation. It has been argued that mispricing as

measured by empiricists is not necessarily an accurate reflection of revenue; for example, the

benchmark may reflect the (expected) auction outcome (Nyborg and Sundaresan, 1996). Indeed,

the measurement may be highly sensible at the time it is done. Some Treasuries have noticed

that secondary market price movements show a V-shape on the auction day, where the V lower

corner represents secondary market prices at the auction closing time (see the examples in

Figure 1 and 2 in the Appendix II). However, in my sample this was not the case, since on

average the overpricing is persistently above zero on the auction day, without showing any

definite trend (see Figure 1 at the end of the previous Section).30 The measurement issue is

strictly related to the presence of market manipulation, if any. Indeed, overpricing could be

nothing else than evidence of market manipulation. Should this be the case, primary dealers

would manipulate secondary market prices around the auction deadline to make primary

issuances appear more costly for them in order to defend their current privileges or even to

enhance their bargaining power when meeting with Treasuries to renew their contract as

primary dealers.31 To this respect, focusing away from those models which consider a stand-

alone auction and considering those which situate strategic bidding in a richer environment, two 29 See p. 452 in Goldreich (2007). 30 An appropriate and reliable measurement, i.e. a là Scalia (1998), should also take into account a profit analysis per bidder considering all the transactions made on the secondary market around the auction day, to see if primary dealers are effectively incurring losses on this dimension of their fixed income business. On the other hand, from a market to market perspective this is indeed the case. 31 For example according to the Agence France Tresor official website “The SVTs (i.e. the French primary dealers) agree to comply with the specifications drawn up annually since 1986. These specifications were the subject of intense discussions with the SVTs, which led to the development of a new charter of the relationship with the Agency France Trésor” (see http://www.aft.gouv.fr/article_788.html?id_article=788).

20

are the main contributions which could address the overpricing issue: Drudi and Massa (2005)

and Bikhchandani and Huang (1989).32 Both of them look at how the interaction between the

auction market and the secondary market may impact on bidding strategies, in particular

inducing bidders to bid more aggressively. Drudi and Massa (2005) provide evidence of price

manipulation which show in bidders selling in the morning before the auction deadline to reduce

prices, then buying in the afternoon and bidding aggressively at the auction. Bikhchandani and

Huang (1989) point to the reverse strategy, namely participants behave aggressively in auction

trying to send signals to the secondary market. This relies on the fact that submitted bids reveal

the private information of the better informed bidders, influencing the resale price in the

secondary market and then creating an incentive for the primary dealers to signal their private

information to the secondary market participants. Hence, the phenomenon of overpricing should

be treated with caution, and it is important to check for anomalous movements of secondary

market prices on the auction day before testing the implications of auction theory (see Figure 1).

b) The bundle product. The primary markets here considered strongly rely on a primary

dealership system involving a rich set of privileges and duties, which in turn greatly affects the

profitability of primary dealers’ business. To this respect, Sareen (2006) makes a tentative

assessment of the impact of changes in the privileges and obligations system on the dealear’s

choice about becoming/remaining a primary dealer or not, showing that they significantly

impact on it, at least on those dealers who have a limited access to significant client-networks:

whenever the advantages-obligations mix moves against them, they simply quit being primary

dealers. Dunne, Moore and Portes (2006) point to European Treasuries deploying the auction-

syndicate structure to guarantee liquidity in their Government bonds as a system which leads to

situations where primary dealers actually make losses at the auction stage (i.e. overpricing)

spurred on by supernormal syndicate profits. In their game-theoretical model profitable

syndicated issuances represent the “carrot” by which Treasuries incentivise dealers to enter the

Government security market. Thus, to the extent the rules set by Treasuries bind the main

privileges, such as the share of securities a primary dealer can purchase at the auction price in

the reserved post-auction reopenings,33 the right to syndicate longer term or index linked

securities, or the access to debt management operations (buy backs, exchanges, swaps…) and

repo facilities, to the primary market performances of primary dealers, the phenomenon of

32 There are several studies addressing the interplay between parallel markets trading the same security or relating derivatives. For example, Chatterjea and Jarrow (1998) show how in the auction, the dealer bids aggressively when there are chances of manipulation: if the dealer corners the market, he may charge a premium price in the post-auction secondary market. Still, Nyborg and Strebulaev (2004) study how a potential short squeeze impacts on bidders' strategies and auction performance, and how the design of the auction affects the incidence of short squeezes. 33 See Coluzzi (2007) which values such an option granted by the Italian Treasury to primary dealers, relying on a time homogenous one factor CIR model.

21

overpricing is nothing else than a visible effect of bundling a number of different commodities.

Indeed, recently, both the Spanish and the Italian Treasury have introduced some changes in the

evaluation system of primary dealers. The reasoning behind these changes is the belief that

overbidding is closely linked with the aim of scaling up in the ranking of primary dealers in

order to improve their chance of being chosen for such profitable debt management

operations.34

c) The discretion in the stop-out price setting. The fact that some EU-12 countries effectively

“choose” the price, since they decide the supply after observing the bids or keep flexible the

relative composition of the durations to be sold only fixing the range of the aggregate supply to

be sold (see previous Table 1), could make such a phenomenon less puzzling. In this case the

overpricing would be the result of the Treasury maximizing its revenues choosing cut-off prices

on the received aggregate demand above the current secondary market prices. Both Rocholl

(2005) and Keloharju, Nyborg and Rydqvist (2005) show that while the Treasuries are

committed to maximize their revenues, they do so not in a given auction but under a long term

perspective, trying to set the auction stop-out prices as much as possible close to the secondary

market prices. Discretion is acted to cope with the potential for underpricing.

Then, the traditional measure of overpricing should be carefully interpreted taking into account the

potentially complex interplay between the when-issued market, the auction and the after market, the

implications of the primary dealership system and features of the institutional set-up such as the

discretion enjoyed by some Treasuries in setting the auction stop-out price. As a consequence, only

after having surveyed and assessed the above issues, then one could focus on auction mechanisms

in themselves, comparing the best practices worldwide with the state-of-the-art of the auction

theory. In the following, I pay more attention to the last two points b) and c), since Figure 1 scales

down the role of what in point a). To this aim, I run a series of regressions considering mis_(-35 to -

34 In the Spanish case, those financial institutions that have been announced as top five/six of the group have been selected as book-runners for the major syndicated deals (10-, 15- and 30-year Obligaciones). As from 2006, the book-runners of the syndicated deals have been instead selected from the top ten of the ranking. The position in the ranking is not communicated to primary dealers. In addition to the aforementioned rule, regularity in the activity of primary dealers is assessed in the qualitative part of the evaluation. Hence, at the end of the year, a comparison of the points obtained monthly by each primary dealer is carried out. Those primary dealers which have been more regular and that stand above the average of the primary dealers receive some points in the qualitative part of the evaluation. This also helps to avoid “excessively aggressive bids” in certain moments of the year. As a complementary measure, the Spanish Treasury analyses the structure of the bids submitted at the auctions and could penalise in the qualitative part of the evaluation those primary dealers which have introduced aggressive prices in the auctions. In the Italian case, since 2005 the Treasury introduced an Auction Aggressivity Index which measures the contribution of the auction strategy of each primary dealer in determining the difference between the auction price and the fair value of the bond (values close to zero in the AAI indicate a strategy that is not very or not at all aggressive). For this parameter a score between 0 and 18 is assigned, a higher score corresponds to low AAI, that is for very low or not aggressive strategies (see the Annex to the Public Director Decree no. 128678 of December 27th, 2006, Ministry of Economy and Finance, Rome).

