WfIS a1ii> POLICY RESEARCH WORKING PAPER 2712 Contract Risks and Credit In infrastructure projects bondholders and Spread Determinants in the shareholdersshareresidual International Pro.ect Bond risks over time despite debt covenants meant to mitigate M arket risk shifting. For projects accessing international bond markets to benefit from Mansoor Dailami longer maturities and lower Robert Hauswald borrowing costs, it is therefore necessary to pay attention to such design features as capital structure, guarantees, ofF-take agreement, and project economics. The World Bank World Bank Institute Governance, Regulation, and Finance Division November 2001 I POLIcY RESEARCH WORKING PAPER 2712 Summary findings International bond markets have become an increasingly perceptions and prices in this segment. Judicio as use of important source of long-term capital for infrastructure an output price-contingent debt service guarar tee by projects in emerging market economies over the past shareholders can significantly reduce project ri,ks, and decade. The Ras Laffan Liquified Natural Gas (Ras Gas) markets reward issuers through tighter credit sareads. project represents a milestone in this respect: its $1.2 Bondholders and shareholders share residual r-isks over billion bond offering, completed in December 1996, has time, despite covenants meant to preempt risk ,hifting. been the largest for any international project. The Ras This type of risk shifting originates from incorr plete Gas project has the right to extract, process, and sell contracts and the nonrecourse nature of project finance. liquefied natural gas (LNG) from a field off the shore of It does not necessarily result from a deliberate attempt by Qatar. The principal off-taker is the Korea Gas management to increase shareholder value at the expense Corporation (Kogas), which resells most of the LNG to of debt holders by pursuing high-risk, low-value the Korea Electric Power Corporation (Kepco) for activities, although project managers and share iolders electricity generation. could still exploit their informational advantag-i by In this clinical study Dailami and Hauswald analyze leaving output supply contracts incomplete in ways the de terminants of credit spreads for the Ras Gas beneficial to their private interests. project in terms of its contractual structure, with a view The results hold important lessons for global project to better understanding the role of contract design in finance. Projects incorporating certain design features facilitating access to the global project bond market. can reap significant financial gains through lower Market risk perceptions have long been recognized to be borrowing costs and longer debt maturities: a function of firm-specific variables, particularly asset * Judicious guarantees by parents that enjoy a value as embodied in contracts. The authors therefore particular hedging advantage enhance a project'; appeal, study the impact of three interlocking contracts on the as reflected in favorable pricing. credit spreads of the project's actively traded global Pledging receivables rather than physical assets as bonds: the 25-year output sales and purchase agreement collateral and administering investor cash flows through with Kogas-Kepco, the international bond covenant, and an off-shore account offers additional security to debt an output price-contingent debt service guarantee by holders. Mobil to debt holders. * Projects should use their liability structure tv create Using a sample of daily data from January 1997 to an implicit option on future private debt financing that March 2000, the authors find that the quality of the off- matches the real option of a project expansion. taker's credit-and, more important, the market's * The finding that bondholders bear residual risks assessment of the off-taker's economic prospects-drive means that shareholders can reduce their risks arising project bond credit spreads and pricing. In addition, from bilateral monopolies and buy insurance against seemingly unrelated events in emerging debt markets unforeseen and unforeseeable events. spill over to project bond markets and affect risk This paper-a product of the Governance, Regulation, and Finance Division, World Bank Institute-is part of a larger effort in the institute to disseminate the lessons of experience and best practices in infrastructure finance and risk management. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact William Nedrow, room J3-283, telephone 202-473-1585, fax 202-676-9874, email address wnedrow@a worlcbank.org. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at mdlailami@worldbank.org or rhauswald@Crhsmith.umd.edu. November 2001. (48 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas aoout development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. Phe papers carry the names of the authors and should he cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or ehe countries they represent. Produced by the Policy Research Dissemination Center Contract Risks and Credit Spread Determinants in the International Project Bond Market Mansoor Dailami World Bank Institute The World Bank Washington, DC 20433 and Robert Hauswald* R.H. Smith School of Business University of Maryland College Park, MD 20742-1815 The authors would like to thank Peter Christoffersen, Tim Crack, Lutz Kilian, Nagpurnanand Prabhala, John Strong, Alex Triantis and Volker Wieland for stimulating discussions and suggestions, Panos Kogkalidis for excellent research assistance, and participants of the 2000 Asia Development Forum Workshop on Project Finance for comments. Special thanks are due to Greg Randolph, Bim Hundal and Ghassan Abdulkarim, Heads of Structured Finance, Capital Markets (London) and Middle East/North Africa at Goldman Sachs, respectively, for talking with us and providing a wealth of background information on the Ras Gas project. Executive Summary The importance of international bond markets as a major source of long-term capital for infra- structure projects in emerging market economies, has increased significantly over the past dec- ade. The Ras Laffan Liquified Natural Gas project (Ras Gas) represents a milestone in this re- spect because its USD 1.2 billion bond offering completed December 1996 is the largest for any international project to date. The Ras Gas project has the right to extract, process and sell lique- fied natural gas (LNG) from a field off the shore of Qatar. The principal off-taker is the Korea Gas Corporation (Kogas) which resells most of the LNG to the Korea Electric Power Corpora- tion (Kepco) for electricity generation. In this clinical study, we analyze the credit spread determinants of the Ras Laffan Liquified Natural Gas project in terms of its contractual structure, with a view to better understand the role of contract design in facilitating access to the global project bond market. Market risk perceptions have long been recognized to be a function of firm-specific variables and, in particular, asset value as embodied by contracts. The view of the firm as a nexus of contracts, first formulated in the seminal papers by Alchian and Demsetz (1972) and Jensen and Meckling (1976), underlies much of modern corporate finance. In particular, it serves as the foundation of many theories of capital structure design and corporate governance, i.e., the allocation of return and control rights. According to this view, the firm is defined in terms of the individual contracts that govern its existence such as labor and other input contracts, financial contracts including covenants and guarantees, supply and output purchase contracts. The nature and interaction of these contracts motivate financing choices, determine corporate governance arrangements, and provide a framework for firm valuation. While the theoretical foundations of project finance have received some attention in the literature there are very few empirical studies of project finance. This paper represents a first at- tempt to fill this gap in the literature. We study the impact of three interlocking contracts on the credit spreads of the project's actively traded global bonds: the 25 year output sales and purchase agreement with Kogas-Kepco, the international bond covenant, and an output price contingent debt service guarantee by Mobil to debtholders. Using a sample of daily data from January 1997 to March 2000, we find that off-taker credit quality and, more importantly, the market's assess- ment of the output buyer's economic prospects drive project bond credit spreads and the:r pric- ing. Also, seemingly unrelated events in emerging debt markets spillover to project bond niarkets and affect risk perceptions and prices in this segment. Furthermore, we document how the judi- cious use of an output price contingent debt service guarantee by shareholders can significantly reduce project risks and that markets reward issuers through tighter credit spreads. Our main contribution consists in showing how the firm as a nexus of contract allocates contracted and non-contracted risks between different stakeholders and how markets assess the latter in the pricing of financial claims. We show that, in the presence of contractual incomplete- ness, bondholders and shareholders share residual risks over time in spite of covenants otherwise meant to pre-empt risk shifting. This type of risk shifting originates from incomplete contracts and the non-recourse (stand-alone) feature of project finance. It does not necessarily result from the deliberate attempt by management to pursue high risk, low value activities in order to in- crease shareholder value at the expense of debtholders (debt agency) although project managers and shareholders could still exploit their informational advantages in leaving output supply con- tracts incomplete in a manner beneficial to their private interests. Our findings hold important lessons for global project finance because they show that market risk perception are a function of a project's contractual structure. In particular, the recep- tion that a project bond will receive in global capital markets depends on the project's ability to address investors' concerns about residual risks so that well-designed projects can reap signifi- cant financial gains through lower borrowing costs and longer debt maturities. We identify five such design features. Judicious guarantees by parents that enjoy a particular hedging advantage and a deliberate attempt to match debt service cash flow profiles with payment ability are recog- nized by the markets as enhancing a project's appeal. Our analysis also reveals that cash flows rather than physical assets are a project's true collateral so that a well-thought out cash flow routing structure with an off-shore account such as Ras Gas' offers additional security to debtholders. Fourth, Ras Gas shows how one can use the project's liability structure to create an implicit option on future private debt financing that matches the real option of a project expansion. Finally, the sensitivity of project credit spreads to contract related risk factors demonstrates that bondholders shoulder ex post part of the residual risks arising from non-contractibilities in the off-take agreement. This risk sharing means that shareholders can reduce their risks arising from bilateral monopolies and buy insurance against unforeseen and unforeseeable events. 