An Eonomi Assessment of Alternative Prodution Pathways for Peruvian Biofuels Prodution JULY 2010 IFPRI
The onlusions given in this report are onsidered appropriate for the time of its preparation. They may be modified in light of further knowledge gained at subsequent states of the projet. The designations employed and the presentation of material in this information produt do not imply the expression of any opinion whatsoever on the part of the Food and Agriulture Organization of the United Nations (FAO) onerning the legal or development status of any ountry, territory, ity or area or of its authorities, or onerning the delimitation of its frontiers or boundaries. The mention of speifi ompanies or produts of manufaturers, whether or not these have been patented, does not imply that these have been endorsed or reommended by FAO in preferene to others of a similar nature that are not mentioned. The views expressed in this information produt are those of the author(s) and do not neessarily reflet the views of FAO.
Introdutory Note Using a Computable General Equilibrium (CGE) Model to analyze the effets of biofuel polies in Peru By Yasmeen Khwaja Introdution In 2007 a blending poliy was established in Peru imposing the following mandates: Biodiesel: obligatory 2% (B2)blending starting in 1 January 2009 obligatory 5% (B) blending starting in 1 January 2011 to replae B2 Ethanol: obligatory 7.8% starting in 1 January 2009. The two main objetives of the biofuel blending mandate are i) to diversify energy soures and ii) to reate growth and employment opportunities. It is this seond objetive that has provoked muh interest within the poliy mahinery of Peru. A key question is how the biofuel mandates in Peru should be implemented to optimize the soial benefits that ould potentially emerge. There are of ourse many avenues through whih biofuels an be developed. However, speifi biofuel pathways may potentially offer greater soial and eonomi benefits ompared to others. Thus, the BEFS projet in Peru ommissioned a Computable General Equilibrium (CGE) analysis in order to quantify the eonomy-wide impats of implementing partiular biofuel pathways. The advantage of the CGE approah is its ability to measure the ultimate impat of a poliy or a number of poliies on aggregate welfare by quantifying the hange in the inome and onsumption of various groups who may be affeted by the poliy either diretly or indiretly. In the ontext of the Peruvian biofuel mandate, the CGE model simulates the expansion of biofuels prodution by prediting hanges in onsumer pries and household inomes. This is then used to estimate hanges in poverty, both at the national level and for different households groups (e.g., rural/urban). The CGE analysis in Peru onsidered nine alternative prodution senarios or pathways to explore the impat of developing the biofuel setor on the eonomy of Peru 1. 1 Five analyze the impat of ethanol prodution and four analyze the impat of biodiesel prodution. Ethanol and biodiesel are produed using a variety of tehnologial pathways using two feedstoks for
Construting a CGE model for an analysis of biofuel poliy on development variables In using CGE models to analyse any developmental impats of biofuels poliy the set-up of a model beomes quite ritial. The set-up refers to the harateristis of the model and the assumptions used to define the model. CGE models an be ountry, regional or globally based. Sine the BEFS projet was interested in looking only at impats within Peru a one-ountry CGE model was developed. A number of assumptions have been made with respet to employment, tehnology and other key parameters. Generally speaking assumptions should relate to the eonomy under onsideration but in all ases they an be relaxed or made more stringent depending on the kinds of questions that a poliy-maker want to answer. Sometimes assumptions are made for analytial simpliity and do not alter the outome of the results generated by the model. The distintion between stati and dynami models is an important one in CGE analysis. The use of a partiular type of model depends on the kind of poliy impats a government is interested in analysing. In some ases a stati model may serve the poliy purpose and in other ases a dynami model is more useful. A stati or omparative-stati CGE model onsiders the effets on an eonomy of a poliy hange of the eonomy at only one point in time. From a poliy analysis point of view, these results show the differene (usually reported in perent hange form) between two alternative future states (with and without the poliy shok). The proess of adjustment to the new equilibrium is not expliitly represented in suh a model. However, while a stati model provides a one-time snapshot of poliy impats some poliymakers may be interested in seeing what happens during the period of transition. For example, there may be unemployment and poverty effets during adjustment whih are not aptured by the analysis. Seondly, soial benefits from a partiular poliy may take time to take effet. A stati model may only partially apture full effets. In addition, any negative shorter term effets may be offset in the longer term and vieversa. For this reason poliy-makers may prefer a dynami analysis. By ontrast, a dynami CGE models expliitly traes eah variable through time often at annual intervals. These models are more realisti, but more hallenging to onstrut and solve they require for instane that future hanges are predited for all exogenous variables, not just those affeted by a possible poliy hange. eah biofuel. Sugarane and molasses produe ethanol, and palm oil and jatropha produe biodiesel. For both biofuels, feedstok prodution in most ases (exept for molasses and one of the palm oil produing senarios) is a 60/40 share between large sale ommerial farms and smallholders, respetively. 4