22

5) weighted by duration as the dependent variable (i.e. wgt_mis) and including a composite set of

regressors (see Table 4 for the related descriptive statistics and Table 5 for the empirical results),

namely:

- l_cover. This is the ratio between the bidders’ total demanded quantity and the Treasury allotted

quantity in the previous auction for the same security and proxies for the expected auction

participation and then expected auction competition. The above mentioned Bikhchandani and

Huang (1989) model implies that auction participation measured by the cover ratio should affect

the auction cut-off price directly, as shown by Spindt and Stolz (1992). However, in a

commodity bundling perspective, it may be seen as a proxy for the primary dealers competition

to gain extra points in the Treasury league table via a higher share of securities won in auction.

Hence, the expected sign of this first explanatory variable on the overpricing is positive.

- vol_mts. This measures the volatility on the secondary market before the auction cut-off time,

that is the standard deviation of secondary market prices during the 105 minutes between the

120th minute and the 15th minute before the bidding deadline. An implication of the

generalized winner’s curse is that a bidder would reduce his overall demand when uncertainty

increases. The implication is that if private information and the winner’s curse are important, we

should see uncertainty having a positive effect on bid shading, a negative effect on quantity

demanded and then a negative effect on overpricing. Hence, when the signals received by

bidders are less precise, namely there is a higher dispersion of bidders’ expectations on the

securities value, the winner’s curse should be more severe and then the auction expected

outcome for the Treasury will be lower (see Ausubel and Cramton, 2002, and Nyborg, Rydqvist

and Sundaresan, 2002). The expected sign of this explanatory variable on the overpricing is then

negative.

- depth. It is a liquidity measure on the secondary market calculated as the average amount of

securities available at the best bid and ask quotes during the 105 minutes between the 120th

minute and the 15th minute before the bidding deadline. It is included in all the specifications of

Table 5 to control for the presence of a liquidity effect (see Nyborg and Sundaresan, 1996), if

any. In this case a lower liquidity on the secondary market may increase pressure on the primary

market bringing about higher auction prices if bidders enter the auction with significant short

positions (see Scalia, 1998, and Nyborg and Strebulaev, 2004), or may indicate a lower level of

information on the security being auctioned which would lower auction prices (see Balduzzi,

Elton and Green, 2001, and Fleming and Remolona, 1999), or may reflect manipulation

strategies by some influential bidders (see previous point a) and note 32). Should these be the

23

cases, the expected sign of this explanatory variable on the overpricing is ambiguous, otherwise

not significant at all.

- num_pd. This is the number of primary dealers admitted in each national primary market in

2004 and is included to check for any size effect of the primary dealership system on the auction

outcomes. By construction, it takes the same value for all the observations related to a specific

country.

- num_synd. This is the number of syndicated issuances carried on by each Treasury in 2004 and

is intended to assess the effect of one of the most important elements of the commodity

bundling on the auction outcomes. As for the previous variable, it takes the same value for all

the observations related to a specific country.

- dummies. The remaining labels on Table 5 identify respectively:

• dummies on countries (from austria to spagna), auction formats (uniform and hybrid) or

institutional details (discretion, option and access) which take value 1 if the observation falls

in the identified category and 0 otherwise. Whereas the dummies on countries and auction

formats are immediate to appreciate, those on institutional details work in the following

way: if the dummy discretion takes value 1 the observation relates to an auction of an issuer

who enjoys the right to intervene in setting stop-out prices; if the dummy option takes value

1 the observation relates to an auction of an issuer who does provide for reserved reopenings

to primary dealers (i.e. the latter are granted an implicit call option to purchase a share of the

security being auctioned the following day at the auction price); and if the dummy access

takes value 1 the observation relates to an auction with free entry, i.e. auction participation is

not only restricted to primary dealers;

• previous regressors (l_cover, vol_mts, and depth) multiplied by the auction format dummy

to test for the presence of structural breaks between different auction formats (see

specifications iv.).

TABLE 4 - Descriptive statistics

Dependent variable mean std. dev. min max obs

wgt_mis 0.0170 0.0125 -0.0197 0.0770 169

Independent variables mean std. dev. min max obsl_cover 2.40 0.88 1.16 6.73 169vol_mts 0.0232 0.0188 0.0000 0.1085 169depth (mld of €) 0.0234 0.0151 0.0026 0.0644 169num_pd 21.76 6.50 7 40 169num_synd 2.62 2.40 0 6 169

24

TABLE 5 - Empirical results

Regressor i. ii. iii. iv. v. vi.l_cover 0.0052 ª 0.0028 ª 0.0058 ª 0.0069 ª 0.0052 ª 0.0052 ªvol_mts -0.2081 ª -0.1630 ª -0.1818 ª -0.1768 ª -0.1464 ª -0.1519 ªdepth 0.1306 ª 0.1040 ª 0.1385 ª 0.1328 0.1326 ª 0.1363 ªaustria 0.0115 ªbelgium 0.0044france 0.0028greece 0.0363 ªgermany 0.0029italy 0.0073 ªportugal 0.0210 ªspain 0.0037uniform 0.0027 0.0131l_cover*uniform -0.0053 ªvol_mts*uniform -0.0509depth*uniform 0.0608hybrid -0.0020 0.0214l_cover*hybrid -0.0074vol_mts*hybrid -0.0414depth*hybrid -0.1745discretion -0.0082 ªnum_pd 0.0002 0.0002option 0.0070 0.0095access -0.0063 ª -0.0096 ªnum_synd 0.0015 ªconstant 0.0063 0.0053 0.0035 0.0006 0.0045 -0.0049num obs 169 169 169 169 169 169F 22.27 ª 18.21 ª 13.64 ª 7.29 ª 14.33 ª 13.32 ªR-squared 0.32 0.57 0.33 0.36 0.46 0.43

ª Statistical significance 5% or better.

