Public Disclosure Authorized Public Disclosure Authorized

GOVERNMENT SECURITIES, MONEY MARKET AND CASH MANAG EWI ENT IN : Key Issues and Recommendations for Development

YlBlN MU AND MICHAEL HORGAN

Public Disclosure Authorized JUNE 2006 Public Disclosure Authorized Financial Sector I \NO41 1) HAPJK TABLE OF CONTENTS

TABLE OF CONTENTS ...... 2 ACKNOWLEDGEMENTS ...... 3 ACRONYMS AND ABBREVIATIONS ...... 5 EXECUTIVE SUMMARY ...... 6 GOVERNMENT SECURITIES MARKET ...... 6 MONEY MARKET ...... 7 CASH MANAGEMENT ...... 8 GOVERNMENT SECURITIES MARKET ...... 10 CURRENT SITUATION ...... 10 POLICY CONSIDERATIONS ...... 19 RECOMMENDATIONS ...... 25 MONEY MARKET ...... 28 CURRENT SITUATION ...... 28 POLICY CONSIDERATIONS ...... 30 CASH MANAGEMENT ...... 33 CURRENT SITUATION ...... 33 POLICY CONSIDERATIONS ...... 35 RECOMENDATIONS ...... 37 ACKNOWLEDGEMENTS

This report was prepared by Yibin Mu (FSE-OPD, the World Bank) and Michael Horgan (Consultant).

Thordur Jonasson (Senior Debt Specialist, FSE-OPD), Dimitri Vittas (FSE-OPD, consultant), Lars Jessen (Senior Financial Officer, BCFBD), Martin Slough (Senior Financial Specialist, ECSPF) all provide valuable comments.

The authors appreciate the guidance and support of Carolyn Jungr (Country Manager, ECCW), Rodney Lester (Business Group Leader, FSE-OPD), Nancy Cooke (Lead Country Officer, ECCUI), Marilou Uy (Sector Director, FSE-OPD), Gerard Caprio (Sector Director, FSE-OPD), Fernando Montes-Negret (ECSPF, Sector Director), and Mike Edwards (ECSPF, Lead Financial Advisor).

The authors would like to thank the Ministry of Finance, , the Securities Commission of Serbia, the CRSD of Serbia, the Exchange, financial institutions and other market participants for their support in the preparation of this paper. ABSTRACT

This paper presents the main issues and key recommendations relevant to the government securities market, money market and government cash management in Serbia.

In terms of the government securities market, the paper analyzes the current T-bill program, euro- denominated frozen savings (FFS) bonds, and dinar T-bonds. The paper also analyzes the pros and cons of future hnding options: (i) international foreign currency borrowing, (ii) Fresh FFS euro bonds, (iii) fixed rate dinar bonds, (iv) floating rate dinar Bonds, and (v) euro indexed dinar bonds.

In terms of the money market, the paper recommends the introduction of interbank repos and further improving the transparency on the interbank market. In both cases, the paper recommends that banks that are not well-capitalized and of poor credit standing should not be allowed to stand in the path of desirable reforms. The paper favours the introduction of reverse repos by NBS.

The paper also points out that NBS has started issuing repos (out to 60 days) in its own bills, thereby segmenting the bill market and undermining T bill issuance. The paper recommends that NBS make best efforts to reduce the scale of its interventions using its own paper and to rely instead on T bonds for repos purposes. MOF, for its part, should commensurately increase T bill issuance and also initiate SRD T bond issuance not beyond 3 year maturity- as the budgeting situation permits- a process that would assist the NBS in its efforts to manage liquidity and help to develop the capital markets.

In terms of government cash management, the paper points out that the current Treasury Department has a well-developed system of budgetary control. However it does not incorporate an efficient cash management system. This entails forecasting future cash flows and, ideally, the ability of the Treasury to transact in the market to manage over (or under) funding. The MOF should improve its cash forecasting capacity. ACRONYMS AND ABBREVIATIONS

CRSD Central Registry, Depository and SRD Dinar (the local currency of Serbia) DvP Delivery versus Payment FFS Euro-denominated "frozen" savings bonds MOF Ministry of Finance NBS National Bank of Serbia PPA The Public Payments Authority EXECUTIVE SUMMARY

GOVERNMENT SECURITIES MARKET

This paper outlines the reasons why the government has a strong interest in fostering the development of the government securities market in Serbia.

A developed government securities market is not just a distributional channel for government debt: its structure will actually affect the price at which the government can expect to sell its securities. A key element is that the government should issue sufficient dinar debt in a way that contributes to the development of government securities markets.

The Ministry of Finance (MOF) has performed well in terms of budget execution, tight expenditure control, an expanded tax base combined with good tax enforcement and a marked increase in GDP. As a result, the stock of government debt in 2004 has been reduced markedly to some 60% of GDP, and the annual budget deficit IGDP ratio is approximately zero. The government is faced annually, however, with refinancing maturing debt of over €200million. Funding is obtained from the new T bill programme and privatisation proceeds. Given that the government's foreign currency denominated debt represents 96% of the public debt, MOF should consider sooner rather than later alternative sources of domestic dinar funding. In selecting from the various options - whlch are addressed in this paper - the government should bear in mind the strategic need to promote the development of the government dinar securities market, since in an underdeveloped (in which buy and sell between themselves), with little or no liquidity, the government will have to pay a premium in regard to its funding costs. The choice of instrument will affect market development.

The Treasury bill (T-bill) issuance program is well established under an auction system that is efficiently operated by the Central Registry, Depository and Clearing (CRSD) electronic system. However, there is virtually no secondary market in T-bills, which are held to maturity; and there are also certain technical deficiencies relating to the in T-bills. The paper offers recommendations to address these issues.

The only government bonds traded in the secondary market are euro-denominated "frozen" savings bonds (FFS bonds - €3.7 billion), which have a zero coupon, with one bond maturing each year between 2005 and 2016. No further FFS bonds are being issued. Because of non- issuance of new FFS bonds and latent demand both at home and from abroad, FFS bonds are traded on the secondary market at artificially high prices and correspondingly low yields of 4.7% to 5.4%per annum. These yields are some 200 basis points over corresponding German Treasury bond yields -which is arguably not an excessive compensation to pay for relative market illiquidity, a poor credit rating and the risks involved. Trading takes place in part on the but mainly in the OTC market, which is totally lacking price transparency. The CRSD is expected to put in place, without further delay, arrangements for making real-time prices of FFS bond transactions available to the market. This would be essential if new Treasury bonds (T-bonds) were to be issued.

The government dinar T-bonds (SRD 19 billion): nearly all are held exclusively by the National Bank of Serbia (NBS) and are not traded. However, between August and October 2005, NBS sold in 5 auctions, a modest amount (SRD 0.75 billion) of the T bond 2006. Overall, the auctions were a success in terms of and cover since they were bid by foreign investors, who are key to the development of government securities market. The bonds held by NBS are used as collateral by the NBS for the purpose of weekly auctions of repos (14 and 28 day) in order to drain interbank liquidity. Regarding future funding options, the paper addresses the following:

1. International foreign currency borrowing; 2. Fresh FFS euro bonds; 3. Fixed rate dinar bonds; 4. Floating rate dinar bonds; 5. Euro indexed dinar bonds.

Costs and other disadvantages associated with international foreign currency borrowing at market rates rule out this option. The case against further euro denominated domestic borrowing is even stronger because such borrowing would, in addition, stifle the medium-term development of the government securities market in dinar. Because confidence in the dinar (SRD) remains low, with the euro being strongly favoured by to hold their investments, household savings are predominantly in euro. This means in turn that demand for dinar assets is not strong, at least beyond a one-year maturity. Nevertheless the dinar T-bill issue, with maturities up to 6 months, has proved to be very successful and, as such, our main recommendation is that it may be timely to test a new dinar T-bond of 2-years maturity in, say, 2006-2007.

Our main views on this issue is that in periods of uncertainty and high inflation as prevails in Serbia, a 2- year floating interest rate T-bond would ofler a relatively attractive option for investors, who would be guaranteed the prevailing interest rate payable every three months based on the average yield in the 3- month T-bill auction. If this course were adopted and were pursued successfully, MOF couldpursue - especially if interest rates were trending downwards -the further option of a 2-3 yearfixed interest rate T-bond, although realistically this will depend on the demonstrable success of the Government's stabilization policies. If neither of the foregoing proved feasible, a situation that seems most unlikely to arise, the option of a T-bond denominated in dinar but index-linked to the euro could be then considered as a final option.

The paper also recommends that MOF should modernize the tax regime applicable to private pension contributions and funds, which will be a valuable source of future medium-to- term government funding.

The need for a comprehensive medium term management strategy is underscored in the final paragraph (1.37) of the Government Securities section of this Report; but as a matter for later consideration: it need not hold up implementation of the recommendations in the paper.

MONEY MARKET

As explained in the paper, an active money market is important to the development of the government securities market. For this reason, the report gives careful consideration to how price transparency on the interbank market could be encouraged. While progress in price transparency has been notable, the authors make certain recommendations to drive the process further. It also recommends making progress on the introduction of interbank repos. In both cases, the strong recommendation is that banks that are not well-capitalized and of poor credit standing should not be allowed to stand in the path of progress.

The paper favours the introduction of reverse repos by NBS. The paper also supports MOF issuing guidelines to local authorities on procedures to be adopted before placing deposits with banks. The paper recommends that government securities be included as eligible assets making up the NBS excess reserve requirement. Any such move could give rise to a further increase in the percentage excess reserve requirement, which is already quite high. NBS has started issuing repos (out to 60 days) in its own bills, thereby segmenting the bill market and undermining T bill issuance. It should be the priority that NBS make its best efforts to reduce the scale of its interventions using its own paper and to rely instead on T bonds for repospurposes. MOF, for its part, should commensurately increase T bill issuance and also initiate SRD T bond issuance not beyond 3 year maturity as the budgeting situation permits: aprocess that would assist the NBS in its efforts to manage liquidity and help to develop the capital markets. The proceeds from add-on government securities for monetary policy purpose should be deposited at the special account with the NBS and not be usedfor budget purposes. Accordingly, coordination between the MOF and NBSshould be strengthened signlJicantly.

CASH MANAGEMENT

The Treasury Department has a well-developed system of budgetary control. It is not, however, a cash management system, which in general terms focuses on malung certain that all revenue is collected as early as possible after, and payments made as late as possible before, the date when they are due. This entails forecasting future cash flows and, ideally, the ability of the Treasury to transact in the market to manage over (or under) funding.

