Energy Economics 34 (2012) 1922–1930
Contents lists available at SciVerse ScienceDirect
Energy Economics journal homepage: www.elsevier.com/locate/eneco
Biofuels and economic development: A computable general equilibrium analysis for Tanzania Channing Arndt a, Karl Pauw b, James Thurlow c,⁎ a b c
Department of Economics, University of Copenhagen, Denmark International Food Policy Research Institute, Washington, DC, USA United Nations University, World Institute for Development Economics Research, Katajanokanlaituri 6 B, FI-00160 Helsinki, Finland
a r t i c l e
i n f o
Article history: Received 22 December 2009 Received in revised form 23 July 2012 Accepted 26 July 2012 Available online 11 August 2012 JEL classification: D58 O13 Q42 Keywords: Biofuels Growth Poverty CGE model Tanzania
a b s t r a c t Biofuels could offer new economic opportunities for low-income countries. We use a recursive dynamic computable general equilibrium model of Tanzania to evaluate different biofuels production options and estimate their impacts on growth and poverty. Our results indicate that maximizing the poverty-reducing effects of biofuels production in countries like Tanzania will require engaging and improving the productivity of smallholder farmers. Evidence shows that cassava-based ethanol production is more profitable than other feedstock options. Cassava also generates more “pro-poor” growth than sugarcane-based systems. However, if smallholder yields can be improved rather than expanding cultivated land, then both sugarcane and cassava out-grower schemes generate similar pro-poor outcomes. We conclude that, in so far as the public investments needed to establish a biofuels industry are consistent with other development needs, then producing biofuels will enhance economic development in countries like Tanzania. © 2012 Elsevier B.V. All rights reserved.
1. Introduction Many low-income countries see biofuels as an opportunity to promote development (Ewing and Msangi, 2009). Tanzania, for example, is considering establishing a domestic biofuels industry in order to stimulate agricultural growth, create jobs and reduce rural poverty (Arndt et al., 2011b). Evidence suggests that optimism about biofuels in developing countries may be justified. In Mozambique, for example, Arndt et al., (2011a) find that proposed large-scale biofuels investments will increase economic growth by half a percentage point each year over the coming decade and lift 5% of the population above the national poverty line. This supports the view held by some that biofuels could help low-income countries overcome their dependence on oil imports while also reducing greenhouse gas emissions and increasing farmers' participation in the growth process (see, for example, FAO, 2008). Optimism over biofuels is countered by uncertainty over possible trade-offs between biofuels and food production, and the effects that declining food supplies may have on poverty and food insecurity. This concern has received considerable attention in the biofuels debate
⁎ Corresponding author. Tel.: + 358 9 615 99219; fax: + 358 9 615 99333. E-mail address:
[email protected] (J. Thurlow). 0140-9883/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.eneco.2012.07.020
and has gained support after the 2008 global food crisis (Headey and Fan, 2008; Rosegrant, 2008). Shifting resources away from food production could increase households' reliance on marketed foods, and biofuels may not raise the incomes of poor households enough to offset higher food prices. Finally, biofuels may not reduce greenhouse gas emissions once the effects of land clearing and fertilizer use are considered (Melillo et al., 2009; Searchinger et al., 2008). Possible trade-offs between biofuels and development have prompted low-income countries to consider a range of biofuel production options, such as smallholder versus plantation systems. In evaluating proposals from foreign investors, governments must decide which feedstocks and farming systems are both economically viable and contribute to national development. Most studies that evaluate biofuel policies use global models, group low-income countries into regions, and/or focus on developed countries' policies (see Kretschmer and Peterson, 2010 for a review). However, biofuel strategies in (smaller) developing economies should be informed by country-specific analysis. To illustrate the benefits of such analysis, we develop a recursive dynamic computable general equilibrium (CGE) model of Tanzania, and use the model to estimate the impact of alternative biofuels scenarios on economic growth and employment. The model is also linked to a microsimulation module that estimates impacts on poverty. Section 2 outlines Tanzania's biofuels production options; Section 3 describes the model and how the various options are simulated; and Section 4 presents the
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
results. The final section summarizes our findings and identifies areas for further work. 2. Options for producing biofuels in Tanzania 2.1. Identifying biofuels production scenarios Tanzania is considering various biofuel production options (see Felix et al., 2010) that differ on three characteristics: (1) the type of feedstock used and biofuel produced (i.e., sugarcane or cassava); (2) the scale of feedstock production (i.e., smallholder versus estate); and (3) the way in which feedstock production is expanded (i.e., increasing yields or harvested area). Table 1 summarizes the six biofuels scenarios considered in this paper. The first three scenarios (Sugar 1–3) refer to ethanol produced from sugarcane. In the first scenario (Sugar 1) all feedstock is produced by smallholder farmers through an out-grower scheme and is supplied to large processing plants. Conversely, the second scenario (Sugar 2) assumes that all feedstock is produced on large-scale commercial farms. These two scenarios allow us to contrast the impacts of small- and large-scale sugar production options. Finally, in the third scenario (Sugar 3), sugarcane is produced via an out-grower scheme, but by raising smallholders' yields of sugar for biofuel feedstock (from 43 to 70 tons per hectare) rather than by expanding the amount of land under sugarcane cultivation. This reduces the amount of land that is currently used for crops and which would have to be displaced by biofuel feedstock production. We also consider cassava as a potential biofuel feedstock. In each scenario, we assume that production is by smallholders through an outgrower scheme and that processing is done by large-scale processing plants. The first two scenarios differ in that Cassava 1 assumes that production is achieved through extensification (i.e., land expansion) while Cassava 2 assumes that cassava feedstock yields are increased (from 10 to 20 tons per hectare) thereby limiting the amount of land displaced to produce biofuel feedstock. The Cassava 3 scenario assumes a mixed production system, with 40% of feedstock obtained from smallholders through yield improvements (i.e., as in Cassava 2) and the rest produced by large-scale commercial farmers situated close to a large-scale processing plant. This mixed farming system offers a compromise between ensuring reliable feedstock supplies (from plantation farms) and reducing poverty (via smallholder participation). The six biofuel production options allow us to compare feedstocks, scale of production, and intensive/extensive approaches. To make the scenarios comparable, we simulate the same biofuels production volumes in all scenarios. More specifically, we model the establishment of a biofuel industry in Tanzania capable of producing 1000 million liters of ethanol per year (i.e., 3 million liters per day) by 2015. This is greater than the capacity currently envisaged by either the Government of Tanzania or private foreign investors; however, it is useful for this exercise as it permits us to identify economywide impacts. 1 2.2. Estimating production costs and technologies The biofuels scenarios in Table 1 contrast the impacts of different feedstocks and processing plants. These scenarios will produce different outcomes because they use different technologies (i.e., factor and intermediate inputs) and generate different profit rates for farmers and 1 Instead of imposing growth of alternative biofuel production technologies on the model through growth in fixed factors (land and capital), it would be possible, in principle, to leave the level of biofuels production as endogenous and allow the model to determine the level of production given the international price. In practice, some response dampening formulation is almost invariably required (hence the ubiquity of the Armington and constant elasticity of transformation functions on imports and exports).
