IFPRI Disussion Paper 00893 August 2009 Managing Future Oil Revenues in Ghana An Assessment of Alternative Alloation Options Clemens Breisinger Xinshen Diao Rainer Shweikert Manfred Wiebelt Development Strategy and Governane Division
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE The International Food Poliy Researh Institute (IFPRI) was established in 1975. IFPRI is one of 15 agriultural researh enters that reeive prinipal funding from governments, private foundations, and international and regional organizations, most of whih are members of the Consultative Group on International Agriultural Researh (CGIAR). FINANCIAL CONTRIBUTORS AND PARTNERS IFPRI s researh, apaity strengthening, and ommuniations work is made possible by its finanial ontributors and partners. IFPRI gratefully aknowledges generous unrestrited funding from Australia, Canada, China, Denmark, Finland, Frane, Germany, India, Ireland, Italy, Japan, the Netherlands, Norway, the Philippines, Sweden, Switzerland, the United Kingdom, the United States, and the World Bank. AUTHOR Clemens Breisinger, International Food Poliy Researh Institute Researh Fellow, Development Strategy and Governane Division Email:.breisinger@giar.org Xinshen Diao, International Food Poliy Researh Institute Senior Researh Fellow, Development Strategy and Governane Division Email: x.diao@giar.org Rainer Shweikert, Kiel Institute for the World Eonomy Kiel, Germany Email: rainer.shweikert@ifw-kiel.de Manfred Wiebelt, Kiel Institute for the World Eonomy Kiel, Germany Email: manfred.wiebelt@ifw-kiel.de Noties 1 Effetive January 2007, the Disussion Paper series within eah division and the Diretor General s Offie of IFPRI were merged into one IFPRI wide Disussion Paper series. The new series begins with number 00689, refleting the prior publiation of 688 disussion papers within the dispersed series. The earlier series are available on IFPRI s website at www.ifpri.org/pubs/otherpubs.htm#dp. 2 IFPRI Disussion Papers ontain preliminary material and researh results. They have not been subjet to formal external reviews managed by IFPRI s Publiations Review Committee but have been reviewed by at least one internal and/or external reviewer. They are irulated in order to stimulate disussion and ritial omment. Copyright 2008 International Food Poliy Researh Institute. All rights reserved. Setions of this material may be reprodued for personal and not-for-profit use without the express written permission of but with aknowledgment to IFPRI. To reprodue the material ontained herein for profit or ommerial use requires express written permission. To obtain permission, ontat the Communiations Division at ifpri-opyright@giar.org.
Contents Aknowledgments v Abstrat vi 1. Introdution 1 2. A new era of oil in Ghana: new hallenges for growth and maro stability 3 3. Modeling alternative oil revenue alloation options 5 4. Impats of alternative oil revenue alloation options 8 5. Spending versus saving oil revenues: onluding remarks 17 Appendix A 18 Appendix B 21 Referenes 27 iii
List of Tables Table 1. Projetion of oil prodution and revenues 3 Table 2. Impats of oil revenue inflows on seleted eonomi variables in the model simulations, without onsideration of produtivity spillovers from publi investments (perentage-point hanges ompared to base-run growth rates) 12 Table 3. Impats of oil revenue inflows on seleted eonomi variables in the model, with onsideration of produtivity spillovers from publi investment in export setors (perentage-point hanges ompared to base-run growth rates) 14 Table 4. Impats of oil revenue inflows on seleted eonomi variables in the model, with onsideration of produtivity spillovers from publi investment in domesti setors (perentage-point hanges ompared to base-run growth rates) 15 Table A.1. Eonomi struture in the base year, Ghana, 2007 18 Table A.2. 2007 Ghana SAM 19 Table A.3. Expenditure shares and elastiities by household groups 20 Table A.4. Growth vs. aumulation: Deviation of GDP and oil fund from base run (100 million edis in onstant 2007 pries) 20 Table B.1. Mathemati presentation of DCGE model: sets, parameters, and variables 21 Table B.3. Mathematial presentation of DCGE model: model equations 24 List of Figures Figure 1. Additional spending and asset aumulation before, during, and after oil inflows (perent of initial GDP) 9 Figure 2. GDP plus interest on oil fund, deviation from base (perent of GDP) 16 iv
ACKNOWLEDGMENTS This paper has been prepared under IFPRI s Ghana Strategy Support Program (GSSP). We thank Kwaku Boateng (Ministry of Energy), Felix Asante (Institute of Statistial Soial and Eonomi Researh [ISSER]), Shashi Kolavalli (IFPRI-Ghana), James Sakey (IFPRI-Nigeria), Lothar Diehl (German Ageny for Tehnial Cooperation [GTZ]), Ian Gary (Oxfam), and Sebastien Dessus (World Bank) for their omments and insightful disussions. We are most grateful to Dr. Ernest Addison (Bank of Ghana) and his team for their invitation to present a draft of this paper at a Bank of Ghana seminar on June 30 in Ara, Ghana. Comments and disussions with Bank of Ghana staff have greatly improved the paper. We also aknowledge the omments from partiipants at the Global Trade Analysis Projet (GTAP) onferene in Santiago, Chile, and from partiipants in a poliy modeling ourse, organized by the University of Ghana and IFPRI, from July 12 to 25 in Ara, Ghana. Finally, we thank two anonymous reviewers for their valuable suggestions. The projet was funded by USAID under its Ghana Strategy Support Program (GSSP), the German Ageny for Tehnial Cooperation (GTZ) and the Poverty Redution, Equity, and Growth Network (PEGNet). v