All the regressions are estimated with ordinary least squares specifying that the Huber/White/sandwich estimator of variance be used in place of thetraditional calculation.

Table 5 reports the estimates for six different specifications. Since my conclusions are based on

the regression results, this is correct if any error caused by using wrong prices to measure the mis-

pricing is residual in the regressions. However, if the mean error of this measurement error is

different from zero, the constant term will incorporate this bias, and the main effect of this

measurement error on the regression results is an increase in the variance of the slope coefficient

estimates. These errors from using wrong prices may then induce heteroscedasticity. On these

grounds, all the regressions are estimated with ordinary least squares specifying that the

Huber/White/sandwich estimator of variance be used in place of the traditional calculation. As

Goldreich (2007), I pool all the auctions together. While all of the specifications include the first

three listed regressors, they differentiate for the inclusion or exclusion of the remaining regressors

and dummies. The specification i. is the starting one, the specification ii. checks for differences due

25

to a country effect, the specifications iii. and iv. investigate if the country effect comes from the

auction formats, whereas specifications v. and vi. if from other institutional details, such as the

discretion enjoyed by the issuer in setting the auction stop-out price, the practice of bundling

together different commodities, the size of the primary dealership system, or the free access to the

auction for financial institutions other than primary dealers.35 Hence, I first comment on the

implications of the overall outcome of each specification to then close on the specific regressors

coefficient estimates.

The estimates obtained from specification i. on the first three regressors turn out to be robust

throughout the subsequent specifications. Specification ii. suggests the presence of a country effect

on the level of overpricing, dividing the sample auctions relating to the nine countries in two

groups: the higher level overpricing group includes the auctions of Austria, Greece, Italy and

Portugal, whereas the lower level overpricing group comprises the auctions of Belgium, France,

Germany, Ireland and Spain.36 Specification iii. excludes that such a difference on the level of

overpricing between countries is due to the actual auction format. Specification iv. further

strengthens the previous result, rejecting the presence of any structural break between different

auction formats, i.e. the behavior of the first three regressors are not conditioned by the auction

formats. The last two specifications finally shed light on the sources of the differences on the level

of overpricing among countries found in specification ii., getting closer to its overall significance

level (see the related F statistics) and goodness of fit (see the related R-squared).

Given the robustness of the estimates on the first three regressors I continue commenting on

them and on the other regressors and dummies. Whereas one of the main implications of auction

theory is not verified, namely that the level of auction prices should be lower than the

contemporaneous secondary market prices (there is overpricing instead), this is not true for the

winner’s curse effect. The proxy for uncertainty on the value of the securities being auctioned, i.e.

vol_mts, has the predicted negative effect on the overpricing. The more the uncertainty, the less

aggressive the bidding behavior, the less the overpricing. Also the results on depth are to be read

in the same way, pointing again to the role of uncertainty in lowering the overpricing. On the other

hand, l_cover has a significantly positive impact on overpricing, highlighting the role of

competition among bidders. This result is even clearer given the significantly positive effect of the

main regressor relating to the commodity bundling, i.e. num_synd (see specification vi.). The more

the number of syndicated issuances, the more primary dealers are incentivised to compete to

35 Specifications v. and vi. differ in including discretion and excluding num_synd and viceversa, since they are highly and significantly correlated (-0.83), then biasing the estimates if used together. 36 Finland, Luxembourg and The Netherlands auctions are not included in Table 4-5 since the first ones did not hold any auctions in the 2004, whereas the last one adopts a tap auction procedure (see note 9).

26

ameliorate their ranking and improve their chance of winning a place in these operations, the more

the overpricing. This clearly indicates that there are forces other than those traditionally suggested

by auction theory, i.e. the commodity bundling, which bias upward the auction prices. On the other

hand other factors mitigate it. This is the case for the discretion enjoyed by the issuer in setting the

auction stop-out price: the dummy discretion shows a significantly negative effect on the

overpricing (see specification v.). This is consistent with what found by Rocholl (2005) and

Keloharju, Nyborg and Rydqvist (2005) about the principles followed by German and Finnish

Treasuries in setting the auction stop-out price, taking into account that their studies were carried

out during periods where the potential for underpricing was the main concern. Interestingly, the

auction format is entirely neutral to the overpricing phenomenon (see specifications iii. and iv.).

This is consistent with a little experiment run by the Spanish Treasury in 2002.37 On the other

hand, the result on access is puzzling, since it takes a significantly negative coefficient. Indeed, this

is not consistent either assuming access as an indicator of the presence of randomness in the supply

faced by strategic bidders such as the primary dealers (see Back and Zender, 1993) or as an

indicator of the possibility for uninformed bidders participation (see Daripa, 2001). Finally, the last

two remaining dummies (option and num_pd) result as not significant at all. The latter result is

reasonable considering that in a context of overpricing the call option is worth less because the

strike price is higher than the price of reference in the secondary market, i.e the option is out of the

money (though an option still has a value, see Coluzzi, 2007).38 I conclude noticing that my results

on the role of winner’s curse (vol_mts) and competition (cover) are remarkably consistent with

those obtained by Goldreich (2007),39 whereas they are at odds with respect to the results on

auction formats. The latter is due to the fact that Goldreich’s sample auctions refer to the same

37 The Spanish Treasury’s auctions are run through a “modified Dutch” procedure by which all bids between the stop-out rate and the average rate are allotted at the submitted rate and rates below the average rate are assigned at the average rate. It was believed that with this system an incentive for overbidding could exist, as those entities bidding aggressively in the auctions would pay the average price, not the bid price. In order to check if this was the case, in 2002, the 10-year Obligaciones auctions were run through the pure multi-price system (all bids allotted at the bid price); however this did not reduce overbidding (see http://www.tesoro.es/en/deuda/mercados/mprimario/pemision_subasta/subasta.asp). 38 The fact that auction prices structurally are higher than the secondary market prices so as to condition the significance of this right is recognized by the same debt agencies. For example, the Belgian Debt Agency states in his annual review for the year 2004: “As a consideration for their active participation in the primary market, the primary dealers are granted the right to take part in non-competitive subscriptions. They can buy securities at the weighted average auction price, based on a fixed percentage of their accepted bids. The right to non-competitive subscriptions for all the primary dealers taken together amounted to 3.55 billions euro, 3.4% of which was exercised (compared to 44.7% in 2003). The exercise of this right depends exclusively on market conditions at the time of the non-competitive round. The rise in interest rates immediately after the March, July and Novemeber auctions meant that the primary dealers renounced their rights to the non-competitive round” (see p. 20 in Federal Government Debt, Annual Report, 2004). 39 Notice that the sample auctions of Goldreich (2007) showed significant underpricing, measured in basis point and as the average auction yield (defined as the average winning yield in multiple price auctions and defined as the market clearing yield in the uniform-price auctions) minus the average when-issued transaction yield in the 10-minute period before each auction. Then the Goldreich’s positive effect of volatility on underpricing is qualitatively the same as my negative effect on overpricing and the Goldreich’s negative effect of competition on underpricing is qualitatively the same as my positive effect on overpricing.

