Policy, Research, and External Affairs| | International Trade International Economics Department The World Bank August 1991 WPS 739 Managing Financial Risks in Papua New Guinea An Optimal External Debt Portfolio Jonathan R. Coleman and Ying Qian Commodity-linked bonds issued with payments linked to the prices of oil and cocoa could significantly improve Papua New Guinea's risk management. The Policy, Research, and Extemal Affairs Complex distributes PRE Working Papers to dissemnuate the findings of work in progress and to encourage the exchange of ideas among Bank staff and all others intcrcsted in dcvclopmcnt issues. nhe.sc papers carry the names of the authors, reflect only their views, and should be used and cited accordingly. The findings, interprceations, and conclusions are the authors' own. They should not bh attributed to the World Bank, its Board of Directors, its management, or any of its member countrics. Ptc,Rsarch, and External Affairs International Trade WPS 739 This paper-- a product of the International Trade Division, International Economics Department-is part of' PRE's research on the use by developing countries of financial instruments linked to commodity prices. Copies are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Julie Carroll, room S7-069, extension 33715 (30 pages, with tables). Papua New Guinea is vulnerable to instability conventional debt denominated in different and uncertainty associated with fluctuating currencies. They judge he hedging effective- commodity priecs. This is because its GDP, ness of this portfolio by ilow much the variance export earnings, and government revenues of expected real imports is reduced. depend largely on sales of a small set of primary commodities whose prices fluctuate substantially The results indicate that commodity-linked oIn lhe international market. Papua New Guinea bonds could play an important role in Papua is also exposed to fluctuating exchange rates. New Guinea's risk management strategy. The The degree of exposure depends heavily on (1) proportion of commodity-linked bonds in the how the currency composition of net export optimal debt portfolio ranges from 20 percent to earnings match the currency composition of net 45 percent for real interest rates of 8 percent to I liabilities and (2) how changes in commodity percent. They show that commodity bonds prices affect exchange rates. issued with payments linked to the prices of oil and cocoa could substantially lower the variabil- Based on these criteria, Coleman and Qian ity of expected future imports. show that Papua New Guinea's assets and liabilitics may be poorly balanced for debt Their results also show that Papua New servicing. Thus, it could benefit substantially Guinea's external debt structure is no, well from active risk management, especially through balanced to hedge the foreign exchange risk bettcr selection of the financial instruments in its from the existing composition of non-U.S. debt portfolio. dollar-denominated liabilities. The debt portfo- lio contains an excess of Japanese yen- and Coleman and Qian present a model and Deutschemark-denominated liabilities, while estimate of an optimal debt portfolio that allows liabilities denominated in British pounds are for the use of commodity-linked bonds and substantially underrepresented. Thc PRE Working Paper Scries disseminatcs the findings of work under way in the Bank's Policy, Rcsearch, and Extemal Affairs Conplex. An objectivc ofthc scries is to get thesc findings out quickly. cven if prcsentations arc Iess than fully polished. The findings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the PRE Dissemination Center Managing Financial Risks in Papua New Guinea: Optimal External Debt Portfolio by Jonathan R. Coleman and Ying Qian Table of Contents 1. Introduction ................................................................... 1 2. Commodity Price Risks ................................................................... 2 3. Foreign Rate Risks ................................................................... 5 4. Theory ................................................................... 7 5. Estimation ................................................................... 13 6. Results ................................................................... 18 7. Conclusions .................................................................... 23 Appendix I ................................................................... 25 Appendix II..................................................................... 27 References ................................................................... 