econstor Make Your Publications Visible.

Transcription

1 econstor Make Your Publications Visible. Service of Wirtscaft Centre zbwleibniz-informationszentrum Economics indseil, Ulric Working Paper Over- and underbidding in central bank open market operations conducted as fixed rate tender ZEI working paper, No Provided in Cooperation wit: ZEI - Center for European Integration Studies, University of onn Suggested Citation: indseil, Ulric (004) : Over- and underbidding in central bank open market operations conducted as fixed rate tender, ZEI working paper, No Tis Version is available at: ttp://dl.andle.net/049/39640 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenscaftlicen Zwecken und zum Privatgebrauc gespeicert und kopiert werden. Sie dürfen die Dokumente nict für öffentlice oder kommerzielle Zwecke vervielfältigen, öffentlic ausstellen, öffentlic zugänglic macen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt aben sollten, gelten abweicend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewärten Nutzungsrecte. Terms of use: Documents in EconStor may be saved and copied for your personal and scolarly purposes. You are not to copy documents for public or commercial purposes, to exibit te documents publicly, to make tem publicly available on te internet, or to distribute or oterwise use te documents in public. If te documents ave been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise furter usage rigts as specified in te indicated licence.

2 Zentrum für Europäisce Integrationsforscung Center for European Integration Studies Reinisce Friedric-Wilelms-Universität onn Ulric indseil Over- and Underbidding in Central ank Open Market Operations Conducted as Fixed Rate Tender

3 Over- and underbidding in central bank open market operations conducted as fixed rate tender Ulric indseil Tis version: February 003 bstract Open market operations play a key role in allocating central bank funds to te banking system and tereby to steer sort-term interest rates in line wit te stance of monetary policy. Many central banks apply so-called fixed rate tender auctions in teir open market operations. Tis note presents, on te basis of a survey of central bank experience, a model of bidding in suc tenders. In teir conduct of fixed rate tenders, many central banks faced specifically an under - and an overbidding problem. Tese penomena are revisited in te ligt of te proposed model and te more general question of te optimal tender procedure and allotment policy of central banks is addressed. JEL classifications: D84, E43, E5 Keywords: open market operations, tender procedures, central bank liquidity management Directorate General Operations, European Central ank. ulric.bindseil@ecb.int. Te views expressed in tis paper are not necessarily tose of te European Central ank. I would like to tank an anonymous referee, ndres Manzanares, Henner sce, Nikolaus artzsc, Elena isagni, Jackie revis, Cris Darkens, Roelf DuPloy, Juergen von Hagen, Pilipp Hartmann, Per Kvanstroem, ndres Manzanares, Dieter Nautz, xel Ockenfels, Francesco Papadia, Takesi Sirakami, Jens Tapking, Tuomas Valimaki; ntoine Veyrassat, Teo Wassner, Flemming Wuertz, Tsatsuya Yonetani, members of te ESC s Market Operations Committee, participants to Seminars in te EC and at te Center for European Integration Studies (ZEI) in onn and two anonymous referees for very useful comments. I also wis to tank Juergen von Hagen and te Center for European Integration Studies at te University of onn for teir ospitality.

4 . Introduction Open market operations play a key role in allocating central bank funds to banks in modern financial systems. To ensure an efficient allocation mecanism and an equal treatment of banks, central bank open market operations ave been normally conducted at least since te late 980s as tenders. However, tese tenders are and were not always specified as genuine auctions in wic banks could bid at different rates, but, even by a majority of central banks, as so called fixed rate tenders. In fixed rate tenders, te interest rate at wic credit is provided is set by te central bank, and banks simply bid for te amount tey wis to obtain at tat rate. However, central banks ten normally do not allot 00% of te bids, but only a certain quota ( allotment ratio ) applicable equally to all bids submitted. Te use of fixed rate tenders instead of genuine auctions was occasionally motivated by central banks as allowing to send a stronger signal wit regard to te stance of monetary policy (i.e. wit regard to te intended sort term interest rate). For instance according to Deutsce undesbank [995, ]: Troug fixed rate tenders, te undesbank can give te market a clear interest rate signal and, in periods of uncertainty, exercise a stabilizing influence on interest rate movements. Economists, wic are less familiar wit te implementation of monetary policy, are often surprised by te fact tat in fixed rate tenders, central banks fix bot te quantity and te price of te good tey offer. Tis note is basically about tis consistency issue and te circumstances under wic it can be solved. On te basis of a survey of central bank experience wit fixed rate tenders, a model of te aggregate bidding beavior of banks is developed wic allows a comparative assessment of fixed rate tenders and genuine auctions under various circumstances. Special focus is given to te European Central ank (EC). Te EC as, since January 999, allocated per year funds wort more tan EUR 3 Trillion troug its regular open market operations, te main type of wic are for collateralised loans wit -week maturity. Te open market operations of te EC are by far te largest tenders in te world ever conducted in terms of yearly total volumes, suggesting tat teir efficiency sould be of igest interest. During te first 8 monts of te euro, tey were specified as fixed rate tenders, afterwards, tey were defined as variable rate tenders (i.e. genuine auctions) wit a minimum bid rate, wic, as will be sown, implies properties similar to fixed rate tenders if markets expect declining central bank interest rates. Te tendering procedures for fixed and variable rate tenders of te Eurosystem are described precisely in EC [00a]. Even toug te implementation of monetary policy by te EC as so far worked rater smootly, te design of its open market operations as also been questioned by financial market participants and academic economists in relation to te penomena of overbidding and underbidding. Overbidding refers to extremely ig bid volumes submitted to fixed rate tenders, implying, ceteris paribus, extremely low allotment ratios. Tis penomenon was observed in te case of te EC in te second alf