27

country (the U.S.), which means the same institutional set up. Moreover, the same author does not

mention any role for those institutional details here so relevant in biasing the functioning of auction

mechanisms. Indeed, both Arnone and Iden (2003) do not document the existence of those

privileges typical of the EMU primary markets within the U.S. primary market under the period of

Goldreich’s analysis, and Sareen (2004) further confirms the absence of any relevant rewards for

primary dealers as the U.S. Treasury’s debt issuance counterparties.40 In general, my results on

auction formats are in line with the findings of Daripa (2001), who proves the decisive influence of

institutional changes on auction revenues, though he focuses the analysis on the introduction of

minimum quantity bid, the possibility of free entry and the right to cancel part of the supply

(random-supply).

The following Sections focus on the commodity bundling hypothesis to further exploring its

relevance in the economics of Treasury auctions and the related main medium-long term

consequences for the Treasuries.

5. A tentavive evaluation of one of the rents enjoyed by the primary dealers in

the light of the overall cost of overpricing

As seen in the previous Paragraph 2.2 of Section 2, an important leg of the primary dealership

system is made up of a number of privileges enjoyed by the financial institutions belonging to it.

Most of them are very profitable debt management operations, such as buy backs, exchanges,

interest rate swaps, syndication and foreign currency issues, or other as much valuable rights such

as purchasing a share of the security being auctioned the following day at the auction price (namely

a call option on the Government securities being auctioned).41 Indeed, the condition of primary

dealer is a prerequisite when the Treasury chooses counterparties for the above mentioned debt

management operations, while a good ranking according to the Treasury’s evaluation criteria is

often required. Since these criteria largely rest on the primary dealer’s compliance with the

obligation of purchasing a certain quantity of securities in auction, the apparent costs incurred in

Treasury auctions due to the overpricing phenomenon, should be contrasted with the profits coming

from such privileges in order to gain a better insight into the economics of the primary dealership

system. To this aim, I calculate the overall cost of overpricing and the profits relating to one of the 40 See the Appendix II p. 63 in Arnone and Iden (2003) and pp. 41-43 in Sareen (2004). 41 Coluzzi (2007) compares the value of the call options granted by the Treasury with the respective overpricing costs for a subsample of the Italian Treasury auctions held in 2004. Such call options are valued on the basis of a time homogenous one factor CIR model. In particular, it comes out an almost perfect balancing between the overall call options value and the overall overpricing costs for the subsample considered.

28

major privileges: the participation in syndication issuances. I choose the mis_(-35 to -5) measure of

overpricing to obtain the overall cost across all the auctions and the countries considered. In

particular, I multiply mis_(-35 to -5) by the nominal amount issued divided by 100 per each auction

and country.42 Table 6 reports the overall cost of overpricing per each country, except for Finland

and Luxembourg which did not hold any auctions in 2004.

TABLE 6 - The overall cost of overpricing in 2004

Nominal amount issued

(mln of €) (mln of €)* %**

Italy 146,750 100.11 0.068%France 89,837 62.77 0.070%Spain 29,327 22.76 0.078%Germany 89,608 51.96 0.058%Belgium 11,843 12.55 0.106%The Netherlands 20,359 10.14 0.050%Austria 9,100 15.45 0.170%Greece 8,800 19.27 0.219%Ireland 2,550 2.08 0.082%Portugal 3,300 5.11 0.155%Eurozone 411,473 302.19 0.073%

* It is equal to the sum of the overpricing cost of each auction, which is calculated by dividing the auction amount by 100 and multiplying the result by mis_(-35 to -5) .** In percentage of the Nominal amount issued.

Overall cost ofoverpricingIssuer

On the other hand, I attempt an estimation of the syndication profits in the following way: once

retrieved all the details of the syndicated issuances conducted in 2004 across eurozone countries,

except for Ireland and Germany which did not provide for this practice yet (see Table 7), I establish

the average fee (in euros per 100 € of face value) granted to each lead manager of the syndicate on

the basis of a survey of the Internationl Finance Review (see the first note of Table 7).

Accordingly, I calculate the overall profits (i.e. the size of the pie) coming from the syndication

rights per each country, knowing the size of the issuances. Then, I contrast these findings with the

overall cost of overpricing per each country and for the eurozone on the whole, by calculating the

ratio between them (see last column of Table 7). It turns out that the syndication profits are equal to 42 Recall mis_(-35 to -5) is measured in euros per 100 € of face value.

29

69% of the overpricing costs in the eurozone, ranging between 15% (France) and 254% (The

Netherlands) among eurozone countries.

TABLE 7 – The syndicated issuances and the relating profits in 2004

Nominal amount

issued(mln of €) (mln of €)* %**

IT0003535157 30-year BTP January 4,000 9.00IT0003625909 10-year BTP€i February 5,000 11.25IT0003532915 5-year BTP€i March 3,250 7.31IT0003644769 15-year BTP March 8,000 18.00IT0003625909 10-year BTP€i April 3,500 7.88IT0003745541 30-year BTP€i October 4,000 9.00

27,750 62.44 62%GR0124024580 10-year Bond January 5,000 11.25GR0338001531 20-year Bond January 1,750 3.94GR0114017420 5-year Bond February 5,000 11.25GR0338001531 20-year Bond April 400 0.90

12,150 27.34 142%NL0000102325 10-year DSL March 5,102 11.48NL0000102309 5-year DSL October 6,327 14.24

11,429 25.72 254%BE0000303124 10-year OLO January 5,000 7.50BE0000304130 30-year OLO May 5,000 13.75

10,000 21.25 169%ES0000012916 10-year OBLE June 5,000 11.25

5,000 11.25 49%FI0001005704 10-year Bond May 5,000 11.25

5,000 11.25 no auctionFR0010050559 15-year OAT€i January 4,000 9.60

4,000 9.60 15%AT0000386073 10-year BUND January 3,000 6.75

3,000 6.75 44%PTOTE2OE0000 4-year Bond July 2,500 5.63

2,500 5.63 110%

* According to a survey of the Internationl Finance Review issued between July 2004 and June 2005, syndication fees for sovereign bonds issuances range between 0.10 € and 0.35 € per security of 100 € face value. Hence the values are calculated multiplying the average fee (0.225 €) by the number of securities with 100 € face value issued. For France and Belgium only was possible to get the exact fee paid to the lead managers, namely 0.24 € for the French 15-year OAT€i, 0.15 € for the Belgian 10-year OLO and 0.275 € for the Belgian 30-year OLO.** In percentage on the overall overpricing cost reported in Table 6 for each country. The percentage in the last row referring to the euro-zone is calculated excluding Finland from the overall syndicated profits and Germany and Ireland from the overall overpricing cost.*** The two bonds are issued through the DDA which is an hybrid format of syndication (see the relating note in Table 1).