1. Introduction Market risk perceptions have long been recognized to be a function of firm-specific variables and, in particular, firm value as embodied by its constituent contracts.' In this paper, we analyze the credit spread determinants and dynamics of the Ras Laffan Liquified Natural Gas Cornpany (Ras Gas for short) in terms of the project's contractual structure. We pursue two objectives with this study. On the one hand, we attempt to provide some empirical evidence on credit spread de- terminants from the perspective of the firm as a nexus of contracts. Prior studies on financial and organizational design based on large samples have focused on one contractual relationship at a time and are unable to identify the precise risk distribution and its evolution over a longer period. On the other, we wish to draw attention to the field of project finance that offers many exciting and unique opportunities to investigate issues of fundamental importance in finance. Indeed, no other practical case corresponds more closely to the standard setting of corporate finance models in terms of time structure with corresponding resolution of uncertainty, small number of inves- tors and classes of financial claims, typical actions taken, a single indivisible investment, etc. The view of the firm as a nexus of contracts, first formulated in the seminal papers by Alchian and Demsetz (1972) and Jensen and Meckling (1976), underlies much of modern corporate finance. In particular, it serves as the foundation of many theories of capital structure design and corporate governance, i.e., the allocation of return and control rights (Zingales, 2000). According to this view, the firm is defined in terms of the individual contracts that govern its existence such as labor and other input contracts, financial contracts including covenants and guarantees, supply and output purchase contracts. The nature and interaction of these contracts motivate financing choices (Fama, 1990), determine corporate governance arrangements (Jensen and Meckling, 1976), and provide a framework for firm valuation (see Kaplan and Ruback, 1995 for an application in terms of discounted cash flows). From a corporate finance perspective, this view of the firm begs the question how finan- cial contracts interact with other contractual relationships, how the latter affect the former, and how capital markets price these interactions. In theory, the firm as a collection of contracts should be worth the sum of its contracts. In practice, firms are very complex webs of contractual relationships, whose intricate interplay does not easily lend itself to empirical investigations. However, there is one particular area where a firm's contractual structure is sufficiently well- documented for such analysis: project finance. This financial technique is defined as the raising of funds to finance a single indivisible large-scale capital investment project whose cash flows are the sole source to meet financial obligations and to provide returns to investors.2 The Ras Gas project has the right to extract, process and sell liquefied natural gas (LNG) from a field off the shore of Qatar. We study the impact of three interlocking contracts on the credit spreads of the project's actively traded bonds: a 25 year output sales and purchase agreement with a dominant output buyer, the bond covenant, and an output price contingent debt service guarantee by shareholders to debtholders. Such contracts are incomplete by nature in that they could not possibly anticipate all future contingencies, including non-enforceability of liens ' See Zingales (2000) for a discussion of the necessary conditions for a firm's value to be the sum of its contracts. 2 Brealey, Cooper and Habib (1996) contains an excellent survey of the economic issues involved in project finance. Contrary to a large company, projects such Ras Gas have only one cash flow stream to meet all debt obligations and pay dividends. For further discussion of project finance, see Finnerty (1996). 2 on assets and receivables. In Ras Gas' case, the contractual incompleteness primarily stems from the very specific nature of the required investment in LNG infrastructure (asset specificity), their location, and the long-term nature of the sales contract creating a bilateral monopoly. In such circumstances, the project's investors bear the costs of unforeseen, i.e., non- contracted, contingencies and potential opportunistic behavior by the output buyers because the LNG supply contract as the major source of revenue effectively secures the debt. Consequently, we would expect capital markets to price non-contracted risks stemming from the supply contract. Using the structural default rate framework of Madan and Unal (2000), we analyze the evolution of Ras Gas credit spreads in terms of firm-specific risk variables, in particular the ultimate output buyer's credit spread (the Korea Electric Power Company, Kepco for short). Any material deterioration in the economic prospects of the output buyer, as measured by K.pco's credit spread, should increase the likelihood of breach of contract and, therefore, drive Ras Gas' spreads. Output prices as a major determinant of revenues are the second important contract- related risk factor. Ras Gas offers the unique opportunity of assessing the contractual dynamics arising from a bilateral monopoly on the basis of market information because both the seller (Ras Gas) and buyer (Kepco) have actively traded global bonds outstanding. Using a sample of daily data from January 1997 to March 2000, we relate Ras Gas credit spreads to their own lags, to current and lagged Kepco credit spreads, to a crude oil reference price used to settle LNG sales (13rent), Korean control variables, and the current and lagged returns on four regional emerging debt market indices (contagion and spillover effects) in a linear regression framework. We repeat the analysis in a simultaneous equation setting in order to distinguish direct effects of the risk factors 3 from indirect ones operating through the output buyer. This market-based approach to gauging risk perceptions allows us to investigate how the three interlocking contracts allocate project risks between shareholders and debtholders and test for residual risk shifting. We find that Ras Gas credit spreads exhibit a very high degree of persistence. By far the most important explanatory variable for both levels and changes in credit spreads is the off- taker's (Kepco) credit spread. Investors rationally anticipate the incidence of the output buyer's financial and economic condition on the riskiness of their bond. However, we also find evidence for over-reaction and market inefficiencies: while Ras Gas spreads widen with contemporeanous Kepco spread movements, they narrow in lagged ones. The output price (Brent) comes out largely insignificant: investors seem to disregard commodity price risk. In light of the debt service guarantee contingent on Brent prices, this result comes as no surprise. Markets do not price contracted risks, as predicted by theory. Further investigation shows that the direct oil price impact on Ras Gas is insignificant but that the indirect impact via Kepco' s financial position is highly significant. In terms of Korean country risk factors, we find evidence of Ras Gas exposure to the Ko- rean currency both directly and indirectly through Kepco credit spreads despite the fact that the off-take agreement is US dollar (USD) based. As Kepco's revenue is almost entirely denomi- nated in Korean Won, any currency depreciation makes USD denominated energy imports more expensive and erodes its financial position, which might call into question contractual commit- ments. Hence, Ras Gas and its investors bear some Korean currency risk. We also find significant evidence of financial contagion. As returns in European, Middle Eastern and Latin American 4 emerging debt markets fall, we find that Ras Gas spreads are predicted to widen considerably. In particular, the impact of contemporaneous and past events in European emerging debt markets stands out. This responsiveness reflects spillovers from the 1998 Russian financial crisis, vvhich heavily affected other emerging debt market segments. While the theoretical foundations of project finance have received some attention in the literature (see, e.g., Shah and Thakor, 1987, Berkovitch and Kim, 1990, Chemmanur and .lohn, 1996) there are very few empirical studies of project finance. This paper represents a first attempt to fill this gap in the literature.3 Esty (1999) describes a comparable crude oil project in Venezuela but the existence of a well-developed oil spot market does not lead to a bilateral monopoly with the ensuing contract risk dynamics. Esty and Megginson (2000), who analyze how political risk shapes the syndication process and pricing of project loans, complement our findings from a private debt perspective. Our analysis is also related to the literature on bond covenants going back to Smith and Warner (1979). We show that other contractual relationships besides covenants impact bondholders so that one cannot abstract from their contracting and enforcement costs. Furthermore, our results highlight the interdependence of debt finance and risk distribution recently identified in the context of hedging by Mello and Parsons (2000). This paper also contributes to the nascent empirical literature on structural models of credit spreads. From a methodological point of view, our analysis draws on the theoretical framework of Madan and Unal (2000) whose