1923
Table 1 Simulated ethanol production scenarios for Tanzania. CGE model scenario
Scale of feedstock production
Feedstock yield (tons per hectare)
Land expansion (% of land from displacement)
Sugar 1 Sugar 2 Sugar 3 Cassava 1 Cassava 2 Cassava 3
Small Large Small Small Small Small/large mix
Low (43 mt/ha) Low (84 mt/ha) High (70 mt/ha) Low (10 mt/ha) High (20 mt/ha) High (20 mt/ha)
Yes (50%) Yes (50%) No (0%) Yes (50%) No (0%) Yes (30%)
Source: own calculations using information from Felix et al. (2010).
downstream processors. Felix et al. (2010) estimate biofuel production costs for Tanzania, as shown in Table 2. Producing ethanol in Tanzania costs US$0.37 per liter under a mixed small and large-scale cassava-based production system (i.e., Cassava 3) and US$0.43 per liter for large-scale commercial sugarcane-based production (i.e., Sugar 2). Both options compare favorably with ethanol production costs in countries such as Brazil (US$0.47), United States (US $0.46) and India (US$0.52). However, the estimated costs of producing ethanol from smallholder-based sugarcane (i.e., Sugar 1) suggest that not all biofuels production options in Tanzania are as competitive as production in other countries. In our analysis, we assume that the domestic ethanol price received by processing plants is US$0.56 per liter, implying that all six options have net operating surpluses although the returns to land and capital dedicated to biofuel production vary and may be below market in some scenarios (e.g., Sugar 1). Using the estimated production costs and crop budgets, we derive production technologies for the six biofuels scenarios (see Table 3). The top half of the table shows the inputs required and outputs generated for 100 hectares of land allocated to feedstock production. From the first two columns, we see that smallholder crop yields (i.e., Sugar 1) are lower than larger-scale farmers' yields (i.e., Sugar 2), implying that small-scale farm land produces about half the output of plantations on the same amount of land (i.e., 4280 versus 8400 tons). Small-scale farms are also more labor-intensive (i.e., 0.4 hectares per worker versus 2.4 hectares per worker on larger farms). Increasing smallholders' sugarcane yields increases production levels per 100 hectares of land (i.e., to 6,999 tons), but requires additional labor for weeding and harvesting. Cassava production is also labor-intensive and requires more land per liter of ethanol than sugarcane. The mixed cassava production system (i.e., Cassava 3) is more labor-intensive than the equivalent smallholder scenario (i.e., Cassava 2) since new commercial farms require additional laborers whereas smallholders increase production by raising yields on their existing farm land. The lower half of Table 3 shows the inputs required to produce 100,000 liters of ethanol. All scenarios use large-scale processing plants
Table 2 Production cost estimates for ethanol scenarios.
Cost per liter (US$) Raw materials Service fluids Labor Maintenance Operating charges General plant costs Administrative costs Capital depreciation Co-products
Sugar 1
Sugar 2
Sugar 3
Cassava 2
Cassava 3
0.567 0.416 0.039 0.001 0.014 0.000 0.007 0.038 0.063 − 0.011
0.434 0.310 0.025 0.001 0.014 0.000 0.007 0.029 0.063 − 0.016
0.529 0.393 0.027 0.001 0.015 0.000 0.008 0.035 0.070 − 0.019
0.469 0.252 0.086 0.000 0.025 0.000 0.013 0.030 0.064 0.000
0.369 0.190 0.079 0.000 0.020 0.000 0.010 0.024 0.045 0.000
Source: own calculations based on Felix et al. (2009).
1924
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
Table 3 Production technologies under alternative biofuels scenarios. Sugar 1 Inputs and outputs per 100 ha of feedstock production Land employed (ha) 100.0 Crop production (mt) 4280 Land yield (mt/ha) 42.8 Farm workers employed (people) 225.2 Land per farm worker (ha/person) 0.44 Capital per hectare (cap. units/ha) 1.76 Biofuels produced (l) 297,078 Processing workers employed (people) 2.33 Inputs and outputs per 10,000 l of ethanol produced Biofuels production (l) 100,000 Feedstock inputs (mt) 1441 Feedstock yield (l/mt) 69.41 Land employed (ha) 33.66 Farm workers employed (people) 75.81 Processing workers employed (people) 0.78 Capital employed (capital units) 105.2
Sugar 2
Sugar 3
Cassava 1
Cassava 2
Cassava 3
100.0 8399 84.0 41.8 2.39 3.98 582,999 3.36
100.0 6999 70.0 81.5 1.23 n/a 485,847 4.18
100.0 1000 10.0 215.7 0.46 0.72 183,328 0.45
100.0 2000 20.0 66.6 1.50 n/a 366,636 0.91
100.0 2000 20.0 153.3 0.65 1.4 366,636 0.91
100,000 1441 69.41 17.15 7.16 0.58 342.6
100,000 1441 69.41 20.58 16.77 0.86 183.6
100,000 546 183.31 54.55 117.66 0.25 214.5
100,000 546 183.31 27.28 18.17 0.25 214.5
100,000 546 183.31 27.28 41.82 0.25 373.7
Source: own calculations based on Felix et al. (2010), Coles (2009), Kapinga et al. (2009) and Rothe et al. (2007).