30 The authors wish to acknowledge the valuable suggestions and comments from Taka Akiyama, Stijn Claessens, Ron Duncan, Chris Jones, Don Larson, and Lloyd Kenward. 1. ~~~~~~~~~~~~~~~~~~ Throughout the 1980s, Papua New Guinea (PNG) increasingly turned to extenal borrowing to finance its current account dcflcits. As shown in Table 1, PNG indebtedness increased more than four- fold between 1980 and 1989. In contrast, the Increase in debt has not been matched by increases in GDP and export earnings, which depend, to a large extent, on sales of a small set of agricutural and mineral commodities. Moreover, given the large fluctuations in primary commodity prices on the interational market during the 1980s, the foreign exchange earnings of PNG from exports have tended to be unstable (Table 1). Table 1.- Measre of lndebteness of Ptapu New Guinea. -1280-8 (Million US Dollar) 1980 1983 1986 1987 1988 1989 Total Debt Stock (TDS) 720 1,860 2,420 2,700 2,270 2,450 Total Debt Service 150 290 495 510 540 530 Gross National Product (GDP) 2,460 2,220 2,450 2,940 3,320 3,650 Export Eaing (EE) 1,090 950 1,190 1,400 1,720 1,670 Current Account Balance -310 -380 -100 -370 -160 -200 TDS/GDP(%) 29 84 99 81 68 67 TDS/EE (%) 66 196 203 192 132 147 Source: World Deb WTables 1989-90, Tbe World Bank. In addition to the instability associated with fluctuating commodity prices, PNG faces additional financial risks associated with sharp movements of exchange rates. This is because a large proportion of its external debt is denominated in non-US dollar currencies and because significant movements of the US dollar vis-a-vis other mabJr currencies in excess of interest differentials have taken place in recent years. Thus the currency valuation effect (i.e., unanticipated deviations from interest rate parity - see 2 section III) on PNG's external debt is considerable. The fluctuations in the level of US dollar measured debt stock are likely to continue, given that the volatility of cross-currency exchange rates is not expected to decline and that the level of new borrowing by PNG in non-US dollar currencies will remain high, given the current pattern of international capital availability. This paper has three major objectives. The irst is to analyze the two financial risks faced by PNG-exchange rate and commodity price risks-and to assess to what extent these risks create serious problems in terms of external debt management. Given that there may be scope to manage these risks better by re-shaping the debt portfolio to include commodity-linked bonds and by altering the currency composition of conventional debt, the second objective of this paper is to present a radonal expectations model-extending Myers and Thompson (1989)-which solves for the optimal debt portfolio (including conventional debt denominated in different currencies and commodity-linked bonds). The third objective of the paper is to compare the optimal debt portfolio (derived from the model) with the actual debt portfolio of PNG, and, on the basis of this comparison, to make some broad suggestions as to how the current debt structure might be altered in order to manage commodity price and exchange rate risks more effectively. The paper is organized into the following sections. The risk exposure to commodity price fluctuations and to exchange rate movements is discussed in sections 2 and 3, respectively. In section 4, a rational expectations model is presented which can be used to solve for the optimal portfolio of extemal debt; and in section 5, the estimation procedure is discussed, and in section 6 the results from estimating the model are reported and discussed. Finally, in section 7, some conclusions are drawn. 2. Commodity Price Risks In recent years copper, gold, coffee, logs and palm oil have accounted for over 80% of total export revenues in PNG. Furthermore, the contribution of these five major export commodities to total 3 export earnings has been increasing over time (see Table 2). This dependence on the price and volume performance of a small set of commodities is likely to continue into the 1990s. "ar instance, the current mineral and petroleum expansion and exploration programs will further increase the dependence of PNG on copper, gold, and crude oil exports'. Table 2. Contribution of Major Primary Commo./ity Exports to- Total Export Earnings. PNG. 1985- 1992. Commodity 1985 1986 1987 1988 1989 19901 19911 1992' Percent Minerals 46.5 60.9 67.7 70.6 69.0 65.5 64.1 64.7 Gold 25.4 40.2 41.5 36.4 25.1 31.2 34.3 40.1 Copper 21.1 20.7 20.2 34.2 43.9 34.3 29.8 24.6 Nomninerals 53.5 39.1 38.3 29.4 31.0 34.5 35.9 35.3 Cocoa 7.5 6.3 5.3 3.5 3.8 3.2 2.6 2.6 Coffee 13.4 15.4 16.5 9.4 10.9 9.2 7.2 6.9 Copra 5.1 2.0 1.0 1.3 1.4 1.5 1.4 1.2 Logs 7.7 5.7 6.9 7.6 6.6 8.9 10.1 9.6 Palm Oil 7.5 3.9 2.8 1.7 3.1 4.3 5.4 5.6 Other 12.3 5.8 5.8 5.9 5.2 7.4 12.2 9.4 Source: Based on Table 11 in Annex VII of IMF Report on PNG, March 1990. 