5 of 999 and te first alf of 000, wit bids surpassing te allotment amount by up to a factor of 00. Underbidding refers to te lack of bids in a fixed rate tender, suc tat te central bank cannot allot te liquidity actually needed by banks to fulfill smootly teir reserve requirements. It was experienced once in pril 999, four times in 00, and twice in 00. Overbidding and underbidding in te EC s operations ave found noticeable academic interest. Te first autors assessing te overbidding experience of te EC were Nautz and Oecsler, wo came in October 999 to te conclusion tat (p.8-9) te auction rules are flawed since tey encourage banks to increasingly exaggerate teir demand for reserves Considering te vanising quota te EC s repo auctions are about to become a farce in view of tese problems our suggestion for te EC would be to employ price discriminating variable rate tenders indseil and Mercier [999] provide a basic teoretical model on ow central bank liquidity management affects bidding beavior in fixed rate tenders and propose a golden rule ensuring tat overbidding does not occur, according to wic allotment decisions sould be suc tat sort term market rates are kept close to te fixed tender rate. Erard [00] provides experimental evidence wic is supposed to sow te contrary, namely tat te EC s fixed rate tender system unavoidably leads to overbidding and tat even accommodate policy cannot prevent increasing exaggeration in te bids, ence confirming te result of Nautz and Oecsler [999]. yuso and Repullo [000], [00] concentrate on demonstrating tat te EC ad an asymmetric objective function, wic made it provide systematically too little liquidity, ence creating te overbidding problem. yuso and Repullo [00] and Nautz and Oecsler [999] bot sare te view tat allotment ratios in fixed rate tenders will eiter be indeterminate (or drawn from a continuum of equilibria) or will tend to infinity (or to a limit). Nautz and Oecsler [999] make use of adaptive expectations building to reconcile te indeterminacy wit te evidence. It is noteworty tat te evidence exposed in section (and also its interpretation in te ligt of te proposed model) suggests a rater different conclusion wit regard to te nature of overbidding in te case of te Eurosystem, namely tat it was linked to rate ike expectations. Valimaki [00] proposes an elaborated model of bidding in fixed rate tenders wic owever does not contain two elements wic appear important according to our survey of experience, namely te cost of bidding and te coordinating role of te central bank wen taking an allotment decision. Finally, Ewerart [00] provides a teoretical model of underbidding specifically in te case of an environment as te one set up by te EC, providing several interesting insigts, but again, witout tese two model elements. Te rest of tis paper is organized as follows. Section provides te survey of overbidding and underbidding experienced by central banks applying fixed rate tenders. Te survey sows tat bot under- and overbidding normally relate eiter to a biased allotment policy, or to expectations of canges of central bank rates. Wile a biased allotment policy can be

6 easily corrected by te central bank, expectations of rate canges can normally be made irrelevant only troug adequate reforms of te operational framework of central banks, suc tat te rate canges are excluded to become relevant in te period until te tender in question matures. Starting from te evidence presented in Section, Section 3 introduces a model of central bank tender operations and of a money market in te case of a one-day reserve maintenance period. s will be explained, te one-day model is sufficient to work out te main equilibrium conditions in te absence of expectations of rate canges. t te same time, it allows to investigate ow under- and overbidding may be triggered in suc an environment by an inadequate liquidity policy of te central bank. It is argued tat te central bank liquidity policy sould be neutral in te sense of being compatible wit a money market rate equivalent to te minimum bid rate. Section 4 focuses on ow rate cange expectations trigger over- or underbidding. Tis requires te setting up of a model wit two days per reserve maintenance period. Tis variant of te model also allows analyzing te properties of genuine auctions under rate cange expectations. It is sown tat fixed rate tender ave, under tese circumstances, specific disadvantages relative to variable rate tenders (genuine auctions) wic sometimes cannot be compensated troug any allotment policy. Finally, section 5 concludes on wat tender procedure and allotment policy te central bank sould coose under different environments and preferences and comes back to te basic question of te nature of under- and overbidding.. Central banks experience wit over- and underbidding.. Introduction Nearly all central banks ave sometimes used fixed rate tenders and in fact it even seems tat te majority of central banks currently prefer it to pure auctions. However, in tis survey, we will focus on a few of tem, wic ave applied te procedure frequently in te 990, namely te Swiss National ank (CH), te Deutsce undesbank (D), te anque de France (F), te ank of Finland (FI), te ank of Japan (J), te EC/Eurosystem (EC), te Sout frican Reserve ank (S), te Swedis Rijksbank (SE), and te ank of England (UK). special focus will be on te Eurosystem s experience, since, due to te features of its operational framework, te penomena of over- and underbidding were more pronounced ere ten elsewere. Over- and underbidding are reviewed successively. efore tat, owever, two more issues need to be exposed. First, te central banks operational framework beyond te specification of te fixed rate tender is important in understanding te extent of over- and underbidding tat may be observed. Especially te following four dimensions appear to be relevant, for reasons tat D, F and FI are considered as up to 998, i.e. up to te introduction of te euro. 3