Source: Sovereign issuers' official websites.

Portugal

Belgium

Spain

Finland

France

Italy

Greece

The Netherlands***

Austria

Overall profit from syndicated issuancesIssuer ISIN code Security Month of

reference

80,829 181.22 68%Eurozone

30

6. The overpricing as a result of Treasuries bundling several commodities

Treasury securities have traditionally faced underpricing: the prices at which Treasury notes,

bills and bonds are sold in auction are lower than the when-issued or secondary market prices.43

However, in recent years, EU-12 Government securities began to be subject to an overpricing

phenomenon: auction prices exceed secondary market prices. The empirical analysis and the

evaluations in the previous Sections make it clear that the commodity bundling practice has a major

role in feeding this phenomenon. Some evidences along this interpretation emerge elsewhere

sometimes surfacing expressly through insiders (see for example Coeuré, 2003)44, sometimes

implicitly within some of the most recent publications on Treasury auctions (see for example

Elsinger and Zulehner, 2007, and Rocholl, 2005), though the latter do not even mention the term

overpricing even if they find evidence of it in their data. This is because the theoretical and

empirical literature was always focused on the underpricing phenomenon, so that even when the

opposite is found, it is always interpreted within the underpricing framework. However, a careful

analysis of such works validates the commodity bundling hypothesis. Take for example Elsinger

and Zulehner (2007) which basically repeat the methodology developed by Nyborg, Rydqvist and

Sundaresan (2002) and Hortaçsu (2002) on Austrian Treasury auction data: their measure of

average bid shading (calculated as the difference between the post auction price and the quantity-

weighted average bid, see their Table 1 p. 118) results to be negative during the period 2001-2006;45

since the quantity-weighted average bid used to calculate the bid shading comprises both winning

and non-winning bids, this means that on average auction prices are higher than current secondary

market prices, i.e. overpricing. Furthermore, looking at their descriptive statistics, in particular

Chart 3 p. 113, one notices that since 1999 the Austrian Treasury started to rely more and more on

syndicated issuances to place its bonds, especially for first issuances which are the largest ones.

This is just an example of a practice which has gained ground throughout the eurozone and caused

the end of the undepricing era to enter the overpricing era. Indeed, primary dealership system and

syndication issues are strongly interrelated, at least since the introduction of the euro, as expressly

stated by the same issuers, like the Belgian agency:

43 See Table 1 page 10 in Scalia (1997) or Table X page 1896 in Keloharju, Nyborg and Rydqvist (2005) for an excellent survey on underpricing evidences worldwide. 44 See the CREST seminar held by the Deputy Chief Executive of Agence France Tresor, Professor Benoît Coeuré, on the 22 May 2003 in Paris (http://www.aft.gouv.fr/IMG/pdf/030522_Auction_Presentation_May_2003.pdf). 45 On the other hand, the same mesure in Nyborg, Rydqvist and Sundaresan (2002) takes a positive value (see Table 4 p. 406). This study indeed refers to 400 Swedish auctions held during the period 1990-1994.

31

“...Finally, the innovations introduced in 1999 included the establishment of a new group

of primary dealers and recognized dealers, as well as recourse to use of syndication for the

initial launch of OLO lines to increase line liquidity and to diversify the investor base…”46

or the Finnish State Treasury:

“...Since the introduction of the euro in the beginning of 1999 the new Serial Bonds have

been launched as syndicated issues. Larger banks and financial institutions including

primary dealers may be invited to the syndication…”47

Other indipendent researchers, as Amor (2002), also noticed:

“…This is where issue syndication or underwriting comes in, a method used successfully in

1999, and more frequently in 2000, by the Treasury Departments of Austria, Portugal,

Finland and Belgium for the first-tranche issues of their new ‘on-the-run’ series.

Participating entities are mandated by the Treasury in the corresponding contracts. The

Spanish Treasury is already following this procedure on a regular basis for its

international and foreign currency issues, and for its open EMTN facilities…”48

However, what are the implications of overpricing in the long term? Should the Treasury consider it

a success of its placing system? The auction overpricing stems from a bundle of obligations and

rewards at the basis of the primary dealership system, namely: the tightening of primary dealers

market requirements, the right to take part to reserved post-auction reopenings, and the practice of

granting the right to syndicate Treasury bonds and to participate in debt management operations

(buy backs, exchanges, interest rate swaps…) to the best primary dealers. Hence, overpricing is not

the right name for such a phenomenon. If the auctions were selling the relevant government

securities only, then the fact of auction prices exceeding the secondary market prices would simply

be viewed as a success of the issuing technique. However, several eurozone Treasuries are

bundling a number of commodities: a) relevant government securities; b) syndication rights; c) debt

management operations rights; and d) implicit call options relating to the primary dealers’ reserved

reopenings. Then, there would be overpricing only if one compares the auction prices to market

46 See p. 11 in the Annual Report 1999 of the Ministry of Finance, Treasury Administration Department, Bruxelles. 47 See p. 8 in the Information Memorandum, prepared in relation to the issuance of Serial Bonds denominated in euro, 23 February 2000, Helsinki. 48 See p. 6 of Amor (2002).

32

prices for (a) only. If one also considers the value of (b-d), then it becomes rather implausible that

bundled commodities are being sold at auction for more than their value. In particular, considering

(b) and (c), some problems could arise from overpricing, if asymmetries lounge among the primary

dealers participating in Treasury auctions. Indeed, all the primary dealers experience negative-

profits from the business of intermediation between primary and secondary markets, whereas only

few selected primary dealers earn positive-profits from the business of syndication and debt

management operation rights. It is likely that the largest primary dealers would get such rights,

whilst the others would never get such a chance. Nevertheless, all the primary dealers compete with

each other in the same auctions for the relevant Government securities. Then, this implies that the

smaller primary dealers will end out of the primary market. This is even more severe in uniform-

price auctions where all the primary dealers pay the same price for the bundle consisting of a

Treasury security and an enhanced probability of winning the syndication or other debt

management operations rights. This enhanced probability has positive value for the largest dealers,

but zero value for the smaller dealers. In other words, the smaller dealers are being asked to cross-

subsidize their larger competitors. The predictable response of the smaller dealers is non-

participation in the Treasury auction. And non-participation reduces the competitiveness of the

Treasury auction, inevitably increasing the Government’s borrowing costs. These medium-long

term developments are at odds with the Treasury debt management objectives, calling for different

solutions, which I will discuss in the conclusions.