As time onstraints did not permit for a dynami analysis, the CGE analysis used in Peru uses a stati model that assumes full employment of prodution fators. Whilst this assumption is quite limiting, the stati model provides a preliminary basis for raising awareness on some poliy and methodologial issues. However, given that the stati model measures impats only at one point in time it is lear that a future analysis should be based on a dynami model to show the trajetory of impats. Moreover, some of the assumptions of the stati model an be hanged to allow for sensitivity analyses in order to shows how impats hange depending on the assumptions and model speifiations. The Peru CGE analysis- a starting point for future analysis The Peru analysis has been arried out using the Standard Stati IFPRI model. This was used to quantify the effets of developing the biofuel setor in Peru and its impat on setoral and eonomy-wide produtivities, resoure alloation, and welfare. The analysis to date should be seen as an exploratory analysis to illustrate how CGE models an be used to onsider outomes from speifi biofuel pathways. As with any analysis, there is always an opportunity to enrih the way a model is onstruted to allow for a more in depth onsideration of speifi effets. Thus the results of this analysis must not be seen as definitive or preditive in any sense. Rather, the model illustrates some important issues that would require a more omprehensive CGE analysis set in a dynami framework without the full employment assumption. This would allow for a better examination of development effets over time and speifially to apture growth and employment effets. The stati model full-employment model is onstrained by only allowing redistributive effets to take plae. The assumption of full-employment means that the growth effets from biofuels poliy annot be fully aptured. A further diretion for future analysis relates to the household setor. In the urrent analysis this is divided into rural and urban households. A future analysis may well onsider rural/urban, regional, and poverty status. Thus households ould be disaggregated for example as: urban/poor and urban/non-poor Selva/poor and Selva/non-poor Coastal/poor and Coastal/non-poor Suh a lassifiation would apture better the rural dimension of poverty in Peru. Depending on government priorities other deisive household harateristis may be inluded within the model suh as female-headed households or indigenous households. The CGE analysis in Peru to date presents a generalized and preliminary onsideration of the impats biofuel poliy on the Peruvian eonomy. For the government wanting to use CGE 5
analysis further, negotiations will be required between the poliy mahinery and CGE analysts in order to determine the preise nature of what is to be analysed. On the basis of this, a model that reflets the areas of interest of any impats from a poliy hange an be set up. In addition negotiations are a neessary part in gaining onsensus in how to define the key harateristis of the eonomy so that the final model reflets the reality and to allow for speifi impats to be identified. It is a serious mistake to use CGE analyses as an eonomi rystal ball. Results generated by CGE models are not foreasts but relate to how one set of results differs from the benhmark equilibrium of the model. In other words, how do the results of a partiular biofuel pathway ompare to the identified benhmark of no biofuel mandates in Peru? It should be noted that the strength of CGE analyses for poliy purposes does not lie in their preditive auray, but in their ability to shed light on the eonomi mehanisms through whih prie adjustments are transmitted in markets. CGE models are often used to onsider what if senarios. Indeed, in Peru the CGE analysis asks: what if feedstoks are produed through a 60-40 partnership of large farms and smallholders, what are the impats on the eonomy. These impats however should to be analyzed in terms of the dynamis of the eonomi interations that generate them. This means, ritially, one needs to distinguish to what extent results are generated by the harateristis and assumptions of the model. CGE analysis often does not aount for these linkages. Thus, sensitivity analysis plays an important role in examining the extent to whih results are driven by the model harateristis. The Peru analysis arried out a number robustness tests to show the impats of hanges in assumptions. 6
An Eonomi Assessment of Alternative Prodution Pathways for Peruvian Biofuels Prodution Perrihan Al Riffai and David Laborde 2 JULY 2010 IFPRI 2 Authors are members of the International Food Poliy Researh Institute 7
Table of Contents I. Introdution and Bakground...14 II. CGE Methodology...16 A. Standard CGE Model Assumptions...16 1. Ativities, Prodution, and Fators...16 2. Institutions: Households, Government, and Rest of the World (RoW)...17 3. Commodity Markets...18 4. Maroeonomi Balanes...20 B. Modeling the Biofuel Setor...21 1. Tehnologial pathway for Ethanol and Biodiesel Produtions...21 2. Biodiesel Prodution...23 3. Feedstok Prodution...23 4. Trade...24 III. The Model s Database: Peru 2002 Soial Aounting Matrix (SAM)...26 A. Peru SAM, 2002...27 1. The Aggregated SAM...27 2. Peru SAM...29 B. Biofuel Setors in the Peruvian SAM...30 1. In the Agriultural Setor...30 2. Non Agriultural Setors...30 C. Struture of the Eonomy...30 IV. Peru CGE Model Results...35 A. Impat on Gross Domesti Produt...35 B. Setoral Prodution...37 C. Farm Value Added...39 D. Inome effets and welfare...41 V. Robustness Cheks...45 A. Land Robustness Chek...45 B. Trade Robustness Chek...47 C. Optimal Feedstok Alloation Robustness Chek...47 VI. Conlusion...48 VII. Annexes...51 A. Tables and Figures...51 B. Standard Model in Mathematial Notation...61 C. Referenes...68 8
Tables and Figures Tables Table 1: Prodution Matrix for Biofuel Prodution in Peru and Resultant Poliy Senarios...22 Table 2: Peru Aggregated SAM, 2002 (Billions of Soles)...28 Table 3: Ativities, Fators and Institution in the Peru SAM...29 Table 4: GDP Composition...31 Table 5: Setoral Struture in Peru, 2002, (%)...32 Table 6: Poverty Headount, % of Total Population...33 Table 7: Soures of Inome and Contribution to Fators of Prodution (%)...34 Table 8: Seleted Household Expenditure Shares (%)...34 Table 9 : Setoral Prodution % Change ompared to the Baseline...38 Table 10 : Setoral Prodution of Biofuels and their Feedstoks Volume Changes Compared to the Referene Situation...39 Table 11 : Value Added per Hetare (Ethanol Prodution), % Change over Baseline...40 Table 12 : Value Added per Hetare (Biodiesel Prodution), % Change over Baseline...40 Table 13 : Equivalent Variation, Value, Millions Soles...42 Table 14 : Additional Land Needed to Stabilize Land Rends in Real Terms, 1,000 Ha...45 9