structural model of the hazard (default) rate implies that credit spreads are linearly related to firm-specific exogenous variables. In contrast to much of the recent theoretical literature (e.g., Duffie and Singleton, 1999), this approach allows us to 5 cast cash asset value and default risk in terms of the risk factors arising from Ras Gas' contrac- tual structure. As a result, our analysis reconciles continuous time corporate default models with the dominant view of the firm in corporate finance and provides evidence in favor of the Madan and Unal (2000) default risk model. Our main contribution consists in showing how the firm as a nexus of contract allocates contracted and non-contracted risks between different stakeholders and how markets assess the latter in the pricing of financial claims. We show that, in the presence of contractual incomplete- ness, bondholders and shareholders share residual risks over time in spite of covenants otherwise meant to pre-empt risk shifting. This type of risk shifting originates from incomplete contracts and the non-recourse (stand-alone) feature of project finance. It does not necessarily result from the deliberate attempt by management to pursue high risk, low value activities in order to in- crease shareholder value at the expense of debtholders4 (debt agency) although project managers and shareholders could still exploit their informational advantages in leaving output supply con- tracts incomplete in a manner beneficial to their private interests. The paper is organized as follows. The next section provides background information on the Ras Gas project and its contractual structure. Section 3 describes the project-specific sources of contractual incompleteness and risk factors. Section 4 contains a description of the data and our methodology. In Sections 5 and 6, we summarize the results of our empirical analysis. Sec- tion 7 concludes. We relegate all tables to the Appendix. 3 See Tuffano (2001) for a discussion of the merits and importance of clinical studies in this respect. 4 See, e.g., Smith and Warner (1979), Green (1984) and John (1987) for more on this point. 6 2. The Ras Gas Project The Ras Gas project, while a typical example of its kind, represents a milestone in the annals of project financing because of its recourse to global bond markets.5 Capital markets debt wvas in- strumental in the successful design and financing of the project because it provided flexibility not otherwise available through the syndicated loan market. Its USD 1.2 billion bond offering com- pleted December 1996 is the largest for any international project to date, the first for a LNG pro- ject, the first capital markets financing for a Qatari issuer, and the first for a Middle Eastern is- suer with a maturity beyond 7 years. To put the Ras Gas financing into perspective, the total amount of project bonds issued in 1996 was USD 4.79b (with Ras Gas accounting for 25% of this amount) while total bank lending to projects amounted to USD 42.83b. By 1999, the propor- tion of project debt raised in bond markets had grown from 10.06% in 1996 to 21.62%.6 Ras Laffan Natural Liquified Gas Company Limited is a joint venture between the Qatar General Petroleum Corporation (66.5%) and Mobil Corporation of the US (26.5%), located in Qatar (Persian Gulf).7 Ras Gas, a Qatari company, has the right to develop lOm tons of liquified natural gas (LNG) annually from Qatar's North Field, the world's largest unassociated natural gas field with about 380b cubic feet of confirmed recoverable reserves (about 9% of world gas reserves). To this end, Ras Gas has constructed a 5.2 MMTA (million metric tons per annum) 5 The following project description draws on its bond offering prospectus (Goldman Sachs, 1996), Standard and Poor's (1996a, 1999, 2000) and Randolph and Schrantz (1997). According to Greg Randolph, Goldman Sachs, Ras Gas, whose structure is much copied in the energy sector, exemplifies state-of-the-art project design and financing. 6 The use of public debt markets for project financing is a relatively recent phenomenon. By 1999, global project lending by banks had increased to USD 72.392b (USD 56.65b in 1998) while global project bond issuance rose from USD 9.979b in 1998 to USD 19.966b (Pepiatt and Rixon, 2000); for earlier data and an excellent overview of the project finance market, see Esty (2000). For a description of the syndicated loan market's role in more traditional project finance, see Esty and Megginson (2000). 7 liquification facility in Ras Laffan consisting of two identical LNG processing trains, offshore drilling platforms, storage facilities, pipelines and port loading facilities. Construction was completed in late 1999 at a cost of USD 3.264b, slightly below the initially projected USD 3.4b. Exhibit 1 summarizes the final capital structure and construction budget. To make the project attractive to debtholders, the parent firms heavily capitalized it (30% equity), signed a long-term supply agreement before the start of construction with a high credit quality off-taker (rated AA-), and provided debt service guarantees contingent on LNG settlement prices. While the project had initially been all equity and bond financed, Ras Gas had reserved the option to fund a second liquefication train with private debt under the bond covenant provided an additional supply agreement (SPA) could be signed with a single 'A' or better rated off-taker. When the Korea Gas Corporation (Kogas) agreed to double its LNG purchases in June 1997, Ras Gas exercised this option to secure the significant economies of scale offered by the second train. Uses of funds Sources of funds % of total (USD millions) Drilling 239 Senior debt 2,285 70.00 Commercial banks 382 11.70 Offshore facilities 453 ECA guaranteed 703 21.50 Bonds due 2006 400 12.30 Onshore facilities 1,670 Bonds due 2013 800 24.50 Venture costs 380 Equity 979 30.00 QGPC 651 19.90 Financing costs, 593 Mobil 260 8.00 interest during construction Itochu 39 1.20 Nissho Iwai 29 0.90 Total costs 3,264 Total funds 3,264 100.00 Exhibit 1. Ras Gas Construction Budget and Capital Structure" 7 The initial stakes were 70% and 30%, respectively, and fell with the addition of two Japanese output buyers as shareholders. Kogas has the option to acquire a 5% equity stake, which is one of the standard devices to overcome contractual incompleteness and hold-up problems (see Noldeke and Schmidt, 1995). 8 Standard and Poor's (1999); ECA refers to bank loans and facilities guaranteed by three export credit agencies: the US Exim Bank, the UK's ECGD and Italy's SACE. 8 The presence of long-dated bonds was instrumental in bringing the project to fruition because of the particular cash flow profile of projects in general, and the large usi-front investment of Ras Gas in particular. As a result, the project would have had insufficienilr debt service capacity in its first 6 to 8 years, which is the maximum available maturity for projects in the syndicated loan market. Only the public debt markets offered longer maturities that could stretch out debt repayment significantly beyond the start-up phase, mitigating liquidity concerns. This dependence of debt finance on cash flow profiles, established through the debt covenants and maturity structure (medium-term bank debt, long-term public debt), echo the intertemporal liquidity aspects of risk management analyzed in Mello and Parsons (2000). In their model, intertemporal liquidity concerns lead to a pairing of hedging with debt financing strategies. The same liquidity effects drive Ras Gas' capital and, indeed, overall contractual structure in the face of buyer default, output price (revenue) and foreign currency risk. In the absence of appropriate hedging instruments for such risks, the parties have recourse to contractual provisions and shareholder guarantees albeit at the price of potential risk-shifting through non-contractibilities. The two Ras Gas bonds proved to be in very high demand. Despite increasing the issue size, they sold out on the first offering day (December 16, 1996) and were twice over- subscribed.9 The long bond due in 2013 has a total size of USD 800m and was priced at ali issue yield of 8.294% or 187.5 basis points above 15 year US Treasury bond yields (interpolated). 5The bonds proved to be so high in demand that Ras Gas could have been funded entirely in the global bond mar- kets. However, the parties decided to keep the bank loan component at an average all-in cost of 9.60%, about 95 basis points above the average all-in cost of the bonds (8.65%), in order to insure easier access to bank debt for fu- ture project expansion in the form of additional liquefication trains (Greg Randolph, Goldman Sachs). 9 Issued as a global bond, i.e., both as an off-shore (Eurodollar) and 144A foreign (Yankee) debt security, it was sold to institutional investors with strong international demand (20% international, 80% US based investors).'" Since the bond trades actively we can use its spread over US Treasuries to gauge market perceptions of changes in Ras Gas' prospects and, hence, its riskiness. According to Goldman Sachs, the smaller USD 400m 10 year bond due in 2006 has been bought up by Middle-Eastern investors and trades infrequently. As is customary in project finance, most of the output was sold through long-term supply contracts before construction started. The principal off-taker, the Korea Gas Corporation (Kogas), is a state-owned company whose shareholders include the Republic of Korea (50%), the Korea Electric Power Corporation (Kepco: 34.7%) and regional governments (15.3%). As such, Kogas shares its credit rating with the sovereign rating of South Korea as does Kepco, which is currently being privatized. Most of the Ras Gas LNG bought by Kogas, who has a legal monopoly of gas sales and purchases in Korea, is for resale to Kepco as fuel for peak-load electricity generation. Consequently, Kepco, which is about to double its existing LNG powered electricity generation in the next years, is Ras Gas' de facto off-taker (Standard and Poor's, 1999). Kogas-Kepco currently account for more than 75% of the project's expected revenue. The following diagram summarizes the project's principal parties and its contractual structure." 