ABSTRACT Contemporary poliy debates on the maroeonomis of resoure booms often onentrate on the shortrun Duth disease effets of publi expenditure, ignoring the possible long-term effets of alternative revenue-alloation options and the supply-side impat of royalty-finaned publi investments. In a simple model applied here, the government deides the level and timing of resoure-rent spending. This model also onsiders produtivity spillovers over time, whih may exhibit a setor bias toward domesti prodution or exports. A dynami omputable general equilibrium (DCGE) model is used to simulate the effet of temporary oil revenue inflows to Ghana. The simulations show that beyond the short-run Duth disease effets, the relationship between windfall profits, growth, and households welfare is less straightforward than what the simple model of the resoure urse suggests. The DCGE model results suggest that designing a rule that alloates oil revenues to both produtivity-enhaning investments and an oil fund is ruial to ahieving shared growth and maroeonomi stability. Key words: oil fund, publi expenditures, growth, Afria, Ghana, CGE analysis vi
1. INTRODUCTION Average growth rates of 5 perent and a halving of poverty over the past two deades make Ghana a reent suess story and a rising star in Afrian development (Aryeetey and Kanbur 2008; World Bank 2008a). Yet the ountry remains dependent on relatively few soures of foreign exhange inflows and retains a high level of external debt (IMF 2008). Therefore, the reent disovery of offshore oil is seen by many as an opportunity to overome persisting strutural weaknesses and has raised Ghana s prospets of beoming a frontrunner in Afrian development. However, experienes from other Afrian ountries suh as Nigeria and Zambia show that properly managing resoure windfalls remains a hallenge for many developing ountries and that misguided alloation strategies an harm the proess of eonomi development instead of aelerating growth (Gelb and Assoiates 1988; Auty 1990; Rodrik 2003). Crossountry evidene onfirms that ountries depending heavily on natural resoures tend to have less trade and foreign investment, more orruption, less equality, less politial liberty, less eduation, less domesti investment, and less finanial depth (Gylfason 2005 and Gylfason 2007). Given this inherent risk, two extreme positions for managing oil revenue inflows an be distinguished. 1 Proponents of a onservative strategy, suh as the Bank of Ghana (2007, 2008), argue that government spending of mineral windfalls often leads to exessive Duth disease effets, 2 where exhange-rate appreiation and ompetition for domesti resoures auses a redution in the ompetitiveness of non-oil setors, and orruption further undermines effetive spending (Gylfason, Herbertsson, and Zoega 1999; Eifert, Gelb, and Tallroth 2002; Gelb and Turner 2007). Moreover, notoriously volatile world oil pries and the physial limitations of mineral resoures ompound the importane of a sound revenue-management strategy. From this perspetive, the Norwegian model (essentially the saving of resoure inflows in an oil fund) onstitutes an important tool to phase in and out of oil-revenue spending and thus to support a balaned budget, a redution in foreign debts, and the aumulation of savings for future generations (see, for example, Gylfason 2007; Matsen and Torvik 2005). Advoates of a big push strategy argue that developing ountries often run fisal and trade defiits in development periods of rapid eonomi growth. Revenues from newly found oil resoures therefore provide an opportunity to inrease government investment to support growth. In fat, publi investments that failitate private-setor-led growth and raise produtivity have been identified as an important omponent for many ountries that transitioned rapidly from low- to middle-inome status (Syrquin 1988; Collier 2006; Breisinger and Diao 2009). investments in agriulture and rural development finaned by oil and opper revenues have played an important supporting role in growth and strutural transformation (Temple 2003). However, ross-ountry empirial evidene suggests that the impat of resoure inflows ritially depends on initial onditions, espeially on the strength of institutions and human apital (Brunnshweiler 2008; Bulte, Damania, and Deaon 2005; Gelb and Grassman 2008). Avoiding the resoure urse and turning future oil windfalls into an opportunity to aelerate eonomi transformation in Ghana will therefore require sound fisal management and a strategy that balanes urrent government spending and savings. In this paper, we fous on the alloation of oil revenues and use a dynami omputable general equilibrium (DCGE) model to assess trade-offs between different spending and saving senarios. The DCGE model inludes different types of publi spending and also introdues an oil fund. To alibrate the model s baseline (2009 to 2027), we draw information from the International Monetary Fund s (IMF s) urrent aount, government balane, and interest 1 3 In ountries suh as Indonesia and Chile, publi 1 For a omprehensive overview of these and intermediate positions, see van der Ploeg (2006). 2 There is broad agreement that strong appreiation of the real exhange rate hurt the ompetitiveness of export setors. However, Matsen and Torvik (2005) show that some real appreiation might be optimal, given the inflow of resoures and the potential to hange the growth path of the eonomy. 3 The ase studies inluded in Collier and Gunning (1998a, 1998b) enter on the savings response of publi and private agents when faed with trade shoks.