Figures Figure 1: Prodution Tehnology...17 Figure 2: Flow of Marketed Commodities...18 Figure 3: Real GDP...35 Figure 4 : Farm Value Added...39 Figure 5 : Household Inome (Before Redistribution of Firm Profits)...42 Figure 6 : Household Food Consumption...43 Figure 7 : Household Fuel Consumption...44 Figure 8 : Rural Household Welfare, Robustness Cheks for Land Markets...45 10
Exeutive Summary This study explores the potential impat of the biofuel promotion poliies implemented in reent years by the Government of Peru on the ountry s maro eonomy, setoral prodution and welfare. The study uses a single ountry stati Computable General Equilibrium Model (CGE), the appropriate analytial tool when quantifying the eonomy-wide effets of poliy hanges. The CGE model relies on the 2002 Soial Aounting Matrix (SAM) for Peru, as the main database and simulates the effets of various senarios refleting the government mandatory blending poliy under alternative assumptions. Relying on a stati model and assuming full employment of prodution fators, this study does not aim to apture exhaustively all the growth mehanisms assoiated to the biofuel poliies. It should be onsidered as an exploratory work that raises awareness on important poliy and methodologial issues. The biofuel poliy is being implemented under blending poliy mandates for biodiesel and ethanol prodution. Nine prodution senarios are analyzed based on the different soures of feedstok and prodution (smallholders and/or ommerial farms) 3. Four types of feedstok are onsidered: sugar ane ethanol, molasses from sugarane, palm oil, and jatropha. Four land markets are onsidered to reflet differenes between two regions (oastal and jungle), and two farm tehnologies (smallholder and ommerial farms) are used to produe the feedstok used for biofuel prodution. Within this framework, it is assumed that the effets of a mandatory blending poliy would lead to the onsumption of 3,600 barrels of petroleum equivalent per day (BPD) of biodiesel and 1,100 BPD of ethanol by 2015. Initially, the prodution of biofuels is assumed to be zero and trade is not allowed to take plae. On one hand, the results indiate that the ethanol mandatory poliy based on sugar ane has no signifiant effets on the Peruvian real GDP, and the biodiesel mandate leads to a minimal effiieny ost (up to 0.07% of real GDP) if palm oil is used. On the other hand, using only molasses for the prodution of fuel ethanol would be more ostly, in partiular if they are urrently used effiiently by other industries, if no demand exists for the additional sugar prodution, or if the sugar setor (and sugar mills) remains loated in the oastal area where land is sare. Under the assumption of onstant taxation of fuel, using biofuels would inrease the ost of the fuel for both, intermediate users i.e. the transportation setor - and final users i.e. the onsumer - thus leading to a deline in fuel onsumption. This effet is very limited when sugar ethanol is used, however, when molasses and jatropha are used as feedstok, pries an rise, in the most adverse ase, by up to 20% and 9%, respetively. Under a mandatory poliy, onsumers bear the ost of suh a poliy. Given their pattern of onsumption, urban households will suffer net real inome losses from suh a poliy. However, in the ase of subsidies or tax redution not onsidered in this study urban households, that are the main taxpayers, may also suffer a similar ost. Agriultural value added inreases signifiantly (more than 1%) only under the biodiesel senarios due to the land expansion. Similar results are found when molasses is used for ethanol prodution beause sugar ane prodution has to inrease on a muh larger sale when molasses is used than when sugar ane juie is used diretly. Non-biofuel agriultural prodution 3 These represent the entral ase for the model s results. They are not to be onfused with results that stem from the different robustness heks performed further on. 11
is not impated when expansion takes plae in the jungle area as there is no ompetition for land amongst the rops. When molasses is used for ethanol prodution, land ompetition and maroeonomi downturns lead to a fall in prodution of nearly 1% in both, ereal and livestok prodution. When sugar ane expansion takes plae in the oastal area, the prodution of rops, other than sugarane, delines by about 0.2%. In general, rural household inome expands as a result of the biofuel mandates, exept under the molasses and the low-yield jatropha senarios. Gains to rural households ome at the expanse of urban households who inur higher osts as a result of higher fuel pries. More speifially, when sugarane is used diretly in ethanol prodution, inome of rural households inreases by 12 million soles (0.04% ) in jungle area and 22 million of Soles (0.08%) in oastal areas. Under the biodiesel mandate, the inrease in inome for rural households ranges between 16 (0.06%) and 47 (0.17% ) million Soles, respetively, the latter attributable to jatropha being supplied by high-yield smallholders. These results suggest that a biofuel poliy that aims to benefit rural households and smallholders 4 would do so to the disadvantage of urban households who would stand to lose more than 500 million Soles. Only ethanol poliies may have positive or neutral overall outomes on both the rural and urban households. The results from this exploratory analysis are quite signifiant: first, imposing a blending target on biodiesel when Peru s omparative advantage is in the prodution of ethanol, makes it a relatively ineffiient tool to redistribute inome and redue poverty in rural areas. Seond, oupled with the upward trend in diesel use, a blending target on biodiesel would lead to high eonomi osts. Finally, the high ost of fuel for road transportation may also lead to adverse onsequenes for the most remote loations and the poor regions aross Peru by inreasing transportation osts nationwide. In order to provide a more omprehensive analysis of omplementary poliy onsiderations to Peru s biofuel strategy, the authors test for robustness by introduing alternative assumptions. Three heks for robustness are onduted: introduing alternative assumptions to the land market (the supply of land