1/ Projected. Figure 1 shows the annual percentage change in export unit values of the five major commodity exports of PNG. The coefficients of variation2 (CV) of these export unit values between 1977 and 1988 'PNO's fust oil project is likely come on stream in t second half of 1992, with a production profile that is heavily front-end loaded. ITe standae deviation aa a proportion of the mean. 4 are: copper 42%, gold 50%, cocoa 37%, coffee 34% and logs 25%. Unstable export prices have resulted in unstable export revenues, which recorded a CV in excess of 27% during the 1980s. This indicates a high degree of instability, even among developing countries, and presents a significant problem in debt servicing, as well as affecting PNG's ability to secure additional loans. While the export performance of the AnnualI Percentage Changes country depends on the price and PNG Export Prices Copper volume of a small set of Gold 100 . ... .. .. .. .. .. .. .. .. . . .. . .- - - - -------------/ -coa commodities, thie mix of impor s has./\ ... comraodities~~ ~~~~~,.te m xo i p r, a 50 -A .\v...... ,,............................... .......Cfe been more diversified and the unit u *-,- -..7$.. values of imports have been relatively more stable. -.00- Various instruments have 1977 1978 1979 1980 1981 19182 193 1984 1985 1986 1987 1988 Year been used by countries attempting to Source: IECCn. The World Banl Figure 1 manage commodity price risk. Among these are: (a) buffer stock schemes, commodity stabilization fund schemes and macroeconomic policies; (b) financial market instruments such as futures, forwards, options, swaps and commodity bonds; and (c) international commodity agreements and compensatory financing schemes. PNG has extensively used instruments listed under (a) and (c) above to manage its risk exposure, with moderate success3. Claessens and Coleman (1991) report that the financial market instruments listed in (b) above have not been used in PNG. ' The Minerals Resources Stabilization Fund (MRSF) was established in 1975 to prevent unstable minera!-based tax revenues fm causing instability in the rest of the economy. However, Claeuens and Coleman (1991) show that the coefficient of variation of out-flows from this fund is only slightly lower than the coefficient of variation of its in-flows, indicating only moderate effectiveness in stabilization. PNO operated conunodity price stabiizaion funds for its major agricultural export commodities (i.e., coffee, cocoa, copra and oil palm). These funds were successful for many yeam and often cited as models of how effective a well managed scheme can be in risk management. However, with the sharp fall in the prices of these commodities in recent yean these funds becaen depleted and new approaches to commodity price risk management for agriculture san being sought. PNO was a member of the Intemational Coffee Organizstion. 5 Among the financial instruments associated with primary commodities listed above, commodity- linked bonds have substantial potential for PNG which relies on commodity export revenues to meet its debt obligations. Coomodity-linked bonds differ f&om conventional bonds in that coupon and/or principal payments are 'nked to a given quantity of a commodity. For example, a gold-denominated commodity bond might require coupons payments to be made annually, equivalent to the value of a pre- specified quantity of gold, with the price set at some average daily price over the preceding 12 months. A decline in the price of gold leads to lower coupon payments and vice versa, with the result that debt servicing requirements are better matched with the ability to pay. 3. Foreign Exchange Risk. In addition to commodity price risks, PNG is vulnerable to risks associated with exchange rate fluctuations4. The degree of exposure to exchange risk depends, to a large extent, on (i) the matching of the currency composition of net export earnings with the currency composition of net liabilities, and (ii) the effects on exchange rates of changes in commodity prices. Based on these criteria, Claessens .t al (1989) showed that the assets and liabilities for debt servicing in PNG were potentially poorly balanced. An appreciation of the Kina vis-a-vis the currency of a country