and so technologies are the same. The scenarios differ with respect to the scale of feedstock production and, hence, the required amount of land and farm workers. The number of workers used in processing biofuels is much smaller than the number of farm workers used in producing the feedstock (e.g., 1 processing worker is needed for every 121 farm workers in the labor-intensive Sugar 1 scenario). It is clear that biofuel's employment effects will come from feedstock production rather than processing, and so we do not consider differences between small- and large-scale processing options. In summary, the six biofuels production scenarios compare different feedstocks; small/large-scale production structures; and intensive/ extensive feedstock production options. Detailed production cost estimates are used to estimate unique technologies for each scenario. We now integrate these technologies within the economywide model. 3. Modeling impacts on growth and poverty 3.1. Structure of the Tanzanian economy Table 4 shows the structure of the Tanzanian economy in 2007. Agriculture accounts for one third of gross domestic product (GDP) and four-fifths of employment. Most farmers are smallholders with average land holdings of 1.6 hectares. They produce most of the country's food, which dominates the agricultural and downstream manufacturing sectors. Tanzania also imports foods (mainly cereals), which account Table 4 Structure of Tanzania's economy, 2007. Exports/ output (%)
Imports/ demand (%)
100.00 6.11 5.83 0.28
9.44 13.23 1.64 63.45
22.01 7.28 10.05 7.08
10.81
0.00
14.98
0.00
0.17 1.46 1.12
25.06 12.83 2.13
4.61 87.88 10.01
82.26 8.26 2.00
72.26 61.42 20.80
3.22
0.35
10.69
77.87
21.79
83.87
10.35 45.05
0.99 14.92
– 27.22
– 6.09
–
Share of total (%)
Total GDP Agriculture Food crops Traditional exports Other agriculture Mining Manufacturing Food processing Other manufacturing Other industries Services
GDP
Employment
Exports
Imports
100.00 31.82 19.06 3.20
100.00 82.46 39.97 12.22
100.00 34.89 2.57 21.50
9.56
30.27
3.94 8.84 5.62
Source: Tanzania 2007 social accounting matrix.
– 1.40
0.72
for 15% of total imports and 20% of processed foods. This dependence on food imports stems in part from smallholders' low crop yields and a reliance on traditional rain-fed farming technologies. Larger-scale commercial farmers are more engaged in nonfood export crops, such as coffee, tobacco and tea, which together account for about a third of total merchandize exports. Gold mining dominates non-agriculture and accounts for a third of merchandize export earnings. Mining does not create much employment or value-added, and most non-farm workers work in services and construction (“other industries”). Wages in many nonfarm sectors, such as trade, are on average only slightly above those in agriculture, reflecting low education levels and a skilled labor shortage — most of the workforce has not completed primary schooling. The CGE model captures Tanzania's initial economic structure. This class of models is often used to examine external shocks and policies in low income countries. Their strength is their ability to measure linkages between producers, households and the government, while also accounting for resource constraints and their role in determining product and factor prices. These models are, however, limited by their underlying assumptions and the quality of the data used to calibrate them. The rest of this section describes the model and the design of the simulations. 3.2. Description of the CGE model Our model belongs to the neoclassical class of CGE models.2 Economic decision-making in the model is the outcome of decentralized optimization by producers and consumers within a coherent economywide framework. Production occurs under constant returns to scale. Intermediate demand is determined by fixed technology coefficients (i.e., Leontief demand), while constant elasticity of substitution (CES) production functions allow factor substitution based on relative prices. Profit maximization implies that factors receive income where marginal revenue equals marginal cost. The model identifies 58 sectors (i.e., 26 in agriculture, 22 industries and 10 services). Based on the 2000/01 Household Budget Survey (HBS) (NBS, 2002), labor markets are segmented across three skill groups: (1) workers with less than primary education; (2) workers with primary and possibly some secondary schooling; and (3) workers who have completed secondary or tertiary schooling. Agricultural land is divided across small- and large-scale farms using the 2002/03 Agricultural Sample Survey (MINAG, 2004). All factors are fully employed and capital is immobile across sectors. This 2 The model's mathematical specification is provided in the appendix and discussed in Diao and Thurlow (2012).
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
Imports
Rest of world Exports
Non-agriculture
Food crops
Feedstock crops
Export crops
National markets
Biofuel processing
Remitted profits
1925
Intermediates
Land Households
Labor
FDI
Capital Fig. 1. Conceptual framework.