inreases to balane demand, and smallholders are also allowed aess to the new lands); liberalizing trade; and freeing the markets of feedstok for fuel blenders. The first test of robustness highlights the role of the land market in Peru. Relaxing the land onstraints in the oastal areas leads to positive results and even the large losses inurred under the molasses senarios are eliminated. Similarly, giving smallholders aess to more land will mitigate the inrease in their prodution osts under the jatropha senarios - where smallholder partiipation is greater and land requirements are higher than in palm oil prodution - and onsequently, redues the total ost of the biodiesel poliy. Removing trade barriers leads to more trade in ethanol and biodiesel. Loal prodution of biodiesel disappears, sine no feedstok appears to be ompetitive, and ethanol prodution and exports inrease. Trade liberalization, under biodiesel promotion, leads to a worse outome than under the autarki strategy that uses palm oil. 4 By imposing the biodiesel mandate. 12
When fuel blenders are allowed to hoose the most ost effetive feedstok the most effiient prodution proess for ethanol is the one that uses sugarane grown in the jungle and that for biodiesel it is the use of palm oil as a feedstok in a 60/40 prodution mix between large firms and smallholders. 13
Introdution and Bakground The eonomy of Peru mirrors its geographi harateristis, and is urrently split between a modern setor in the oastal area, and less-developed inland areas of the Sierra and Selva regions. Extreme poverty ontinues to be higher in the Sierra and Selva regions in partiular in rural areas. As of 2001, about 61% and 41% of the population in these regions, respetively, did not have enough inome to over their essential food needs. Although the share of the agriultural setor in GDP has generally been following a downward trend, the main inome soure for rural households ontinues to be from agriulture, diretly and indiretly. Consequently, it is important for governments to onsider strategies for rural development. This projet analyzes whether biofuel development may be onsidered an instrument toward this objetive. For Peru the main rops for biofuel prodution are sugar ane (inluding molasses), palm oil and jatropha. A blending poliy was established in 2007 imposing a mandate for biofuel blending. Starting January 1 st of 2009, there is an obligatory biodiesel of 2% (B2) that would rise to a 5% blending by January 1 st, 2011. Ethanol has an obligatory blending of 7.8% that starts in January 1 st, 2009. The objetive behind the biofuel blending mandates is to diversify the energy soures and reate growth and employment opportunities for the Peruvian eonomy. Biofuel development is also seen as part of the ountry s anti-narotis initiatives, where the development of biofuel feedstoks, espeially in the Amazon region, is viewed as an alternative to drug ultivation. The objetive of the study is to analyze, within a general equilibrium framework, the eonomywide effets from the development of the biofuel industry in Peru. Alternative senarios and tehnologies are assumed to illustrate the ost and benefits of different options. The main database used for the Peru Biofuel Model is the 2002 Soial Aounting Matrix (SAM) for Peru. Other data is inluded to supplement the SAM data for the prodution and pries of key setors in the eonomy. Detailed ost strutures for eah feedstok prodution proess, as well as for ethanol and biodiesel prodution, are also used to omplement the data. There are four land markets, land for ommerial farms and land for smallholders in both the oastal and jungle areas of Peru. In order to disuss more omprehensively the impat of the biofuel poliy on rural livelihoods and welfare, the household setor is divided into rural and urban households who have different demand and inome strutures and onsequently would be affeted differently by these poliies. The SAM also distinguishes between skilled, semi-skilled and unskilled labor. This study uses a single ountry stati CGE model adapted for Peru to analyze the biofuel promotion poliies implemented by the Government of Peru. Two types of feedstok - sugar ane ethanol and molasses from sugar ane are onsidered for ethanol prodution and two feedstoks - palm oil and jatropha- are onsidered for biodiesel prodution. Initially, the prodution of biofuels is assumed to be zero and trade is not allowed to take plae. Within this framework, it is assumed that the effets of a mandatory blending poliy would lead to the onsumption of 3,600 barrels of petroleum equivalent per day (BPD) of biodiesel and 1,100 BPD of ethanol by 2015. 14
The Peru Biofuel Model operates under the full employment assumption to analyze nine prodution senarios that differ based on the different feedstoks and prodution tehnologies used to produe ethanol and biodiesel 5. For eah type of biofuel under onsideration, the model distinguishes between the prodution pathway of ommerial farmers who operate at a high level of tehnology (inputs, fertilizer, et.) and that of smallholders who operate at a more traditional level. In all but one senario, the feedstoks are grown and supplied to the fuel blenders under a fixed ratio of 60% originating from ommerial farms and 40% from smallholders, aording to the mandate set by the Government of Peru to ensure smallholder partiipation in the national biofuel poliy. Results from the nine senarios are ompared to the baseline ase of no biofuel mandate, highlighting the effets on Peru s maro eonomy, setoral prodution and welfare. In order to provide a more omprehensive analysis of omplementary poliy onsiderations to Peru s biofuel strategy, three robustness heks are onduted. The results of these robustness heks shed light on future researh that may be needed to enrih the urrent exploratory work. The first robustness hek introdues alternative land assumptions in order to highlight the role of the land market in Peru and the seond hek involves removing trade restritions from the model by allowing new exports and imports to take plae. The final robustness hek allows fuel blenders to hoose the most ost effetive feedstok for the prodution of the relevant mandated biofuel. The report is strutured as follows: Chapter II disusses the general CGE methodology used and the adaptations introdued to the standard CGE model to represent the Peruvian eonomy and the biofuel setor. Chapter III presents the database used for the model: the 2002 SAM for Peru. Chapter IV disusses simulation results from nine biofuel promotion senarios. Chapter V disusses the different robustness heks undertaken, and finally, Chapter VI onludes, highlighting the different poliy options available as a result of this exploratory undertaking. 5 These represent the entral ase for the model s results. This is not to be onfused with results that stem from the different robustness heks performed further on. 15