An efficient payments system is also essential for effective cash management. The Treasury is working on a new payments system - a permanent Treasury system - with the active involvement of an external consultancy firm. The new system is expected to include a cash management module, one of the benefits of which will be to reduce excesslidle cash balances in the consolidated Treasury accounts of the various public authorities, for example, the indirect budget users. The Public Payments Authority (PPA) invests the surplus in the main (but not wholly) with the NBS. Total net balances amounted to SRD 49.6 billion at the end of 2004. Ths was greater than the government's borrowing requirement, while the Republic's surplus of SRD 4.5 billion exceeded 2004 net T-bill funding. NBS pays 2.55% per annum on deposits in comparison with borrowing costs latterly of the order of 15% per annum, which is a compelling reason for MOF to pursue more efficient cash management.

There is not a single consolidated account for all public bodies. The Treasury needs to take control of and to manage the overarchng off-set account that aggregates end-of-day surpluses of the various consolidated Treasury accounts of the Republic and of other public entities. Management of over-funding could be tackled to a limited extent by restricted T-bill issuance -but not to the extent that would undermine the T-bill programme. Investment in commercial bank deposits (which can be undertaken on behalf of local authorities but not the government) is not a feasible course as long as there is a structured liquidity surplus in the interbank market, since such extra deposits would aggravate the existing liquidity surplus and would have to be frozen by NBS. Investment in -term paper of a AAA sovereign, such as, Germany, would be feasible although the additional yield over what NBS pays may not be worth the effort. A more pragmatic approach would be for NBS to increase interest rates closer to market rates: specifically, NBS should pay the average overnight lending rate in the inter-bank market ("Belibor").

More accurate monthly forecasts are necessary, but success may depend on the successful operation of the new permanent Treasury payments system together with the supply of regular up-to-date information flows from the different agencies in order to up-date the historically-based trend forecasts produced by the new Treasury system. Even with the new system, it remains open as to whether or not it will in fact be possible to produce the accurate one to five day cash flow forecasts that are the cornerstone of efficient cash management. Nevertheless, it is important that the Treasury follow up routinely all significant forecasting errors in order to rectify systemic problems that may be inherent in the forecasting (monthlylquarterly). Finally, the Treasury should make transparent the forecasting process, by publishing quarterly forecasts and outturns of revenue and expenditure, together with funding details, so that the market is not taken unawares by MOF'S funding intentions.

Apart from improvements in forecasting of cash flow and in the payments system, which are the cornerstone of efficient liquidity management, active management of over funding by government could be undertaken. The paper addresses the possibilities of ad hoc T bill issuance, placement of excess funding with commercial banks or in AAA rated sovereign paper and rules them out. Instead, it recommends that MOF considers opening standard credit lines with a few select banks, thereby reducing the cash surplus needed at NBS; and also that NBS should pay, on government funds, the average of the overnight lending rates on the Belgrade interbank market (see paragraph 2.2 below).

Once the new permanent Treasury payments system is operational, it would be desirable for MOF to consider establishing a cash management taskforce that would document the various elements of cash management activities and determine the strategic measures required to optirnise the cash management arrangements by, for example, (i) smoothing the pattern of revenue and expenditure; and (ii) separating cash management (settlement and physical payment of government transactions) from budget appropriation (authorising of and accounting for government finances). The taskforce would have representation from MOF, NBS, PPA and agencies reporting to it, together with the Revenue agencies.

In sum, the early priority issue of cash management should focus on improving the short-term cash flow forecasting, the better management of cash surpluses, and the separation of operational controls. The latter of which is responsible for the authorization of accounting of Government finances and the settlement and physical payment of Government cash management transactions. I. GOVERNMENT SECURITIES MARKET

CURRENT SITUATION

1.1 Importance of Developing the Government Domestic Securities Market. Developing an efficient and liquid government securities market in Serbia is essential for several reasons: a) an efficient government securities market enables the government to hnd budget deficits in a non-inflationary manner and at a lower long-term cost than would otherwise be possible; b) the existence of a bond market would mean that the government need not rely on domestic banks for its financing; c) a diversified and efficient domestic bond market can help to buffer the effects of domestic and international shocks on the real economy; d) a well developed money market and bond market would facilitate the independent operation of debt management policy from that relating to monetary policy; e) government benchmark yield curve(s) become valuable reference points for pricing other public and private financial contracts (e.g. corporate bonds), thereby reducing price uncertainty; and f) a market would facilitate the development of a range of fixed income products (such as interest rate swaps, repos, future and options) which can be used by market players to contain financial risk. Each of these factors is a compelling reason to foster the development of the government securities market in Serbia

1.2 Serbia - Some Key Economic Indicators:

GDP (% change)-in real terms GDP in USD (millions)-Nominal GDP per Capita (USD) Employment (millions) Unemployment (millions) Dependency ratio (%) Price inflation (%) Savings with banks (euro million) Dinar-Euro exchange rate Current alc deficit (% GDP) -- Source: Ministry of Finance (MOF): published Economic quarterly.

Despite steady economic growth, Serbia's standard of living remains low, with a high dependency ratio and high unemployment. Inflation declined between 2001 and 2004; but in the first nine months of 2005, it was 11.8% and, as such, remains stubbornly high. Due to the financial system's increased stability and the greater confidence of citizens in most banks, household savings have increased dramatically during last few years. Due to substantial remittances by Serbian citizens abroad, together with ongoing low confidence in the dinar (SRD), the bulk of household savings are denominated in euro - thereby reducing potential demand for SDR assets. The 2004 current account deficit of the balance of payments was high at 12% of GDP, while between 2001 and 2004 the dinar depreciated by around 33% against the euro. Against this background, it is not surprising that monetary policy remains restrictive and that the government maintains a tight fiscal policy stance, in order to contain inflation, interest rates and balance of payments deficits', as well as to promote economic growth.

I despite these weaknesses macro-economic stability, prudentfiscalpolicy and ongoing economic reforms and EU integration prospects resulted in Serbia's credit rating being upgraded by S&P to BB (July'O5).

10 1.3 Fiscal situations and borrowing: The public budget deficit as a percentage of GDP is estimated, by MOF, at less than 2 percent in 2004~and a modest surplus in 2005 is expected, Government is maintaining tight control on public expenditure, which is necessary, given that it represents 43 percent of GDP. The improved budgetary situation is also helped considerably by a widening of the indirect tax base and improved enforcement by the tax authorities. The ratio of Serbia's government debt to GDP has accordingly declined dramatically from 1 10% in 200 1 to some 60% in 2004, reflecting privatisation receipts, tight budgetary policy, economic growth and a debt write off under the October 2004 London Club Re-scheduling Agreement.

1.4 Funding the deficit: The 2004 original Budget plan -which is on a cash basis, rather than an accrual accounting basis - was overtaken by a supplementary Budget which replaced the original Budget. The outturn deficit was funded as follows: SRD Billion Privatisation proceeds 9.7 International concessional credits and grants 8.4 Net domestic borrowings 4.0

It is clear, nevertheless, that budget deficit funding heavily depends on the proceeds of privatisation - some 44 percent of funding last year. Net domestic securities accounted for only 20 percent. Maturing Euro FFS savings bonds each year (up to 2016) is the major funding item and, as such, is of key concern to MOF. Up to the present, privatization proceeds have been buoyant and have proved adequate to meet most of the government's annual funding requirements. However, there is a question-mark over the medium-term viability of privatization receipts.

1.5 Privatization proceeds: Apart from any uncertainty overhanging funding of maturing FFS Euro bonds, there is a particular uncertainty over the medium-term prospects for privatization proceeds. These receipts have in recent years been used in an effort to refinance much of maturing FFS bonds. There are two separate privatization agencies: One dealing with the privatization of banks and the other with all non-bank enterprises. About eight banks remain to be privatized. Three banks were privatized ths year, generating proceeds of €272.5 million. It is expected that the remaining banks will be privatised over the two years or so. In the case of non-bank state-owned enterprises, or socially owned companies, the privatization agency involved expects to credit MOF with proceeds of around €153 million. This is in addition to the €1 50 million from bank privatization. The pace of future privatization of socially owned entities is more difficult to judge, but the agency involved feels that the program should wind down within the next two years, and that the benefit accruing to the Budget may not be significant. For significant proceeds to arise in the non-bank sector, it will be necessary for political decisions to be made in relation to the large state-controlled corporations such as telecoms, electricity, gas, railways, etc. Experience suggests that such decisions will not be forthcoming quickly or readily.

1.6 External debt represents about 56 percent of total public debt; but when account is taken of domestic frozen foreign currency (euro) bonds (FFC bonds), as well as the small amount of euro Rehabilitation Bonds, foreign currency denominated public debt is 96 percent of total public debt. Even though the external element is undertaken on concessional terms, the fact that the dinar has depreciated since 2001 against the euro by approximately one-third represents a substantial extra borrowing cost that must be borne by Serbia's taxpayers.

If adopting IMF methodology, the 2004 deficitlGDP would be even more favourable, about zero. 1.7 Government securities: As of end-2004, the outstanding stock of debt, according to figures published by the Treasury at MOF, are as follows:

1. T-bills - SRD 6.4 billion 2. T-bonds - SRD 19.4 billion 3. FFS bonds - €3.7 billion (SRD 292 billion equivalent) 4. Rehabilitation ~onds~= €40 million (SRD 3 15 million equivalent)

These securities represent 20 percent of GDP and, as such, can be said to represent an important milestone on the road to development of the government securities market.

1.8 T-bonds of SRD 19.4 billion are held exclusively in the portfolio of the National Bank of Serbia (NBS). These bonds were issued in December 2004 in exchange for previous government debt held by the NBS consequent upon its financing of the government budget deficits. They were issued and paid for at par, since there was no market in T-bonds. Effectively, MOF and NBS agreed that the following three bonds should be issued in lieu of miscellaneous debt obligations of the government held by the NBS:

SRD billion 8%% coupon, 2006 = 3 8%% coupon 2007 = 6 8%% coupon 2010 = 10.7 19.4

Throughout 2005, NBS used these bonds as repos to drain liquidity. As a result, debt management and monetary policy management became segregated activities. Despite the circumstances under whch these bonds were issued they are potentially marketable. However, given that the coupon is 8.5% p.a. while interest rates are in double digits (see 2.2. below) the potential market value today would be significantly lower than the face value-which would imply a sizable loss to the NBS if it sold some of the bonds to private investors.