7 will appear in te next paragrap. () Te frequency of te fixed rate tender operation varies between at least daily (CH, J, UK) to once a week (D, EC, F, FI, S, SE), wereby for instance F and SE do additional daily fine tuning operations. () Te maturity of te operations varies between day (SE), one week (CH, F), two weeks (D, EC) and one mont (FI, S), wile some central banks (J, UK) seem to frequently operate at various maturity. (3) Regarding te collateralisation requirements, one needs to distinguis te case in wic only allotments at te moment of settlement need to be covered by collateral (J, EC, S, SE, CW, UK) or te entire bid amount needs to be covered at te moment of bidding (DE). France switced from te first to te second approac as a reaction to overbidding (see below). (4) Wile D, EC, F, J, S ave a reserve requirement system wit averaging and a one mont reserve maintenance period, SE, CH and UK do not impose relevant reserve requirements. Second, te basic logic underlying under- and overbidding sould be explained. s mentioned in te introduction, bot result from an imbalance between te fixed tender rate and te corresponding market rate. To understand te determinants of te market rate, assume for a moment tat tere is no uncertainty concerning liquidity factors in te remainder of te reserve maintenance period. Tis being te case, and assuming a perfectly efficient interbank market, reserves are obviously eiter sort in relation to reserve requirements, in wic case overnigt rates sould rise to te emergency rate carged by te central bank (e.g. te marginal lending rate in case of te Eurosystem), or long, in wic case tey drop to a possible floor (zero, or, if available, te rate of an excess liquidity absorption facility offered by te central bank). Now, consider te more interesting case in wic liquidity supply until te end of te maintenance period is uncertain. Te overnigt rate on any day will ten correspond to te weigted average of te two standing facilities rates, te weigts being te respective probabilities tat te market will be sort or long at te end of te maintenance period. Ex ante, overnigt rates are constant witin te reserve maintenance period, i.e. te expected overnigt rates witin te maintenance period can never diverge from one day to te next. Te reason for tis property is tat any anticipated cange in te overnigt rate witin te maintenance period would trigger attempts on te part of banks to rescedule teir fulfillment of reserve requirements witin te maintenance period, wic would ultimately lead to an equalization of rates. In fact, te goods reserves eld today and reserves eld tomorrow are perfect substitutes in tis model, since tey bot contribute equally to te fulfillment of reserve requirements. Hence, denoting by T te last day of te maintenance period and by i ml T id, T,, te rates of te marginal lending and te deposit facilities, respectively, at te end of te reserve maintenance period, we obtain te following core equation explaining te market overnigt rate i t (see also for instance artolini et al. [00]): 4

8 i t Et ( it )... Et ( it ) Et ( iml, T ) P(" sort") Et ( id, T ) P(" long") () Inter-bank rates wit te maturity of te tender operation can be assumed to be implied from expected overnigt rates using te expectations ypotesis of te term structure of interest rates. s was noted for instance by Perez-Quiros and Mendizabal [00], a necessary teoretical condition for te martingale property to old is tat te no-overdraft constraint for reserve oldings of banks wit te central bank is not binding, eiter because reserve requirements are so ig tat liquidity absorbing autonomous factor socks (net of open market operations) never pus banks into overdraft at te end of te day, or because overnigt overdrafting is allowed. Wile te latter condition is not met for any major central bank, te former indeed seems to be verified in cases of money markets like te Japanese or te euro area ones due to te substantial reserve requirements imposed. In te case of te euro area, reserve oldings in te first four years of te euro fell from teir required aggregate level of EUR billion once in Marc 999 to a minimum of EUR 63 billion (te second lowest figure being EUR 76 billion in January 999; reserves ave not fallen below EUR 00 billion since pril 00), wic still appears to be far above working balances. Indeed, Wurtz [003] sows tat te time series properties of te euro area overnigt rate (EONI) speak cleary against te relevance of te no-overdraft constraint. 3. Overbidding Survey of central bank experience Overbidding generally occurs if market interest rates are, for te maturity of te tender operation, above te fixed tender rate, suc tat all banks rus to tis arbitrage opportunity, and ence submit large bids. ll central banks in our sample experienced at least sometimes in teir fixed rate tenders bids being considerable above intended allotment amounts of te central bank, leading ence to low allotment ratios. However, te strengt of te penomenon, te reason beind it, te factors limiting its extent, and te central bank policy towards it differed considerably. Te penomenon was probably least relevant in Japan, were an ample liquidity policy and rate cut expectations ave prevailed since sortly after te introduction of tender procedures in te early 990s. Te penomenon is also of little relevance in S were allotment ratios are normally between 60 and 00% wit a minimum of around 0%. pparently, even under rate ike expectations and altoug only allotted funds need to be collateralised, scarcity of collateral is sufficient in S to deter banks from overbidding. In SE, CH and UK, allotment ratios ave not dropped below a minimum of 0% 3 s Perez-Quiros and Mendizabal [00] sow teoretically, te no-overdraft constraint, if relevant, would lead to a spike of te overnigt rate at te end of te reserve maintenance period. However, in te case of te euro area, Wurtz [003] reveals an opposite tendency of te overnigt rate to drop by a few basis points on te last days of te maintenance period. 5