CGE Methodology 6 Standard CGE Model Assumptions A stati omputable general equilibrium (CGE) model is the appropriate tool to apture the eonomi relationships and links between the maro and miro setors of the Peru eonomy, with a speifi fous on the Biofuel setor. This CGE model is used to quantify the effets of developing the biofuel setor in Peru and its impat on setoral and eonomy-wide produtivities, resoure alloation, and welfare. One advantage of CGE models is that they apture the prie and resoures alloation effets of poliy hanges after market adjustment in a onsistent maroeonomi framework. Typially, a CGE model onsists of set of linear and non linear equations that are solved simultaneously, and the resultant solution is a general equilibrium in all markets providing a omplete and onsistent piture of the irular flow in an eonomy, and at the same time aounting for all market-based interations among eonomi agents (Robinson 1989). The behavior of ommodity and fator markets is based on standard miroeonomi theory where households maximize utility from onsumption under budget onstraints, and produers maximize profits given the existing tehnology. The model traes the impat of an exogenous shok on growth and inome distribution through its effet on fator wages and employment whih in turn affet the inomes and expenditures of the various household groups given their initial struture of fator ownership. CGE models portray the household level impats of poliy hanges by having a representative household (RH) that represents all the individuals in a partiular lass or group. The model used for Peru follows the IFPRI Standard stati CGE model (Lofgren et al. 2002). The model has been extended to expliitly inlude disaggregated agriultural and food proessing setors, aounting for different tehnologies in the prodution of rops that provide feedstoks to the Peru s biofuel setor. The model simulates various senarios that aim to quantify the magnitude and diretion of alternative biofuel poliy diretives on the Peruvian eonomy. The following setions desribe four key omponents of the model in Lofgren et al. (2002): (1) ativities, prodution, and fators; (2) institutions; (3) markets; and (4) model losures or maroeonomi balanes. Ativities, Prodution, and Fators Eah prodution setor (ativity) is assumed to maximize profits subjet to onstraints from the assumed prodution tehnology and fator market employment rigidities. The tehnology struture is shown in Figure 1, where it is represented by nested CES (onstant-elastiity-of-substitution) and Leontief (fixedoeffiient) funtions. At the top level, the tehnology is speified by a Leontief funtion that would determine the mix between quantities of value added and aggregate intermediate input. Value added is itself a CES funtion of primary fators 7 whereas the aggregate intermediate input is a Leontief funtion of disaggregated intermediate inputs. Eah ativity produes one or more ommodities aording to fixed yield oeffiients and any ommodity may be produed by more than one ativity. For instane, the ativity Sugar produes, both, the Sugar and the Molasses ommodities and the Ethanol ommodity is produed by 5 ativities; Ethanol1, Ethanol2, Ethanol3, Ethanol4 and Ethanol5. Similarly biodiesel an be produed by Biopalm6, Biopalm7, Biojatropha8 and Biojatropha9 ativities, eah one representing 6 The struture of this setion follows Lofgren et al, 2002. 7 With a value of 0.8 for all setors 16
a different tehnology, inluding fixed ratio ombinations of feedstoks produed by smallholders and large farms. The revenue of the ativity is defined by the level of the ativity, yields, and ommodity pries at the produer level. As part of its profit-maximizing deision, eah ativity uses a set of fators up to the point where the marginal revenue produt of eah fator is equal to its wage 8. For agriultural setors, the model allows returns by unit of land to differ among rops. Full employment of the prodution fators is assumed. Labor and apital supplies are fixed and these fators are perfetly mobile aross setors but land is not, remaining only in the agriultural setor. Figure 1: Prodution Tehnology Commodity outputs Ativity level ( Leontief ) Value added ( CES ) Intermediate ( Leontief ) Primary fators Composite ommodities Imported Domesti Soure: Lofgren et al, 2002 Institutions: Households, Government, and Rest of the World (RoW) Households reeive their inomes from fator earnings, and transfers from other institutions. More speifially, eah household reeives fixed shares of inome flowseither from fators, the government or the rest of the world. Households use their inome to pay diret taxes, save, onsume, and make transfers to other institutions. It is assumed that diret taxes, and transfers to other domesti institutions are defined as fixed shares of household inome whereas the savings share is flexible for seleted households. Remaining inome, after taxes, savings, and transfers to other institutions is spent on onsumption. Household onsumption is for marketed ommodities purhased at market pries adjusted for ommodity taxes. Household onsumption is alloated aross different ommodities aording to Linear Expenditure System (LES) demand funtions, derived from the maximization of a Stone-Geary utility funtion 9. 8 Unless additional fator-related onstraints are imposed. 9 For details, see Dervis et al. 1982. 17