10 According to Goldman Sachs, about 70 institutional investors and banks excluded from the syndicated and guaran- teed loan tranches participated in the bond offerings with typical investments ranging from USD 1 5m to 20m (Greg Randolph; the largest single block bought was USD 125m). " Typical webs of contracts in project finance comprise joint venture agreements, equity claims, debt contracts in- cluding covenants, construction, input supply and operating contracts, and output supply (off-take) agreements. 10 State of Qatar Mobil Corp. 100% 100% Qatar General ti Petrole P| Power Corp | Govermnents| Perlum Corp. Moi MGs n.Rpbi fKra KrAElCtrcRoa 70% 1 30% 1 50% 34.7% 15.3% Ras Laffan Liquefied Natural Gas Co. Ltd. * Korea Gas Corporation Onshore EPC Platforms EPC Pipelines EPC Korea Electric Contract Contract Contract Power Corp. JGC Corp./ The McDermott-EPTM M.W.Kellogg East, Inc./ Chiyoda Saipem S.p.A. Company Corp. Exhibit 2. Ras Gas Project Participants The two sales and purchase agreements (SPAs) with Kogas stipulate a fixed o ff-take quantity of 4.8 MMTA of LNG. Since August 1999, Kogas is receiving LNG shipments for 25 years on a take-or-pay basis. Under such an agreement, the purchaser (Kogas on behalf of Kepco) is obligated to pay for the gas whether or not they take delivery. Hence, Kogas can make a cash payment in lieu of delivery, which is credited against charges for future deliveries. The off-taker can vary gas shipments by deferring about 5% per annum up to a total of 10% which must be paid for within 5 years whether Kogas accepts delivery or not. The remaining LNG produced is for sale on the nascent LNG spot market and two small off-take agreements with Japanese customers. The Kogas SPAs effectively index LNG prices to world crude oil prices. Following market conventions for LNG pricing, one of two crude spot reference prices (the Japan Crude 11 Cocktail or Brent) serves as the monthly settlement price for the LNG shipments in terms of energy equivalents.'2 The other products sold, in particular condensate, a crude oil substitute that naturally occurs in the liquefication process, and some spot sales are similarly priced. To reassure bondholders, Mobil has given an effective minimal price guarantee in form of a USD 200 million credit line for debt service payments triggered at an oil price somewhere below USD 11 per barrel.'3 The following figure relates Ras Gas' contractual arrangements to its cash flow structure. The two bond issues represent senior secured debt and rank pari passu (same seniority) with the bank and ECA guaranteed debt.'4 The Kogas off-take agreement serves as undivided security interest for all debtholders under New York law. Debtholders hold all the rights to the receivables from Kogas and also have a security interest in the Ras Gas assets under Qatari and New York law. In order to minimize moral hazard in payments, Kogas and other output buyers make payment for shipments directly to an off-shore trust account whose administrator then services public and private debt and remits the balance to Ras Gas for operational expenses and dividends. The superscripts denote the order in which payments are made. 12 One metric ton of LNG has the energy content of about 8.68 barrels of crude oil (with minor variations depending on the crude oil reference used) and is priced accordingly. " Standard and Poor's (1999) estimate that the average break-even Brent oil price triggering the guarantee is about USD 10.15 per barrel. However, in individual years, especially before 2003, a Brent oil price of USD 14/bbl might suffice to activate the guarantee. From a hedging perspective, this arrangement makes a lot of sense. For a large en- ergy company such as Mobil it might be easy to find a low-cost natural hedge for the guarantee in its activities or through its balance sheet while individual investors would be hard pressed to find appropriate hedging instruments. 14 The bond and loan covenants are virtually identical; indeed, the former are based on the latter (Greg Randolph, Goldman Sachs). 12 Other LNG KoreaGas Corp. dCondensat N Buperscrip I Buners f e ne s Payments oays g Royalties and Taxes, Offshore Account wt Debt Service e wae of Oatar l the provsion ofst t Facilits Af.inc 100% .10%li srQGPC c th Moil pi c|onte holders Banks T 70% 3% Oprtng CoStS2 L Ras Laffan LNG Co. Ld Note: Superscript numbers indicate payment order of pfiofity. Exhibit 3. Ras Laffan Contract and Cash Flow Structurefi The nexus of contracts that we study consists of the Ras Gas - Kogas/Kepco long-term supply agreements, the Ras Gas bond contract with its covenant, and Mobil's implicit LNG price guarantee to debtholders. At its heart lies the fact that the Kogas off-take agreement effectively collateralizes the project's debt and its cash flow profile. Ras Gas forcefully illustrates the point made in Fama (1990) that a firm's capital structure depends on all contracts with stakeholders, including output purchase agreements and financial guarantees. Since the firm is essentially a web of interlocking contracts, the provisions of the long-term supply contract drive its financi.al structure including the oil price contingent debt service guarantee by shareholders. T he corresponding financial transactions reflect this reality. They attempt to find an optimal balance of the various parties' rights and obligations and serve to allocate risks to the entities best suited to bear them. 1 3 3. Contractual Incompleteness and Risk Factors A large-scale project such as Ras Gas typically requires huge up-front investments with a high degree of asset (physical infrastructure) and relationship (output buyer) specificity. By their very nature, the necessary physical assets such as pipelines, storage facilities, LNG ship terminals, etc. cannot readily be removed and utilized elsewhere. As a result, there is a danger that Ras Gas and its financial backers suffer opportunistic behavior such as unilateral renegotiation of contracts or the redefinition of property rights. In the absence of a well functioning legal system that is willing to define and enforce property rights and contractual clauses, the physical assets - always subject to hold-up problems - are of limited value as security to investors. Hence, the location of the assets in Qatar and the lack of credible legal institutions (enforcement) render them inadequate for creditor protection. Instead, the sales and purchase agreements with Kogas provide the only effective security to debtholders. However, the output supply contract as collateral suffers from contractual incompleteness. From the off-taker's perspective, commitment to such a long-term contract poses the difficulty of not knowing at the time of contracting the future value of the output, i.e., future settlement prices (Brent crude oil reference), the availability of re-contracting opportunities and alternative suppliers. Hence, a project such as Ras Gas faces the danger that its dominant buyer reneges on the long-term contract as alternative sources of LNG supplies are more cheaply available elsewhere than through the SPA. Put differently, the off-taker has always an implicit real option through breach of contract. In addition to opportunistic behavior, the output buyers might experience exogenous shocks such as a severe demand reduction in electricity or a 15 Bond offering prospectus (Goldman Sachs, 1996). 14 liquidity crisis that might force them to cut back on their LNG purchases. In the presence of a well-developed LNG spot market, such off-take risk would hardly matter. It is its absence that exacerbates the consequences of contractual incompleteness and nan- enforceabilities. At the heart of the problem lies the lack of transportation capacity"6 and the huge up-front investment in receiving facilities (terminal, storage, regasification plant, pipelines). In 2000, only 39 out of more than 2000 LNG cargoes were for true spot delivery (less than 234X of the total market). Together with short-term secondary trading of LNG, whereby an off-taker sells a cargo to a third party rather than defer delivery, they accounted for 4MMTA out of a total of 104MMTA of LNG produced in 2000 (up from 2% in 1996; Tusiani, 2001). Consequently, the parties often build dedicated vessels for LNG transportation tied to a specific project"7 and, in an attempt to protect their investments in physical infrastructure, sign long-term off-take agree- ments. 8 Hence, the most important hold-up risk for Ras Gas and its investors consists of breach of contract or unilateral renegotiation of the SPA by Kogas, the off-taker. Such risks are directly passed through to debtholders. They are locked into the project and, hence, vulnerable to 16 With the availability of LNG tankers not tied to specific projects, the nascent LNG market for immediate delivery is expected to develop into a full-fledged spot market over the next decade. However, Standard and Poor's (1999) reckon that "[I]ong-term contracts for LNG still continue to dominate the LNG trade because of expense and sccpe of dedicated systems for delivering, receiving, and using LNG. A true short-term spot trading market remains elu.sive for the foreseeable future." For more on current LNG trading trends, see Banaszak (2001). 17 The off-take agreements with Kogas-Kepco stipulate the construction of landing, storage and regasification facili- ties in Korea as well as 7 dedicated LNG vessels (costing around USD 200m each; LNG tanker prices are down 40% from mid 1990s level (Tusiani, 2001)) to be completed by 2002 when the project produces at peak capacity. To date, Kogas and Kepco have invested about USD lOb in tankers and LNG infrastructure (Standard and Poor's, 1999). 18 Tying transportation capacity to particular projects through the off-take agreements in turn inhibits the emergence of a true spot market. Another problem are the substantial LNG infrastructure investments required on the receiving end that become economically viable only once a source of long-term supply has been secured. The current state of the LNG world market is reminiscent of crude oil trading in the 1950s and 1960s when the solution to hold-up and unilateral renegotiation threats between bilateral monopolists was vertical integration. A spot market for crude oil only emerged around 1970 with the availability of excess shipping, receiving and storage capacity. 