Nevertheless, between 12.08.05 and 07.10.05 NBS sold, in five competitive auctions, SRD 750million of the T bond 2006 held in its portfolio. Notable features of these auctions were: 9 Amounts offered for sale were modest only: SRD 200m each (€2.4 million). 9 The first two auctions were undersubscribed but the last three were heavily oversubscribed: cover in the last auction was 1.9 1. 9 Average auction yield, for 15-17 month paper, was 16.4%, which compares very favourably with the costs of T bills and interbank rates for much shorter maturities. 9 Success in these auctions is traceable in no small way to the fact that some domestic banks bid the auctions on behalf of non resident investors, whch are key to the development of all government securities markets. Clearly, non-residents were happy to accept the yields on offer despite nominal depreciation of the dinar against the Euro, although in real terms the dinar appreciated against the euro during 2005.

1.9 T-bills: MOF began issuing T-bills by auction in 2003. Six month bills are auctioned on the first Tuesday of each month and three month bills are auctioned on the second and fourth Tuesdays of every month4. Issuance, which is in dematerialised form, is carried out electronically at the Central Registry, Depository and Clearing (CRSD). The system operates real-time DvP and is widely acknowledged to be

' Rehabilitation Bonds are held by Serbian citizens but secondary market trading is negligible. Occasionally, six-month paper has been "re-openedn4after one month safe and efficient. The amount of T-bills outstanding amounts to over Dinar 7 billion and they are widely regarded by MOF and the market to be a success. To submit electronic bids at a T-bill auction, a bidder must be a member of the CRSD. Other bidders may submit bids indirectly, using the services of a member. Some 100 banks and brokerage houses are members of the T-bill auction.

Participants in auctions are obliged to pay at least 25 per cent of the total discount value of submitted bids into their cash account at CRSD by 10 a.m. on the auction day. Minimum bid is SRD 100,000; but there is no ceiling per bidder. Auction is open for bids between 10 and 11a.m. MOF must decide on the lowest acceptable bid price by 11:30a m. It announces the auction results, via CRSD, before lpm. MOF also publishes data on the volume of bill issuance, average discount rate bid and average discount price achleved in the auction. Secondary market trading in T bills is very thin, with bills tending to be held to maturity.

1.10 Treasury Bills (3 months) Auction Results (2003) [Millions of Dinar]:

Auction date Amount Total Bids Accepted Cover Ratio Discount

Offered (2) Bids (2) + (1) Rate (5)

29 ~pr03 6 May 03 13 May 03 20 May 03 27 May 03 10 Jun 03 24 Jun 03 15 Jul03 29 Jul03 12 Aug 03 26 Aug 03 9 Sep 03 23 Sep 03 14 Oct 03 28 Oct 03 11 Nov 03

Note: coverage ratios larger than 1 shown for emphasis. Source: MOF

Clearly, it took time for the T bill (3 month) auction to take off in 2003. Initially bids fell short of the amount on offer (SRD 500 million), so from June onwards the amount on offer was reduced to SRD 250 million. By September 2003 the cover (i.e. total bids divided by amount on offer) hit 2.1 times and during October 2003, the cover exceeded 4 times. The discount rate (i.e. the effective annual discount to yield) in 2003 averaged 18- 19 percent per annum. 1.11 Treasury Bill (3 month) Auction Results (2004) - millions of dinar

Auction Date Amount Total Bids Accepted Bids I Cover Average Offered (2) (3) (2) - (1) Discount Rate (%)

13 Jan 04 19.94 27 Jan 04 20.47 10 Feb 04 20.99 24 Feb 04 21.35 9 Mar 04 21.20 23 Mar 04 2 1.34 13 Apr 04 2 1.06 28 Apr 04 22.23 11 May 04 21.72 25 May 04 21.39 8 Jun 04 21.10 22 Jun 04 21.41 13 Jul04 22.18 27 Jul04 22.57 10 Aug 04 22.65 24 Aug 04 21.30 7 Sep 04 20.65 21 Sep 04 2 1.25 12 Oct 04 20.40 26 Oct 04 20.59 9 Nov 04 21.09 23 Nov 04 2 1.03 7 Dec 04 21.10 21 Dec 04 2 1.20

I I I I Source MOF

In 2004, the amount on offer during the year was increased quite substantially. As a result, the auctions remained relatively under-subscribed. Thus, the cover was still not satisfactory - only half of the auctions were covered. Furthermore, the discount rate achieved was higher in 2004, averaging between 2 1-22 percent per annum. Both the cover and the average discount factors were, it appears, adversely influenced by the amounts on offer. 1.12 Treasury Bill (6 month) Auction Results (2004) - millions of dinar

Auction Date Amount Offered Total Bids Accepted Bids Cover Average

(1) (2) (2) + (1) Discount l----- Rate (%) 3 Feb 04 2.Mar 04 6 Apr 04 4 May 04 1 Jun 04 6 Jul04 3 Aug 04 3 1 Aug 04 5 Oct 04 2 Nov 04 30 Nov 04 Dec 04

I Source: MOF

"Cover" in the case of 6 month bills was somewhat better. In 7 out of 12 auctions in 2004, total bids equalled or exceeded total offers. This is due in no small way to the fact that the amounts offered in the 6 month T bill auctions were substantially less than the amounts on offer in the 3 month auctions. Of the total amounts offered at T-bill auctions in 2004, only 15 percent was attributable to 6 month bills. Average yields in the 6 month bills were also less volatile than those of 3 months, and trended around 2 1.5 percent per annum.

1.13 The situation changed in the course of 2005. Between March and October T bills outstanding fell from SRD 6.3 billion to SRD 5.9 billion and yields also fell to 14.71% for 3 month T bills and 15.67% for 6 month T bills. T bills auctions have been undersubscribed more recently as most banks invest in alternative NBS repos of its own paper, on which yields are 16.5% (see 2.6 below). The following graph illustrates the consequential decline in T bills. Offer. Demand and Average D$scounrRate or1 3-month Tbllls lri 2005

1.14 T-bills ownership @ 21.03 05: @ll.ll.O5 'Yo 'Yo Banks 90.82 73.68 Corporates 9.18 26.32

Source: CRSD& MOF

T bills have been a valuable source of funding in the primary market - although the amount of T bills outstanding has declined markedly during 2005 (from SRD6.3 billion to 5.9 billion) due to substantially reduced funding requirements. The secondary market in T-bills is insignificant: for example, a mere 330 transactions amounting to SRD 1.4 million were effected in 2004. In this context, it is significant to note that non-residents cannot hold T-bills; whereas they are allowed to invest in government securities longer than 1 year.

1.15 Frozen Foreign Exchange Savings Bonds (FFS bonds): The process of converting frozen foreign currency bank deposits into FFS bonds (denominated in euro) commenced in 2001 and is virtually complete. The bonds are zero coupon obligations, one bond maturing each year between 2005 and 201 6. No further FFS bonds are being issued. Interest on the bonds was fixed, arbitrarily, at 2 percent per annum and this was capitalised annually on the face value of each bond. Thus, the nominal outstanding in each bond will increase marginally by the interest factor. Outstandings in the bonds at end 2004 amounted to euro 3.7 billion. Amounts (in euro) redeemed on maturity in recent years were: 2002 - 93.7 million 2003 - 179.8 million 2004 - 194.9 million 2005 - 2 16.0 million Frozen Foreign Currency Savings Bonds outstanding (in Euro): Year Outstanding 2006 208million 2007 203million 2008 204million 2009 209million 2010 2 18million 201 1 230million 2012 245million 20 13 263million 2014 285million 2015 306million 2016 3 19million

(Source: Ministry of Finance & CRSD)

Annual refinancing of these bonds is of key concern to the MOF. It should be noted that the amounts shown in the above table do not reflect the additional cost in terms on dinar of any future exchange rate depreciation against the Euro. For example, the rate of depreciation between 2001 and 2005 was approximately 40%.

1.16 FFS bonds traded on the Stock Exchange (20 percent) and the OTC market (80 percent). Total turnover in 2004 of €600 million,is well up on 2003, but is still less than 20 percent of the stock of bonds outstanding (€3.7 billion). There is nevertheless latent demand from residents and non-residents for the bonds, and since there appears to be no fresh supply in the market, annual yields are relatively flat,, ranging from 4.7percent in 2006 to 5.4 percent in 2016. Demand at the long end is higher since those bonds are used in payment for privatisation shares. Thus, the yields on these euro FFC bonds bear no relationship to money market rates in dinar or to T-bill rates (see paragraph 2.2 below). In other words, there are two separate yield curves, one for short-term dinar and the other for euro-denominated bonds.

1.17 Settlement: OTC FFS bond transactions are settled on a DvP basis in T + 0 whereas the Stock Exchange transactions are effected on T + 3. All bond trades are settled on the CRSD on a DvP basis.

1.18 Price transparency: A major drawback to the existing secondary market structure is the lack of price transparency in respect of OTC bond transactions. In contrast, the Stock Exchange transactions are in real time and, in turn, form the only basis for pricing the 80 percent of transactions that take place on the OTC market.

1.19 Limited presence of domestic institutional investors: Modem government securities markets are dominated, not by individual investors, but by institutional investors. But in Serbia there are a limited number of institutional investors. In particular, there are but six life insurance companies, funded pension arrangements are at an embryonic stage, and there are no investment funds (either open or closed). Banks are the main institutional investors in government T-bills and also in FFS (euro) bonds. However banks hold FFS bonds primarily for trading purposes: they are the main drivers of secondary market activity via over the counter trading with their customers and, occasionally, doing large deals for non-resident institutions. Domestic institutions that are natural holders of long-term government paper are life assurance companies (and pension funds), since long-term insurers are very dependent upon the availability of asset classes that will match their long-term liability structure.

Contractual savings and insurance products in particular, are important for financial sector development and can contribute directly to real sector development. The OECD has pointed out that insurance represents the largest pool of long term savings in Europe and that contractual savings support the development of capital markets, reduce bank term transformation risk and help keep yield curves flat. Serbia has a long history of developed and balanced financial markets and the challenge for the insurance sector is one of rebuilding rather than starting from scratch. The life business was relatively healthy at the beginning of the 1990's. It was then effectively destroyed by a succession of disasters including hyperinflation and sanctions-based economic collapse. The sector is now only beginning to re-emerge after the hyperinflation and scandals of the 1990's and has the potential along with the pension funds, to generate a significant pool of long-term savings over the next decade. In 2004, life insurance accounted for only 7% of total insurance premium income. However it is growing rapidly, the equivalent 2003 figure having been 4%. A significant proportion of life business has initially been written either in competition with bank term deposit contracts or as mortgage endowments associated with the developing mortgage market. It is virtually all written with euro-denominated benefits or indexed benefits and premiums. Accordingly Serbian life businesses invest in the only available listed government security, which is denominated in euro also, that is, FFS bonds with maturities out to 2016. Life companies and pensions funds need assets with 20 years plus maturity and the lack of a well rated and liquid long-term securities market inhibits the development of long term intermediation. Other issues are on-going improvements to the regulatory and supervisory system and legislation to effect the privatization of the two major, socially owned, insurers.