The government ollets taxes (or pays subsidies) and reeives transfers from other institutions. The model by default assumes that all taxes are at fixed ad valorem rates. The government uses this inome for its own onsumption, and for transfers to other institutions. Typially, government onsumption is assumed fixed in real (quantity) terms whereas government transfers to domesti institutions (households and enterprises) are CPI-indexed. Government saving is the differene between government inome and spending in the model. The rest of the world represents the foreign setor. Transfer payments from the rest of the world to domesti institutions and fators are all fixed in foreign urreny. Foreign savings (or the urrent aount defiit) is the differene between foreign urreny spending and reeipts. Commodity trade with the rest of the world is disussed in the following setion. Commodity Markets Figure 2 shows the physial flows for marketed ommodities along with the assoiated quantity and prie variables as defined in the model equations disussed in Lofgren et al (2002) 10. Domesti output may be sold in the market or onsumed at home. For marketed output, aggregated domesti output (QX) for eah ommodity is omputed as the omposite output of different ativities that produe the ommodity (QXAC). These outputs are imperfet substitutes beause of differenes in timing, quality, and loation between different ativities. A Constant-Elastiity-of-Substitution (CES) is used as an aggregation funtion. The demand for the output of eah ativity is derived from minimizing the ost of supplying a given quantity of aggregated output subjet to a CES funtion. Ativity-speifi ommodity pries (PXAC) lear the impliit market for eah disaggregated ommodity. Figure 2: Flow of Marketed Commodities. Soure: Lofgren et al, 2002 The model relies on the Armington speifiation to allow imperfet substitutability between domestially produed and imported ommodities. That is, imports (QM) and the demand for domestially produed goods (QD), within the same setor are treated as differentiated goods whose demand is haraterized by a speified onstant elastiity of substitution. The same Armington speifiation is applied to domesti output (QX). It is assumed that there is imperfet transformability of domesti output (QX) to exports (QE) and to domesti supply of domestially produed goods (QD). The degree of transformability is the elastiity of transformation. 10 Modified equations for this study are presented in appendix B. 18
Under the small ountry assumption, Peru's imports have an infinitely elasti world supply and its exports fae an infinitely elasti world demand, thus world pries of imports pwm and of exports pwe are exogenously determined. The domesti pries of imported and exported produts are given by: and PM = pwm ( 1+ tm ) EXR PE = pwe ( 1 - te ) EXR respetively, where EXR is the exhange rate (domesti urreny per unit of foreign urreny), and tm and te are the impliit tariff and export tax rates that aount for legal tariffs and export taxes, as well as any quantitative trade restritions and diret prie ontrols that affet the disparity between the domesti and border pries of traded goods. The Armington assumption implies that onsumers fae ommodities that are omposite goods (QQ), and their demand is a onstant elastiity of substitution (CES) aggregation funtion of imports (QM) and domesti sales (QD), with a speified substitution elastiity. Consumers maximize utility, whih in the model is the same as maximizing onsumption (QQ), so that the desired ratio between (QM) and (QD) is a funtion of their relative pries. Similarly, produers provide a omposite ommodity (QX) that is a onstant elastiity transformation (CET) funtion of exports (QE) and domesti output (QD) and maximize profits so that their desired ratio is a funtion of their relative pries. and QM PM = CES* QD PDD QE PE = CET* QD PDS Where CES* and CET* refer to the first-order onditions for utility maximization and profit maximization. Setoral omposite good pries are the weighted averages of the domesti pries of their omponent produts: PQ = (PDD QD + PM QM ) QQ = CES (PDD, PM) and PX = (PDS QD + PE QE ) QX = CET (PDS, PE) 19
where the CES and CET funtions refer to ost funtions that relate omposite pries to their omponent pries. They reflet the first-order onditions desribed above. The assumptions of imperfet transformability and imperfet substitutability permit the model to more appropriately reflet the realities of most ountries. The assumptions used give the domesti prie system a degree of independene from international pries and prevent unrealisti export and import responses to eonomi shoks. At the disaggregated ommodity level, these assumptions allow for a ontinuum of tradability and two-way trade, whih is ommonly observed even at very fine levels of disaggregation 11. This model speifiation determines the role of the exhange rate in ahieving equilibrium. The model speifies a funtional relationship between the balane of trade and the real exhange rate i.e., the relative prie of tradable and semi-tradables. Maroeonomi Balanes Maroeonomi balanes or System onstraints are onstraints that have to be satisfied by the eonomi system, but are not onsidered in the optimizing deision of any miro agent (Robinson 1989). There are three maroeonomi balanes: the (urrent) government balane, the external balane (the urrent aount of the balane of payments, whih inludes the trade balane), and the savingsinvestment balane. The mehanism by whih the model satisfies these onstraints is referred to as the losure rules. It is worth noting that the hoie of a losure rule has no effet on the base-line solution generated by the urrent CGE model but would typially influene the results for other poliy simulations. The Government losure assumes that government savings are a flexible residual with all tax rates fixed for domesti institutions. The presene of the rest of the world in the model requires an expliit treatment of how foreign exhange outflows and inflows are equilibrated. Typially, the real exhange rate, defined as the relative prie of traded to non-traded goods, is the equilibrating variable, an assumption that is followed for Peru. The losure rule adopted for the model is an investment-driven saving investment balane whih assumes that investment demand is fixed and the value of savings adjust (a Johansen losure). A Johansen losure assumes no link between the maro variables and aggregate employment, as a result, in the presene of the full employment assumption 12, a shok introdued to the system would affet the omposition of aggregate demand, but not GDP. 11 Robinson and Lofgren (2005) note that introduing a degree of substitutability and transformability is theoretially onsistent with the Salter-Sawn model whih assumes a rigid dihotomy between tradable and nontradable ommodities. 12 It is assumed that smallholders, who are of great interest for the Government of Peru, are not unemployed. 20