payments projetions. We then use the oil revenue projetions of the IMF and the Institute of Soial and Statistial Eonomi Researh (ISSER), a loal think tank, to assess the trade-offs between maroeonomi stability and produtivity-enhaning publi investments. As two extreme ases, we onsider a senario in whih the government spends all oil revenues it reeives annually and a senario in whih the government saves all oil revenues by reating an oil fund and spends only the interest earned from the fund. We also introdue an alloation rule, whih allows smoothing of oil-revenue spending, whereby a part of the oil revenue in eah period is saved for future government spending, thereby balaning urrent and future government spending. The rest of the paper is organized as follows. Setion 2 provides an overview of the size of the future oil setor in the Ghanaian eonomy and disusses related opportunities and hallenges in terms of balaning growth aeleration and maroeonomi stability. Setion 3 introdues the DCGE model used for this study, and Setion 4 presents the alloation of oil revenues and the potential impat of this alloation together with the model simulation results. Setion 5 summarizes and onludes. 2
2. A NEW ERA OF OIL IN GHANA: NEW CHALLENGES FOR GROWTH AND MACRO STABILITY Oil was disovered off the oast of Ghana in 2007, with total reserves estimated at between 500 million and 1.5 billion barrels and the potential for future government revenues estimated at US$1 1.5 billion annually (Table 1). Even at a modest long-term oil prie of US$60 per barrel over the next 20 years, oil revenues will add around 30 perent to government inome and will onstitute between 6 and 9 perent of gross domesti produt (GDP) annually over the period of exploitation. While the relative amount of expeted oil revenue is smaller than in other resoure-rih ountries, the positive shok does provide new opportunities to further aelerate growth and speed up eonomi transformation. Prospets are further raised by Ghana s sound institutional reord. Table 1. Projetion of oil prodution and revenues 2010 2015 2020 2025 2030 Barrels per day (in 1,000s) 120 250 250 250 250 Barrels per year (365 days) 43,800 91,250 91,250 91,250 91,250 Oil value (per day, in 1,000s) US$60 per barrel 7,200 15,000 15,000 15,000 15,000 US$80 per barrel 9,600 20,000 20,000 20,000 20,000 Oil value (per year, in 1,000s) US$60 per barrel 2,628,000 900,000 900,000 900,000 900,000 US$80 per barrel 3,504,000 7,300,000 7,300,000 7,300,000 7,300,000 Government revenue per day (in 1,000 edis) US$60 per barrel 2,750 5,730 5,730 5,730 5,730 US$80 per barrel 3,667 7,640 7,640 7,640 7,640 Government revenue per year (in 1,000 edis) US$60 per barrel 1,003,896 1,343,800 1,343,800 1,343,800 1,343,800 US$80 per barrel 1,338,528 2,788,600 2,788,600 2,788,600 2,788,600 Soure: Osei and Domfe (2008) Compared with other Afrian ountries in whih oil or other natural resoures have been disovered in the past, urrent onditions in Ghana seem favorable to avoiding the resoure urse. First, Ghana has experiene in managing resoure windfalls. Gold and ooa have been the most important export ommodities throughout the ountry s entire modern history. After the strutural adjustment program implemented in the mid-1980s, the ountry has finally reahed maroeonomi stability, and these favorable maroeonomi onditions, together with other pro-growth and pro-poor strategies, have led to steady growth and rapid poverty redution over the past 20 years. Seond, politially, Ghana has beome a stable demorati state, as demonstrated by peaeful transitions of power in two onseutive free and fair eletions, in 2000 and 2008. Third, the governane indiators reported by the World Bank show that Ghana has been steadily improving its governane, and in 2007 the ountry ranked above the regional averages for Asia, Latin Ameria, and Afria in most important governane indiators, suh as government effetiveness, regulatory quality, and ontrol of orruption (World Bank 2008b; Kaufmann, Kraay, and Mastruzzi 2008). 3
Yet several growth-, equity-, and marostability-related hallenges must still be addressed before Ghana an ahieve its development goals. The ountry aims to beome a middle-inome ountry by 2015, and ahieving this development goal will require annual growth rates of around 7 perent over the next 10 years (NDPC 2005; Breisinger, Diao, and Thurlow 2009). While this goal seems reasonable given the size of expeted oil revenues, this growth requirement is higher than the growth rates that the ountry has ahieved in reent years under favorable international onditions. In addition, the large share of agriulture in GDP (about 40 perent), the high share of agriulture-related proessing in manufaturing (about 60 perent), and the high share of the population working in agriulture (about 70 perent) indiate that without Green Revolution type agriultural growth as a main driver, Ghana may well fail to ahieve suh rapid growth (Breisinger et al. 2008). The pattern of urrent growth reveals ertain weaknesses in promoting private investment, generating more employment opportunities, and enouraging eonomi diversifiation. The distribution of growth benefits has also started to show warning signs, as inome growth in lagging northern regions does not math the fast growth in the oastal regions (Aryeetey and Kanbur 2008). Lessons from other ountries, and from Ghana s own history, show that maintaining maroeonomi stability is ruial for sustainable growth (Bank of Ghana 2007, 2008). Before implementation of the strutural adjustment program in the mid-1980s, ineffiient publi expenditure shemes, overvalued exhange rates, trade protetion, and an oversized publi setor held Ghana bak from transforming its eonomy (Agyeman Duah, Soyinka, and Kelly 2008). While Ghana also benefited from the Highly Indebted Poor Countries (HIPC) debt relief in 2002 to restore maroeonomi balane (IMF 2008), new debt has started to reaumulate reently due to rapidly inreasing fisal defiits. The fisal defiit onstituted about 3 perent of GDP in 2005, yet it is expeted to reah more than 10 perent in 2009 (IMF 2008). The IMF estimates that publi debt will rise to more than 50 perent of GDP in 2009, and other soures estimations are even higher (EIU 2009). Osei and Domfe (2008) argue that the food and energy risis is the reason for part of the additional spending. Other soures emphasize the sharp inrease in reurrent spending (espeially for ivil servants wages) and the stagnation of the share of investment in spending (EIU 2009). Reduing these high defiits to 5 perent of GDP would require an average annual growth rate of 7.7 perent to stabilize total publi debt at 65 perent of GDP. Therefore, oil revenues must help sustain a high spending-to-gdp ratio and will not be available for additional spending. 4 Hene, using these revenues to spur produtivity-led growth is ritial to ahieving growth and sustainable debt levels in the long run. Striking the right balane between growth and maroeonomi stability in the spending of oil revenues will therefore be a key hallenge. In the following setion we desribe a model to address the question of what this balane might look like. 4 For more details, see Breisinger et al. (2009). 4