means that as new biofuel sectors expand they generate additional demand for factor inputs, which then affect economywide factor returns and production in other sectors by increasing resource competition. Substitution possibilities exist between production for domestic and foreign markets. This decision of producers is governed by a constant elasticity of transformation function, which distinguishes between exported and domestic goods. Profit maximization drives producers to sell in markets where they achieve the highest returns based on domestic and export prices. Further substitution possibilities exist between imported and domestic goods under a CES Armington specification. This takes place in both final and intermediates usage. Under the small-country assumption, world demand and supply is perfectly elastic at fixed world prices, with the final ratio of traded to domestic goods determined by the endogenous interaction of relative prices. Production and trade elasticities are drawn from Dimaranan (2006). The model distinguishes between 15 representative households (i.e., rural farm, rural nonfarm and urban nonfarm groups by per capita expenditure quintiles). Households receive income in payment for producers' use of their factors of production, and then pay direct taxes, save and make foreign transfers (all at fixed rates). Households use their remaining income to consume commodities under a linear expenditure system (LES) of demand. The model includes a micro-simulation module with each respondent in the HBS linked to their corresponding
representative household in the CGE model. Changes in commodity prices and households' consumption spending are passed down from the CGE model to the survey, where total per capita consumption and poverty measures are recalculated. The government receives revenues from direct and indirect taxes, and makes transfers to households and the rest of the world. The government purchases consumption goods and services, and remaining revenues are saved (budget deficits are negative savings). All private, public and foreign savings are collected in a savings pool from which investment is financed. The model includes three macroeconomic accounts: government, current account, and savings-investment. To balance these macroaccounts, it is necessary to specify a set of “macro-closure” rules that provide a mechanism through which macroeconomic balance is maintained. A savings-driven closure is assumed in order to balance the savings-investment account. This means that households' marginal propensities to save are fixed, and investment adjusts to income changes to ensure that the level of investment and savings are equal in equilibrium. For the current account, it is assumed that a flexible exchange rate adjusts in order to maintain a fixed level of foreign savings. In other words, the external balance is held fixed in foreign currency terms. For the government account, direct tax rate rates are fixed and the fiscal deficit adjusts to equate total revenues and
Table 5 Core macroeconomic assumptions and results, 2007–2015. Initial, 2007 Average annual growth rate, 2007–2015 (%) Total GDP 100.00 Labor supply 56.07 Capital stock 17.53 Livestock stock 2.20 Land supply 24.20 Small-scale 22.48 Large-scale 1.72
Baseline scenario 4.61 2.12 2.52 1.00 1.00 1.00 1.00
Deviation from baseline final year value, 2015 (%) Real exchange rate 1.00 – Real food prices 1.00 – Source: results from the Tanzania CGE model.
Sugar 1
Sugar 2
Sugar 3
Cassava 1
Cassava 2
Cassava 3
4.86 2.12 2.62 1.00 1.24 1.26 1.00
4.97 2.12 2.62 1.00 1.12 0.87 4.08
4.98 2.12 2.63 1.00 1.29 1.00 1.00
4.86 2.12 2.59 1.00 1.38 1.41 1.00
4.97 2.12 2.59 1.00 1.38 1.00 1.00
4.99 2.12 2.61 1.00 1.27 0.87 3.95
− 6.96 0.66
− 6.01 − 0.05
− 6.49 0.32
− 6.53 1.23
− 6.07 0.39
− 5.06 0.51
1926
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
Table 6 Agricultural production results, 2007–2015.
Biofuels (1000 l) Crop land (1000 ha) Biofuels crops Food crops Export crops Production (1000 mt) Biofuels feedstock Food crop: Maize Food crop: Rice Food crop: Cassava
Initial value, 2007
Baseline value, 2015
0 8207 0 7236 970 0 2354 1084 5284
Deviation from Baseline final value, 2015 Sugar 1
Sugar 2
Sugar 3
Cassava 1
Cassava 2
Cassava 3
0 8887 0 7711 1175
1000 168 337 23 − 191
1000 86 172 65 − 150
1000 0 0 163 − 163
1000 273 545 − 85 − 187
1000 0 0 142 − 142
1000 50 99 45 − 127
0 2713 1268 5873
14,407 18 −1 26
14,407 41 15 73
14,407 60 19 137
5455 − 16 − 18 − 56
5455 52 15 123
5455 21 4 33
Source: results from the Tanzania CGE model.
expenditures. Finally, the producer price index is chosen as the model's numéraire, and so all product and factor price movements are relative to this fixed price index. The model is “recursive dynamic”, implying that it is solved as a series of static equilibriums, with updating of key parameters between periods. Unlike full inter-temporal models, which include forwardlooking expectations, the recursive dynamic model used in this paper adopts a simpler set of adaptive rules, under which investors essentially expect prevailing price ratios to persist indefinitely. Under this specification, sectoral capital stocks are adjusted each year based on previous investment levels, net of depreciation. The model adopts a “putty-clay” formulation, whereby each new investment can be directed to any sector in response to differential rates of return, but installed equipment must remain in the same sector. Unlike capital, growth in labor and land supply is determined exogenously. Sectoral productivity growth is also exogenous, but may vary by factor. Using these simple relationships to update key variables, we can generate a series of growth paths based on different biofuel investment scenarios.
3.3. Modeling biofuels production Biofuels were not produced in Tanzania in 2007; and so there is initially no biofuels sector in the 2007 social accounting matrix used to calibrate the CGE model. The production cost information in Table 2 and farm crop budgets provide the technology vectors needed to create new biofuel sectors in the model. Based on these different biofuels technology vectors, we create feedstock and processing sectors with effectively zero production in the base.3 The CGE model is first run forward over the 2007–2015 period assuming no expansion in biofuels production. This produces a baseline “without biofuels” scenario. Then in the biofuels simulations we expand production in the feedstock and processing sub-sectors to produce 1000 million liters of biofuels. A conceptual framework for these simulations is shown in Fig. 1. Biofuels expansion is assumed to be driven by foreign direct investment (FDI) and so all profits generated by capital in the biofuels sectors are remitted abroad (after applying average corporate tax rates). Biofuels producers must, however, compete with other sectors for intermediate inputs, and for land and labor resources, whose returns are distributed to households within Tanzania. In the CGE model we assume full employment, which means that total labor supplies are fixed and increasing labor demand per unit of land raises workers' wages, as would be expected when the supply of foreign 3 An anonymous reviewer correctly noted that it is possible to have exactly zero production in the base solution if biofuels is produced using a Leontief production function. Our model adopts a CES production function with low substitution elasticities, thus requiring small nonzero base production levels. However, the two approaches produce virtually identical results, especially as we are primarily concerned with deviations from the baseline.