15 opportunistic and strategic behavior not only from shareholders, but also from Ras Gas' dominant customer. While the final Kogas off-take agreement includes deferral options, '9 meant to pre-empt breach of contract, the demand risk arising from their exercise is directly transmitted to investors and, especially debtholders, given the lack of an LNG spot market and alternative sources of revenue. Long-term supply contracts such as the 25 year SPA between Ras Gas and Kogas only offer an imperfect remedy to contracting problems arising from a bilateral monopoly. As economic circumstances change, the absence of enforceable, complete contracts means that investors must constantly reassess their initial financing decisions. Ras Gas' bond prices and, hence, credit spreads (over US Treasury yields) should then reflect the capital market's collective assessment of the evolution of contractual risks. The inherent incompleteness of the interlocking contracts, therefore, leads to a structural relation between credit spreads and risk factors as the project passes residual risks on to both debtholders and shareholders.20 Using Ras Gas' simple contract structure we can identify their precise sources and test how they shape market sentiment. We now turn to several key risk variables in the supply and purchase agreement that have a direct incidence on the project's financial prospects. The first variable behind the postulated chain of contractual risks are output prices, which effectively determine Ras Gas' revenue because annual off-take quantities are (almost) fixed. We 19 In the aftermath of the Asian financial crisis, Korean electricity demand declined in 1998 by about 3.6% after pre- viously growing by 10% annually. As a result, Kepco reduced purchases of LNG, its marginal fuel, from Kogas by as much as 22%. However, electricity demand has recently picked up (8.1% increase in 1999), and, while demand growth is expected to fall short of initial forecasts, Kepco still plans to add about 20,000 MW of generation capacity including LNG fired power stations over the next years (Standard and Poor's, 1999, 2000). 16 use the logarithm of the price of Brent (BRENT) - one of two commonly used crude oil reference prices for LNG2' - to analyze the incidence of output prices on the riskiness of Ras Gas.22 The contractual provisions of the off-take agreement permit us to separate demand volume from price risk because Kogas, by and large, has committed to buying a fixed amount of output per annum. Hence, demand risk essentially translates into breach of contract risk. Since Kepco is Ras Gas' effective off-taker and Kogas only an affiliated intermediary,23 we take -he mid-closing yield spread of the Kepco 7.75% global (Eurodollar and Yankee) bond maturing in April 2013 (KORELES) over 10 year US Treasury yields to measure the economic and financial prospects of the LNG buyer as assessed by capital markets. From a statistical perspective, using Kepco credit spreads has the added benefit that they are an instrumental variable for Kogas spreads, which should be simultaneously determined with Ras Gas spreads because of the bilateral monopoly relationship between the two firms. Ras Gas' fortunes also depend on Korea's macroeconomic environment through its impact on Kepco and Kogas. A severe recession might cast serious doubts on Kogas-Kepco's ability to honor their contractual commitments. We use the logarithm of the Korea Composite Stock Index (KOSPI) as a proxy for the incidence of the Korean macroeconomic environmeni on electricity and gas demand. To control for Kepco's idiosyncratic (operational, regulatory and financial) risks, we include KEPCO, the logarithm of its stock price. A further risk factor is the 20 See Zingales (2000) for a discussion for situations in which there might exist other residual claimants besides shareholders. Projects rarely issue publicly traded equity so that in their absence project riskiness is best assesseJ by the price of publicly traded debt, whenever available. 21 Gas prices turn out to be statistically non-significant when included in the regressions together with Brent prices, which is not really surprising given that about 0. I metric tons of LNG are priced as one barrel of crude oil. 22 Diagnostic testing reveals that logarithms offer superior fit over levels for several of the explanatory variables. 23 See Standard and Poor's (1999) for Ras Gas' financial dependence on the Korean electricity market and Kepco. 17 credit quality of the off-taker, which might reflect both systematic changes in the Korean macroeconomic environment, the industry structure (i.e., loss of monopoly, privatization) or purely idiosyncratic risks. Its importance to debtholders can be seen from the fact that the Ras Gas bond covenant restricts additional SPAs to buyers rated single 'A' or better, a condition Kogas and Kepco satisfied until December 1997. However, their credit rating (shared with the Republic of Korea) has varied from 'AA-' to 'B+' back to 'BBB' over the sample period. According to average yearly transition probability estimates by Brand and Bahar (2001), 'AA' rated borrowers maintain an 'A' or better rating with 96.14% probability while credit migration such as Kepco's occurs only with 0.09% probability, which appears to be a negligible risk.24 To gauge these effects, we construct a rating index (KRR) that reflects not only the changes in S&P credit ratings but also their magnitude. Foreign currency might appear to be of relatively minor concern as all revenues and costs accrue in USD in the case of Ras Gas. However, by the very nature of the off-take agreements, the customer still poses a subtle indirect currency risk. Both Kogas and Kepco generate their revenue in local currency so that an adverse currency movement (devaluation or depreciation of the Korean Won against the USD) might imperil their ability to honor the SPA. The 1997 Asian financial crisis was a stark reminder of this fact: as the Korean Won depreciated against the USD the effective cost of LNG to Kogas and Kepco doubled in local currency terms. Hence, exchange rate risk when borne by the off-taker has a tendency to transform itself into a credit risk. To measure this effect, we include KRW, the logarithm of the KRW-USD exchange rate. 24 To be precise, Brand and Bahar (2001) estimate that the average yearly transition probability from 'AA' to 'B' is 0.09% while the cumulative average default probability over 15 years, the weighted average life of the 2013 Ras Gas bond, is 1.07% for a 'AA' rated entity. Standard and Poor's rated Ras Gas 'BBB+' and maintaining its rating during the Asian and Russian financial crises. For comparison, the estimated 15 year cumulative default probability for 'BBB' rated borrowers is 4.48%. 18 Finally, we need to control for financial contagion and other "guilt by association" characteristics of emerging debt markets. To gauge the incidence of such shock propagation mechanisms on Ras Gas credit spreads, we use the JP Morgan emerging market bond regional indices (EMBI family), i.e., Asia, Middle East, Europe and Latin America. The following table summarizes the predicted direct and indirect effects acting through Kepco of the various variables on Ras Gas yield spreads: Dependent Variable Ras Gas Yield Spread Changes in RG Spread Effect Direct Indirect Direct Indirect Variable Description Coefficient Coefficient Coefficient Coefficient RGS(-I) Lagged Ras Gas spread persistent BRENT (Log) oil price indeterm. + indeterm. + BRE.NT<14 (Log) oil price below USD 14 insign. + indetern. + BRENT: 14-23 (Log) oil price: USD 14 to 23 indeterm. + + + BRENT>23 (Log) oil price above USD 23 indeterm. + indeterm. + KORELES Kepco yield spread + KORELES(-1) Lagged Kepco spread insign. persistent persistent KEPCO (Log) Kepco stock price insign. - indeterm. KRW (Log) Korean Won FX rate + + insign + KOSPI (Log) Korea Stock Price Index insign insign KRR Korean country rating index + + insign. + ASIA Emerging debt returns Asia EUR Emerging debt returns Europe LAT Emerging debt returns Latin Am. MEA Emerging debt returns Middle East Exhibit 4. Explanatory Variables and Their Coefficients' Predicted Sign 19 4. Data Description and Methodology Our analysis relies on daily data that covers the period from January 1997 to March 2000 and is drawn, for the most part, from Bloomberg, IDC and Baseline. All market related data (e.g., oil and stock prices, bond yields, and emerging debt market returns) are based on daily closing prices. The bond yield reflects, as far as we can tell, actual transaction data. Whenever we found missing observations, we cross-checked the time series with other news sources and filled in the missing data or, if this was not possible, deleted the observation leaving 725 observations before taking lags. As a robustness check, we repeat the analysis with weekly closing data (140 observations) but report the results only for major specifications (Table 8 in the Appendix). In terms of structural modeling, we avail ourselves of the results in Madan and Unal (2000) who derive credit spreads as a fumction of firm-specific variables in a hazard rate framework. In this setting, the hazard rate, i.e., the instantaneous probability of borrower default, governs the arrival of a sudden loss driven by structural parameters such as cash asset value or, in our case, the value of the supply and purchase agreements to Ras Gas investors. By expressing the hazard rate as a first-order approximation in terms of exogenous variables, we obtain Ras Gas credit spreads as a linear function of loss inducing risk factors, neglecting higher order terms. Consequently, we take as our dependent variable the mid-closing spread of the 2013 Ras Gas bond yield over the 10 year benchmark US Treasury yield.25 The explanatory variables are the risk factors affecting the contractual relationships at the heart of the Ras Gas project that we 25 According to Bim Hundal of Goldman Sachs, the bond is quite actively traded contrary to the 2006 one and, there- fore, constitutes a much better measure of investor and market sentiment regarding the project's prospects. 