7. Conclusions

Considering the recent debate in a number of European countries about the effectiveness,

efficiency and operation of Government security issuance practices, I study the overall functioning

of the eurozone issuing systems with respect to the overall objective of financing the public debt at

low costs with acceptable risks. However, one of the higher concern of an issuer is the full

placement of their debt (see Sareen, 2006).49 Then eurozone countries places great emphasis on

practices such as granting the right to syndicate specific issuances and participate in profitable debt

management operations to the best auction participants or the option to buy Government securities

at the auction price the day after the same auction, in order to attract financial institutions to enter

their primary dealership system and make them accept participation obligations. This of course

49 An auction failure is not a so remote possibility. Take the example of the Japanese Government bond market: it happened two times, in September 2002 and July 2003, that two benchmark 10-year JGB auctions went undersubscribed.

33

ensures the full placement of their debt, but entails a sort of commodity bundling between the

securities being auctioned and the privileges attached to them, which may bias the auction pricing

mechanism. A good ranking is often needed to have a chance for a place in the syndicated

issuances and access to other profitable debt management operations. This has gained ground

throughout the eurozone, making primary dealership and syndication issuances strongly

interrelated, at least since the introduction of the euro (see the excerpts reported in the previous

Section). However, I show how this practice has caused the end of the undepricing era to enter the

overpricing era.50 Indeed, the main contribution of this paper rests on bringing evidence for the first

time of significant overpricing in Treasury auctions and secondly on ascribing it to the practice of

bundling different commodities in Treasury auctions.51 Auction prices have been persistently

higher than secondary market prices throughout 2004, being typical of the EU-12 auctions and not

specific to a single country. I document how the adoption of a primary dealer model which bundles

advantages with frequent participation, distorts the pricing function of the auction, further extending

the empirical findings of Sareen (2006) on participation and placement issues to Government

securities auction pricing. Whereas Sareen (2006) shows how the efficacy of the primary dealer

model locates in enticing buyers with low reserve value for frequent participation via reducing their

purchase price offering advantages as compensation, I show how the same model disrupts the main

function for which an auction mechanism is chosen, i.e. its price discovery function. Moreover,

those countries which enjoy discretion in setting the auction stop-out price, have a lower

overpricing. This is consistent with Rocholl (2005) and Keloharju, Nyborg and Rydqvist (2005)

which show how the issuer uses its discretion to set auction prices as close as possible to market

prices. These conclusions come from the empirical results on the relative dummies variables which

clearly indicates that there are forces other than those traditionally suggested by auction theory

which bias upward the auction prices (see num_synd), whereas other dummies mitigate it (see

discretion). On the other hand, the auction format is entirely neutral to the overpricing

phenomenon. Finally, I further validate the hypothesis on the role of the commodity bundling in

feeding the overpricing phenomenon by comparing the overall cost of overpricing with the overall

profits from one of the major primary dealers privileges (the syndication rights), which result to

balance out. To my knowledge there is only one study which mentions overbidding in Treasury

50 See Table 1 page 10 in Scalia (1997) or Table X page 1896 in Keloharju, Nyborg and Rydqvist (2005) for a comprehensive survey on underpricing evidences, and Rocholl (2005) and Elsinger and Zulehner (2007) for first evidences on overpricing (though not significant). Notice that the former refer to auction held in the ’90, whereas the latter in the new millenium. 51 Moreover, this is in sharp contrast with standard multi-unit auction theory which shows that when bidders have a pure common value and possess private information (assuming they receive strictly-affiliated signals of that value), the issuer should expect revenues less than the expected value of the securities being auctioned (i.e. underpricing), since bidders can extract informational rents from their private information (see Ausubel and Cramton, 1998).

34

auctions, i.e. Dunne, Moore and Portes (2006), pointing to the auction-syndication model in

creating the potential for overbidding, though, therein, the crucial role is played by the opaqueness

of the primary market.52

Where will overpricing lead Treasuries? This may have potentially damaging consequences for

the issuer, since it creates a rift between primary and secondary markets, pushing end investors to

hardly ever participate in auctions and making primary dealership a not deemed profitable activity

any more for smaller primary dealers, leading to lower human capital investment and ultimately to

concentration among a few intermediaries. Bidding strategy depends on the primary dealer’s

characteristics, whether it is or not a global market player and whether it is (or is not) able to track

and manage cross-subsidies among activities. The strategic behavior of participants should be

studied in the broader context of the economics of primary dealership. Even if the profits and costs

for the primary dealership system on the whole balance out, the implications for the single primary

dealer may be really different. While all the primary dealers experience negative-profits from the

business of intermediation between primary and secondary markets, only few selected primary

dealers on the basis of the league table drawn up by Treasuries earn positive-profits from the

business of syndication and debt management operation rights. It is likely that the largest primary

dealers would rank higher and get such rights, whilst the others would never get such a chance.

Nevertheless, all the primary dealers compete with each other in the same auctions for the relevant

Government securities. This then implies that the smaller primary dealers will end out of the

primary market, lowering the competitiveness of the Treasury auctions and likely increasing the

Government’s borrowing costs. These medium-long term developments are at odds with the

Treasury debt management objectives and require some remedies.

Since the behavior of overbidding seems to be the result of distorting incentives in the

evaluation criteria for primary dealers, one possible solution implies that the syndication and debt

management operation rights were no longer bundled with the securities at auction. With that,

participation by smaller primary dealers would not drop, and competition in the Treasury auction

would be maintained.53 The other possible solution is to explore alternative auction mechanisms

such as the clock auctions, and in particular those which put up different items simultaneously, in

this case the Government securities and the syndication and debt management operation rights.

Auctioning them simultaneously enables bidders to submit bids based on the substitution

52 Also Bikchandani and Huang (1989) deal with overbidding, however within a model which gives prominence to information asymmetries between the primary and secondary markets. 53 See Grimm, Pacini, Spagnolo and Zanza (2006) which focus on the question whether to bundle contracts or to auction them separately, discussing the optimal division into lots. Even if the paper relates to a procurement context, nevertheless it lays down the economic principles behind the optimal auction design whenever there are different lots/commodities to auction.