capital increases. Feedstock production also displaces lands used for existing crops, since these lands will be assigned to new biofuels investments and smallholder farmers will also reallocate resources towards feedstocks. Thus, while new lands may be available to feedstock producers, we expect that at least some existing lands will be displaced by biofuels crops. Table 1 shows that for most scenarios we assume that half of the lands used by biofuels feedstocks come from lands already in use by smallholder farmers. 4 There is no land displacement in the Sugar 3 and Cassava 2 scenarios since feedstocks are produced entirely through intensification (i.e., raising yields). The grey shaded areas in Fig. 1 represent new foreign capital and crop land resources, which cause national production to expand in the simulations and raise aggregate labor demand and average wages. We assume that all biofuels will be exported. However, it is possible that some of the ethanol produced in Tanzania may be blended with imported petroleum for domestic use (see Felix et al., 2010). However, if the Government of Tanzania does not subsidize domestic ethanol, and if biofuels and petroleum are near-perfect substitutes (as we assume), then the difference between increasing biofuels exports or reducing petroleum imports is very small (i.e., the effect on the balance of payments is symmetric). The model includes co-products produced during the sugarcane-based production processes, the sale of which helps reduce ethanol production costs. We do not, however, explicitly model markets for co-products, but assume that they are used to reduce fuel and electricity inputs used during processing.
4. Model results 4.1. Baseline scenario We first calibrate the model to track observed trends in key demographic and macroeconomic indicators (see Table 5). Population growth is set at 2.5% per year during 2007–2015. Total labor supply grows at 2.1% per year in all scenarios, reflecting our assumption of full employment (perfectly inelastic labor supply). Skilled labor supply grows faster than unskilled labor reflecting gradual improvements in educational attainment. Livestock stocks and agricultural land expand at 1% each year, capturing rising population density, especially in rural areas. In order to achieve recently observed total GDP growth rates, total factor productivity growth is set at 2.7% per year during the simulation period.
4 We assume that half of the land that will be used for biofuel feedstock production is currently uncultivated. This is broadly consistent with the amount of arable land in Tanzania that is considered suitable for biofuels feedstock production (see Salvatore et al., 2010).
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
1927
Table 7 Sector growth results, 2007–2015.
Total GDP Agriculture Food crops Traditional exports Biofuels crops Other agriculture Mining Manufacturing Food processing Biofuels processing Other manufacturing Other industries Services
Share, 2007 (%)
Baseline growth (%)
Deviation from Baseline growth rate (%-point) Sugar 1
Sugar 2
Sugar 3
Cassava 1
Cassava 2
Cassava 3
100.00 31.82 19.06 3.20 0.00 9.56 3.94 8.84 5.62 0.00 3.22 0.31 45.05
4.61 2.20 1.88 2.49 0.00 2.71 7.17 5.49 3.82 0.00 8.03 8.03 5.35
0.25 0.21 0.00 − 1.49 – − 0.27 − 0.02 0.00 − 0.12 – − 1.16 − 0.97 0.32
0.35 0.34 0.15 − 0.90 – − 0.12 − 0.02 1.25 − 0.04 – − 1.05 − 0.91 0.16
0.37 0.50 0.22 − 1.11 – − 0.10 − 0.02 0.53 0.03 – − 1.03 − 0.87 0.28
0.25 0.25 − 0.17 − 1.61 – − 0.33 − 0.02 0.57 − 0.18 – − 1.01 − 0.79 0.16
0.35 0.55 0.19 − 0.97 – − 0.08 − 0.02 0.71 0.03 – − 0.94 − 0.77 0.17
0.37 0.38 0.04 − 0.92 – − 0.14 − 0.01 1.52 − 0.03 – − 0.71 − 0.57 0.14
Source: results from the Tanzania CGE model.
4.2. Changes in agricultural production In the biofuels simulations, we increase the amount of land and foreign direct investment allocated to biofuel sectors. We assume that only half of biofuels' land requirements will displace land already being cultivated. We therefore expect an increase in the total amount of land under cultivation. This is shown in the third column of Table 5, where the rate of land expansion for smallholders increases from 1.00% per year in the Baseline to 1.26% under the Sugar 1 scenario. Conversely, as we shift towards larger-scale feedstock production in Sugar 2 the expansion rate of smallholder lands falls below one. This is because we assume that it is smallholders' lands that are displaced when large-scale plantations expand feedstock production. No smallholder land is displaced in the Sugar 3 and Cassava 2 scenarios since production is achieved by improving yields. There is some land displacement in the mixed cassava production scenario (Cassava 3), because the portion that is produced by commercial farmers requires additional lands, half of which comes from smallholders. Displacement of lands to produce biofuels feedstocks causes production of certain other crops to contract (see Table 6). The biofuels debate in low-income countries centers on possible negative effects on food production. Our findings indicate that real consumer food prices will rise in response to biofuel production. This occurs because food producers must compete against biofuel producers for land and labor, which will as a result command higher returns and drive up production costs and food prices. However, our findings also suggest that, in the case of Tanzania, it is export crops that experience declines in production. This is because in our simulations biofuels
account for a third of total merchandize export earnings by 2015. Since we assume that the current account balance is fixed in foreign currency, the increase in export earnings causes the real exchange rate to appreciate relative to the Baseline (see Table 5). This reduces the competitiveness of traditional export crops, such as coffee, tobacco and tea, and these exports decline. This releases land for biofuels and food production. For example, the amount of land allocated to export crops falls by 191,000 hectares in the Sugar 1 scenario. In the same scenario the land allocated to food crops increases slightly, as farmers reallocate land away from export crops and rising incomes raises food demand. Food crop production therefore increases under most biofuels production scenarios, even though higher land and labor input costs mean higher overall real food prices. The only exception is the land-intensive Cassava 1 scenario, where a large amount of land is needed to produce the same amount of biofuel, causing food production to fall. However, even in this scenario, the trade-off between food production and biofuels remains small, with export crops much more severely affected. The same amount of ethanol exports are produced under Sugar 1 and Sugar 2, causing a similar appreciation of the real exchange rate in these two scenarios (see Table 5). This suggests that moving to larger-scale feedstock production does not remove the negative impacts for nonbiofuels exporters. Larger-scale feedstock production technologies do, however, favor food crop production, since the higher yields of largescale farmers mean that less land is needed to produce biofuels feedstocks and hence more land previously used by traditional export crops is reallocated to food crops. This finding suggests that any tradeoffs that do exist between biofuels and food production are likely to be
Table 8 Employment results, 2007–2015.