Pending the passage of special legislation to regulate funded supplementary occupational pensions5, private pensions are written through 4 specially licensed life insurers. Such business is, as noted above, still at an embryonic stage. At present the few pension schemes involved hold roughly 60% of their assets in Euro FFS bonds, 20% in euro-linked corporate bonds, 15% as equities and 5% as cash.

Overall, demand for government securities by institutional investors in Serbia is very weak but there is the potential for substantially increased growth in the short to medium term. In thls context, taxation has been identified (see paragraph 1.38 below) as a special issue.

PERCENTAGE OWNERSHIP OF FFS (EURO) BONDS AS OF 05.05.2005

BROKERAGE FIRMS 1 0.63 I INSURANCE COMPANIES 1 0.99 I DOMESTIC LEGAL ENTITIES 1 2.79 I FOREIGN LEGAL ENTITIES 1 5.81 I 1 DOMESTIC NATURAL PERSONS 1 82.94 I FOREIGN NATURAL PERSONS 1 1.41

1 TOTAL 1 100.00

A new pensions Act is due to become operational in 2006. Source: CRSD *Note: During the exchange crisis of the 1990 S , the Government imposed,foreign currency constraints, and depositors 'foreign exchange deposits in commercial Banks were frozen and the proceeds used by the Government. Euro- denominated savings bonds were issued in 2001 in compensation to individual account holders: see 1.14 above.

Modest FFS bond secondary market annual turnover of €600 million compared with a total outstanding stock of €3.7 billion suggests - as does the above table - that most individual bond holders intend to hold their bonds until the maturity. The market tends to be predominantly one-way - a seller's market - with poor depth due primarily to a lack of bond supply by the government, even though demand for euro- denominated paper remains strong.

POLICY CONSIDERATIONS

1.20 CRSD: A safe, reliable and efficient settlement and custody system is crucial to a well-developed market in government securities. The CRSD meets these criteria.

1.21 Cost: The cost of transacting in T-bills and T-bonds using CRSD was reduced recently to 0.1 percent flat (both sides) and SRD 500 per repos transaction. These costs are, however, in addition to any brokerage firm fees payable by insurance companies and other non-members of the CRSD. When transactions are effected via the Stock Exchange, for price transparency considerations, a further 0.1 percent flat is charged. MOF has the power in the context of government securities auctions to require auction participants to inform the Ministry about the fees chargeable to clients. On the basis of this power, and in an effort to ensure that the purchasing and any subsequent trading of government securities is not impeded unduly by transaction costs, MOF should critically review of the costs involved, in consultation with market participants.

1.22 Bill auction access: In principle, it is in the interests of taxpayers and investors alike to admit to the auctions an base that is as wide as possible, subject to meeting certain credit standards, thereby fostering competition. Non-members of the CRSA cannot access the auction directly: thus, life insurance companies are excluded even though CRSD rules entitle "companies for management of funds" to membership of the CRSD. Indirect participation may be less conducive to competition among bidders because the intermediary would inevitably know the terms of the bids submitted on behalf of his clients. Since a substantial element of life insurance business itself involves funds management, there is a case for widening the direct participant base to include insurance companies.

1.23 Risk of non-compliance in T-bill auctions: The solution adopted by MOF was until very recently to require 25 percent cash deposits. For auction participants this was a particularly costly way of securing compliance with buying requirements, and it could discourage participation. An alternative, and less costly, procedure was adopted: should non-compliance occur, the defaulter could be excluded from future auctions, for one to three years, and he could be charged a penalty also.

1.24 Auction ceilings In some countries ceilings on the maximum amount per bidder are operated as an anti-cornering device in respect of T-bond auctions. However, this restriction is not normally applied to T-bill auctions, and its usefulness even at the theoretical level is debatable. The preferred course to deal with unrealistically low bids is for MOF to use its right to reject any bid. It seems clear that MOF has done this in the past where prices bid were too low. However, MOF should guard against offering an amount for auction and then refusing to accept it solely on the grounds that the funding is not required immediately. This would create uncertainty for investors 1.25 Publicity: The lag between the deadline for presenting bids ((9-10.45 am) ) and the announcement of results plus action details (about 1 p.m.) should be shortened as much as possible, in order to reduce uncertainty and also to reduce the market risk to which auction bidders are exposed. Given that the CRSD operates a state-of-the-art electronic auction system, it is recommended that all announcements relating to an auction be made by 11:30a.m. and certainly not later than noon. The appropriate balance between the published information on the auction details seems to favour aggregate data along the lines already supplied by MOF. In Serbia's thin money market this appears appropriate. Nevertheless, in an effort to spread auction information as widely as possible, so as to facilitate the process of true pricing, but without facilitating cartel-like behaviour, it is recommended that the MOF consider also making public the highest and lowest accepted bid price and yield in each T-bill auction.

1.26 Amount of T-bills offered in auction: In the light of the analysis in paragraphs 1.10 to 1.12 above, and in order to reduce cost of funding, it is recommended that the MOF should, target a more even distribution of T-bill offer amounts during the course of the year. For example, a target "cover" of 1.5 or so would, if adopted for the 3 month T-bill auctions in 2004, have averaged €500 million (rather than €650 million) or SRD 12 billion in total. This practice would help the market in forming expectations on equilibrium prices and would enhance transparency. Too large a volume on offer may prove difficult to absorb and too small a volume may not attract sufficient market attention. It is recognised that for cash flow funding reasons, MOF will not always be able to adhere to the same offer amount, but it should strive to do so as much as possible while retaining flexibility.

1.27 Non-Competitive Bids: Non-competitive bids involve a commitment to sell a certain amount of T-bills at the average competitive auction price. Non-competitive bidding is appealing to bidders that do not have specific views on expected auction prices, or those who wish to avoid the risk of buying at above-average prices, e.g. corporates. Permitting this form of auction participation can, therefore, widen the scope of the T bill market. MOF should consider introducing non-competitive bids, subject to capping the amount of the non-competitive element, say, to 10-15 percent of the element accepted in the competitive price. This would help to avoid determining the average auction price over an unacceptably narrow range of the market.

1.28 Non-resident investors in T-bills: One of the key driving forces in developing a government securities market, and giving it liquidity, is to make such investment as attractive as possible to non- resident investors, who will carry the exchange rate risk.

In this context, by increasing potential demand for T-bills it would serve the interests of future debt management in Serbia to have foreigners buying T-bills (from which they are presently excluded) both at auction and in the secondary market. Their impact would drive down the cost of T-bill funding. NBS, on the other hand, appears to be opposed to permitting such investment, which they regard as falling into the category of "hot money". It should be noted that the capital inflow arising would be in addition to already very buoyant inflows arising from foreign borrowing by banks and corporates. What is at issue is a matter of timing: foreigners will be allowed, over time, into the T-bill market. 1.29 T bonds in the NBS ~ortfolio: As was noted above (paragraph 1.8) nearly SRD 19 billion T bonds are held by the NBS and SRD0.75 billion which the bank sold, is held by the market. The latter sale demonstrated that, despite double-digit inflation and nominal exchange rate deprecation, modest amounts of T bonds of 1-2 year maturity can be sold on the open market at competitive rates. On the other hand the transactions involved appear to be of an ad hoc nature (with no consequences for liquidity management by NBS) rather than part of a planned issuance under a well-publicised debt management policy developed by MOF. Consequently, the paper strongly recommends that future T bond issuance be camed out in the context of national debt management, by MOF; and that NBS reach agreement that further T bond sales to the market will not take place except under the aegis of MOF.

The prevailing government budget is in surplus and privatisation receipts are still strong. Refinancing the SRD3 billion T bond maturing in 2006 may not, therefore, be a serious problem. The government may not be in a after 2007, as previously noted, to count on worthwhile privatisation proceeds. Should this prove to be the case and given that, in 2007, SRD6 billion T bonds (held by NBS) and €203 million Euro FFS bonds will mature, MOF may well need, for funding purposes to re-instate historical levels of T bill issuance and also issue a new T bond (in dinar). Funding needs will be all the greater between 2008 and 20 10 when SRD 10 billion T bond and €63 1million FFS bonds mature.

One option is for MOF to issue in 2007 by competitive auction, a new 2 or 3 year T bond. Prior to the auction steps would need to be taken to ensure price transparency on the OTC market (see 1.30 below). Depending on MOF's funding capacity, MOF may, or may not, be in apposition to redeem outright in 2007, the entire SRD 6 billion T bond. To the extent that it is unable to do so, (£rom privatisation, sales of T bills and a new T bond), then MOF could swap the unredeemed balance of the SRD6 billion bond for a new 2009 or 2010 bond at the average price in the T bond auction. Thereafter, however NBS should aim to withdraw from financing (or refinancing) the governments borrowing requirements, in accordance with the Maastrict Treaty principle that the central bank must not fund budget deficits viz. by purchasing government securities in the primary market. There is nothing, however, to stop NBS buying T bonds in the secondary market.

1.30 FFS Euro bonds - lack of price transparency: Eighty percent of trading in FFS Bonds occurs on the OTC market and the market is worried by the resultant lack of price transparency. An immediate solution lies with the CRSD: it envisages issuing price/volume details in real time confined to CRSD membership, together with end-of-day summary details on its general website. This would represent a major advance in the present situation and, as such, it should be acted upon without delay. It has been noted above that there is strong demand for euro-denominated paper, and, given the lack of bond supply, yields at around 6 per cent in Serbia are quite low, although high by eurobond standards owing to poor credit rating and lack of liquidity. This issue is addressed below.