35

possibilities or complementarities among the items at various price vectors. Indeed, theory predicts

- and practice supports - the notion that dynamic auctions yield better prices than sealed-bid

auctions.54

With respect to avenues for future research, I believe it could be fruitful to follow the repeated

game approach, as in Daripa (2001) and Fabra (2003), to further investigate the issue of competition

on the Government bond market and understand why there is so much competition and so little

collusion among banks, from the point of view of the bidding strategies on the primary market and

of the manipulating behaviors on the secondary market, eventually checking for an increase of

concentration in the primary dealership business, if any. Should the latter not be the case, it would

be valuable to understand why.

54 See Ausubel and Milgrom (2002), Ausubel (2006) Ausubel, Cramton and Milgrom (2006) and Ausubel and Cramton (2004). For a quick survey of these auction formats see Ausubel (2003).

36

Bibliography Abbink, K., Brandts, J., and Pezanis-Christou P. (2006), "Auctions for government securities: A laboratory

comparison of uniform, discriminatory and Spanish designs", Journal of Economic Behavior & Organization, vol.

61, no. 2, pp. 284-303.

Amor, J. M. (2002), Government Bond Markets in the Euro Zone, Wiley Finance.

Arnone, M., and Iden, G. (2003), “Primary dealers in Government Securities: Policy Issues and Selected

Countries’ Experience”, IMF working paper, WP/03/45.

Ausubel, L. (2003), "Auction Theory for the New Economy", in D. C. Jones (ed.), New Economy Handbook,

Elsevier Academic Press.

Ausubel, L. (2006), "An Efficient Dynamic Auction for Heterogeneous Commodities", American Economic

Review, 96(3), pp. 602-629.

Ausubel, L., and Cramton, P., (1998), "Auctioning Securities", Working Paper, University of Maryland.

Ausubel, L., and Cramton, P., (2002), “Demand Reduction and Inefficiency in Multi-Unit Auctions”, Working

Paper, University of Maryland.

Ausubel, L., and Cramton, P. (2004), “Auctioning Many Divisible Goods”, Journal of the European Economic

Association, 2, 480–493.

Ausubel, L., Cramton, P., and Milgrom, P. (2006), “The Clock-Proxy Auction: A Practical Combinatorial

Auction Design, in Cramton”, Y. Shoham and R. Steinberg (Eds), Combinatorial Auctions, MIT Press.

Ausubel, L., and Milgrom, P. (2002), “Ascending Auctions with Package Bidding”, Frontiers of Theoretical

Economics, 1, pp. 1–45.

Back K. and Zender J. (1993), "Auctions of Divisible Goods: On the Rationale for the Treasury Experiment",

Review of Financial Studies, vol. 6, no. 4, pp. 733-764.

Back, K., and Zender, J., (2001), "Auctions of Divisible Goods with Endogenous Supply”, Economics Letters,

73, pp. 29-34.

Bagella, M., Coppola, A., Pacini, R., and Piga, G., (2007), "Euro Debt Primary Market Report", Unicredit

Markets & Investment Banking (ed.), Milano.

Balduzzi, P., Elton, E., and Green, C. (2001), "Economic News and Bond Prices: Evidence from the U.S.

Treasury Market", Journal of Financial and Quantitative Analysis, vol. 36, no 4, pp. 523-543.

Bartolini, L., and Cottarelli, C., (1997), "Treasury Bill Auctions: Issues and Uses" in Macroeconomics

Dimensions of Public Finance, Essays in honor of Vito Tanzi, edited by Blejer and Ter-Minassian, Routledge, pp.

267-336.

Bikchandani, S., and Huang, C. (1989), “Auction with Resale Markets: An Exploratory Model of Treasury Bill

Markets”, Review of Financial Studies, vol. 2, no. 3, pp. 311-339.

Bikchandani, S., and Huang, C. (1993), “The Economics of Treasury Securities Markets”, Journal of Economic

Perspectives, 7, pp. 117-134.

Brenner, M., Galai, D., and Sade, O. (2007), “Auctioning Sovereign Bonds; A Global Cross-Section

Investigation of the Price Mechanism”, Working paper, Jerusalem School of Business, Hebrew University of

Jerusalem.

37

Coluzzi, C. (2007), "The Pricing of the Option implicitly granted by the Italian Treasury to the Primary dealers in

the Reserved Auction Reopening", University of Rome Tor Vergata, Mimeo.

Coeuré, B. (2003), "Treasury Bonds Auctions: The Issuer’s Viewpoint", Deputy Chief Executive, Agence France

Trésor, Prepared for the CREST seminar, Paris, 22 May 2003.

Chari, V., and Weber, R. (1992), “How the U.S. Treasury Should Auction its Debt,” Federal Reserve Bank of

Minnesota Quarterly Review, 16:4.

Chatterjea, A., and Jarrow, R. (1998), “Market Manipulation, Price Bubbles, and a Model of the US Treasury

Securities Auction Market”, Journal of Financial and Quantitative Analysis, vol. 33, pp. 255-289.

Daripa, A. (2001), “A theory of treasury auctions”, Journal of International Money and Finance, vol. 20, pp.

743–767.

Department of the U.S. Treasury, the Securities and Exchange Commission, and the Board of Governors of

the Federal Reserve System (1992), “Joint Report on the Government Securities Market”, Washington.

Drudi, F., and Massa, M. (2005), "Price Manipulation in Parallel Markets with Different Transparency”, Journal

of Business, vol. 78, no. 5, pp. 1625-1658.

Dufour, A., and Skinner, F., (2004), “MTS Time Series. Market and Data Description for the European Bond

and Repo Database”, ISMA Centre Discussion Papers in Finance 2004-07.

Dunne, P., Moore, M., and Portes, R. (2006), "European Government Bond Markets: transparency, liquidity,

efficiency", Centre for Economic Policy Research, London.

Elsinger, H., and Zulehner, C. (2007), “Bidding Behavior in Austrian Treasury Bond Auctions”, Monetary

Policy & the Economy, Q2/07, OeNB, Vienna.

Fabra, N. (2003), “Tacit Collusion in Repeated Auctions: Uniform versus Discriminatory”, The Journal of

Industrial Economics, vol. LI, no. 3, pp. 271-293.

Fleming, M., and Remolona, E. (1999), “Price Formation and Liquidity in the US Treasury Market; The

Response to Public Information”, The Journal of Finance, vol. 54, no. 5, pp. 1901-1915.

Goldreich, D. (2007), “Underpricing in Discriminatory and Uniform-Price Treasury Auctions”, Journal of

Financial and Quantitative Analysis, vol. 42, no. 2, pp. 443-466.