9 and normally, tey are above 0%. It appears tat in tese countries, due to te absence of reserve requirements, te central banks can, even under rate ike expectations, easily provide incentives against overbidding by allotting excess funds, suc tat excess liquidity puses overnigt rates below te tender rate at least until te potential rate ike. However, in case of te UK, te central bank neverteless felt tat it was necessary to ave two tools in and against overbidding: counterparties are not permitted to bid for more tan te size of te liquidity sortage (an average daily sortage being in te region of bn) and te central bank reserves te rater discretionary rigt to reject individual bids and to accept individual bids in part and at levels determined by te ank to ensure tat access to te liquidity provided by te ank of England is available as smootly as possible to a wide range of market participants. In DE, average allotment ratios declined trougout te 990s and ended at around 5% in 998. Incentives to overbid originated apparently from te fact tat troug a somewat tigt allotment policy, market rates were kept on average in te period February 996 to December 998 (te last period of permanent use of te fixed rate tender) basis points above te tender rate. In contrast, for instance, te average spread in te euro area in te period January 999 to June 000 was only 7 basis points (altoug tis period was dominated by rate ike expectations). However, overbidding was limited in te undesbank s case by te requirement to cover te entire bid at te moment of bidding by collateral. Since collateral is always scarce, tis limits te scope of overbidding. Te case of F was somewat similar, wereby te spread between market and tender rates was even muc iger, namely 39 basis points on average from 994 and 998. Te spread owever varied considerably over time during tis period, te yearly averages amounting to 7, 40, 3, 9 and 7 basis points, respectively. efore F introduced in 994 te same requirement regarding marketable collateral as D, namely to entirely cover bids at te time of bidding, tis lead to a decline of allotment ratios to less tan %. fter te cange, allotment ratios stabilized around 0-30%. It is also interesting to note tat in F, tere were in te 990 always two types of auctions in parallel: auctions for wic marketable paper was accepted as collateral, and tose were bank loans were accepted. llotment ratios for te latter were normally around 50%, reflecting te lower total amounts of available eligible bank-loans, and te requirement to cover bids fully. noter country tat needed to react to acute overbidding was FI, wic switced in 996 from pro rata allotments to systematic 00% allotments. It continued wit tis system until te end of 998. Finally, te EC also experienced intense overbidding wic also forced it to abandon te standard fixed rate tender, as described in detail in te following subsection. One may conclude tat apparently, central banks wit reserve requirement and averaging in a banking system wit plenty collateral are te most vulnerable to overbidding. Ten, sufficient conditions for overbidding to emerge are eiter expectations of possible rate ikes witin te same reserve maintenance period (e.g. te EC), or a tendency of te Central ank to steer liquidity conditions in a tigt way suc tat market rates are driven troug tis cannel above te fixed tender rate. Te only limit to 6

10 overbidding is ten to require bids to be covered by collateral (as D and F did) or to switc to a 00% allotment rule (FI) or to abandon te fixed rate tender (EC). Te model presented in sections 3 and 4 will be designed in a way to fully capture tese relationsips. closer look at te Eurosystem s experience Te EC s overbidding episode as been described repeatedly in te literature. Te EC itself summarized its experience after announcing te switc to te variable rate tender in a press release as follows: Te strong rise in bids in te first alf of 000 was due to te fact tat, during most of tat period, tere were market expectations of interest rate ikes and sort term money market rates were significantly above te main refinancing rate. Tis made it attractive for banks to bid for large amounts of liquidity from te central bank. To provide illustrative evidence, Cart draws related time series. s explained at te beginning of tis section, te inter-bank overnigt rate sould depend on te expected central bank rates and on te expected liquidity conditions, bot at te end of te reserve maintenance period. Terefore, part (a) of cart sows te total net use of te marginal lending facilities on te last tree days of eac of te relevant 8 reserve maintenance periods ( net means tat te use of te deposit facility is deducted from te use of te marginal lending facility; 999- is te first reserve maintenance period of 999, ending on 3 February of tis year, ended on 3 Marc, etc.). It appears tat te EC was biased more to te loose side and tat, wit te symmetric corridor of standing facility rates set around te tender rate, liquidity management of te EC sould ence, according to equation (), not ave explained te relatively ig average positive spread between te market and te tender rates. Part (b) of te cart displays canges of te EC s fixed tender and standing facilities rates. Rates declined once in pril 999, but were ten iked 5 times in te period November 999 to June 000. s Gaspar; Perez-Quiros and Sicilia [00] ave sown, te rate canges of te EC ave, at least in te first two years of te euro, been very well anticipated by te market, and indeed, tis is confirmed by part (c) of Cart, in wic te -week EONI Swap spread systematically increased before rate ikes. Te - week EONI swap rate is generally considered to be te most liquid money market instrument at tis maturity. Finally part (d) of Cart provides te total bid amounts in billion of euro submitted in tenders (te allotment volume was on average EUR 69 billion during te first 8 monts of te euro). Te correlation of increase of bid amounts wit te -week rate spread is impressive, as te arbitrage explanation of overbidding would suggest. id amounts tended to increase especially in periods of a ig positive differential between te market and te tender rate. Te igest bid amount was reaced on 7 June 000 wit EUR 8,49 billion, implying, taking into account te allotted amount of EUR 75 billion, an allotment ratio of 0.88%. 7