1.31 Future debt issuance options: Medium-term reliance on privatization receipts and primary fiscal surplus to fund maturity bonds and other deficit elements is highly uncertain. MOF would, therefore, need to consider alternative funding options, viz:

a) International foreign currency borrowing:

b) Domestic (i) FFS euro bonds; (ii) Fixed rate Dinar bonds; (iii) Floating rate Dinar bonds; (iv) Euro indexed Dinar bonds 1.32 Advantages / Disadvantages of international foreign currency borrowing: Among the advantages of option (a) above are: (i) comparatively low nominal interest rates, which makes loans in foreign currencies appear cheaper than domestic borrowing (in dinar); (ii) a seemingly unlimited international market in which the government runs little risk of crowding out other borrowers; (iii) the chance of gaining international experience and making useful contracts in the financial world; (iv) foreign currency borrowing could put downward pressure on Serbian interest rates below their prevailing levels. As against this, there are a number of strong, and indeed compelling, reasons why such a course of action would not be prudent:

a) Domestic interest rates, nominal and real, are high in Serbia at present because of exchange rate pressure and not because of excess domestic funding of the government. At times of pressure on the dinar Serbian interest rates have to be above foreign rates so as to shake off . If foreign borrowing by the government is excessive, Serbian interest rates may fall too far. This would lead, in turn, to an outflow of funds, necessitating more foreign borrowing to support the external reserves at NBS. The outflow will continue as long as it is financed by higher borrowing and in the end it will be necessary to stop the drain by allowing domestic liquidity to be tightened to the point where Serbian interest rates are again pushed back to their market-determined relationship with relevant international rates. b) Serbia has a uniquely high foreign debt component (96 percent of Public Debt), so arguably, the government does not have the luxury of increasing foreign borrowing, particularly in circumstances where it is likely to be counter-productive. c) Further foreign borrowing could question the country's credit standing (B,) and would increase the future drain on the economy from foreign debt service payments. In this context, average interest rate costs of existing foreign debt will, it is understood, rise further from 2007 and onwards. d) The dangers of escalating foreign debt are well known: it can be very expensive if raised in the wrong currency or if the local currency has, at a future date, to be devalued; depreciation of the exchange rate is not easy to hedge against over a prolonged period; e) It would do nothing to assist the development of the domestic market in government securities and would be a bad omen for the credibility of the Dinar.

The authors are of the view that, given tight fiscal policy, the benefits of contracting further foreign currency borrowing at international market rates are strongly outweighed by the costs referred to at (a) to (e) above.

1.33 Further euro-denominated domestic bonds There is, as already noted, latent demand for the existing euro savings bonds, and they trade at exceptionally, if not artificially, low yields. If the government were to make further issues, they could do so by tranching6 into the existing bonds and thereby building up large liquid issues in, say, 2010 and 2015: that is 5 and 10 year maturities which are a common feature of government bond issuance in OECD countries. The impact of such action on the price of the bonds, that is the price at which the government could sell them, would depend in large measure on the amount to be issued and whether or not it was part of an intended programme of future issuance or just a one-off measure. The former would weaken the bond prices and increase yields; whereas, the latter

"Tranching" means issuing a fresh slice of an extant bond, bearing all of the characteristics of the original bond and be fungible with it. might have little impact. A further advantage is that the new euro bond issuance would enhance the market in that paper. The disadvantages are broadly similar to those listed at paragraph 1.32 above in respect of international foreign currency borrowing. In addition, "domestic" Euro bonds could crowd out the medium-term prospects of developing a market in government dinar bonds and leave that market seriously neglected, if not "killed". In other words, euro domestic borrowing represents, arguably, the worst possible option from a domestic debt management viewpoint.

1.34 . Domestic fixed rate bullet SRD bonds: The downside on raising such debt on the domestic market is that the pool of investible funds is limited: especially since the dinar is widely regarded as being risky - and the bulk of household savings are already in euro. The risk of crowding out the private sector is therefore not insignificant. Interest rates for 3 and 6 month bills are around 15% per annum; and confidence in the government's anti-inflationary policies is relatively low among the financial community, thereby making it difficult to contemplate seriously raising medium-term funds in dinar. Nevertheless, the successful auction of 15-17 month T bonds by NBS has already been noted above although it was on a modest scale. The longer MOF defers issuing a fixed rate SRD T-bond, the longer Serbia's money and capital markets will remain underdeveloped and the cost of borrowing, including the exchange risk attaching to foreign currency denominated borrowing, will remain unduly high. A well- developed government securities market would enable the government to reduce these costs and risks to the taxpayer. Thus, most OECD governments accord a high priority to the goal of developing their capital markets and have encouraged domestic currency borrowing in its different forms to achieve that objective. Countries that did not pursue this goal have had to pay a premium in yields owing to market illiquidity.

In present budgetary circumstances, it may not be feasible to contemplate a government dinar T-bond. However, assuming interest rates and inflation trend down somewhat between now and, say, 2007 when the second NBS T-bond (in the amount of SRD 6 billion) matures in addition to €203 billion of FFS bonds, it may prove necessary for MOF to auction a new T-bond with a 2 to 3 year maturity, as recommended at 1.29 above. MOF could, if it were necessary, tranche further issuance of the new bond to NBS in exchange for all or part of the NBS paper, at the average price in the auction. Furthermore, NBS and MOF should agree that the bank will hold T bonds on its portfolio to redemption, unless MOF agrees to buy-backs ahead of redemption.

1.35 Floating Rate T bond (dinar): As a prelude to issuing fixed interest rate T-bonds, MOF could, in the first instance, issue a floating interest rate T-bond, with the rate of interest determined every three months based on the average yield in the most recent 3-month T-bill auction. Again, it would appear advisable to limit the maturity to two years. In effect, this proposal would involve an extension of T-bills from 3-6 months to two years, which would enhance government funding and still pay current interest every three months -which should be attractive to investors. From the taxpayers' viewpoint, the hope would be that over the two year period inflation and interest rates would decline. The floating rate T- bond could be auctioned around May 2006, when the first NBS-held T bond matures. MOF could exchange the floater, at the average accepted auctioned price, for the maturing paper, or for a proportion of it.

1.36 Index-linked Dinar T-bond: This option is essentially similar to further FFS euro bond issuance, insofar as the Dinar bonds, principal and interest, would be indexed linked to the euro. Thus, the taxpayers would be taking the exchange risk. The main points for and against auctioning this instrument may be surnrnarised as follows: FOR - (i) Indexation to the euro could be taken as a statement of willingness on the part of the govemment to maintain a stable exchange rate with the euro.

(ii) Servicing costs on the index-linked bonds should be substantially reduced in the early years compared with a conventional fixed rate Dinar T-bond. This assumes that the yield at which indexed bonds would be sold would be less than if sold without indexation.

(iii) The availability of index-linked instruments may encourage an increase in the level of dinar savings.

(iv) Inflows from abroad may be induced by the availability of the index-linked bonds, depending on an investor's view of exchange rate and inflation prospects.

AGAINST - (i) Indexation to the euro may weaken the incentives for government and investors to invest in dinar paper; and it could set a most undesirable precedent for institutionalizing euro indexation of wages, tax allowances, etc.

(ii) At a time when incomes policy in Serbia is restricting increases in real wages, it would be invidious for the government to guarantee rates of return in euros to holders of index-linked stock, thus effectively exempting them from the difficult adjustments necessary to the economy.

(iii) The total cost of index linked debt might prove as high or higher than that for a conventional fixed rate dinar T-bond (if one existed!) depending on the coupon and depreciation of the exchange rate (if any). At best, it may lead to a postponement to maturity of the real cost.

1.37 Conclusion re future funding options: MOF will continue to issue, and rely upon, T-bills as a source of budget financing, as well as for cash management purposes. Developments further out along the yield curve, that is, those affecting T-bonds, raise complex issues. Nevertheless, it is clear that MOF should avoid increasing its foreign currency denominated borrowing (at market rates) both on international market and, in particular, on the domestic market which would be most damaging to the medium-term development of the govemment securities market in dinar. In periods of uncertainty and high inflation such as those that characterise the current situation in Serbia, floating interest rate T-bonds in dinar, (say, 2 year maturity) would offer a relatively attractive option to investors, who would be guaranteed prevailing interest rates payable every 3 months in line with 3 month T-bill auction results. As well as obtaining 2 year dinar funding for the first time, the government could then gain if inflation and interest rates reduce. If this course were successfully followed, MOF could consider, especially if interest rates were reducing, the further option of a twolthree-year fixed rate dinar T-bond, although realistically this would depend on the demonstrable success of the Government's stabilization policies. If neither of these two recommended courses of action prove to be feasible, or if there were a sudden and serious funding problem in the course of the next twelve months (neither of which seems likely) the option of an index-linked (to the euro) T-bond denominated in Dinar could be then considered - but only as a last resort.

1.38 Pension fund development: NBS needs to develop clear investment rules and regulations appropriate to pension fund investment. Moreover, MOF should streamline the tax code to meet pension fund needs. In most OECD countries the tax regime is designed in such a way as to foster development of private pension schemes for good social and economic reasons. Such funds in turn play a major role in the expansion of the government securities market, in particular by investing in medium and long-term T. bonds.

In Serbia, there are a number of taxation obstacles to private pension fund development viz:

(a) Employee contributions are not allowed as a deduction for personal income tax; (b) Employer contributions are treated, for tax purposes, as wages and are deemed to be taxable (by employer); (c) Pension funds are chargeable to tax on all realised gains.

In most developed economies (a) to (c) do not apply; in the case of (c) tax becomes payable only when payments are made from a fund e.g. pension payments.

MOF should examine the situation and take appropriate action to modernise the tax regime as it applies to pension contributions and to pension funds.

1.39 Coordination between the MOF and NBS on debt policy and monetary policy

In Serbia, the central bank issues its own bills and bonds to mop up currency inflows at a time when the government has no need of cash. The central bank uses its securities as a draining mechanism for its monetary market management. But the use of different but similar instruments for monetary policy and debt management purposes incurs the risk of market fragmentation and the loss of the liquidity and cost benefits offered by a larger securities market. There is an alternative approach where the MOF issues additional Government securities as an add-on to the normal auction, but sterilizes the proceeds by holding them in a separate account at the central bank, which will be remunerated at the discount rate set in the securities auction. This arrangement will allow the central bank to preserve the optioil of using monetary policy tools, but meanwhile prevent the risk of market fragmentation. However, the achievement of the desired outcome is contingent on the satisfaction of certain conditions, including (i) the transparency to the market of this arrangement and the amount of each auction, and (ii) identity between securities issued at the request of the central bank and the rest of the outstanding government securities.'

1.40 Comprehensive Strategy for National Debt Management:

This paper addresses the issues that need to be focused on for the early development of the government securities market. Ideally, implementation of these recommendations should be set in the context of a comprehensive strategy that takes account of the medium-term aspects of future debt management such as the evolution of the government borrowing requirements; sources of funding; mix of foreign /domestic borrowing; relationship to monetary policy implementation, development of a modern properly resourced debt management unit either inside or outside MOF; risk management and implications for the government securities market of potential membership of the European Union. These are among the issues for later consideration, and should not hold up implementation of the measures recommended in this paper.