3. MODELING ALTERNATIVE OIL REVENUE ALLOCATION OPTIONS The ability to apture synergies, trade-offs, and linkages between maroeonomi balanes and growth at the setor and household level have made general equilibrium models an important tool to analyze the impats of resoure booms. In this paper, we therefore use a reursive DCGE model for Ghana. While this model does not attempt to make preise preditions about the future development of the Ghanaian eonomy, it does measure the trade-offs between the alternative options of saving and spending oil revenues. The DCGE model is onstruted onsistent with the neolassial general equilibrium theory. The theoretial bakground and the analytial framework of CGE models have been well doumented in Dervis, de Melo, and Robinson (1982), while the detailed mathematial presentation of a stati CGE model is desribed in Lofgren, Harris, and Robinson (2002). A full desription of the DCGE model from whih our Ghana model is developed an be found in Thurlow (2004). The equations and parameters are presented in Appendix B. The Ghana DCGE model is an eonomywide, multisetoral model that solves simultaneously and endogenously for both quantities and pries of a series of eonomi variables. On the supply side, the model defines speifi prodution funtions for eah eonomi ativity. Assumptions that are made before alibrating the model to the data inlude onstant returns to sale tehnology with onstant elastiity of substitution (CES) between primary inputs. This is a neessary assumption for the model to reah a general equilibrium solution. For the substitution between primary and intermediate inputs in the prodution funtions, we assume a Leontief tehnology. The demand side of the DCGE model is dominated by a series of onsumer demand funtions. This demand system is derived from well-defined utility funtions. In our model, the onsumer demand funtions are solved from a Stone-Geary type of utility funtion in whih the inome elastiity deviates from 1 (whih is a typial assumption in a Cobb-Douglas type of utility funtion), and hene the marginal budget share of eah good onsumed differs from its respetive average budget share. As in other general equilibrium models, onsumers inome that enters the demand system is an endogenous variable in our model. Inome generated from the primary fators employed in the prodution proess is the dominating inome soure for onsumers, while the model also onsiders inomes from abroad (as remittane reeived) or the government (as diret transfers). The DCGE model expliitly models the relationship between supply and demand, whih determines the equilibrium pries in domesti markets. To apture the linkages between the domesti and international markets, the model assumes prie-sensitive substitution (imperfet substitution) between foreign goods and domesti prodution. 5 While the linkages between demand and supply through hanges in inome (an endogenous variable) and produtivity (often an exogenous variable) are the most important general equilibrium interations in an eonomywide model, prodution linkages also our aross setors through intermediate demand and ompetition for primary fators employed in prodution setors. The model has a neolassial losure in whih total domesti investment is determined by the sum of private, publi (budget surplus), and foreign savings (urrent aount defiit), net of publi savings abroad in the natural resoure funds. Publi investment is assumed to be a fixed proportion of overall domesti investment, while private investment is onstrained by total savings net of publi investment, where household savings propensities are exogenous. This rule, broadly onsistent with onditions in developing ountries where unrationed aess to world apital markets is virtually zero and domesti private saving is relatively interest inelasti, means that any shortfall in government savings relative to the ost of government apital formation, net of exogenous foreign savings, diretly rowds out private investment (and any exess in government savings diretly rowds in private investment). 5 Appendix Table A.1 provides seleted indiators of the export orientation of individual setors and the import dependene of domesti demand, together with information on setoral prodution and employment struture. 5
The model has a simple reursively dynami struture. Eah solution run traks the eonomy over the period 2007 to 2027, in whih eah period also orresponds to a fisal year. While publi and private apital stoks are fixed within eah year, they aumulate over time. Capital aumulation is affeted both by savings (partiularly for private apital aumulation) and government deisions regarding the alloation of publi funds. Investment (and hene apital aumulation) is also affeted by the foreign inflow of apital, in whih new oil revenues beome an important omponent in the simulations. Distribution of inreased apital aross setors is determined by the relative return to setor apital, and suh returns are the endogenous variables in a general equilibrium model. Speifially, the aumulation of setoral apital stok is defined as follows: K i,t = K i,t-j (1-μ i ) + ΔK i,t-j (1) where K i,t is the apital stok, μ i is the rate of depreiation that is setor speifi, and t-j measures the gestation lag on investment. In the simulations presented below, the default setting is j = 1, although the effets of assuming that publi investment augments the stok of infrastruture apital only with a longer lag may also be examined. The model also onsiders the effets of publi investment on produtivity as an externality fator resulting from publi investment in infrastruture. Publi investment is assumed to generate a Hiksneutral improvement in total fator