Gray, S. (1997), “Government Securities: Primary Issuance”, Handbook in Central Banking, no. 11, Centre for

Central Banking Studies.

Grimm, V., Pacini, R., Spagnolo, G., and Zanza, M. (2006), "Division into Lots and Competition in

Procurement", in N. Dimitri, G. Piga e G. Spagnolo (eds.), Handbook of Procurement, Cambridge University

Press, 2006.

de Haan, J., and Wolswijk, G. (2005), “Government Debt Management In The Euro Area Recent Theoretical

Developments And Changes In Practices”, Occasional Paper Series, No. 25/2005, European Central Bank.

Hortaçsu, A. (2002), “Bidding Behavior in Divisible Good Auctions: Theory and Evidence from the Turkish

Treasury Auction Market”, Mimeo.

Kremer, I., and Nyborg, K. (2004b), “Divisible Good Auctions: The Role of Allocation Rules”, RAND Journal

of Economics 35, 147-159.

Kremer, I., and Nyborg, K. (2004a), “Underpricing and Market power in Uniform Price Auctions”, Review of

Financial Studies 17, 849–877.

38

Keloharju, M., Nyborg, K., and Rydqvist, K. (2005), “Strategic Behavior and Underpricing in Uniform Price

Auctions”, Journal of Finance, 60, 4, 1865-1902.

Lengwiler, Y. (1999), “The Multiple Unit Auction with Variable Supply”, Economic Theory, vol. 14, pp. 373–

392.

Li Calzi, M., and Pavan, A. (2005), "Tilting the Supply Schedule to Enhance Competition in Uniform-Price

Auctions", European Economic Review, vol. 49, pp. 227-250.

McAdams, D. (2006), “Uniform-Price Auction with Adjustable Supply”, Working Paper, MIT.

McConnachie, R. (1996), “Primary dealers in Government Securities Markets”, Handbook in Central Banking,

no. 6, Centre for Central Banking Studies.

Miller M. (1991), “Treasury’s Troubled Auctions,” New York Times, September 15, F13.

Nyborg, K., and Sundaresan, S. (1996), “Discriminatory versus Uniform Treasury Auctions: Evidence from

When-Issued Transactions”, Journal of Financial Economics, vol. 42, pp. 63-104.

Nyborg, K., Rydqvist, K., and Sundaresan, S. (2002), “Bidder Behavior in Multiunit Auctions: Evidence from

Swedish Treasury Auctions”, Journal of Political Economy, vol. 110, pp. 394-424.

Nyborg, K., and Strebulaev, I. (2004), “Multiple Unit Auctions and Short Squeezes”, Review of Financial

Studies, vol. 17, pp. 545-580.

Rocholl, J. (2005), “Discriminatory auctions with seller discretion: evidence from German treasury auctions”,

Discussion Paper, Series 1: Economic Studies, No 15/2005, Deutsche Bundesbank.

Sareen, S. (2004), “Cross-Country Comparison of Models for Issuance of Government Securities”, Unpublished

manuscript, Department of Economics, Duke University.

Sareen, S. (2006), “Commodity Bundling in Government Securities Auctions”, Unpublished manuscript, Bank of

Canada.

Sade O., Schnitzlein C. and Zender J. (2006), “Competition and cooperation in divisible good auctions: An

experimental examination”, Review of Financial Studies, Vol. 19, No. 1, 195–235.

Scalia A. (1998), "Bidder Profitability under Uniform Price Auctions and Systematic Reopenings: the Case of

Italian Treasury Bonds", The Journal of Fixed Income, vol. 7, no. 4, pp. 47-61 (also in Bank of Italy, Temi di

Discussione, No. 303, 1997).

Spindt, P., and Stolz, R. (1992), “Are US Treasury Bills Underpiced in the Primary Market”, Journal of Banking

and Finance, vol. 16, pp. 891-908.

Umlauf, S. (1993), “An Empirical Study of the Mexican Treasury Bill Auction”, Journal of Financial Economics,

vol. 33, no. 3, pp. 313–340.

39

Appendix I – Treasury securities gross issuances and outstanding in the EMU

TABLE 1

AUS* no data available 100% € 16,325 3% 77% 20% € 141,329 BEL 67% 33% € 71,537 11% 89% 0% € 246,129 FIN 59% 41% € 12,198 8% 92% 0% € 60,044 FRA 65% 35% € 315,213 11% 89% 0% € 877,000 GER 33% 67% € 218,000 16% 80% 4% € 901,000 GRE 8% 92% € 25,872 1% 81% 19% € 215,428

IRL* no data available 100% € 1,600 2% 84% 15% € 38,200 ITA 54% 46% € 391,663 10% 83% 7% € 1,213,032 NTL 68% 32% € 83,030 8% 91% 1% € 220,390 POR 69% 31% € 25,031 12% 66% 22% € 101,758 SPA 45% 55% € 65,476 4% 83% 14% € 319,162

* Since data on T-Bills issuances were not available for Austria and Ireland, the total amount in the last column refers only to the sum of issuances of T-Bonds/T-Notes and the shares refer only to medium/long-term issuances.** Marketable euro-denominated central Government debt.*** "Other" securities are either securities sold within specific programs (e.g. Global Bond Program, Medium Term Note Program, Commercial Paper Program, etc.) or retail instruments.

Source: Sovereign issuers' official websites.

T-bills (%)

T-notes/T-bonds

(%)

Total amount (mln of €)

Sovereign issuer

Gross issuances (in 2005) ** Central Government debt outstanding (at the 31/12/2005)

T-bills (%)

T-notes/T-bonds

(%)

Total amount (mln of €)

Other ***

(%)

40

Appendix II – Possible manipulations on the secondary market

FIGURE 1

V-effect in Italy (10-year BTP, auction day 29-05-2003)

102.35

102.65

102.95

103.25

103.55

103.85

104.15

104.45

09.1

6.39

14.5

7.06

09.2

6.28

10.2

5.45

15.1

5.56

16.3

6.26

08.5

3.36

10.1

3.33

11.4

5.17

15.1

6.15

09.2

2.07

09.4

5.36

10.1

3.11

10.4

5.20

10.5

5.38

11.1

1.07

12.0

7.36

13.3

4.16

15.4

7.03

09.2

2.09

10.3

7.52

14.3

6.26

16.3

6.15

14.5

0.10

17.1

7.23

12.2

5.05

14.4

5.26

15.4

1.36

17.2

2.57

h9 Auction deadline h17

Stop-out price

Source: Italian Treasury Public Debt Division.

FIGURE 2

V-effect in France (5-year BTAN, auction day 15-05-2003)

Source: Coeuré B. (2003), Deputy Chief Executive, Agence France Trésor.