RECOMMENDATIONS

(a) Priority issues:

In the UK, the DM0 and the Bank of England have agreed to such an arrangement. As part of the reform of its cash management arrangements, the Brazilian government passed the Fiscal Responsibility Law, which in effect from May 2002, prohibits the central bank from issuing its own securities. 1. Given that the refinancing of FFS (euro) bonds is a demanding annual funding requirement (€200-€250 million a year), and since privatization proceeds available to the Budget in the future are likely to be less than during the years 2001 -2006, MOF needs to consider a number of funding options, viz:

1. International foreign currency borrowing; 2. Fresh FFS euro bonds; 3. Fixed rate dinar bonds; 4. Floating rate dinar bonds; 5. Euro index-linked dinar bonds.

Costs and other disadvantages associated with international foreign currency borrowing at market rates effectively rule out this option. The case against further euro denominated domestic borrowing is equally strong and, indeed, even more so since it would, in addition, stifle the medium-term development of the government securities market in Dinar.

In periods of uncertainty and high inflation a 2-year floating interest rate T-bond would offer a relatively attractive option for investors, who would be guaranteed the prevailing interest rate payable every three months based on the average yield in the 3-month T-bill auction. If this course were adopted and were pursued successfully, MOF could pursue (especially if interest rates were trending downwards) the further option of a 2-3 year fixed interest rate T-bond, although much would depend on the demonstrable success of the Government's stabilisation policies. If neither of the foregoing proved feasible, a situation that seems most unlikely to arise, the option of a T-bond denominated in dinar, but index-linked to the euro, could be then pursued as a last resort.

2. The coordination between MOF and NBS on debt policy and monetary policy should be strengthened. The fragmentation of the market caused by the MOF and NBS issuing their own securities could be avoided by adopting the following mechanism:

The MOF issues add-on Government securities for monetary policy purpose at the request of the NBS, r The proceeds are sterilized by holding them in a separate account with the NBS, which will be remunerated at the discount rate set in the securities auction. r The proceeds from the add-on government securities shall not be counted as traditional government debt.

3. The need for a medium-term strategy is underscored, in paragraph 1.40 above. However it is a matter for later consideration and need not hold up implementation of the recommendations in this report.

(b) Other issues r MOF should review critically all of the costs involved in purchasing government securities at auction and in transacting such securities, in consultation with market players.

There is a case for CRSD to arrange to admit to membership the insurance and pensions companies that manage funds, so that they may compete directly in T bill auctions. MOF, instead of requiring from bidders at auction 25% cash deposits, could arrange for bidders to present suitable documentation - to be submitted together with a bid - guaranteeing that funds are available for each bid if successful: in July 2005 provision was made to ban a defaulter from future auctions for a period.

Auction ceilings on the maximum award per bidder at a T-bill auction are not recommended. The preferred course is for MOF to reject unrealistically low bids. MOF should guard against offering an unduly large amount at auction and rejecting bids solely because the funding is not required immediately, since this will create uncertainty for investors.

MOF should make public the results of T-bill auctions, including all supporting information, by 11 :30a.m. or, at the latest, noon on the day of the auction. Consideration should be given to publicizing the highest and the lowest accepted bid price and resulting yield.

MOF should aim for T-bill auction offers of a more even pattern throughout the year in order to contain cost of funding, subject to retaining ultimately flexibility in the interests of cash management.

MOF should consider introducing a non-competitive T bill tender based on the average price in the competitive price auction, subject to a limit of 10-15% of the accepted bids in the competitive auction.

MOF and NBS should tease out the justification for sustaining the prohibition on non-residents investing in T bills and, even if it would be premature at present to permit such investment, the authorities should settle on a future timeframe for permitting such investment by non-resident investors.

NBS's portfolio of T-bonds: If a new 2-3 year T-bond were issued at auction in 2007 when the second bond falls due for redemption, NBS could be allocated, by agreement, an appropriate amount of the new bond at the average auction price (i.e. at market rates). NBS and MOF could agree that bank will hold such T. bonds on its portfolio to redemption, unless MOF agrees to pre-redemption buy-backs.

CRSD should put in place arrangements, without delay, to provide real-time prices of FFS bond transactions to the OTC market.

MOF should modernise the tax regime applicable to private pension contributions and to pension funds. 11. MONEY MARKET

CURRENT SITUATION

2.1 An active money market is a prerequisite to developing the government securities market. Moves by investors out or back along the yield curve are usually signalled in the money market (or swaps market). It is instructive to note that the relationship between the money market end of the yield curve (the short end) and the longer T-bond end is reflected in the organization of treasury units in banks and of investment funds: they are organised as an integrated team at dealer level, with each handling a separate part of what is essentially one yield curve. An active money market supports the bond market by increasing the liquidity of securities and makes it easier for financial institutions to cover short-term liquidity needs. Moreover, efficiently determined com etitive interbank rates are the benchmark for pricing new products such as repos, interest rate swapsB and derivatives. This, in turn makes it less risky and indeed cheaper to warehouse government securities for subsequent on-sale to investors and to fund portfolio trading positions. All of these elements foster depth and liquidity in the securities market.

2.2 Liquidity surplus: Serbia's small and underdeveloped interbank money market means that banks incur high liquidity management costs. The interbank yield curve, which extends to only 6 months, is steep. Thus, on 28-09-05 "BELIBOR" rates were as follows:

Overnight Jy -2w -lm 2m 3m -6m 7.08-8.5 1% 14.28% 15.18% 15.79% 16.30% 16.81% 17.34%

The rates reflect the average of the lending transactions undertaken in the Belgrade inter-bank market.

This steepness reflects inflation presently running at an annual rate of some 15 percent a year.. Consequently, there is little activity beyond one week, with the great bulk of activity confined to the period from ovemight to one week. Banks need to hold liquidity because it is difficult and expensive for them to tap liquidity if it runs out. This can occur owing to: (a) the demanding "excess" reserve requirement of the NBS, the effect of which is that a bank must maintain every day a minimum reserve the equivalent of at least 80 per cent of the reserve requirement of 21 per cent of "free" reserves plus balances at NBS; (b) weak market price transparency; (c) poor credit risk among many banks and corporates which can reduce lending opportunities; and (d) a widespread desire among the population to hold euros. As a result, banks tend to hold excess Dinar liquidity from day-to-day. If they run short, NBS charges 2 1.25 percent per annum interest and the borrowings must be repaid on the following day.

2.3 Liquidity shortage: Despite a structured liquidity su lus in the money market, periods of liquidity shortage do occur at regular intervals - notably between 10%' and 15th of each month when VAT and other payments are made.

2.4 Limited trading: The current price-quoting system by banks is predominantly a "quote-on-request" system. On a bilateral basis, bids and offers are displayed by a small number of banks to each other. Despite the structured liquidity referred to above, spreads in the case of corporates can be as high as 300

For example, an investor may temporarily sell his bond holding by way of sale and repurchase agreement, so that the purchaser who is short of the bonds in question say, ovemight can quickly acquire the paper by paying a rate of interest usually below money market rates; and the investor who temporarily sells the bond under the repos-agreement earns additional interest. to 500 basis points due to poor credit standing. While spreads in the case of banks may be less overall, they can also be quite high in case of some banks due to poor credit standing and under-capitalisation.

Before a well-capitalised bank would lend to a weaker credit bank, it would go through the process of lodgng a loan agreement and a promissory note at the NBS which would then block the amount involved in the account of the weaker bank until the sum owing was paid. In short, interbank trading can be expensive and cumbersome. The interbank market is effectively segmented between well-capitalised banks and other banks, thereby reducing its potential depth and liquidity and malung it more costly than necessary for banks to fund themselves in that market.

2.5 NBS intervention: The overall framework for conducting monetary policy is good, but some components needed strengthening. In particular, the issue of NBS bills for draining liquidity was questioned, since it tended to segment the market between NBS bills and T-bills issued by MOF. The position subsequently greatly improved. In early 2005, NBS ceased issuing its own bills and instead conducted 14 and 28 day repos in weekly uniform price auctions; the collateral for these repos is Treasury bonds.

8%% T-bond - 2006 = SRD 3 billion - 2007 = SRD 6 billion -2010 = SRD 10.7 billion

These bonds were issued in exchange for miscellaneous government debts held by the NBS as a result of NBS funding past budget deficits. Repos settlement is effected on a real time DvP basis through the RTGs cash system and CRSD.

The foregoing situation changed subsequently. Between 23-09-05 and 07-10-05, NBS using its own bills held seven repos auctions (i.e. in addition to T. bond repos). Notable features of the auctions were as follows:

Maturities were in most cases for 60 days, and were in competition with T. bills. Offered amounts were larger: from SRD600 million to SRD3 billion, or in total SRD 9.2 billion. However, submitted bids amounted to 5.9 billion only. The last two auctions for 2.5 and 3 billion were heavily undersubscribed - which questions the efficacy of NBS's issuance policy. Issuance pushed up yields: average accepted repos rate was 15.5 percent, with most recent rates being nearly 17 percent. This exceeds the T. bill yield of 15 percent.

If NBS maintains its own NBS bill auctions it is likely that the government T. bill issuance program will be undermined and, also, future T. bond issuance could be set back.

2.6 Separation of functions: To avoid conflicts of interest money market management and national debt management should be kept separate - as was the case under 2.5 above where the only paper used by NBS was government paper (See 1. 39 ).

2.7 Interbank repos: In an attempt to strengthen interbank activity, NBS invited the banks in 2004 to respond to the possibility of entering a repos agreement based broadly on the PSA Master Agreement. However, when the NBS did not receive a generally positive response from the banks, they let the matter drop. NBS appears not to be pursuing this matter further. 2.8 Taxes: Removal of all taxes on T-bills and other government securities has assisted in the repos development, as well as sales of T-bills, most of which are held by banks.

POLICY CONSIDERATIONS

2.9 Price transparency: A number of the well-capitalised banks are anxious to develop the interbank market. This could be affected through screen-based continuous bid-offer prices among at least core banks. If this were done it would, arguably, drive forward the process of interbank price transparency, which is key to further development of the interbank market. This initiative could be followed up with action to adopt a master repos agreement between an even wider group of banks9. Approximately ten banks have in fact, considered arrangements for bid-offer pricing, the expectation being that positive results would emanate from this initiative in a matter of months. In addition, there appeared to be concern at the level of the NBS and of the Association of Bankers that banks be part of the transparency system from the outset.