20 discuss in the preceding section. If markets are informationally efficient, as we henceforth assume, then non-contracted risk factors should contribute to explaining Ras Gas credit spread.s as a measure of project riskiness. Hence, we gauge non-contractibilities and ensuing risk shiftiig in terms of the statistical significance of contract-related explanatory variables. Ras Gas Spread KEPCO Spread---- Brent 12 35 30 10 I 25 8 Sj20 X I 11 K~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ C,~~~~~~~~-k O - i- r - T-,O ~~~~ c~ m ''5 ' 't }e m c^ m m Exhibit 5. Ras Gas Credit Spreads, Kepco Credit Spreads and Oil Prices As Exhibit 5 suggests, the data is quite volatile. Table 1 in the Appendix contains summary statistics for the entire sample period that confirm this point. The pairwise correlation matrix reveals that some of the variables are highly correlated suggesting potential collinearity problems, which we will address through parsimonious specification and using the logarithrn of affected variables. Note that the preceding diagram clearly indicates the two defining events during the sample period: the Asian financial crisis that engulfed Korea in December 1997, and 21 the Russian financial crisis that shook emerging debt markets again in August 1998. We estimate variants of the following empirical specification by Ordinary Least Squares: RGS, =,60 + Ea_,RGS,, + ABRENT + E /2 ,KORELES,-, 0oV<,< 0o1L + /83KEPCO, +,84KRW, + 85KOSPI, + /J6KRR, + J [Y1 ,MEA,, + Y21 ASIA,, + Y3 EUR,, + Y4 ,LAT,, ] + 8, 0:515L where L indexes maximal lag length, RGS, is the spread of the Ras Gas bond over 10 year US Treasury yields, BRENT, the logarithm of the Brent blend oil price index, KORELES, the spread of the 2013 7.75% Kepco global bond over 10 year US Treasury yields, KEPCO, the logarithm of the Kepco stock price, KRW, the logarithm of the Korean Won - US Dollar spot rate, KOSPI, the logarithm of the Korea Composite Stock Price Index, KRR, a shared credit rating index for Korea, Kepco and Kogas, and MEA,,ASIA, ,EUR,j, LA T the continuously compounded daily returns of the JP Morgan regional total return indices in USD for emerging markets in the Middle East-Africa, Asia, Europe and Latin America, respectively. In terms of estimation strategy, we start with the two key contract variables depicted in Exhibit 5, the output (Brent oil reference) price and Kepco bond yield spreads, and successively add explanatory variables to the regression. First, we focus on the supply contract specific variables of oil price and Kepco bond yield. Next, we will add contemporaneous and lagged emerging debt market returns to analyze systematic effects such as spill-overs and contagion. We then include variables related to Korean country risk before estimating models with all risk factor categories. 22 It turns out that the regression residuals exhibit high serial correlation for any number of explanatory variables and their lags (specification 1, Table 2). Including a lagged depende-it variable in the various specifications fixes this problem as evidenced by Durbin and Watson d statistics close to 2.00 or the results of our robust test for serial correlation (see Table 2). Given the high frequency of the data, it comes as no surprise that daily credit spreads exhibit a large degree of persistence: the coefficient on lagged spreads is close to unity (Table 2). However, tests for unit roots (see Table 3) appear inconclusive given the very low statistical power of such tests26 so that we treat the time series as stationary, albeit highly persistent. Comparison of the coefficients on the lagged dependent variable from the weekly estimation results (Table 8) with the corresponding daily ones (Tables 2 and 5) further point to persistence rather than a unit root. Nevertheless, we also estimate our basic model in first differences to address potential non-stationarity problems. Section 6 repeats the analysis in a simultaneous equation framework to explicitly take into account the bilateral monopoly and to separate direct from indirect risk effects acting through their impact on Kepco. Throughout, we eliminate highly insignificant control variables through diagnostic testing in the interest of parsimonious specification. 5. Credit Spread Dynamics and Contractual Risks As conjectured, Ras Gas spreads vary positively with Kepco credit spreads: the second specification in Table 2 indicates that a 100 basis point increase in Kepco spread widens the Ras 26 Campbell and Perron (1991) have pointed out that unit root tests are biased in favor of the null hypothesis (exis- tence of a unit root) if the time series suffers from structural breaks such as the emerging market crises of 1997-1998. 23 Gas spread by about two basis points. Consistent with the provisions of the bond covenant and the nexus of contracts view of the firm, the perceived credit worthiness of the output buyer, a non-contractible risk, feeds through immediately to Ras Gas yield spreads. By pricing such non- contracted off-take risk, debt markets indicate that they recognize the incomplete nature of covenants and output supply agreements and that, at least in part, risk is shifted from Ras Gas owners to its bondholders. Including the lagged Kepco spread reveals the following time pattern of credit spread adjustments. Initially, Ras Gas spreads widen by 15.5 basis points for every 100 basis point increase in Kepco spreads. On the next trading day, they narrow by 13.6 basis points (coefficient on the lagged Kepco yield spread) all other things being equal. A comparison between the second and third regressions reported in Table 2 shows that the previously identified two basis points spread widening is the net reaction over a two-day period.27 Further lags of the Kepco spread are statistically insignificant. The results in Table 2 indicate that this pattern is stable across all specifications and, therefore, does not stem from any omitted variable effects. The reversal of the initial spread reaction is reminiscent of positive stock return reactions after large one-day declines. Cox and Peterson (1994) conclude that bid-ask bounce and liquidity effects rather than short-term overreaction explain short-term reversals. This analogy is all the more pronounced that Exhibit 5 clearly shows both high daily volatility and a short-term reversal pattern. However, given the high degree of volatility and uncertainty in emerging bond markets from 1997 to 1999 and the widespread fears of a prolonged severe recession in Korea, we cannot 27 With weekly data, the net effect is about 10 basis points with a contemporaneous impact of +28.3 and a (one week) lagged reversal of -18.9 basis points (specification 2, Table 8), which closely corresponds to the daily results. 24 exclude the possibility that Ras Gas bondholders over-reacted to news about the off-takers. Buying patterns as communicated to us by Goldman Sachs (Greg Randolph, Ghassanl Abdulkarim) suggest a competing explanation based on liquidity and clientele effects. In late 1997, liquidity in emerging bond markets disappeared and the only buyers of Ras Gas bonds were presumably better informed Middle-East based investors who perceived them As underpriced and, in the process, completely bought up the 2006 bond. As markets stabilized and yield spreads fell in 1999, liquidity improved and other institutional investors showed renewed interest in the more liquid 2013 bond. The weekly contagion pattern's positive relation between Ras Gas spreads and Middle-Eastern bond returns also offers support for such clientele based explanations (specifications 3 to 6, Table 8). Since we use mid-point closing yields we can exclu-de bid-ask bounce effects as a factor. Regarding output prices, we find that the BRENT coefficient is marginally significant at best. LNG settlement prices do not significantly impact Ras Gas credit spread levels and, hence, the bond's riskiness as priced in global markets. It seems quite remarkable that markets view output prices as irrelevant although they determine Ras Gas' revenue. However, in light of tle implicit price guarantee by Mobil, it is perfectly rational for bondholders to disregard price risk. This finding provides further evidence for our hypothesis that markets will not price risks that are explicitly part of the projects' contractual arrangements, in this case through the output price contingent debt service guarantee by shareholders to debtholders. To control for contagion effects, we test contemporaneous and up to five lags of daily regional emerging bond market returns for their statistical significance, successively eliminating 25 the least significant variables. We obtain the emerging market propagation pattern reported in specifications 4 and 5 of Table 2. As predicted, Ras Gas spreads vary negatively with emerging debt market returns.28 The significant lags hint at the time structure of the shock propagation mechanism behind the contagion effects. While Ras Gas spreads show a particularly strong contemporaneous reaction to European emerging debt markets (dominated by Russian debt) the other debt markets' impact is delayed and Asian debt market factors insignificant in the presence of Kepco spreads. Once again, we think that portfolio rebalancing and liquidity effects are responsible for these patterns. Ras Gas bonds belong both to the energy and emerging market segment of global fixed income markets. Based on information from Goldman Sachs (Greg Randolph, Bim Hundal), the Russian financial crisis impacted Ras Gas bond trading twice. Investors negatively reassessed Korea's and, hence, Kepco's prospects after Russia's partial debt default (sovereign spillover). Also, they viewed potentially increased oil and gas exports from the former Soviet Union as a financial threat to Ras Gas (sector spillover). The lagged reaction might be due to the lesser informational transparency of emerging markets as well as portfolio rebalancing that often takes place with time zone induced delays.29 Adding Korea-specific variables (specifications 6 to 8 in Table 2) to the regression reveals that the Korean rating index KRR is insignificant, while the Kepco share price KEPCO is significant at the 1% level. Exchange rate exposure as measured by the KRW coefficient comes 28 Longstaff and Schwartz (1995), and Duffee (1998) similarly find a negative relation between credit spreads and short-terrn interest rates. 29 After all, US investors, who might face a one-day delay in reacting to European or Asian events, initially held 80% of the 2013 bond. The positive relation between Ras Gas credit spreads and lagged Middle Eastern bond returns in the weekly analysis (Table 8) lends further credence to our portfolio rebalancing interpretation. 