produtivities. Speifially, the shift parameter in the prodution funtion, A i, hanges orresponding to the aumulation of publi apital, that is, A s,t = A s Π g {(K g t/k g 0)/(Q s,t /Q s,0 )} ρ sg (2) where g denotes a set of publi apital stoks generally defined over infrastruture, health, and eduation, K g and Q s are the publi apital stoks and setoral output levels under the simulation experiment, and K g 0 and Q s,0 are the orrespondingly defined publi apital stoks and output levels in the base period. The terms ρ sg determine the extent of the spillovers of publi investment to total fator produtivities. If ρ sg = 0, there is no spillover from publi investment in infrastruture, health, and eduation. The higher ρ sg, the higher the spillover effet. We alibrate this model to a new 2007 soial aounting matrix (SAM) for Ghana, whih is based on a 2005 SAM doumented in IFPRI 2007. To update this SAM to 2007, we use national aounts provided by Ghana Statistial Servies (GSS) for 2007, balane of payments data provided by the Bank of Ghana, and government budget data provided by the Ministry of Finane. The newly developed SAM provides information on the demand and prodution struture of eight setors, inluding two agriultural subsetors (domesti and export), four industrial subsetors (mining, manufaturing, utilities, onstrution), and two servie subsetors (private and publi). The SAM and hene the DCGE model onsider the existene of three different types of fators labor, apital, and land from whih rural and urban households derive their inome (see Appendix Table A.2). In addition to the SAM, the main elastiities inlude the substitution elastiity between primary inputs in the prodution funtion, the elastiity between domestially produed and onsumed goods and exported or imported goods, and the inome elastiity in the demand funtions. We use the same CES elastiity of 0.75 in the prodution funtion of all individual setors and for all pairs of prodution fators, whih is drawn from the CGE literature on other Afrian ountries. The other parameters or oeffiients in the prodution funtions of the model (e.g., the marginal produt of eah input) an be diretly alibrated using the ountry data of the Ghana SAM (e.g., the share of value-added for eah input used in the total value-added of this setor). For the use of intermediate inputs in the prodution funtion, we use a Leontief tehnology. With this assumption, a set of fixed input-output oeffiients an be diretly derived using the data of the Ghana SAM. With a Stone-Geary type of utility funtion applied in the model, the marginal budget share (MBS) is the parameter applied in the demand system, whih an be derived from the SAM given that the inome elastiity of demand is known. The inome elastiity is estimated from a semi-log inverse 6
funtion suggested by King and Byerlee (1978) and based on the data of Ghana Living Standard Survey (GLSS5 2005/06). The estimated results, together with the average budget share (ABS) for eah individual ommodity onsumed by eah individual household group diretly alulated using the data of the Ghana SAM, provide a series of MBSs that are applied in the model (see Appendix Table A.3). For ommodities that are sold both domestially and abroad, a onstant elastiity of transformation (CET) funtion is applied, while for ommodities that have both domesti and foreign supply, a onstant elastiity of substitution (CES) or Armington funtion is used. To reflet the relative openness of Ghana s eonomy, we hose high values, of 4.0, for both the CET and Armington elastiity for all traded goods. 7
4. IMPACTS OF ALTERNATIVE OIL REVENUE ALLOCATION OPTIONS We use this model to assess the medium- and long-term impats of four alternative oil revenue alloation options. This setion first desribes the senarios in greater detail and then disusses the ore results and sensitivity results. Senarios The DCGE model is first applied to a senario (the base run) in whih the setoral-level growth rate is onsistent with the growth trends observed in reent years (between 2001 and 2007). Newly found oil is not onsidered in this senario. Along this business-as-usual growth path, Ghana s eonomy will ontinue to grow at an annual rate of 5.6 perent until 2027. Oil extration will be onduted offshore, and the linkages to the domesti eonomy are expeted to be very weak. Bakward linkages will be restrited by the lak of loal apaity to provide the highly tehnology-, apital-, and skill-intensive inputs required for setting up and running oil extration businesses (Seminar at the Bank of Ghana on June 30, 2009, and personal interview with Ministry of Energy, July 1, 2009). Forward linkages are also likely to be limited, sine Ghana does not urrently have refining apaities and oil is expeted to be exported as rude oil. The most important effet from oil extration in Ghana will therefore be the royalties and taxes paid by the onsortium of oil ompanies. 6 Aordingly, we model the oil boom as an inrease in foreign exhange revenues to the government in the model. We then develop four poliy senarios in whih oil revenue as part of new foreign inflows to the 7 government aount is equivalent to 8.5 perent of GDP in 2007 (the base year in the model). This projetion of oil revenues is based on Osei and Domge from ISSER (2008), whih has been summarized in Table 1 above. Total oil revenues are then modeled as foreign inflows, whih are either exlusively used to finane inreased publi investment (senario 1) or are alloated as savings into an interest-earning external oil fund (senario 2). 8 Interests and revaluation gains (and losses) earned from suh savings are used to finane publi investments in senario 2. 