2.10 How price/volume transparency is to be displayed is critical to the success of any new system. The firmer and the more continuous the prices, the more likely it is that, in the first instance, only the better equipped banks will fully participate. Each of these banks will have confidence that the others will meet their agreed obligations. For this purpose the Reuters FX information system is available in the banks generally, and is used extensively on the electronic foreign exchange market. Thus, adopting Reuters (at little, if any, cost) appeared to be the most straightforward way of installing continuous two- way firm pricing among at least a limited number of stronger banks. In October 2005 some 20 banks signed up for the new arrangement but in reality only the 10 most active banks are quoting continuous indicative prices, with size being a matter for individual banks. Despite its limitations, the World Bank team regards the foregoing as a significant step in the development of the inter-bank money market.

2.11 Action envisaged by the Bankers Association: The Association is keen that the new system be organised around its own IT platform, which is used for foreign exchange fixing. The Association's screen would display amounts only for lending by different banks and would be seen by all banks. Prices would not be displayed on a common screen: transactions would be subsequently settled by phone on a bilateral basis. The World Bank team would still urge strongly that steps should be taken to produce continuous two-way finn pricing with tight spreads: even if among a limited number of stronger banks. The two courses of action can be acted on since they are not mutually exclusive. The NBS should pursue these issues vigorously, together with an enhancement of the professional capacity of bank treasuries.

2.12 Interbank Repos: Interbank repos would upgrade the efficiency of the wider market and make the holding of government securities more profitable,(e.g. T-bills). More particularly, it would greatly assist in overcoming the current segmentation in the market and allow full interbank pricing between all banks with T-bills to take place. Once the course recommended at 2.1 1 is bedded down, it should be possible to make progress with the introduction of interbank repos, based on the PSNISMA master agreement. If introduced, this would drive the T-bill market. Consequently, MOF has a strong vested interest in ensuring that interbank repos are launched. Progress on this front, therefore, is strongly recommended; and should not remain dependent on the agreement of each and every bank.

2.13 NBS Bill Repos: As noted above, NBS is now using its own bills for up to 60 days Repos- in addition to T Bonds- in order to drain liquidity from the money market. This tends to segment the market in bills and has undermined T-bill demand. This paper recommends that NBS make best efforts to reduce the scale of its interventions using its own paper and rely instead on T bonds for Repo purposes. MOF,

The CRSD is understood to be equipped to handle interbank repos for its part should commensurately increase and also initiate SRD T bond issuance not beyond 3 years as the budgetary situation permits: a process that would assist the NBS in its efforts to reduce liquidity (see 1.40).

2.14 Reverse Repos: The NBS should consider introducing regular reverse repos of maturity of, say, 1 week or less, at the time each month when structural liquidity shortages occur.

2.15 Inclusion of government securities in eligible assets constituting the excess reserve. NBS should give consideration to the inclusion in the excess reserve requirement constituents of T-bonds, when issued, in order to afford positive assistance to T-bond development. Logically, T-bills would be also included. On the other hand NBS might deem it necessary, in consequence, to increase commensurately the existing reserve requirement, although they are already quite high.

2.16 Deposits in money market by local authorities: Municipal/local authorities have substantial unspent cash balances, some of which they deposit with commercial banks. Under new Public Debt law Local Authorities are restricted in placing deposits with banks when the yield is less than NBS's discount rate. It is unclear what other criteria, including risk considerations, the local authorities adopt in placing deposits with different commercial banks. Transferring large sums to, say, a small number of banks will give those banks a funding advantage. It is recommended therefore that MOF issue guidelines to local authorities on the procedure that they should follow before placing deposits with commercial banks, with particular regard to the tender process for seeking deposit quotations and allowing for the credit status of different banks. The aim should be to diversify placements, spread risk and maximize returns.

RECOMMENDATIONS

(a) Priority issues:

1. NBS Repos: NBS should make every effort to curtail the use of its own securities to drain liquidity. To assist the Bank in this task and also to help develop the , MOF for its part should, start the process of issuing SRD T. bonds as quickly as the budgetary situation allows.

2. Price transparency: The most active banks are displaying continuous indicative bid-offer prices electronically. The NBS should actively encourage this development. If banks generally wish to display separately the volume of their lending on the IT platform of the Association of Bankers, they could also do so. Both courses of action can be adopted since they are not mutually exclusive.

3. Interbank repos: Once the foregoing process are satisfactorily bedded down, it is strongly recommended that repos based on the PSA 1 ISMA master agreement be introduced among as many banks as possible, in order to widen and deepen interbank lending activity, as well as the T- bill market. Progress in launching interbank repos must not be permitted to be held up by the weaker banks.

(b) Other issues:

NBS reverse repos: NBS should introduce regular reverse repos of one week or less maturity when structured liquidity shortages occur in the money market. Government securities in NBS reserve requirement: NBS should afford consideration to government securities being included in the constituents of the excess reserve.

Local authority deposits with banks: MOF should issue guidelines to local authorities generally on tender procedures to be adopted in placing cash with banks, having regard to the need for competitive pricing, risk aversion (depending on the credit status of a bank) and diversification of placements, together with the latest legal constraint on local authorities which cannot place deposits with banks unless the yield obtained exceeds the NBS discount rate. 111. CASH MANAGEMENT

CURRENT SITUATION

3.1 What is cash management and why it is important: Efficient cash management is the series of processes used by an organization to obtain the maximum benefit from its flow of cash funds. The underlying objective of cash management is having enough cash available as and when it is needed. Sound cash management involves better timing of expenditure decisions, earlier collection and banking of revenue, and more accurate forecasts of cash flows. Thls helps minimize the cost of any borrowing that is necessary and, if possible, facilitates investing surplus funds to achieve the best return overall.

3.2 Degree of accuracy - MOF and NBS needs: The degree of accuracy required of cash management - daily, weekly, monthly, and quarterly - depends on the needs of both the MOF and the NBS. There is no fundamental conflict in these needs at similar stages of evolution toward a market-based system. Three policy issues influence the desired degree of accuracy in cash management: (i) the MOF's need to avoid over- and under-funding; (ii) the NBS method of operating monetary policy; and (iii) the desire to develop a liquid money market (which is usually closely dependent on (ii))..

3.3 MOF's needs: The Treasury Department of the MOF requires cash management to minimize idle balances (over-funding) and to minimize use of central bank credit (under-funding). MOF's Budget Department, which monitors budgetary expenditure and revenue, may see cash management as simply an out-growth of a system of budgetary control. The tighter the time-scale on which cash can be managed the greater will be the gains in terms of efficient funding, investment of surplus funds, and real-time budget control. However, there is an administrative cost involved in terms of more advanced information systems, staffing, and developing behaviour within the spending and revenue departments of government. MOF may decide, for example, that the costhenefit for its purposes, given the degree of systems development and inter-departmental coordination, is to have accurate cash management monthly, rather than weekly or daily. Notwithstanding this, fine-tuning the evolution of government cash balances at the Central Bank may still not be feasible. MOF may need to resort to short-term market placement of cash, especially in Serbia where Central Bank overdrafts are proscribed and the Bank pays only 3% on deposits in comparison with the yield on T bills of some 20%.

3.4 NBS's needs: The NBS's need for cash management depends on the method it is using to implement monetary policy. Using direct methods of monetary control (credit ceilings, reserve ratios, interest rate controls) the central bank is unconcerned about the accuracy of cash management. However, when the central bank shifts to indirect methods of monetary control (IMC) as is the case in Serbia, it has an interest in the accuracy of cash management as its method of implementing policy requires the use of to control some definition of base money. Government's cash flows to and from accounts at the central bank affect the quantity of base money. The government's cash flows to and from accounts at NBS have a much bigger impact on the supply of excess reserves than they do on the larger quantity of base money. Accurate daily cash management is therefore very important for accurate daily control of excess reserves by the NBS.

3.5 Development of a liquid and stable money market: The behaviour of the money market depends on the interaction between the accuracy of the central bank's liquidity management; the design of the central bank's accommodation policy; and the length of the reserve maintenance period. However, unless liquidity management is accurate, fluctuations in excess reserves will result in fluctuations in the overnight rate which, in extreme cases, cause uncertainty. The ability of the central bank to set accommodation and reserve policies to encourage market development is constrained by the quality of cash management. 3.6 The cornerstone of efficient government liquidity management is: (i) accurate short-term forecasting of revenuelexpenditure (ii) a single government account and (iii) efficient payment system. These considerations are addressed below.

3.7 Revenue-expenditure forecasts: For budget execution purposes, MOF operates a system of control in which a key role is played by quarterly forecasts of expenditure and revenue. Within each quarter's forecasts, the Treasury Department of the MOF is responsible for budget execution, which the Treasury carries out each day: it authorises routine expenditure commitments of 160 entities if the appropriations sought are within the relevant budgetary provision. Approval is notified to the Public Payments Agency (PPA) which credits electronically the bank accounts involved. The PPA controls all publicly endowed bank accounts. Underpinning the quarterly forecasts are rolling monthly forecasts of expenditure and revenue. This enables the Treasury to plan disbursements month-to-month within each quarter. These arrangements afford tight control over government expenditure from its centralized payment process. However, it is not a system that focuses on increasing certainty around the timing and value of receipts and payments - which is the essence of cash management.

3.8 Bank Accounts are not maintained by ministries. Instead, all ministerial disbursements are made, electronically, via the PPA to the single consolidated budgetary account (at the PPA). The thtnking behind this measure is that - apart from exercising control over expenditure - it is intended to ensure that payments are made only when due and not before. The situation in the case of "indirect" budget users is, however, different. These entities (for example, schools, hospitals, local authorities, etc) have their own bank accounts that are fed by PPA with authorised appropriations, as sought by the indirect users. The Treasury does not have a mandate to control actual commitments of the 160 spending entities with respect to reports for local authorities. In this process, large unspent balances, or idle balances, can accumulate in the various accounts in anticipation of disbursement.

3.9 A consolidated account for the Republic's budget, together with separate consolidated treasury accounts for other public entities, such as local authorities, the social insurance fund and other public bodieslfunds, are held by the PPA at the NBS. There is not a single consolidated account for all the public bodies. It appears, however, that the net amounts remaining in the different "consolidated accounts are treated in effect as one set-off account (e.g. for interest purposes). Some of the proceeds are invested in commercial banks (e.g. local authorities) but the bulk of the funds are invested in the NBS. The government is prevented from placing deposits with commercial banks due to opposition from NBS: see 3.12 (iv) below. In this connection, it may be noted that the PPA separates by the close of play each day's balances owned by the Republic and the other public entities and, as a result, the Authority has been able to transfer successfully all government cash balances fiom the commercial banks to the NBS. The aggregate of the net balances in the different consolidated accounts at 31.12.04 was SRD 39.6 billion; and, by 15.03.05, it had increased to SRD 49.6 billion - which is greatly in excess of the government's 2005 borrowing requirement. The idle balances of the Republic were SRD 2.73 billion as of 31.12.04 and, by 15.03.05, SRD 4.5 billion. Again, this amount is greater than 2004 net T bill funding, at an average cost of the order of 20 percent. Whtle T. bill costs have fallen recently to around 15 percent, this cost is greatly in excess of the rate of return (2.55 percent a year) earned on deposits at the NBS, which is a compelling reason for MOF to pursue more efficient liquidity management.