11 Cart : Main factors explaining overbidding in Eurosystem case 5 0 (a) Impact of liquidity: Net total recourse to marginal lending facility, last tree days of te maintenance period (in billion of euro) (b) Impact of rate cange expectations: Standing facilities and fixed tender rate in % Jan/99 Feb/99 Mar/99 pr/99 May/99 Jun/99 Jul/99 ug/99 Sep/99 Oct/99 Nov/99 Dec/99 Jan/00 Feb/00 Mar/00 pr/00 May/00 Jun/00 Jul/ (c) Spead week Swap vs. fixed tender rate in percentage points Jan/99 Feb/99 Mar/99 pr/99 May/99 Jun/99 Jul/99 ug/99 Sep/99 Oct/99 Nov/99 Dec/99 Jan/00 Feb/00 Mar/00 pr/00 May/00 Jun/00 Jul/ (d) id amounts in billion of euro Jan/99 Feb/99 Mar/99 pr/99 May/99 Jun/99 Jul/99 ug/99 Sep/99 Oct/99 Nov/99 Dec/99 Jan/00 Feb/00 Mar/00 pr/00 May/00 Jun/00 Jul/ In te quotation given above, te EC did not furter spell out wy te very low allotment ratios were regarded as a problem. However, it seems clear tat tendering wit extreme overbidding as to be regarded as a special type of allocation of funds troug queuing, instead of an allocation troug a pure price mecanism. Queuing is known to be a less efficient allocation mecanism, compared to te price mecanism. Queuing always occurs wen a price is kept at a level wic is below te market value of te good sold, at least for te quantity tat is offered. Te queuing equilibrium is caracterized by a marginal condition 8

12 under wic te marginal cost of queuing exactly fills te gap between te fixed price at wic te good is offered and te market price. Te queuing costs implied by overbidding are special if compared to classical cases of queuing in two respects: first, te relevant queuing cost function seems to be unstable, i.e. over time, bidders can lower teir costs of overbidding troug certain investments. Secondly, te queuing costs take to a large degree te form of risk taking, wic is less tangible tan oter costs. How do we ave to imagine te nature and dynamics of overbidding costs? t te start, i.e. wit moderate overbidding, costs of bidding sould be negligible as long as banks own enoug low-opportunity cost collateral (e.g. non-jumbo Pfandbriefe for wic no repo market exists, or bank loans see e.g. EC [00]) to cover teir bid in case tey would obtain te full allotment. If tey bid for more, tey may envisage to use more expensive, so-called general collateral for wic oter uses exist (e.g. unds). If tey bid even beyond tat, tey could envisage to get te collateral in te market after te allotment decision is made public and before te settlement of te operation, wic owever certainly implies furter cost. Indeed, te efficiency of te collateral market is not comparable to te efficiency of e.g. te interbank money market. 4 However, in addition, banks may start to find teir bidding risky since full allotment would imply tat tey receive so muc cas tat tey could ave problems to place it in te market due to credit limits. Finally, if te overbidding becomes extreme, banks are well aware tat tey would be unable to get enoug collateral in te market in time 5, and tat tey are ence taking a speculative stance of wic it can even be doubted tat it is legally sound. Risk managers in banks typically attac a ig cost to suc strategies, and it is common practice in banks tat a cost is attaced to risk taking, suc tat te incentives of risk takers are adequately influenced. One may summarize tat te marginal cost of bidding sould increase wit te extent of overbidding. Tis will be modeled in te present paper in te simplest way possible, namely by assuming tat bidding is free up to a certain amount, but ten exibits increasing marginal costs. Tis will be sufficient to derive relatively simple and intuitive beavioral equilibria 4 First of all, te only standardized liquid market for obtaining collateral is te repo market for expensive collateral suc as Government bonds. If a trader wants to obtain collateral in tis market at very sort notice for a maturity wic is not too common (e.g. weeks), e can be sure to pay an additional markup. Repoing of non-standard collateral is in principle also possible, but terms need to be negotiated individually and te settlement of te collateral leg may be more complex. It sould also be noted tat repos are usually settled at T+, like te main refinancing operation, suc tat tere is very little time for te trader to conduct te deals providing te collateral. Furtermore, central banks may require te collateral to be made available early on T+, before te time te interbank repo operations are settled. Te borrowing of collateral witout a cas lag is not common. So-called securities lending transactions normally involve equity wic is not accepted as collateral by most central banks (see ox in EC [00c], wic also contains a general description of te euro area repo market). Finally, it sould be noted tat settlement of collateral transactions is always costly, wic explains wy unsecured interbank lending still outweigs te repo market for maturities of less ten one mont, i.e. te iger transactions costs of repos outweig te benefits in terms of reduction of credit risk for sort term transactions. 5 Indeed, in te ot overbidding pase in te spring of 000, banks bid more tan te total amount of eligible collateral for Eurosystem operations. 9