26 out negative which is puzzling: one would have expected that breach of contract and, hence, default risk increases as the Korean Won depreciates against the US Dollar, i.e., KRW rises. Instead, a depreciation of the Won seems to reduce Ras Gas risk perceptions, an issue that we will take up in the next section. The last regression reported in Table 2 combines the three components of Ras Gas bond riskiness: contract-related risk variables, Korean country exposure and emerging debt market spill-over effects. The results confirm our earlier findings. The net impact of Kepco yields is still 2 to 3 basis points, the significant Korea variables do not change in either identity or magnitude and the same is true for the emerging debt market contagion structure. Given potential non-stationarities in the dependent variable, we replicate the preceding analysis for changes in Ras Gas credit spreads to assess the previous results' robustness. IBy taking first differences in the dependent variable, we address potential non-stationarities in the data and arrive at the following specification: ARGS, = ,60 + /, BRENT, + E/62,_IKORELES,, + /83KEPCO, +/4 KRW, + /5KOSPI, +/36AKRR, + +[y,,MEA,, y21,ASIA,, + 73,EUR,, +y3-ILA T7]+ El O,]+c, where the dependent variable RGS is the Ras Gas credit spread, BRENT the logarithm of the Brent blend oil pr ce index, KORELES the Kepco credit spread, KEPCO the logarithm of Kepco's stock price, KRW the logarithm of the Koreanr Won - USD spot rate, KOSPI the logarithm of the Korea Composite Stock Price Index, KRR a rating index for Korea, and MEA, ASIA, EUR, LA T the continuously compounded daily retums of the JP Morgan regional total return indices in USD for emerg- ing markets in the Middle East-Africa, Asia, Europe and Latin America. Rho as a robust test for serial correlation (SC) with a lagged dependent variable reports the coefficient and p-value for the t-test of p = 0 (absence of SC) in the regression of re- siduals from the original specification on the same explanatory variables and lagged residuals, i.e., = XO + p, +t j. Specification 1 2 3 4 5 6 7 8 Dep. Variable RGS RGS RGS RGS RGS RGS RGS RGS Variable Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient P-Value P-Value P-Value P-Value P-Value P-Value P-Value P-Value Constant 1.9684 -.1300 -.0980 0.01273 -.1132 .8682 .6125 .4158 .0004 .0374 .1011 0.1515 .0488 .1637 .0115 .0778 RGS(-1) .9755 .9819 0.99812 .9834 .9679 .9650 .9817 .0000 .0000 0.0000 0.0000 .0000 .0000 .0000 BRENT -.3427 .0424 .0309 .0368 -.0089 .0038 .0348 .0571 .0340 .1064 .0462 .7345 .8518 .0584 KORELES .5972 .02496 .1558 .1499 .1631 .1609 .1523 .____________ .0000 .0000 .0000 0.0000 .0000 .0000 .0000 KORELES(-1) -.1365 -.1312 -.1230 -.1213 -.1282 .0000 0.0000 .0000 .0000 .0000 KEPCO .0704 . ________ __________ .0016 KRW -.1934 -.1248 -.0762 l_________ __________ .0219 .0006 .0207 KOSPI .0717 l_________ ___________ ___________ .0238 KRR .0050 .3450 _ - EUR -2.13609 -2.0387 -2.0053 -1.9789 -1.9841 0.0007 .0005 .0005 .0006 .0006 LAT -2.27699 .0129 _ LAT(- 1) -2.09425 -1.5725 -1.5412 -1.5594 -1.5663 0.0218 .0624 .0660 .0621 - .0626 MEA(-2) -3.09962 -2.7622 -2.8062 -2.8438 -2.8012 0.0040 .0058 .0048 .0042 .0051 EUR(-3) -3.07837 -3.2586 -3.3701 -3.2883 -3.2260 0.0000 .0000 .0000 .0000 .0000 LAT(-5) -2.57983 -2.1537 -2.0545 -2.1006 -2.1405 0.0050 .0114 .0151 .0127 .0116 Obs. 724 724 724 719 719 719 719 719 Adj. R2 .63995 .99558 0.99596 0.99569 .99631 .99636 .99638 0.99633 DW d Stat. .02340 1.90390 2.00086 1.89391 2.05283 2.05791 2.05229 2.06149 Rho (robust 0.9883 -0.0481 -0.0008 0.0517 -0.0294 . -0.0328 -0.0300 -0.0340 test for SC) .0000 .1987 .9822 .1697 .4375 .3929 .4350 .3698 39 Table 3: Unit Roots and Cointegration Tests Testing for unit roots we use Augmented Dickey-Fuller tests (ADF: correcting for serial correlation in the errors) of the form: Ay, == & + (y - I)y,, +Ay,-, + , where the dependent variable y is either RGS, the spread of the Ras Gas bond over 10 year US Treasury yields, KORELES, the spread of the 2013 7.75% Kepco global bond, or BRENT, the logarithm of the price of Brent oil. Similarly, we appeal to Augmented Engle-Granger tests (AEG) for cointegration because of the presence of serial correlation in the residuals of the cointegration equation. Specifically, we test for unit roots in the residuals drawn from the corresponding cointegration equa- tion, i.e.. y, = f6o + Ax,r + r, A, = ao + (a-) + A,_, + it, The one-sided asymptotic P-values for the zr statistic (both in bold face) under the null hypothesis (existence of unit root or cointegrated time series) are computed by the methods described in MacKinnon (1994). Test: Daily Data ADF ADF ADF AEG AEG AEG Dependent ARGS, AKORELES, ABRENT, Residuals: Residuals: Residuals: Variable RGS RGS RGS KORELES BRENT KORELES BRENT Variable Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient r -Statistic r - Statistic r - Statistic z- - Statistic r - Statistic z- - Statistic Constant .0099 .0318 .0094 .00003 .0015 .00005 RGS(-I) -.0028 -1.126 DRGS(-I) .0846 KORELES(- 1) -.0085 -2.029 DKORELES(-1) .2079 BRENT(-I) -.0033 DBRENT(-I) .0174 __________________ .467 E(-1) -.0132 -.0052 -.0124 _=______________ -2.261 -1.376 -2.186 DE(-I) .0671 .0851 .0641 P-value under Ho 0.7046 0.2740 0.9838 0.3932 0.8056 0.6478 Observations 723 723 723 723 723 723 DW d Stat. 2.01473 1.92602 1.99971 1.98719 2.01310 1.98945 Test: Weekly Data ADF ADF ADF AEG AEG AEG Variable r -Statistic T - Statistic r - Statistic r - Statistic r - Statistic r - Statistic RGS(- 1) ~-1.651 KORELES(-I -1.597 BRENT(- 1) -1.027 E(- 1) -1.739 -2.040 -1.744 P-value under Ho 0.4565 0.4851 0.7433 0.6589 0.5077 0.8405 Observations 138 138 138 DW dStat. 2.09283 2.01190 1.97969 1.97774 2.03566 1.97971 40 Table 4: Changes in Ras Gas Credit Spreads ARGS, = +8 /ABRENT, + E 82 ,KORELES,, +/3KEPCO, + fiKR -Rw +5KOSPI, + /3AKRR, OShSL + E I,lMEAI,+ r2.,AS14, y3-EURl-ly-LAT-l]+ef oIL where the dependent variable ARGS is the first difference of the Ras Gas bond spreads over 10 year US Treasury yields, BRENT the Brent blend oil price index in logarithms, KORELES the spread of the 2013 7.75% Kepco global bond and DKORELES its first difference, KEPCO the Kepco stock price in logarithms, KRW the Korean Won - USD spot rate in loga- rithms, KOSPI the Korea Composite Stock Price Index in logarithms, KRR a rating index for Korea, and MEA, A',A, EUR, LA T the continuously compounded daily returns of the JP Morgan regional total return indices in USD for emerging markets in the Middle East-Africa, Asia, Europe and Latin America. Specification_ 1 2 3 4 5 6 7 8 Dep. Variable DRGS DRGS DRGS DRGS DRGS DRGS DRGS DRGS Variable Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient C oefficient P-Value P-Value P-Value P-Value P-Value P-Value P-Value _P-Value Constant -.1319 .0008 .0189 .3312 .1364 -.1448 .0048 .0626 ____________ .0283 .8438 .6575 .2437 .2534 .0121 .2475 .8187 BRENT .0365 -.0064 .0560 .0617 .0423 l____________ .0594 .6701 .0336 .0058 .0233 DBRENT -.0185 -.9211 .9073 .9524 KORELES .1597 .1550 .1429 .1530 .1455 .0000 .0000 .0000 .0000 .0000 KORELES(-I) -.1512 -.1468 -.1386 -.1443 -.1326 .0000 .0000 .0000 .0000 .0000 DKORELES .1565 .1566 .1499 l___________ .0000 .0000 .0000 KEPCO -.0149 -.0659 .6858 .0290 KRW -.0406 -.0506 _7_ _ __S_ _ _ _ _.2438 _.1261 KOSPI -.0282 -.0506 .0730 .5983 .0118 .0642 DKRR .0380 .0384 _____________ __________ .0397 .0302 EUR -2.0585 -2.0501 -2.057 .0005 .0006 .0005 LAT(-1) -1.8165 -1.8228 -1.755 .0331 .0346 .0391 'MEA(-2) -3.1005 -3.0727 -3.058 .0022 .0027 .0025 EUR(-3) -3.1628 -3.1450 -3.195 .0000 .0000 .0000 LAT(-5) -2.4008 -2.4026 -2.317 ______________ .0052 .0057 .0069 Obs. 724 724 724 724 724 719 719 719 Ad'.R2 12408 .10983 0.1100 .13066 .13514 .20389 .18877 .20901 DWhd Stat. 1.98276 1.94776 1.9487 2.00345 2.00340 2.03588 1.99468 2.05788 41 Table 5: Simultaneous Equations: Ras Gas Credit Spread Levels RGS, = ,+ , + RGS1+ 11 BREN7T +/21 KORELES, +Y ,[1Y 1 ME41 + 72,, -AS4f + y34 1, EUR, -, + Y4 -,1 LA 7T,] + Os1 L 011l KORELES, = 102 + P12BRENT + L2I2KORELES, + /32KEPCO, + J4,KRW, + /54K0SP1, + 864gAKRI + E2 0lL where the dependent variable RGS is the spread of the Ras Gas bond over 10 year US Treasury yields, BRENT the Brent blend oil price index, KORELES the yield on the 2013 7.75% Kepco global bond, KEPCO the Kepco stock price in loga- rithms, KRW the Korean Won - USD spot rate in logarithms, KOSPI the Korea Composite Stock Price Index in logarithms, KRR a rating index for Korea, and MEA, ASIA, EUR, LAT the daily returns of the JP Morgan regional total return indices in USD for emerging markets in the Middle East-Africa, Asia, Europe and Latin America. We estimate by full information Maximum Likelihood. Specification 1 2 3 4 Dep. Variable RGS KORELES RGS KORELES RGS KORELES RGS KORELES Variable Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient P-Value P-Value P-Value P-Value P-Value P-Value P-Value P-Value Constant -.0962 -.6018 -.1037 .3452 .5383 .6506 .5397 .6536 .1072 .2639 .0832 .5649 .0288 .2924 .0193 .2894 RGS(-I) .9785 .9782 .9653 .9642 .0000 .0000 .0000 .0000 BRENT .0335 .1276 .0359 .1883 .0024 .1816 .1796 .0809 .0110 .0613 .0009 .9069 .0014 .0013 KORELES .0206 .0210 .0370 .0373 .0000 .0000 .0000 .0000 KORELES(-1) .9586 .9446 .9469 .9469 .0000 .0000 .0000 .0000 KEPCO .2191 .0625 .2382 .0633 .2380 .0047 .0047 .0023 .0012 .0023 KRW .2696 .2636 -.1090 .2195 -.1083 .2193 .0003 .0003 .0028 .0034 .0024 .0034 KOSPI -.2354 -.5073 -.5145 -.5137 .0000 .0000 .0000 .0000 DKRR .2431 .0717 .2898 .0695 .2899 .0000 .0000 .0000 .0001 .0000 EUR -2.1776 -1.3301 -2.0705 -1.9705 -2.0013 .0003 .3363 .0004 .0006 .0005 ASIA(-I) -.9312 -1.0715 -.92728 .0754 .3682 .0605 MEA(-2) -2.8146 .1324 -2.8355 -2.9679 -3.0781 .0070 .9554 .0045 .0025 .0017 EUR(-3) -3.3677 -.0418 -3.4154 -3.4000 -3.4034 .0000 .9762 .0000 .0000 .0000 LAT(-5) -2.2099 -.0261 -2.1757 -2.1589 -2.1215 .0127 .9897 .0105 .0096 .0111 Obs. 719 719 719 719 719 719 719 719 Log-Likelihd 554.8793 554.8793 576.4339 576.4339 591.7214 591.7214 589.9580 589.9580 DWdStat. 