Total (1000 s) Agriculture Food crops Traditional exports Biofuels crops Other agriculture Mining Manufacturing Food processing Biofuels processing Other manufacturing Other industries Services
Employment, 2007
Baseline employ., 2015
19,010 15,675 7597 2323 0 5,754 33 278 212 0 66 188 2836
22,487 18,565 8977 2901 0 6,686 54 320 209 0 111 240 3308
Source: results from the Tanzania CGE and micro-simulation model.
Deviation from baseline final employment, 2015 Sugar 1
Sugar 2
Sugar 3
Cassava 1
Cassava 2
Cassava 3
0 − 95 − 76 − 535 758 − 243 −5 − 20 −2 1 − 19 11 109
0 − 76 221 − 327 72 − 42 −5 − 16 0 1 − 17 14 82
0 − 103 178 − 403 168 − 46 −5 − 14 2 1 − 17 12 111
0 − 23 − 269 − 573 1,177 − 358 −4 − 21 −4 0 − 17 9 38
0 − 68 146 − 356 182 − 39 −4 − 13 2 0 − 15 11 73
0 − 50 −4 − 342 418 − 122 −3 − 12 −1 0 − 12 12 54
1928
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
Table 9 Household per capita equivalent variation results, 2007-2015.
Rural Quintile Quintile Quintile Quintile Quintile Urban Quintile Quintile Quintile Quintile Quintile
1 2 3 4 5 1 2 3 4 5
Per capita consumption, 2007 (US$)
Baseline growth, 2015 (%)
Deviation from baseline growth rate (%-point) Sugar 1
Sugar 2
Sugar 3
Cassava 1
Cassava 2
Cassava 3
372.4 109.8 198.6 283.7 433.7 967.4 903.2 120.6 211.3 307.6 470.3 1614.2
1.32 0.82 0.97 0.99 1.17 1.31 1.94 1.22 1.28 1.38 1.52 2.08
0.41 0.31 0.32 0.37 0.40 0.44 0.38 0.35 0.44 0.54 0.52 0.34
0.59 0.18 0.19 0.24 0.28 0.59 0.38 0.27 0.42 0.53 0.52 0.35
0.53 0.58 0.56 0.60 0.59 0.55 0.41 0.43 0.50 0.58 0.56 0.37
0.27 0.29 0.29 0.32 0.32 0.28 0.16 0.22 0.23 0.26 0.25 0.13
0.45 0.59 0.54 0.59 0.55 0.45 0.28 0.25 0.33 0.41 0.40 0.25
0.49 0.22 0.22 0.25 0.26 0.47 0.21 0.22 0.26 0.31 0.30 0.19
Source: results from the Tanzania CGE model.
smaller when feedstocks are produced by larger-scale farmers who achieve greater yields per hectare. Alternatively, when smallholders' yields are increased there is no displacement of land and so traditional export crop lands are reallocated entirely to food crops (see Sugar 3 and Cassava 2). These scenarios indicate that the exchange rate effect is more important than heightened resource competition when determining the overall effect of biofuels investments on food production in Tanzania. Arndt et al. (2011a) reported similar findings for Mozambique, although biofuels investments reduced food crop production because Mozambique does not have a large export crop sector and so at least some lands under food crops are displaced by biofuel feedstocks. 4.3. Impacts on economic growth and employment Table 7 shows the impact of biofuels investments on sectors' real GDP growth rates. Foreign direct investment in the biofuels sectors expands agriculture's capital stock and also brings new lands under cultivation. This expansion in productive resources causes agriculture's growth rate to increase in all of the biofuels scenarios. Larger-scale production of sugarcane feedstocks (i.e., Sugar 2) generates larger gains in agricultural GDP than production through smallholder out-grower schemes (i.e., Sugar 1). There are also larger gains in the manufacturing sector under the Sugar 2 scenario, due to its smaller impact on food crops and downstream food processing. However, all sugarcane scenarios reduce processed food production because the appreciated exchange rate heightens competition in this import-intensive sector. Ultimately, the trade-offs from biofuels production are smaller than the gains from new investments in the biofuels industry. As a result, national GDP growth rates increase in all the biofuels scenarios. The injection of new foreign capital into the biofuels sector and the introduction of previously-unused lands into cultivation directly increase GDP and serve to increase the marginal product of existing labor stocks. Generally speaking, the more profitable the biofuel processing technology is the larger its impact on national economic growth. Thus, in terms of economic growth, there is strong overlap in private and social interests. Improving crop yields rather than displacing existing cultivated lands also generates large economywide gains. This is because these sectors enhance the returns to agricultural resources without greatly reducing food production. However, economic growth and profits do not necessarily translate into employment opportunities and reduced poverty. Table 8 reports impacts on labor employment. The number of new jobs created in the biofuels sector varies greatly across scenarios. The low labor-intensity of large-scale sugarcane production means that only 72,000 farm jobs are created in the Sugar 2 scenario. Conversely,
out-grower schemes employ far more farmers (see Table 3), with 758,000 additional workers producing sugarcane in the Sugar 1 scenario. 5 Sugarcane is less labor-intensive than cassava production, and it is the Cassava 1 scenario that engages the most workers. Moreover, while improving crop yields amongst smallholders does not require additional lands in the Sugar 3 and Cassava 2 scenarios, it still requires additional workers, especially during harvesting. For example, doubling cassava yields in the Cassava 2 scenario draws an additional 182,000 farmers into cassava production. This result emphasizes an often overlooked dimension of the biofuels debate, which focuses on land displacement (especially for food crops) and typically ignores labor competition or “displacement”. Even if all feedstock production were to take place on new lands (i.e., no land displacement), non-feedstock crop production would still decline due to the reallocation of labor. The downstream processing of biofuels creates very few jobs, with almost all employment effects from biofuels investments coming from feedstock production.6 Moreover, unlike feedstock production, jobs in processing plants are largely reserved for semi-skilled and skilled workers, most of whom have to be sourced from other manufacturing sub-sectors as the biofuels sector grows. Lower-skilled feedstock farmers or laborers mainly come from within the agricultural sector itself. However, both sugarcane and cassava have lower-than-average labor-land ratios. This means that reallocating land to these crops effectively reduces demand for agricultural labor overall. Excess farm workers therefore migrate to the nonfarm sector, especially into less skill-intensive trade and transport services. Establishing a biofuels industry in Tanzania will therefore create new job opportunities for some farmers, but will also impose significant adjustment costs on other workers, especially those in export agriculture.