13 between te central bank and te market place. dynamic adaptation of te cost of bidding curve is not modeled explicitly, but is an obvious extension witin te proposed model..3. Underbidding survey of central bank experience Underbidding refers to a submission of bids wic is insufficient to allow te central bank an allotment decision ensuring banks to fulfil teir reserve requirement (or zero current accounts, if no reserve requirement system is adopted) as wised by te central bank until te maturity of te operation. gain, first te experience of a panel of central banks is briefly reviewed. Tree types of underbidding may be distinguised: () Underbidding relating to excess liquidity supply. Tis case appears owever to be limited to te very special case of J in 00/00, in wic te central bank aimed at providing large excess liquidity to te market practically at a zero rate. However, since te value of tis excess liquidity was also practically zero, tere were not necessarily incentives to participate sufficiently to te tenders. () Underbidding due to scarcity of eligible collateral: tis was te case a few times in S and SE during te last years, wereby in te case of SE it related to special settlement dates of interbank-operations involving collateral. (3) Underbidding due to some peculiarity of a tender wic make it unattractive for banks. Tis was apparently relevant for te underbidding in te euro area of 7 December 00 (see below). (4) Underbidding due to rate cut expectations. Tis was te most frequent case occurring at least one time for eac central bank in te panel. relevant parameter for te actual tendency to underbid in suc an environment is te reaction of central banks to underbidding, i.e. specifically weter tey tended to give banks te cance to catc up in teir fulfillment of reserve requirements, eiter troug offering an additional operation, or by providing correspondingly more liquidity in te regular operations before te end of te reserve maintenance period. Te role of tis tendency of central banks to bail out underbidding banking systems will be modeled in section 4. Wile S and J tended to bail out te banks systematically in case tey underbid, D, CH, EC and UK normally did not and ence imposed some costs to banks as a consequence to underbidding. CH also tends to solve te issue by sortening te maturity of operations in suc circumstances to one day, wic, witout reserve requirements and averaging, makes rate cut expectations irrelevant for single open market operations. Te most explicit description of underbidding and of te central bank s reaction oter tan te ones relating to te EC experience is te following one by te Deutsce undesbank, wic fits well wit te EC experience (undesbank [994, 65]), especially in so far as te co-ordination issue between banks is igligted: t te end of January 994, for instance, expectations of interest rate reductions were so strong tat te banks failed to bid for te liquidity tey actually needed at te rate set by te undesbank for te fixed rate tender. Some of te credit institutions apparently relied on oters bidding at te rate, wic tey considered to be too ig. In te event, te liquidity 0

14 available fell considerably sort of wat was needed for minimum reserve compliance. fter tis bidders strike te undesbank was not prepared to provide assistance by means of fine-tuning measures; te banks resorted to lombard borrowing to plug te gap at te end of te mont closer look at te Eurosystem s experience Te EC applied from July 000 on variable rate tenders wit minimum bid rate. Suc as specified by te EC, tey were quasi-equivalent to fixed rate tenders wen interest rate cut expectations prevail since in tat case, te minimum bid rate became a binding constraint. s summarized in te following table, underbidding in MROs as occurred so far seven times in te euro area, eac time but once in an environment of rate cut expectations. Table : Te five cases of underbidding on te Eurosystem (all amounts in billion of euro) Date of (a) Rate cut (b) id (c) llotment (d) Reserve (e) Net recourse to settlement of expectation wen volume (= volume tat would fulfillment te marginal lending te MRO bidding: tree actual ave allowed a deficit tat facility before end of monts EURIOR allotment) smoot fulfillment accumulated reserve rate (minimum of reserve until next maintenance period bid) tender rate requirements MRO 07/04/99-0 bp /0/0-8 bp /04/0-0 bp /0/0-7 bp //0-8 bp //0 - bp 6 5 -* 8//0 +9 bp ** *Since tere was a second occurrence of underbidding witin te same reserve maintenance period, te end of maintenance period recourse to standing facilities is only attributed to te last operation. **EUR 8 billion would ave accumulated until te end of te reserve maintenance period. ctually, a fine tuning operation was conducted to inject in accumulated terms EUR 60 billion, wic is te difference between columns (d) and (e) for te MRO of 8 December. Disregarding for a moment te very special underbidding on 8 December 00, column (a) reveals tat expectations of rate cuts appeared in te case of te Eurosystem to be a necessary condition for underbidding to occur 6. Te sortfall of bids relative to te neutral allotment amount (te difference between column b and c) varied between EUR 8 billion ( pril and 7 November 00) and EUR 4 billion (4 December 00). Tese sortfalls implied te accumulation of a deficit in te fulfillment of reserve requirements of up to EUR 3 billion (column d; assuming tat no recourse to standing facility would ave taken place). Te costs in terms of net recourse to te marginal lending facility before te end of te reserve maintenance period (column e) varied substantially according to weter or not te EC decided to increase te allotment amounts in te subsequent tenders to allow banks to catc up wit te fulfillment of teir required reserves before te end of te reserve maintenance period on te 3 rd of eac mont (i.e. to bail out te market). Wile te EC rescued (at least from an ex post perspective) te market in November 00, te bail out 6 Te spread as defined ere as te advantage to go sufficiently beyond te current reserve maintenance period, suc tat te importance of underbidding on market rates of tis maturity is limited.