Modeling the Biofuel Setor Tehnologial pathway for Ethanol and Biodiesel Produtions The FAO and the Government of Peru are interested in analyzing nine alternative prodution senarios to explore the impat of developing the biofuel setor on the eonomy of Peru (see Table 1). Of the nine senarios, five analyze the impat of ethanol prodution and four analyze the impat of biodiesel prodution. Ethanol and biodiesel are produed using a variety of tehnologial pathways using two feedstoks for eah biofuel. Sugarane and molasses produe ethanol, and palm oil and jatropha produe biodiesel. For both biofuels, feedstok prodution in most ases (exept for molasses and one of the palm oil produing senarios) is a 60/40 split 13 between large sale ommerial farms and smallholders, respetively. As a working assumption in this study, we onsider that without the mandatory poliy, no prodution of ethanol or biodiesel takes plae. All the biofuel tehnologies are onsidered as latent tehnology in the model. 14 13 A ratio set by the Government of Peru. 14 However, for alibration purpose, we will assume that an infinitesimal prodution exists for eah tehnology initially. 21
Table 1: Prodution Matrix for Biofuel Prodution in Peru and Resultant Poliy Senarios Senario Feedstok Biofuel Prodution Other Prodution Feedstok Prodution Level of Tehnolog y Feedstok onsiderations 1 Sugarane Ethanol* oast 2 Sugarane Ethanol* oast 3 Sugarane Ethanol Jungle Co-generated eletriity Smallholder/ Commerial High - Co-generated eletriity Commerial High - Co-generated eletriity 4 Molasses Ethanol Sugar 5 Molasses Ethanol Sugar 6 Palm Oil Biodiesel Edible oil Smallholder/ Commerial Medium - Smallholder/ Commerial Smallholder/ Commerial Smallholder/ Commerial High High Opportunity ost high Opportunity ost low Medium/ High - 7 Palm Oil Biodiesel Edible oil Commerial High - 8 Jatropha Biodiesel - 9 Jatropha Biodiesel - Soure: FAO, 2010 Ethanol Prodution Smallholder/ Commerial Medium Low yield Smallholder/ Commerial Medium High Yield There are five possible pathways to produe ethanol either diretly through the sugarane feedstok, or, indiretly through its derivative, molasses (Table 1). As for other setors, eah biofuel tehnology will ombine inputs, inluding feedstoks and value added. We assume Leontieff tehnology to maintain the physial yield onstant during the simulations. In addition, for eah tehnology, the Leontieff parameters will also allow us to define the share of one feedstok, e.g. sugar ane, supplied by different types of agents (ommerial farms or small holders). Eah of these pathways is analyzed in the senarios undertaken. For the first three senarios where ethanol is produed diretly from sugarane, eletriity is a generated o-produt. The first two senarios assume that sugarane prodution takes plae in the oastal region using higher level tehnology. The first senario uses the 60/40 prodution split between ommerial farmers and smallholders and the seond assumes that only ommerial produers produe sugarane in the Coastal region. The hoie of the Leontieff funtional form ensures that the 60/40 prodution ratio between ommerial farmers and smallholders is maintained. In the absene of this fored relationship, the prodution of biofuel feedstoks would be overtaken by the ommerial farms due to their more ompetitive ost and yield strutures. In the oastal regions, total osts inurred in growing sugarane for ommerial farms are US$12.32 per hetare, for smallholders it is US$26 per hetare. Yields are also higher for ommerial farms than smallholders. The former have yields of 140 tons per hetare vis a vis 130 tons per hetare for smallholders (see Prodution Cost Tables in the Annex). As a result, a leontief funtion ensures the imposed ommerial/smallholder mix is maintained and yields are onstant. In senario 3, the sugarane is produed in the Selva region again maintaining the 60/40 prodution split in prodution. The last two senarios onsidered for ethanol prodution use molasses. Molasses, a produt of the food proessing setor, is urrently only used for human onsumption and for animal feed. In senario 4, a higher opportunity ost for prodution is assumed and in senario 5, a lower opportunity ost is used. Opportunity osts are modeled through a hange in the prie of eah unit of molasses used. 22
For example, a high opportunity ost in senario 4 means a higher prie per unit of molasses used 15 due to an additional mark-up. It is important to keep in mind that the high opportunity ost does not reflet a hange in the energy ontent of the molasses or an inrease in the share of wasted molasses. Physial yields are not modified. Biodiesel Prodution Biodiesel prodution in Peru may be produed either using palm oil or jatropha. Currently in Peru, palm oil is an edible oil and jatropha is a new feedstok that has not yet been produed. For eah of these feedstok options, two senarios are explored. For palm oil, senario 6 assumes the same 60/40 prodution split between ommerial farmers and smallholders but senario 7 assumes that only ommerial produers grow the oil palm. For jatropha prodution, both senarios (8 and 9) assume the 60/40 prodution split, however, the former assumes a yield of 4 tons per hetare for smallholders whereas senario 9 assumes a higher yield of 6.5 tons per hetare. One of the o-produts of biodiesel prodution from jatropha is glyerol. The model does not expliitly onsider a market for glyerol, instead it is aggregated with the Oil and Fat setor (setoral disaggregation of the model will be disussed in the data setion). Biofuels and Fuels blending All different types of ethanol are merged, i.e. blended, into one generi ethanol setor assuming a CES tehnology with very large degree of substitution aross produts (elastiity of substitution equals to 20). The same methodology is applied to biodiesel. Finally, a fuel setor blends fossil fuel, ethanol and