4.4. Changes in household incomes and poverty The injection of foreign capital and new land resources causes labor demand and wages to rise concomitant with higher labor productivity. 7 This means that, while the profits earned by foreign capital in the biofuels sectors are repatriated, there is an increase in 5 Reported employment numbers do not adjust for underemployment and include unpaid family members. 6 About 620 biofuels processing jobs are created in Sugar 1–2; 860 in Sugar 3; and 248 in Cassava 1–3. 7 The full employment assumption explains why labor wages rise rather than employment levels. If labor supply was assumed to be perfectly elastic then biofuels investments would lead to higher employment levels (at a fixed wage). This would generate larger increases in national GDP and smaller displacement of land and labor resources in the non-biofuels sectors.
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
1929
Sugar 3 Cassava 2
0.6 0.5 0.4
Sugar 1 Cassava 1 Cassava 3 Sugar 2
0.3 0.2 0.1
Change in annual per capita equivalent variation growth rate (%-point)
0.7
0.0 Quintile 1
Quintile 2
Quintile 3
Quintile 4
Quintile 5
Source: Results from the Tanzania CGE and micro-simulation model. Note: Equivalent variation is a measure of household welfare that controls for changes in commodity prices. Expenditure quintiles are based on per capita consumption spending. Fig. 2. Change in equivalent variation from Baseline by quintile, 2007–2015.
the returns to labor, and to a lesser extent land, that causes household incomes to rise. While the increase in household incomes occurs in all of the biofuels scenarios, there are significant differences in the distributional impacts across household groups. Table 9 reports changes in households' equivalent variation, which is a welfare measure that controls for changes in prices. All rural quintiles benefit from the introduction of a biofuels industry in Tanzania. However, higherincome rural households benefit more under larger-scale production scenarios, such as Sugar 2 and Cassava 3, since most large-scale farmers fall into the higher expenditure quintiles. Lower-income households, on the other hand, benefit more under smallholder outgrower schemes, especially when these schemes are combined with improvements in crop yields. Urban households also benefit from an increase in the economywide returns to labor and capital, and from the higher overall level of economic growth in the country. However, it is typically the middle of the urban income distribution that benefits the most, since these quintiles rely more heavily on labor wages for their incomes. Moreover, these households are typically endowed with semi-skilled labor, which is used fairly intensively in the biofuels processing sectors (i.e., as operators and technicians). The national distributional effects of biofuels investments on households' equivalent variation are shown in Fig. 2. While larger-scale sugarcane-based biofuels production benefits all households, it is higherincome households that benefit more than lower-income households (i.e., the curve for Sugar 2 is upward sloping). By contrast, the welfare
gains are more evenly distributed across expenditure quintiles when out-grower schemes are used to produce sugarcane (i.e., Sugar 1). Increasing smallholders' crop yields produces the most pro-poor welfare outcomes. This is reflected in the figure by the higher and downward sloping curves for the Sugar 3 and Cassava 2 scenarios. The mixed cassava production approach (i.e., Cassava 3) is least effective amongst the cassava scenarios in raising household welfare, with higher-income households benefiting the most in this scenario. This is because the displacement of existing farm land in order to establish commercial farms to produce this land-intensive crop is particularly severe for smallholders. Table 10 reports changes in national poverty rates for the various biofuels scenarios. The headcount rate, which measures the share of the population under the poverty line, declines in all scenarios. It declines the most under the two yield-improvement scenarios. Poverty reduction is also more pronounced for technologies that more heavily engage smallholder farmers (i.e., Sugar 1 and Cassava 1) rather than larger-scale commercial farm-based systems (i.e., Sugar 2 and Cassava 3). Finally, investments to enhance crop productivity generate much larger poverty reductions. 5. Conclusions Considerable debate exists concerning the gains from establishing biofuels industries in low income countries, particularly over possible trade-offs between biofuel and food production. It is therefore crucial that governments in countries like Tanzania understand how different
Table 10 Poverty results, 2007–2015.