with which PNG has negative trade balance causes PNG goods to become more expensive. This can lead to a decline in the export revenues from this country. However, if PNG borrows from this country, then debt servicing requirements also fall with a depreciation, so that the overall impact of an exchange rate change may be diminished. For this reason, Claessens Z_dL argued that PNG should secure conventional loans 'Movemens in os-curmncy exchange rats can be expeced to compensae for nominal inrst tste differtals (uncovered intres rate parity), except for risk prmia (ox-ante devations from uncovered inters rate parity). However, oxkno , deviations from uncovered inrest rate parity are unobserved. Therefore, Active curency managemcnt should be employod to reduce risks of deviations through optimal curency conposition. In th model, all variables (denominated in nominal US dollars) am defled by PNO's iport price index (based on US dollars). Thus the model meaue the currency valuaion effect in tenns of units of inpoita. 6 denominated in the currency of those countries with which it has a positive trade balance. As well, it should try to borrow from countries whose currencies move positively with the prices of the major exports of PNG. Perhaps after PNG starts exporting oil it should obtain loans from those countries whose currency value is positively linked to the price of oil, such as the United Kingdom since it exports oil. As the price of oil increases, the pound sterling tends to get stronger, leading to higher debt obligations in terms of PNG's currency (Kina). However, as a result of higher oil prices, revenues from oil exports also rise, allowing the higher debt repayments to be met. In contrast, lower oil prices tend to strengthen the Yen, leading to higher debt repayment, which must be met by lower oil export reveaues, making the Yen a poor currency in which to borrow. As demonstrated in Table 3, the trade pattern and currency debt composition of PNG are not well matched in terms of hedging risk. The most striking imbalance is in the shares of the United States. If the US dollar appreciates, PNG's debt repayments will be higher without any significant offsetting effect of increased foreign excbnge earnings from additional sales of exports to the United States. 7 Table 3.CmparIson of Shajre of Key Currencies an PNGL Public and.-Publicly Guaranteed Debt with Shar of Ttal Eots by Malor EXDnot Destination 1982-87 Country 1982 1983 1984 1985 1986 1987 Percent Export Earnings Australia 8.4 7.5 8.4 10.3 4.8 8.0 Germany, F.R. 26.9 25.7 21.3 30.0 34.4 27.2 Japan 34.0 35.3 29.3 22.5 25.6 28.5 United Kingdom 5.8 5.8 11.3 7.5 4.5 4.7 United States 1.8 2.2 2.7 3.9 3.3 2.1 Total 76.9 76.5 73.0 74.2 72.6 705 - Percent Debt shares Australia Germany, F.R. 3.1 1.9 1.9 2.1 2.3 2.3 Japan 3.4 3.0 8.1 12.9 14.5 16.2 Switzerland 3.4 2.5 4.2 7.2 9.4 10.3 United Kingdom 2.4 1.7 1.2 1.2 1.8 2.0 United States 40.6 47.1 45.1 38.9 35.6 28.6 Total 52.9 56.2 60.5 62.3 63.6 59.4 Source: World Bank Debt Reporting Service; Papua New Guinea Onortunities and Challenis for Accelatd Development, World Bank Report No. 7707 PNG, April 1990. 4. Theoq The model used to determine an optimal hedging strategy for PNG is that of aggregate consumer portfolio choice for a given output. It is an adaptation of the model presented by Myers and Thompson (1989). In conformity with the model, PNG is a small open economy with all its external debt issued by the government. The government has a utility function, u(m,), where m, is the real value of imports of goods and services in period t. The utility function is assumed to satisfy the Von Neuman- Morgenstern axioms, as well as u'(mn:)0 and u"(mJn0. 8 In the absence of borrowing, the current account must balance in each period, with the value of exports equal to the value of imports and debt servicing requirements. However, the government can borrow externally by taking out a conventional loan at a real rate of interest r. In addition, the government can borrow by issuing commodity bonds which mature in one period and require a payment at the beginning of the next period equal to the price in the next period of the commodity to which the bond is linked. Given these conditions, the government faces a budget constraint given by: m +(I +r)(ed;l_ +du-s) +,ptlbtl sx,+(ed 'd +d ` Xs a w, bt 1 where: m, = real value of imports, r = real interest rate5, e, = column vector of exchange rates of non-US dollar denominated debt in units of real US dollUars per other currency, (el,, e2,, ..., e,)', de = column vector of conventional debt of each (non-US) currency, (dl,, d2,, ..., d,)', dt = conventional debt denominated in real US dollars, p, = column vector of real prices of the underlying commodities, (Pht P2t, ---, PJ%)' b, = column vector of the quantity of bonds sold which are denominated in physical units of the commodity, (blt, ba, ..., b)', x, = real value of exports, and w, = column vector of real prices of the comunodity bonds, (wl,, wa, ..., wwy. Equation (1) can be rearranged as: m,+(1 +r)dt-l +[(I +)/g (2) 59 z+d;+[e/,w,][ or, 0~~~~~ ~~(3) Mt+*.! *rd pbixt+dtn +w, b, where: p, = [(1 +r)e',,p'J' 'The model proposed here assumes uncovered interest parity holds. Covered interest parity is not assumed because long term (i.e., more than a year) forward exchange markets do not exist. 9 WI = dd't,1, b'W,J', and w*, = [Oo',w'J'. The governmenat also faces transversailty conditions{: IIm(1 +r)-'d"=1Ilm(1 .r)'w~ 4b,=0 (4) The agent's problem is to choose a portfolio of commodity-linked bonds and conventional debt in different currencies to maximize the expected life-time utility function (5): Eoz 'ug(^t) (S) i.0 subject to (1) and (4), where ,B is the subjective discount rate. The associated Euler equations are.: us '(m,)w - r E,(u k(m1)'pte1) = (C) Solving this maximization problem requires that strict assumptions be placed on the fbrm of the utility function, as well as on the probability distributions of prices and exports. Taking the approach of Myers and Thompson, a solution can be found by using the permanent income theory of consumption. The optimal import path can be defined as?: ftin eonJition ensure that bornwing will not increase the present value of net wealth. This condition is used also in Appendix 'A drivation and diseusion of this equation is given in Appendix 1. 10 M,=-[s (1 +r)*E(x.,i-pt ba,,-(i .r)d, "D (8) The equation 8 is not a decision rule because the Et(x,+i) terms cannot be observed. To retrieve it, EA(x,+,) can be expressed as a function of variables which the government can observe in period t. One procedure is to set up a vector y, in which the first element is x,. That is, yt = (x,, p,, st)', where p, is a vector of commodity prices, and s, is the set of other state variables useful for predicting future exports. The vector y, is assumed to follow the autoregressive process: A(L)yt=et (9) where A(L) is a matrix polynomial in the lag operator and e, is a zero mean, serially uncorrelated error vector, with covariance matrix G. The optimal projection of the future income stream through exports can be defined as (see Hansen and Sargent (1980)): (l e-- -It(X, ,) =Y IYt+B(L)Yt ls' (10) -so with, 'Y =X '4A(-1-Y (11) and, BtL)=4A( +-l (l +pY AJLS 1 (12) where: < = a row vector with a one in the first column and zeros elsewhere. Substituting (10) into (8) gives the operational decision rule: I -ps b,l -(I +r)drjl (13) with imports as a function of observable variables. However, while this formula gives the optimal level of imports and therefore the optimal level of debt, it does not provide the optimal portfolio of conventional and commodity-linked debt. To obtain the optimal debt portfolio further derivations are needed. Rearranging the Euler equations (6) and (7), and assuming that the expected real return on holding bonds is equal to the real interest rate; gives: 1 .r E,am,l(w. --)=O O 14 -) 1 ; r Which implies: COVQ4 '(m )ae/) ( (16) In order to calculate this covariance matrix, it is necessary to obtain an expression for u'(m,+1). Following Myers and Thompson, a linear approximation of the first derivative of the utility function gives, coV(M,,,pt*,1) (17) Leading (13) one period and computing the relevant covariance (17) gives: 12 CO1V(( r (y 'y, +B(L)yt-p +ib;-(I+r)d ))p)f1)=l= (18) 1 +r r+ & Recognizing that B(L)y, is known at time t, the covariance expression is: co lYt.I -Pt'."Io) + ) = (19) Rearranging (19) gives': ,, =0 (20) Where: p, = covariance operation between vector pe and y, and Gp.l = covariance operation between elements in the pe vector. Solving for b,' gives: b9= aX DPyY (21) While the optimal portfolios have been computed, it is important to determine whether the variance of real imports is reduced in order to evaluate the hedging effectiveness of commodity-linked bonds and conventional debt in different currencies. Leading (13) one period, the conditional variance of m,+, at time t gives: 0M(^+)=(Xt)2(V.yAytl .)+MM tes VAAR(ml.1) (r)2V ( ye~) VAL&pg*+lbg -2C0;V(y't (22) =(__t_ 2ylatqy+&, GiP .b.Ea;2!lt;aP.Y) I+r Rearranging (20): 8For simplicity, subscript t+ I has been dropped. 13 Qp,YY= p Obto (23) Substituting (23) into (22): Ii +r; (2 =( r 2(Y/a, y_