15 was especially limited in February and pril 00, wen te bail out coefficient, wic may be defined as -(e)/(d), was only 50% and 74%, respectively. Te underbid tender of 8 December 00 was completely different from all oter underbid MROs since it took place in an environment in wic te minimum bid rate (of.75%) appeared very ceap compared to market rates (for week maturity at around.90%). Te main reason for tis underbidding was, according to te financial market press and wire services, te fact tat commercial bank staff in some euro area countries wanted to avoid participating to tis tender and refinance instead troug interbank operations since te tender matured on 3 December 00, a bank oliday for instance in Germany. 7 Of course, if banks would ave known tat tey could ave obtained MRO funds at.75%, and not only at te expected marginal tender rate of around.90% 8, tey would ave participated, instead of relying erroneously on oter bidders. In so far, te underbidding on 7 December 00 illustrates tat if banks assign a cost to participating in a tender (in tis case, te cost of coming to te office on te maturity date, a bank oliday), ten it can never be excluded tat co-ordination failure can exceptionally lead to underbidding. To better understand te entire nature of underbidding, te following table displays for 00 and 00 9 information on te news content of te announcement of allotment decisions for operations tat were underbid and for te operations tat followed te underbid tenders witin te same reserve maintenance period. Wit regard to te latter, it sould be noted tat in te first tree cases in 00, tere was only one furter operation witin te maintenance period, wile tere were two in te fourt case. Te news content was defined as te impact of te announcement of te allotment decision at :0 on overnigt rate quotations made by money market brokers, wereby te average mid point of quotations between 8:00 and :0 is subtracted from te same figure for te period :00-8:00. Te table suggests tat bot te announcement of underbid tenders and of te allotments in subsequent tenders ad extraordinary news content. Te weakest reactions to underbidding were observed in te tender of 4/0/0, probably since money market players did not yet anticipate te toug stance te EC would take wit regard to te bail out, and in te tender of 8//0, since 7 ccording to Dow Jones Newswires/Elena Logoutenkova of 7 December 00: Euro-zone overnigt money-market rates are sarply up midday Tuesday after te allocation announcement for te main refinancing tender revealed underbidding. round 00 GMT, overnigt money was around 3.05%-3.0%, up from.88%-.89% Monday. Te EC managed to allocate only EUR03.50 billion at te main refinancing operation tis week, draining EUR8.9 billion from te market. Money-market traders were surprised by te outcome because normally banks prefer to secure ample liquidity aead and over te year-end. Underbidding probably as to do wit te unfortunate timing of refi flows as tis week's tender will mature on Dec. 3 wen many small banks will be closed, a trader from soutern Germany said. Te small banks wanted to save costs of running operations on Dec. 3, oping to get necessary liquidity from te market, but now tey ave to pay a dear price, anoter trader from Germany said. ut participants are oping for a quick additional refinancing operation from te EC to make up for some EUR5 billion in missing liquidity. 8 Te weigted average rate of successful bids in tis tender corresponded to.88%, i.e. 3 basis points above te minimum bid rate of.75%. In all oter underbid tenders, te weigted average rate corresponded to te minimum bid rate.

16 te EC ere issued togeter wit te announcement of te allotment a message saying tat it would supply additional liquidity, wic it ad never done before. lso, te signs of te reactions to te publication of te allotment decision in te underbid and following tenders were mixed. mong te underbid tenders, rates always increased after underbidding became known, wit te exception of te tender of 3 December 00, were tey fell because strong underbidding was expected but occurred only to a limited extent. 0 Signs are even more mixed for te tenders following underbid tenders: apparently, counterparties were in te period under consideration unable to infer a consistent bail out policy of te EC, suc tat tey were surprised eac time by te policy reaction. Te average absolute value of te news content of underbid tenders (8.5 basis points) is similar to te news content of subsequent allotment decisions (8.7 basis points). In contrast, te average absolute news content of te oter Eurosystem main refinancing operations in te period June 000 to December 00 as been negligible (.7 basis points), suggesting tat, as long as underbidding did not occur, te allotment decisions of te EC were well anticipated. Table : News content of allotment decisions as measured troug te difference between average overnigt quotation during :00-8:00 and during 8:00-:0, in basis points Date of settlement of te MRO Underbid tender Subsequent tenders in te same reserve maintenance period 4/0/ /04/ /0/ // / 8. 04// /.7 8//0.7 - verage absolute value for all tenders of respective category verages absolute value for all oter tenders June 000- December 00.7 Unsurprisingly, statistical tests fully confirm tat underbid and subsequent tenders are very different from normal tenders wit regard to teir average absolute news content. Te underbidding episodes also ave been discussed by te financial market press. For instance 9 For 999, te data was not available. 0 Indeed, most market participants interpreted te allotment as revealing tat tere was no underbidding. ccording to Dow Jones Newswires/Elena Logoutenkova of 3 December 00: Traders in te euro-zone money market were caugt wrong-footed Tuesday wen it became clear, following te European Central ank s weekly refinancing auction, tat tere was no underbidding Te sock was all te greater because tere as been strong rate-cut speculation aead of an EC meeting Tursday. Traders ad figured te auction would terefore be underbid by banks opting to wait until after te meeting in order to get teir money more ceaply. I am speecless said a Frankfurt-based trader In a knee-jerk reaction to te allocation announcement overnigt rates slumped to 3.5%-3.6% and kept falling trougout te day, ending up below 3% by Tuesday afternoon. No news content was assigned to te tender following te one of 4 December 00. Tis is consistent wit te fact tat banks did not really consider tat underbidding ad appened on 4 December, suc tat tey did not feel tat anyting special was to be expected from te subsequent tender. Indeed, te allotment in te following tender was neutral. 3