3.10 Treasury System: MOF expects that the new permanent treasury system will (with the ongoing active assistance and advice of an EY funded Belgian consultant) be operational in 2006 or early 2007 at the latest. It is expected the new system will: i) introduce a more efficient payments system by streamlining operational arrangements between the PPA and the Treasury Department - incorporating full electronic data reporting, together with full integration of the payment and encashrnent reporting process, to the Treasury; ii) have a much improved "understanding" of cash movements to identify fixed or variable and seasonal or periodical components; iii) enable control to be exercised over the present inexorable growth in idle cash balances of "indirect" budget users by abolishing their separate bank accounts at DPP (i.e. as was done already in the case of ministries).

In short, the new system is expected to have a strong cash management module and will allow of PPA activities in preparatory work of government payments and government accounting to be fully integrated with the Treasury Department of MOF.

POLICY CONSIDERATIONS

3.11 New Treasury system to enhance cash management: The new system is expected to improve cash management and make it more effective than at present. In particular, direct transfer from MOFIPPA to all final recipients, including "indirect" budget users, without intermediate transfers between various accounts or sub-accounts should result in a marked reduction in outstanding idle cash balances. However, the Treasury needs to take formal control of, and manage, the overarching net account at the PPA which consolidates end-of-day balances of the various consolidated Treasury accounts of the different public entities.

3.12 Active management of over-funding shown up by the forecasting process or, ultimately, movements in idle surplus balances would enable the MOF to reduce T-bill issuance or, alternatively, to invest surplus funds on a temporary basis. In other words, the Treasury would, ideally, have the ability to transact in the market to manage over-funding. There are a number of considerations surrounding the use of T-bills in Serbia for government cash management: i) 3 month or 6 month T-bills as a tool of daily or even monthly cash management is likely to be effective to a limited extent only; ii) Active cash management would require the Treasury to issue T-bills of not more than one-month maturity on a periodic basis and, then to conduct daily money market type operations. The Treasury is not technically in a position to conduct such operations at this juncture. iii) While some reduction in the existing T-bill programme must remain open as an option for cash management by MOF, reductions should not be camed out in an apparently arbitrary or capricious manner. To do so would damage the T-bill programme in terms of development and cost of funding. Thls is because the market prefers a steady and predictable flow of securities throughout the year. This even flow makes it easier for the investor market to absorb the supply of debt. It is, therefore, less risky for the MOF. iv) Issue ad hoc short-term T bills, with their maturity and issuing calendar adopted to immediate cash or liquidity needs. These so-called "cash management Treasury bills" have been used in France and the USA: they do not follow a pre-announced calendar and are not for standard periods. In the absence of a deep and active T bill market, ad hoc paper could cut across, if not undermine, regular debt management T bill issuance. The foregoing factors would need to be balanced against the undoubted savings from a reduction in over- funding using T-bills.

Alternative courses involving short-term market placement would be to invest surplus funds in either commercial bank deposits or in other sovereign debt. The former is probably not on the agenda since, if it were done, NBS would feel constrained in the context of an on-going liquidity surplus in the banking sector to freeze any commercial bank deposits made by government. Investment in short-term securities of a sovereign, such as Germany The rate of return earned in comparison with the 2.55 percent payable by NBS might not justify the costs and risks entailed. Politically also it might prove unacceptable. In the short run, the most straightforward course is, for NBS to pay interest rates closer to market rates, i.e. the average of the overnight lending rate on the inter-bank market, even though this would reduce the Bank's profits. The advantage of using the NBS deposit is that it can be speedily accessed and changed. Consideration should also be given to opening standard credit lines with selected banks, which could be used to meet urgent Government cash needs.

3.13 Improve cash flow forecasting: Current monitoring of the Budget is restricted to monthly intervals. More accurate monthly cash flow forecasts would benefit NBS in its management of money market liquidity and MOF in traclung potential movements in its surplus balances and also in the potential use of T bills to reduce or to contain the over-funding. However, more accurate monthly cash flow forecasts will depend heavily on two factors: (i) introduction of the new permanent Treasury system and (ii) because of the of expenditure and of revenue flows, up-to-date information obtained from the different agencies to supplement the application of historical patternsltrends to annual Budget estimate forecasts and also to monthly and quarterly forecasts.

3.14 Forecasts for one to five days ahead are essential in a well-developed cash-management system, that is, in order to take account of variances in cash flows that occur during the course of each month. However, such short-term forecasts are notoriously prone to error and to achieve a reliable degree of accuracy would require the application of very considerable, expertise and organisation. Serbia's Treasury is making steps to progress matters including the introduction of forecasting under a new permanent Treasury system. Nevertheless, whether or not the new system, when operational, will in fact enable the production of accurate one day cash flow forecasting remains open at this stage

3.15 The Treasury should routinely follow up all significant errors in an effort to identify and remedy systemic problems. Based on past experience of forecasting errors, agency performance could be improved through training that stresses the importance of the work of processing staff, and a more thorough approach to documenting:

key processes and payment dates so that agencies are not caught unawares by staff absences;

the information flows and lead times in the making of payments by various mechanisms;

implications of public holidays and of payment dates falling on weekends; and

new administrative arrangements that will be necessary immediately after amendments to legislation affecting significant payments e.g. VAT.

3.16 Public transparency of cash flow forecasting process: The market is at present left unaware of how the government's funding is progressing during the year. As a result, there is market uncertainty in the current year, for example, surrounding whether or not a T-bill issuance programme will be maintained. This is potentially damaging to the development of the securities market and could make future funding more difficult and costly. The MOF should, therefore, publish at the start of each quarter forecasts of revenue, expenditure and net borrowing for the upcoming quarter, whlch it should roll-over, in the light of each quarterly outturn by way of revised forecasts for the ensuing quarter(s). Outturns should be published without undue delay.

3.17 Smooth the pattern of revenue and expenditure: This is a longer-term project and in the case of revenue would require amending legislation. A basic objective of better cash management is smoothing revenue and expenditure flows, so that the MOF's need to borrow is spread more evenly over the year. This task could, it is recommended, be put in train once the new permanent Treasury system is operational.

3.18 Separation of operational controls: Again, this is another long-term project. There is an increasing trend internationally for governments to separate the functions of "authorization and financial control" (i.e. the accounting and financial controller functions) from "confirmation and settlements" (i.e. the back office functions). In Serbia, the Treasury is required to authorize disbursements under appropriation, then to authorize the transfer or payment, and finally to sign-off on the financial statements. While in no way wishing to attenuate the huge progress made in recent years by the Treasury, the fact is that the foregoing practice would be totally unacceptable in a financial institution and is no longer considered sound practice for a government treasury or debt management operation. What is recommended is a clear segregation of functions responsible for the authorization and accounting of government's finances and the settlement and physical payment of government's cash management transactions.

3.19 Strategic cash management project: Once the proposed permanent treasury system is operational, it would be desirable for MOF to consider establishing a taskforce that would: (i) document in detail the up-to-date position of its cash management arrangements, including the sectors in the various agencies that are contributors to or beneficiaries from the Budget; and (ii) establish the strategic measures required in order to optimise day-to-day cash flow management over the long-term, including the last two recommendations referred to above. Clearly, such a project would need to be conducted on a cooperative basis in partnership with the NBS, PPA and the many agencies reporting to it, together with the Revenue agencies. The project would entail considering organizational, procedural and legal changes but also cultural change in organizations to drive home the goals and value of efficient cash management.

RECOMENDATIONS

(a) Priority issues:

1. Accurate short-term cash flow forecasting (for, say, 1 to 5 days), is essential for efficient and effective cash management. The proposed new Treasury system may meet this need, although at this stage the issue remains open.

2. Management of ongoing over-funding shown up by the forecasting process could be contained by restricting the use of T-bills, and by investing surplus balances in, say, commercial banks or short-term paper of a AAA rated sovereign. Alternatively, MOF could open standard credit lines with a few select banks, thereby reducing the cash surplus required at NBS; and NBS should pay market related interest to the government on its surplus balances with the bank: specifically, it should pay the average of the overnight lending rules in the interbank. As a risk avoidance measure and to sustain government securities market development, a regular T-bill programme should be maintained. 3. There should be a separation of operational controls: segregation of functions responsible for (a) authorization of accounting of Government finances, and (b) settlement and physical payment of Government cash management transactions.

(b) Other issues:

Treasury needs to take formal control of and manage the overarching net account (at the PPA) that consolidates end-of-day surpluses on various consolidated treasury accounts of the Republic and other public entities. This may require amending legislation.

More accurate monthly cash flow forecasts - depending, critically, on the operation of the new permanent Treasury system and continuous and up-to-date information flows from different agencies to supplement the application of historical patterns1 trends to annual Budget estimate forecasts and to monthly and quarterly forecasts.

Treasury should routinely follow-up all significant forecasting errors in an effort to identify and remedy systemic problems in accurate forecasting.

Treasury should publish at the start of each quarter forecasts of revenue, expenditure and net borrowing for the upcoming quarter, which it should roll-over in the light of each quarterly outturn by way of revised forecasts for the ensuing quarter(s). Outturns should be published. r Efforts should be made to smooth the pattern of revenue and expenditure: This is essentially a long- term project involving amending legislation.

A strategic cash management project should be undertaken: Once the new permanent Treasury system is operational, it would be desirable for MOF to consider establishing a taskforce that would document, in detail, the up-to-date cash management arrangements, and would establish the strategic changes required to optimize day-to-day cash management in the long-term, including the last two recommendations referred to above. The task-force would have representatives of MOF, NBS, PPA and agencies reporting to it, together with the Revenue agencies. Financial Sector IIiL W31ILU U/\NK

1818 H Street NW Washington DC 20433 www.worldbank.org/finance

The findings. ~ntrepretations,and conclus~onsexpressed in this paper are entirely those of the authors and should not he attributed in any manner to the World Bank, to its affiliated organizations, orto members of ~tsBoard of Executive Directors or the countries they represelit The World Bank does not guarantee the accuracy of the data included ~nthis publ~cat~on a~idaccepts rlo respons~t~~l~tywhatsoever for any consequence of thelr use