9 Senarios 3 and 4 examine the ombination of these two extreme ases, in whih only part of the royalties are saved in the oil fund following different alloation rules. Finally, variations of these four basi senarios investigate the ases in whih publi investment not only inreases the overall apital stok, but in addition raises total fator produtivity in export-oriented setors (senarios 1a 4a) or domesti setors (senarios 1b 4b). Designs of these senarios are based on some other ountries praties. In order to guard against the destabilizing impats of swings in publi expenditure, ertain fisal rules have proved useful in anhoring long-term fisal poliy and in ensuring that windfall revenues are saved as a ushion against future adverse shoks. The best-known example of suessful fisal rules is Norway (Larsen 2006), where spending effets are ontrolled by the government shielding the eonomy through fisal disipline and investments abroad. Aumulating oil revenues in an oil fund would allow the fisal budget to be supported by a moderate but permanent inome stream stemming from interest on these assets. Saving at least part of the oil revenues would therefore provide some support to the budget, while at the same time moderating real appreiation and building up assets for buffering future shoks to foreign exhange inflows and/or fisal revenues. Aordingly, the government of Ghana has proposed the establishment of some form of permanent inome fund. However, the details of this plan, in terms of what proportion to 6 The onsortium of oil ompanies omprises Tullow Oil, Kosmos Energy, and several other small-sale operators. 7 We adopt the from the urrent perspetive more optimisti view that oil pries will average US$80 per barrel over the simulation period. However, assuming a lower oil prie will not hange the general diretion of the results. 8 We do not onsider the ase in whih royalties are used to finane additional reurrent publi expenditure, although a higher publi apital stok might provide a ase for additional expenditure. 9 Sine we assume that oil revenues are invested abroad, interest revenue also inludes revaluation gains and losses. 8
save and what proportion to spend, do not yet seem to be determined, and the publi onsultation proess has remained limited. To help in understanding the design of the senarios, we first provide a simple model to formalize the alloation between government savings and spending, assuming that the government follows a fisal rule to alloate oil revenues either to the fisal budget or to an oil fund. In this ase, total (additional) government spending in the urrent period is given by totgovspen d t = a 1 oilrev t + withdraw t = a 1 oilrevt + a2 oilfund t-1 (3) With 0 < a 1, a 2 < 1, the two parameters determine how muh of urrent oil revenues (oilrev) are alloated into the fisal budget (a 1 ) and how muh are added from aumulated assets (withdraw) out of an oil fund (a 2 ). The aumulation of assets in the oil fund are, in turn, given by the stok of assets in the previous period that inrease due to savings out of urrent oil revenues and derease due to withdrawals beause of additional government spending: oilfund t = oilfund t-1 + (1- a 1 ) oilrev t a 2 oilfund t-1 = (1- a 2 ) oilfund t-1 + (1- a 1 ) oilrev t. 10 (4) The impliations of this simple rule are shown in Figure 1, where we assumed an inflow on oil revenues of 10 eah period, starting in period 4 and lasting until period 20. As an be seen, the fisal rule smoothes the impat of oil revenues on the fisal budget where the spending parameter a 1 determines the (positive) shok at the beginning of oil revenue inflows and the strething parameter a 2 determines the live time of the oil fund. For our model simulations we adopted an intermediate parameter onstellation (0.5/0.15) with a spending parameter alloating 50 perent of urrent oil revenues into the fisal budget, an amount that is topped up by withdrawals of 15 perent of the oil fund stok in the previous period. Figure 1. Additional spending and asset aumulation before, during, and after oil inflows (perent of initial GDP) Spending (perent of GDP) 12 10 8 6 Spend All 0.5 / 0.15 0.75 / 0.15 0.25 / 0.05 4 2 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 10 Equations (3) and (4) imply that the government deides on net savings out of oil revenues in eah period. For the formulation of the alloation rule in terms of net savings, see Breisinger et al. (2009). For an appliation to aid inflows, see Adam et al. (2008). 9
Figure 1. (Continued) Assets (perent of GDP) 90 80 70 60 50 40 30 20 10 0 0.5 / 0.15 0.75 / 0.15 0.25 / 0.05 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Years Soure: Authors alulation for a simulated inflow of oil revenue of 10 units from period 3 to 20. Simulations show parameter ombinations of a 1/ a 2 (see equations 3 and 4 in the text). In terms of equations (3) and (4), we evaluate the following alternative senarios: OIL1: all oil revenues are spent, i.e., a 1 = 1, and a 1 = 0 OIL2: all oil revenues are saved, i.e., a 1 = 0, and a 2 = 0 OIL3: a 5 perent budget support, i.e., a 1 = 0.05 x GDP/OIL, and a 1 = 0 OIL4: smoothing oil-revenue spending, in whih a 1 = 0.5, and a 2 = 0.15 In the next step, we design four senarios to measure the diret and indiret effets of oil revenues that stem from additional publi spending or savings, ignoring possible spillover effets of publi investments on produtivity growth in the eonomy. Thus, we develop two additional sets of senarios to evaluate the joint effet of an inrease in oil revenues that leads to produtivity growth. In senarios 1a 4a the produtivity spillover effets are assumed to our in the export-oriented setors, while in senarios 1b 4b suh spillover effets are assumed to our in the domesti setors. For example, publi investment in teleommuniations infrastruture, trade fairs, and so on is likely to affet the prodution of tradable goods, while the provision and maintenane of rural roads, the funding of agriultural researh, or the establishment of loal marketplaes inreases produtivity in the domesti setors. Given that there is very little empirial onsensus on the size of the produtivity effets of infrastruture investments in developing eonomies, we assume a value of 0.5 for the spillover parameter in equation (2), that is, ρ sg = 0.50 in both ases. 11 This value is omparably higher than the values estimated by Hulten (1996), who studies the relationship of infrastruture apital and eonomi growth. This higher value reflets in part the expetation of a higher marginal produt of publi apital for ountries with a severely depleted apital stok and in part the likelihood that the ontemporary marginal produtivity of publi infrastruture expenditure in Ghana may be higher than the historial point estimates suggest. For eah senario, the average annual hanges for seleted variables over three periods (2007 2009, 2007 2013, and 2007 2027) are reported. To simplify the presentation, we report only a small number of key aggregate variables in tables: the real exhange rate, total and setoral exports, real GDP, total and government fixed investment, and real onsumption of rural and urban households. At given 11 This implies that a 10 perent rise in total publi apital stok relative to setoral output indues a 5 perent inrease in total fator produtivity. 10