17 te oersenzeitung of pril 00, p., commenting on te second case of underbidding as follows, insists also on te surprise element of underbidding (own translation): Te publication of te tender result [of 0 pril 00] came as a bad surprise for money market participants. Despite te fact tat traders ad expected underbidding beforeand, te extent of it triggered growing astonisment, and later on panic demand for funds. Traders of banks wo ad submitted bids to te tender suggested tat te EC sould remain toug, as in February, and sould not rescue te market troug a quick tender. Tis would be te only way to teac speculators an orderly bidding beavior. Traders tat ad remained absent from te tender expressed teir dissatisfaction regarding te current regime in te money market. Since te minimum bid rate would avoid te possibility to submit bids at low rates, similar incidents would appen again and again in te coming weeks. For te model to be developed in sections 3 and 4, two points need to be retained from tis article (wic is very similar to articles written after te oter underbidding cases) and te above-described evidence regarding te news content of underbid tenders: First, an important feature of te model to be developed will be te problem of bidders to make effective use of te information on liquidity needs and to coordinate teir bidding beavior, suc tat in te case of underbidding and ence of impossibility for te central bank to take te role of co-ordinator of te total liquidity injection, te publication of te allotment as substantial news content. Secondly, te model sould allow to analyze and ideally to provide an answer to te question weter underbidding can be stopped troug a tigt allotment policy in subsequent tenders (i.e. a low bail out co-efficient). pparently, tere are two scools among money market participants wit tat regard: one tat feels tat te central bank can cope wit te penomenon of underbidding by coosing te appropriate liquidity management strategy (i.e. mainly not to bail-out te market after underbidding occurred), wile te oters feel tat a minimum bid rate, or, equivalently a fixed rate tender, does not allow any reasonable equilibrium in an environment of rate cut expectations. Starting from te fact tat central banks today define teir operational target in terms of te sort term interest rate and from te observation tat no oter money market event, not even September 00 or te euro cas cangeover, created in total so muc overnigt rate volatility, it appears tat underbidding was te most important issue in te monetary policy implementation of te EC since te launc of te euro in 999. It is terefore not surprising tat te EC on 3 January 003 eventually addressed te issue (togeter wit overbidding) troug a fundamental reform of its operational framework. ccording to te press release publised on tat day, te following two measures were decided by te EC for implementation in spring 004 to improve te operational framework for monetary policy : First, te timing of te reserve maintenance period will be canged so tat it will always start on te settlement day of te main refinancing operation following te Governing Council meeting at wic te montly assessment of te monetary policy stance is pre-sceduled. Furtermore, te implementation of canges to te standing facility rates will be aligned wit 4

18 te start of te new reserve maintenance period. Second, te maturity of te MROs will be sortened from two weeks to one week. s te EC explains in its press release, te combination of te two measures will elp remove expectations of interest rate canges during any particular maintenance period, given tat canges in te EC's key interest rates will only apply, in general, to te fortcoming reserve maintenance period and tat liquidity conditions will no longer spill over from one reserve maintenance period to te next. Hence, te measures will contribute to stabilizing te conditions in wic bidding in te MROs takes place. 3. stylized model of aggregate bidding beavior: a one-day maintenance period witout rate cange expectations 3. Te model Te model developed in tis and te following section focuses on te bidding equilibrium under rational expectations, i.e. under te assumption of a full understanding by market participants of te central bank s allotment strategy. Te model is furtermore built in a way to capture all aspects appearing relevant from te sort survey of central bank experience and a more detailed look of te case of te Eurosystem. Te modeled reserve maintenance period consists in tis section of only one day. Six events are distinguised: () Te reserve maintenance period begins wit te opening of te settlement accounts of banks wit te central bank. t te moment of te opening of te accounts, te funds eld on te current accounts are still determined by te previous maintenance period s open market operation. However, all outstanding open market operations mature on te same day. () Te new open market operation takes place. Te banks submit teir bids, and te central bank takes its allotment decision on te basis of its forecast of liquidity needs and possibly its liquidity target. Te allotment amount may be restricted by te available bids. Te allotment decision is made public. t te same time, te central bank s private information on liquidity needs is revealed. Te operation is settled. (3) Te interbank market session takes place and a market clearing overnigt rate i is determined. (4) Te realization of te autonomous liquidity factor sock takes place (autonomous liquidity factors are all tose factors affecting te banking system s reserve oldings wic are not monetary policy operations, suc as te circulation of banknotes see e.g. indseil and Seitz [00]). (5) Finally, te banks take recourse to standing facilities to cover any liquidity imbalance. Tis recourse is purely mecanic, i.e. it fills te gap between te counterparties reserves and reserve requirements (set to be zero). Te model assumes a perfect interbank market and omogenous banks, suc tat eiter all banks will ave recourse to te marginal lending facility, or all will ave recourse to te deposit facility, but tere is no simultaneous recourse to bot facilities by 5