biodiesel. Here, we assume a leontieff tehnology among inputs (the different type of fuels) and does not onsider value added for this setor. The leontieff tehnology allows us to represent expliitly the mandate and to have onsumption targets independent of the relative pries of the different fuels. Capital Market For the biofuel proessing setors the rate of return of apital may differ between alternative tehnologial pathways. In the prodution of ethanol, the highest return to apital is assumed to arue to proessing plants that use feedstok supplied only from the ommerial farms in the oastal areas (Senario 2). The lowest return to apital is from the prodution of ethanol using high opportunity ost molasses. Given that we assume the same prie of ethanol for all tehnologial pathways, profit margins are redued for tehnology assoiated with higher ost of intermediate inputs. Feedstok Prodution Agriultural feedstoks for biofuels There are four feedstoks onsidered for the prodution of the biofuels; sugarane, molasses, palm oil and jatropha. For eah feedstok, prodution an take plae in one or two regions, involving ommerial farms and/or small holders. For all but the last feedstok, prodution is ative and is re-direted from their original use to the prodution of biofuels. In the baseline, palm oil is used only in the food prodution industry and sugarane is used for the prodution of sugar and molasses. Molasses is used to produe, both, ethanol for human onsumption and feedstuff for the livestok industry. For jatropha, the model assumes that the tehnology for jatropha prodution exists, however, it is not ativated until the mandates beome 15 This prie gap generates rent for all households. 23
effetive. In other words, jatropha prodution is lose to zero in the baseline but inreases as a result of the biofuel development senarios whih use it as feedstok for produing biodiesel (senarios 8 & 9). Land Market There are four land markets portrayed in the model. In both regions (oastal and jungle) there are two markets: one for smallholders and one for ommerial farms. In the jungle region, all land alloated to the prodution of new biofuel feedstok is assumed to be endowed to ommerial farms. Smallholders, on the other hand, do not reeive any new alloations of land in the entral ases. At equilibrium the land market insures that the total demand for land by the different ativities is equal to the total supply of land for eah region and for eah type of land user/owners. Land supply in the Jungle region is not fixed and land extension is allowed in the senarios. Due to the Leontieff tehnology assumed in the prodution of feedstok and the initial assumption of land distribution among rops and regions (see data setion), this mehanism is needed to allow the prodution expansion in the jungle region for whih the initial amount of land is insuffiient to expand feedstok prodution. Differenes in land quality exist between the two regions onsequently determining the average prie of land in eah region. The prie of land in the oastal areas is roughly 6.5 times higher than the average prie of land in the jungle areas. Within eah region returns to land are differentiated aording to the different agriultural ativities. For instane, the returns to land from planting high value rops suh as fruits, vegetables and legumes are assumed to be three times higher than the average returns per region. Returns from Other Crop ativities are assumed to be twie as muh as the average regional land pries. Cereals as a rop are assumed to bring in a lower return to land of slightly less than half the return to land for the two regions. Finally, and in order to highlight the important role of the land market on the overall biofuel poliy, a land market robustness hek is introdued analyzing two variations. The first variation assumes that the relative prie of land for ommerial farms and smallholders is onstant. In the seond variation, the amount of land needed to maintain the real prie of land at this initial level is endogenized to equilibrate the land market. 16 Trade In the entral ase, no trade ours so all the inrease in demand is assumed to be satisfied by domesti prodution. In addition, the world pries of ethanol and biodiesel are assumed onstant, i.e. no growing demand from large importers like the US and/or the EU for instane. Therefore, the simulated biofuel mandates would lead to inreases in imports if trade is allowed. Relative to the baseline whih assumes no prodution or trade of biofuels, the poliy shoks would inrease domesti demand and the prodution pries of ethanol and biodiesel. Sine world pries remain onstant, an inrease in the ost of prodution leads to an erosion of Peru s ompetitiveness for ethanol and biodiesel and an inrease in imports. For fuel ethanol, this result may appear unrealisti sine Peruvian export osts are nowadays below the domesti prie of ethanol for several trade partners (at least the EU). In this situation, the model was unable to explain why, in the referene situation, Peru does not produe and export ethanol. To orret for this, an additional export ost is introdued in the baseline in order to explain the initial equilibrium. 16 Eah region is assumed to have an infinite land supply elastiity. 24
As a further robustness hek, trade is allowed to take plae and the previously imposed trade ost on ethanol is removed. It is also important to note that the model does not assume any learning urve nor does it assume a redution in prodution osts through the amortization of fixed osts. Indeed, nasent industry onsideration in a dynami model may lead to a deline in prodution ost and the domesti mandate may help to boost exports on the long run by developing a loal market to ensure eonomies of sale for the loal industry. In our stati model, the tehnology is assumed to be mature 17 and we assume perfet ompetition for all setors. We exlude the possibility of eonomies of sale. In this situation, the flows of exports/imports would be impated by the domesti mandates (e.g. the biodiesel mandate ould be reahed through imports instead of domesti prodution). 17 Based on the Produt life yle theory, a tehnology is onsidered to be mature if it reahes the stage where hanges in prodution ost are only assoiated with hanges in fator pries. 25