onstant inome-tax rates and given savings rates, the hanges in real onsumption reflet hanges in real disposable inome. OIL1: A Spend All Strategy Fosters Growth, yet Leads to Duth Disease Effets and Hurts Rural Households In senario 1, the primary impat of oil revenues is an inrease in publi investment, whih leads to a higher level of real GDP growth in both the short and medium term ompared to the baseline (see Table 2). The real exhange rate appreiates modestly in the short run, whih is mainly due to the assumed high adjustment flexibility in foreign trade and domesti fator markets. However, inreases in investment demand indue signifiant hanges in the terms of trade and a sizable ontration in exports in favor of higher prodution of domesti goods. These Duth disease effets weaken over time, yet the effet of relative prie hanges on the ost of apital goods persists. This implies that although these hanges moderate over time, the initial deline in export performane does not reverse drastially, and hene initial welfare gains inrease only slightly in the long run. Total household inome inreases in real terms from the base run, yet the gains are not distributed equally aross household groups. While the inome levels of urban households inrease, rural household inomes fall in the short run and remain largely unhanged in the long run. The main reason for these inreasing disparities between rural and urban households is demand-side effets. Additional government investment is primarily spent on apital goods and onstrution, raising the relative pries for these mainly urban setors. Moreover, the bakward linkages from these urban industrial setors to the agriultural setors, from whih many rural households earn their inome, are weak. In addition, the agriultural export setor is hurt by exhange rate appreiation, whih exaerbates the negative effets for rural households. As later results show, these demand-side effets may be largely offset when oil revenues are used for produtivity-enhaning investments but may reemerge and inrease when relative prie effets turn against agriulture. OIL2: A save all and Invest Interest Only Strategy has very Limited Growth Effets In senarios 2 4, all or part of the oil revenues are saved in an oil fund (as foreign assets) and only interest earned from the fund is used to finane investments. In senario 2, a moderate but permanent inome stream generated from interest earnings finanes additional investment of a modest 1.7 and 1.4 perentage points in the short and long run, respetively. Thus, there is now less umulative growth in GDP both in the short and long term and a marked redution in total investment ompared to senario 1. As a onsequene, relative prie hanges are less pronouned in the short run but inrease over time with inreasing inome from interest. Total exports deline in the short run but inrease in the long run, ompared with the baseline. However, the hanges are signifiantly smaller than in senario 1. The short- and long-run effets on household inomes are signifiantly smaller than in senario 1. Savings to reate the oil fund largely avoid the short-run redution of rural households inome. In the long run, rural households inome is not affeted by the deision to either spend or save oil revenues, indiating the need for more targeted spending to raise rural households inome. 11
Table 2. Impats of oil revenue inflows on seleted eonomi variables in the model simulations, without onsideration of produtivity spillovers from publi investments (perentage-point hanges ompared to base-run growth rates) Experiment Period BASE OIL1 OIL2 OIL3 OIL4 Spending Saving Budget Support Exhange rate to T = 2009 0 to T = 2013-0.1-0.2 0-0.1-0.2 to T = 2027-0.2-0.2-0.1-0.1-0.2 Exports to T = 2009 5.7 to T = 2013 5.9-4.6-0.7-1.9-3.5 to T = 2027 7.1 1.1 0.3 0.7 1 Agriulture to T = 2009 7.5 to T = 2013 6.6-5.3-0.7-2.5-3.9 to T = 2027 3.3-3.3-1.7-2.7-3.1 Mining to T = 2009 4.3 to T = 2013 5.9-4.3-0.9-1.7-3.5 to T = 2027 9.9 2.1 0.9 1.5 1.9 Servies to T = 2009 4.2 to T = 2013 3.9-3 -0.5-1.4-2.2 to T = 2027 2.9-1.3-0.7-1.1-1.2 Real GDP to T = 2009 4.6 to T = 2013 4.8 0.6 0.1 0.3 0.4 to T = 2027 5.4 0.7 0.4 0.6 0.7 Investment to T = 2009 6.1 to T = 2013 6.2 9 1.7 4.4 7.1 to T = 2027 6.8 2.1 1.4 2 2.1 Government investment to T = 2009 6.3 to T = 2013 6.4 16.8 17 16.9 16.9 to T = 2027 7.1 3.4 3.6 3.6 3.5 Real inome Rural to T = 2009 3.9 to T = 2013 3.9-0.4 0-0.2-0.3 to T = 2027 4.1 0 0 0 0 Urban to T = 2009 5.2 to T = 2013 5.4 0.8 0.1 0.4 0.6 to T = 2027 6 0.5 0.3 0.4 0.5 Real wage to T = 2009 1.9 to T = 2013 2.2 1.2 0.1 0.5 0.9 Soure: DCGE model results to T = 2027 2.8 0.8 0.4 0.6 0.7 Smoothing 12
OIL3: Supporting the Budget with Oil Revenues Stabilizes the Debt Ratio, yet Only Modestly Aelerates Growth Due to Duth Disease Effets While senarios 1 and 2 represent two extreme ases of oil revenue alloation, senarios 3 and 4 onsider a strategy in whih the government deides to save only part of oil revenues in the oil fund. In the ase of senario 3, we assume that, starting with the inflows of oil revenues in 2010, an oil revenue equivalent of 5 perent of GDP is retained to support the publi budget. Compared to senario 2, this alloation rule results in both inreasing savings in the oil fund and inreasing investment until 2013. As expeted, an inrease in investment demand results in a worsening of the terms of trade in the short term, with the above-mentioned onsequenes. After 2013, with stagnant oil revenues but inreasing GDP, savings deposited into the oil fund derease steadily. However, inreasing interest revenue from the oil fund still allows for an inrease in investment and GDP growth over the whole simulation period. Impats are omparable to those of senario 1, yet with less short-run adjustment ost for rural households. OIL4: In the Long Run, Smoothing Oil-Revenue Spending Balanes Growth, Distribution, and Stability Targets The alloation rule in senario 4 implies that savings in the oil fund inrease with rising oil revenues and derease with a rise in savings stoks. Thus, given the parameterization indiated above, savings in the oil fund inrease between 2010 and 2013 and start dereasing thereafter, until 2027. The long-run maroeonomi, setoral, and distributional outomes are omparable to those of senario 1, in whih all oil revenues are immediately invested as they flow in. However, ompared to spending oil revenues immediately (senario 2), saving part of the revenues in an oil fund has the obvious advantage of providing inome from interest, even after the oil boom. In addition, the short-run adjustment osts with respet to exports and rural inomes are signifiantly lower than in the spending senario. Robustness Chek: Produtivity Improvements Due to Investments Aelerate Growth and Can Offset Duth Disease Effets In senarios 1a 4a and 1b 4b, publi investments raise the produtivity in the eonomy. The absene of produtivity inreases in domesti prodution (in 1a 4a) leads to a stronger appreiation of the real exhange rate ompared to senarios 1 4 (see Table 3). Hene, although the export performane of traditional ash rops and servies is signifiantly stronger beause of the produtivity bias, mining exports are hit relatively hard, with no produtivity effet. When the produtivity gain is biased toward prodution of domesti goods (in 1b 4b), outomes are markedly different (Table 4). The bias in prodution (whih inreases the supply of nontradables and import substitutes) partly offsets the demand effets of the inreased foreign exhange inflows so that the initial real exhange rate appreiation is reversed, even in the short run. Yet the effets on exports are symmetrial with those in senarios 1 4; ash rop exports are hit even harder, while servies exports are less affeted and mining exports reover less than in earlier experiments. Overall export performane is weaker with a domesti bias than with an export bias. The domesti-biased supply response also leads to a larger improvement in the long-run fisal balane, refleting favorable relative prie movements as well as the effets of higher growth and investment than in either the ase without produtivity effets or export-biased forms of produtivity growth. The most striking differene between senarios 1a 4a and 1b 4b is the effet on households disposable real inome. Compared to the ase of no produtivity effets, produtivity effets indued by publi investment lead to higher real inome growth for both household groups. However, the inome gain is spread somewhat differently aross household groups, with rural households benefitting less than urban households when produtivity effets are biased toward domesti prodution. This ontrasts sharply with the export-biased supply response, whih generates a lower aggregate real inome gain in the long run but favors rural households. 13
Table 3. Impats of oil revenue inflows on seleted eonomi variables in the model, with onsideration of produtivity spillovers from publi investment in export setors (perentage-point hanges ompared to base-run growth rates) Experiment Period BASE OIL1 OIL2 OIL3 OIL4 Spending Saving Budget Support Smoothing Exhange rate to T = 2009 0 0.1 to T = 2013-0.1-0.5 0-0.2-0.3 to T = 2027-0.2-0.1-0.1-0.1-0.1 Exports to T = 2009 5.7-0.2 to T = 2013 5.9-1.5-0.7-0.7-1.3 to T = 2027 7.1 0.9-0.2 0.4 0.7 Agriulture to T = 2009 7.5-1.6 to T = 2013 6.6-1.8-0.9-1.2-1.6 to T = 2027 3.3 2.9 2.5 2.7 2.8 Mining to T = 2009 4.3 0.9 to T = 2013 5.9-1.2-0.5-0.5-1.1 to T = 2027 9.9-0.1-1.4-0.7-0.2 Servies to T = 2009 4.2 0.7 to T = 2013 3.9-1.9-0.6-0.2-1.3 to T = 2027 2.9 2.6 0.3 2.4 2.5 Real GDP to T = 2009 4.6 0 to T = 2013 4.8 1.2 0.1 0.6 0.9 to T = 2027 5.4 0.9 0.3 0.7 0.8 Investment to T = 2009 6.1 0 to T = 2013 6.2 9.1 1.7 4.8 7.2 to T = 2027 6.8 2.3 1.4 2.1 2.3 Government investment to T = 2009 6.3 0.1 to T = 2013 6.4 16.9 17 16.9 16.9 to T = 2027 7.1 3.5 3.7 3.7 3.6 Real inome Rural to T = 2009 3.9 0 to T = 2013 3.9 0.4 0 0.2 0.3 to T = 2027 4.1 0.5 0.1 0.4 0.5 Urban to T = 2009 5.2 0 to T = 2013 5.4 1.3 0.1 0.6 0.9 to T = 2027 6 0.4 0 0.3 0.4 Real wage to T = 2009 1.9 0 to T = 2013 2.2 1.8 0.1 0.8 1.3 to T = 2027 2.4 1.4 0.8 1.2 1.4 Soure: DCGE model results 14
Table 4. Impats of oil revenue inflows on seleted eonomi variables in the model, with onsideration of produtivity spillovers from publi investment in domesti setors (perentagepoint hanges ompared to base-run growth rates) Experiment Period BASE OIL1 OIL2 OIL3 OIL4 Spending Saving Budget Support Smoothing Exhange rate to T = 2009 0 0 to T = 2013-0.1 0 0.1 0.1 0.1 to T = 2027-0.2 0 0 0 0 Exports to T = 2009 5.7-0.1 to T = 2013 5.9-5.1-0.8-2.2-3.9 to T = 2027 7.1 0.8 0.1 0.4 0.7 Agriulture to T = 2009 7.5 0.1 to T = 2013 6.6-5.8-0.8-2.7-4.2 to T = 2027 3.3-3.5-1.9-2.9-3.4 Mining to T = 2009 4.3-0.1 to T = 2013 5.9-5.2-1.1-2.1-4.2 to T = 2027 9.9 1.7 0.6 1.1 1.6 Servies to T = 2009 4.2 0.1 to T = 2013 3.9-2.7-0.3-1.2-2 to T = 2027 2.9-1.2-0.6-0.9-1.1 Real GDP to T = 2009 4.6 0.1 to T = 2013 4.8 0.9 0.1 0.5 0.7 to T = 2027 5.4 1 0.6 0.8 1 Investment to T = 2009 6.1 0 to T = 2013 6.2 9 1.7 4.7 7.1 to T = 2027 6.8 1.1 1.5 2.1 2.2 Government investment to T = 2009 6.3 0.1 to T = 2013 6.4 17.1 17.1 17.1 17.2 to T = 2027 7.1 3.7 4 3.9 3.8 Real inome Rural to T = 2009 3.9 0 to T = 2013 3.9-0.1 0 0-0.1 to T = 2027 4.1 0.3 0.2 0.2 0.3 Urban to T = 2009 5.2 0.1 to T = 2013 5.4 1.1 0.2 0.6 0.8 to T = 2027 6 0.9 0.5 0.7 0.8 Real wage to T = 2009 1.9 0.3 to T = 2013 2.2 1.7 0.2 0.8 1.2 to T = 2027 2.8 1.2 0.7 1 1.2 Soure: DCGE model results 15
Comparing Strategies: The Oil Fund will Provide Revenues Long after Oil Resoures are Depleted Figure 2 summarizes the differene in oil-related revenues between a situation with and without an oil fund. We ompare the simulation results with the base run and show the deviations in terms of real GDP adjusted for interest inome from the oil fund on the basis of an interest rate held onstant at 5 perent. This omparison shows that the loss in terms of inome between senario 1 and senarios 2 or 3 is modest. For senario 2, GDP plus interest is about 1.5 (1.2) perent lower in 2013 (2027), indiating that the gap loses over time. For senario 3, in whih oil revenues worth about 5 perent of GDP are alloated to eah year s fisal budget, the gap is even smaller and aounts for only 0.2 perent in 2027. For senario 4, in whih oil-revenue spending is smoothed over time aording to two parameters for urrent spending and for drawing out of the oil fund, the GDP figures adjusted for additional interest inome are even larger than in senario 1. Given that the oil fund allows the ountry to ontinue to enjoy the inome generated from oil even after the end of the oil era, this senario appears to be the best option for long-run growth and stability. Figure 2. GDP plus interest on oil fund, deviation from base (perent of GDP) Soure: Authors alulation from Appendix Table A.4 16