1.9353 1.5691 1.9411 1.7267 2.0017 1.7684 2.0025 1.7684 42 Table 6: Simultaneous Equations: Ras Gas Credit Spread Changes ARGS, = 0+ E &,1RGS,, +A,BRENT + /J2,KORELES, + I[y1,-ME4,,E_+y2,,ASI4,+y3,1,EUR-,+y4,lLAT,-]+1, O23 .0540 .0196 .0722 .0717 .0285 .2239 .0914 .1538 .0915 .5399 .0245 .0193 .3834 .0035 .0090 .0463 KORELES .1551 .1510 .1582 .1415 .0363 .0000 .0000 .0000 .0000 .0000 KORELES(-I) -.1360 -.1124 -.1487 -.1317 .9470 .0000 .0000 .0000 .0000 .0000 -I DKORELES .1387 __________ .0000 KEPCO .0713 .0626 .2376 .0347 -.0141 ______________ .0015 .0050 .0025 .2459 .8307 KRW -.1201 -.1017 .2243 .0035 -.0027 .0015 .0074 .0033 .9019 .9661 - KOSPI -.5077 -.1034 -.0696 .0000 .0066 .4081 DKRR .0382 .0413 .0716 .2897 .0382 .2910 .0289 .0195 .0000 .0000 .0374 .0000 EUR -1.9365 -2.0216 -1.9817 .0008 .0006 .0005 ASIA(-I) -.92218 -1.0567 -.92898 .0653 .0365 .0601 _ MEA(-2) -2.9791 -3.2279 -2.9539 .0027 .0013 .0026 EUR(-3) -3.3475 -3.2658 -3.4137 .0000 .0000 .0000 _ _ _ LAT(-5) -2.1336 -2.3750 -2.1167 .0115 .0055 .0113 Obs. 724 719 724 719 719 719 719 719 Adj. R2/Log-L .99596 .99638 .12835 .21364 592.4560 592.4560 535.7771 535.7771 DW d Stat. 2.00686 2.05491 1.99629 2.05796 2.0029 1.7697 2.0080 1.7923 44 Table 8: Weekly Estimation Results for Major Specifications As a robustness check, this table reports the estimation results for major specifications with weekly data. The variables and specifications are as in the preceding tables (Original Specification) with one obvious difference. Since the emerging market returns are now weekly continuously compounded returns, the contagion pattern cannot be expected to carry over from the daily data so that we specify and test for an appropriate weekly emerging debt market return lag structure. This propagation pattern is then used instead of the original lag structure. The other explanatory variables remain the same. S pecification .2 3 4 5 6 Original Table 2 Table 2 Table 2 Table 4 Table 5 Tasle 6 Specif-ication Spec. 2 Spec. 3 Spec. 7 Spec. 8 Sp z.4 Spec. 4 Dep. Variable RGS RGS RGS DRGS RGS KORELES DRGS LDKORELES Variable Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient _________ P-Value P-Value P-Value P-Value P-Value P-Value P-Value P-Value Constant -.5722 -.4438 2.9571 .2331 2.6035 4.1721 .9083 1.1216 _________ .0886 .1648 .0083 .8628 .0066 .1231 .0091 .1292 RGS(-1) .8768 .9155 .8460 .8518 i __________ .0000 .0000 .0000 .0000 BRENT .1869 .1376 -.0194 .6646 .1438 .4103 _________ .0825 .1794 .8323 .0064 .1580 .0575 DKORELFS .1765 .0000 KORELES .1257 .2829 .2707 .2159 .1632 .0000 .0000 .0000 .0000 .0000 KOREI,ES(-I) -.1896 -.0913 -.1649 .8240 .0001 .0275 .0001 .0000 KEPCO .3145 -.2199 .2679 .8098 .0014 .1313 .0006 .0152 KRW -.5719 -.1847 -.5019 .5272 .0008 .2444 .0007 .0933 KOSPI .2589 -1.8297 -.1962 -.3473 l_______ .1836 .0002 .0130 .0346 DKRR .1026 .1069 .2511 .9909 .2857 I________ __________ .0076 .0010 .0005 .0069 .0002 ASIA -1.8922 .0340 _ EIJR -4.3878 -4.6133 -4.6897 -4.4447 .0000 .0000 .0000 .0000 LAT -3.7872 -5.6687 -5.5776 -11.251 -5.3509 -14.180 l__ .0146 .0011 .0001 .0003 .0011 .0000 LAI(- I) -3.8421 -4.4369 -3.7012 -4.3350 .0101 .0060 .0071 .0048 MEA(-3) 7.5506 5.7465 7.3800 5.5789 .0001 .0057 .0000 .0048 LAT(-5) -4.3054 -6.7406 -4.7676 -6.5664 .0038 .0000 .0005 .0000 Ohs. 139 139 134 134 134 134 134 134 \idj. R2 .97702 .97936 .98731 .53476 Lo--Likelihd -48.509 -48.509 -72.901 -72.901 DW d Stat. 1.40219 1.65359 1.88652 1.83247 1.7453 2.0886 1.7461 2.0958 Rho (robust 3134 .1769 .0515 .0920 test lor SC) | .0003 .0429 .6084 .3475 45 REFERENCES Alchian, Armen A. and Harold Demsetz (1972), "Production, Information Costs, and Economic Organization," American Economic Review 62: 777-795. Banaszak, Sara (2001), "World Trade of LNG," in Fundamentals of the Global LNG Industry Yearbook 2001, Petroleum Economist March 2001, London, UK. Brand, L. and R. Bahar (2001), "Corporate Defaults: Will Things Get Worse before They Get Better?" Standard & Poor's Structured Finance January 2001. Brealey, R.A., I. A. Cooper, and M.A. Habib (1996), "Using Project Finance to Fund Infrastructure Investments," Journal of Applied Corporate Finance 9: 25-38. Campbell, J.Y. and P. Perron (1991), "Pitfalls and Opportunities: What Macroeconomists Should Know about Unit Roots," in O.J. Blanchard and S. Fischer (eds.), NBER Macroeconomics An- nual: 141-201. Chemmanur, T.J., and K. John (1996), "Optimal Incorporation, Structure of Debt Contracts, and Limited-Recourse Project Financing," Journal of Financial Intermediation 5: 372-408. Cox, Don R. and David R. Peterson (1994), "Stock Returns following Large One-Day Declines: Evidence on Short-Term Reversals and Longer-Term Performance," Journal of Finance 49: 255- 267. Duffee, G.R. (1998), "Treasury Yields and Corporate Bond Yield Spreads: An Empirical Analysis," Journal of Finance 53: 2225-2242. Duffie, D. and K.J. Singleton (1999), "Modelling Term Structures of Defaultable Bonds," Review of Financial Studies 12: 687-720. Esty, B.C. (1999), "Petrozuata: A Case Study of the Effective Use of Project Finance," Journal of Applied Corporate Finance 12: 26-42. Esty, B.C. (2000), "An Overview of the Project Finance Market," note, Harvard Business School. Esty, B.C. and W.L. Megginson (2000), "Syndicate Structure as a Response to Political Risk in the Project Finance Loan Market," mimeo, Harvard Business School. "Evolution and Development of the World's LNG Industry," in Global LNG Map, Petroleum Economist March 1998. Farna, E.F. (1990), "Contract Costs and Financing Decisions," Journal of Business 63: 71-91. F'nnerty, J. (1996), Project Financing. Asset-Backed Financial Engineering, Wiley. Green, Richard C. (1984), "Investment Incentives, Debt, and Warrants," Journal of Financial Economics 13: 115-136. Goldman Sachs (1996), Ras Laffan Liquified Natural Gas Co. Ltd, Offering Prospeotus, Goldman Sachs and Credit Suisse First Boston, December 1996. Jensen, M. and W. Meckling (1976), "Theory of the Firm: Managerial Behavior, Agency Costs, and Capital Structure," Journal of Financial Economics 3: 305-360. John, K. (1987) "Risk-Shifting Incentives and Signalling through Corporate Capital-Structure", Joturnal /f Finance 42: 623-641. Kaplan, S.K. and R. Ruback (1995), "The Valuation of Cash Flow Forecasts: An Empirical Analysis,' Journal of Finance 50: 1059-1093. MacKinnon, James G. (1994), "Approximate Asymptotic Distribution Functions for Unit-Root and1 Cointegration Tests," Journal of Business and Economic Statistics 12: 167-176. Longstaff, Francis A. and Eduardo S. Schwartz (1995), "A Simple Approach to Valuing Risky Fixed and Floating Rate Debt," Journal of Finance 50: 789-819. Madan, D. and H. Unal (2000), "A Two-Factor Hazard Rate Model for Pricing Risky Debt and the Term Structure of Credit Spreads," Journal of Financial and Quantitative Analysis 35: 43-65. Mello, A.S. and J.E. Parsons (2000), "Hedging and Liquidity," Review of Financial Studies 13: 127-153. Nkldeke, G. and K. Schmidt (1995), "Option Contracts and Renegotiation: A Solution to the Hlokl-Up Problem," Rand Journal of Economics 26: 163-179. Pepiatt, S. and S. Rixon (2000), "Breaking the Bond Barrier," in Legal Guide: The Devil's in the Deiails, Project Finance March 2000: 2-4. Randolph, G. and A. Schrantz (1997), "The Use of the Capital Markets to Fund the Ras (,-as Project.".Journal of Project Finance 3 (2): 5-15. Shah, S., and A.V. Thakor (1987), "Optimal Capital Structure and Project Financing," Journa.' of Economic Theory 42: 207-243. Smith, Clifford W., Jr., and Jerold B. Warner (1979), "On Financial Contracting: An Analysis of Bond Covenants," Journal of Financial Economics 7: 117-161. 47 Standard & Poor's (1996), "Ras Laffan Liquefied Natural Gas Co. Ltd.," Global Project Finance December 5, 1996. Standard & Poor's (1999), "Ras Laffan Liquefied Natural Gas Co. Ltd.," Infrastructure Finance October 1999. Standard & Poor's (2000), "Ras Laffan Liquefied Natural Gas Co. Ltd. 'BBB+' Rating Affirmed," Infrastructure Finance April 2000. Standard & Poor's (2001), "Alliance Pipeline," Infrastructure Finance February 2001. Tuffano, Peter (2001), "Introduction. HBS-JFE Conference Volume: Complementary Research Methods," Journal of Financial Economics 60: 179-185. Tusiani, Michael (2001), "The Commercialisation of LNG Markets," in Fundamentals of the Global LNG Industry Yearbook 2001, Petroleum Economist March 2001, London, UK. Zingales, Luigi (2000), "In Search of New Foundations," Journal of Finance 55: 1623-1653. 48 Policy Research Working Paper Series Contact Title Author Date for paper WPS2693 Helping People Help Themselves: David Ellerman October 2001 B. Mekuria Toward a Theory of Autonomy- 82756 Compatible Help WPS2694 Financial Development and Financing Inessa Love October 2001 K. Labrie Constraints: International Evidence 31001 from the Structural Investment Model WPS2695 Trade, Credit, Financial Intermediary Raymond Fisman October 2001 K. Labrie Development, and Industry Growth Inessa Love 31001 WPS2696 Firms as Financial Intermediaries: Asli DemirgGc-Kunt October 2001 K. Labrie Evidence from Trade Credit Data Vojislav Maksimovic 31001 WPS2697 Regional Integration and Industrial Dorsati H. Madani October 2001 L. Tabada Growth among Developing Countries: 36896 The Case of Three ASEAN Members WPS2698 Foreign Bank Entry: Experience, George Clarke October 2001 P. Sintim-Aboagye Implications for Developing Countries, Robert Cull 38526 and Agenda for Further Research Maria Soledad Martinez Peria Susana M. Sanchez WPS2699 Benefits and Costs of International Pierre-Richard Agenor October 2001 M. Gosiengfiao Financial Integration: Theory and Facts 33363 WPS2700 Business Cycles, Economic Crises, Pierre-Richard Agenor October 2001 M. Gosiengfiao and the Poor: Testing for Asymmetric 33363 Effects WPS2701 Trade and Production, 1976-99 Alessandro Nicita November 2001 L. Tabada Marcelo Olarreaga 36896 WPS2702 Productivity versus Endowments: Hiau Looi Kee November 2001 L. Tabada A Study of Singapore's Sectoral 36896 Growth, 1974-92 WPS2703 Integrating Independent Power Fiona Woolf November 2001 Energy Help Desk Producers into Emerging Wholesale Jonathan Halpern 30652 Power Markets WPS2704 Regulatory Governance and Chile's Ronald Fischer November 2001 G. Chenet-Smith 1998-99 Electricity Shortage Alexander Galetovic 36370 WPS2705 Concession Contract Renegotiations: Antonio Estache November 2001 G. Chenet-Smith Some Efficiency versus Equity Lucfa Quesada 36370 Dilemmas Policy Research Working Paper Series Contact Title Author Date for paper WPS2706 Household Income Dynamics Jyotsna Jalan November 2001 C. Cunanan in Rural China Martin Ravallion 32301 WPS2707 Financial Intermediary Development Thorsten Beck November 2001 A. Yaptenco and Growth Volatility: Do Mattias Lundberg 38526 Intermediaries Dampen or Magnify Giovanni Majnoni Shocks? WPS2708 Accountability and Corruption: Daniel Lederman November 2001 P. Soto Political Institutions Matter Norman Loayza 37892 Rodrigo Reis Soares WPS2709 Explaining Leakage of Public Funds Ritva Reinikka November 2001 H. Sladovich Jakob Svensson 37698 WPS2710 Breaking up the Collective Farm: Martin Ravallion November 2001 C. Cunanan Welfare Outcomes of Vietnam's Dominique van de Walle 32301 Massive Land Privatization WPS2711 Identifying Class Size Effects in Miguel Urquiola November 2001 H. Sladovich Developing Countries: Evidence from 37698 Rural Schools in Bolivia