Headcount (P0) Rural Urban Gap (P1) Rural Urban Squared gap (P2) Rural Urban
Deviation from final Baseline poverty rate, 2015 (%-point)
Poverty rate, 2007 (%)
Baseline poverty, 2015 (%)
Sugar 1
Sugar 2
Sugar 3
Cassava 1
Cassava 2
Cassava 3
40.00 44.72 20.18 13.23 15.01 5.76 6.10 6.97 2.46
36.77 41.34 17.52 12.00 13.70 4.89 5.49 6.31 2.07
− 1.36 − 1.37 − 1.32 − 0.54 − 0.60 − 0.32 − 0.27 − 0.31 − 0.13
− 1.05 − 1.05 − 1.05 − 0.34 − 0.36 − 0.25 − 0.17 − 0.18 − 0.10
− 2.18 − 2.32 − 1.60 − 1.00 − 1.12 − 0.48 − 0.52 − 0.59 − 0.21
− 1.28 − 1.34 − 1.00 − 0.52 − 0.58 − 0.27 − 0.27 − 0.30 − 0.12
− 2.21 − 2.36 − 1.57 − 1.04 − 1.18 − 0.46 − 0.54 − 0.63 − 0.20
− 1.15 − 1.20 − 0.94 − 0.44 − 0.49 − 0.25 − 0.23 − 0.25 − 0.11
Source: results from the Tanzania CGE and micro-simulation models.
1930
C. Arndt et al. / Energy Economics 34 (2012) 1922–1930
biofuel technologies contribute to achieving national development objectives. Drawing on detailed production cost estimates, this study developed a recursive dynamic economywide model of Tanzania to estimate the growth and distributional implications of alternative biofuels production scenarios. These scenarios differed in the feedstock used to produce biofuels (sugarcane and cassava), the scale of feedstock production (small-scale out-grower versus larger-scale plantations), and the way in which feedstock production is increased (yield improvements versus land expansion). Model results indicate that while some individual farmers may shift resources away from producing food crops, there is no nationallevel trade-off between biofuels and food production in Tanzania. Rather, it is traditional export crops that will be adversely affected via the mechanism of a sizable appreciation of the real exchange rate alongside competition for labor and land. It is the relatively large size of Tanzania's agricultural export sector that prevents food production from contracting. This is because the amount of land displaced by biofuel feedstock is smaller than the lands released by declining traditional export crops. As a result, food production increases slightly under most biofuel investment scenarios. Overall, national GDP rises and new employment opportunities are created in biofuels sectors. This leads to welfare gains throughout the income distribution, albeit following a possible period of adjustment in which prices, farm workers and non-biofuel exporters adapt to new market conditions. Our findings suggest that, while all biofuels production scenarios improve household welfare, it is small-scale out-grower schemes, especially for typical smallholder crops like cassava, that are most effective at raising poorer households' incomes. Tanzania should therefore explore opportunities to engage smallholders in the production of biofuels, possibly through mixed small- and large-scale production systems, which provide some security of feedstock supply for downstream processors through the large scale component and the impetus to poverty reduction from the small scale component. If these mixed systems can help to improve feedstock yields in small scale, they become even more desirable. Our findings confirm the growth and welfare gains from yield improvements for feedstocks rather than land expansion. Overall, given its strong pro-poor outcomes and greater profitability, our findings favor a cassava-based biofuels industry for Tanzania. There are, however, a number of limitations to our analysis. Most importantly, while the scenarios based on yield improvements generated the highest levels of pro-poor growth, we only accounted for the private costs involved in establishing the biofuels industry. We did not include public sector costs, such as the provision of irrigation and farm inputs to improve farmers' productivity. Given the difficulties that Tanzania's government has faced in the past in raising smallholders' crop yields, some of the yield-oriented biofuels scenarios may prove overly optimistic. Moreover, in all of the biofuels scenarios we did not consider the cost of providing infrastructure or tax incentives that may be demanded by foreign investors to produce biofuels. If public sector support of biofuels is not consistent with national development plans, then they will incur opportunity costs for Tanzania (see Peters and Thielmann, 2008). In summary, excluding public sector costs, our results indicate that establishing a biofuels industry in Tanzania can contribute to achieving the country's development objectives of enhancing economic growth and reducing poverty. Acknowledgments This paper contributed to the Bioenergy and Food Security (BEFS) project of the Food and Agriculture Organization (FAO) of the United
Nations. We are especially grateful to Irini Maltsoglou and Erika Felix for their valuable comments, and to Heiner Thofern for his stewardship of the BEFS project. Appendix A. Supplementary data Supplementary data to this article can be found online at http:// dx.doi.org/10.1016/j.eneco.2012.07.020. References Arndt, C., Benfica, R., Tarp, F., Thurlow, J., Uaiene, R., 2011a. Biofuels, poverty, and growth: a computable general equilibrium analysis of Mozambique. Environ. Dev. Econ. 15 (1), 81–105. Arndt, C., Msangi, S., Thurlow, J., 2011b. Are biofuels good for economic development? An analytical framework with evidence from Mozambique and Tanzania. Biofuels 2 (2), 221–234. Coles, C., 2009. Processed cassava sub-sector and value-chain analysis in the Mtwara and Lindi regions. Report prepared by Match Maker Associates for the One UN JP1 program, Dar es Salaam, Tanzania. Diao, X., Thurlow, J., 2012. A Recursive Dynamic Computable General Equilibrium Model. In: Diao, X., Thurlow, J., Benin, S., Fan, S. (Eds.), Strategies and Priorities for African Agriculture: Economywide Perspectives from Country Studies. International Food Policy Research Institute, Washington, DC. Dimaranan, B. (Ed.), 2006. Global Trade, Assistance, and Production: The GTAP 6 Data Base. Center for Global Trade Analysis, Purdue University, Indiana. Ewing, M., Msangi, S., 2009. Biofuels production in developing countries: assessing tradeoffs in welfare and food security. Environ. Sci. Policy 12 (4), 520–528. FAO, 2008. The State of Food and Agriculture. Biofuels: Prospects, Risks and Opportunities. Food and Agriculture Organization of the United Nations, Rome. Felix, E., Cardona, A.C., Quintero, J.A., 2010. Technical and economic viability of biofuel production chains. In: Maltsoglou, I., Khwaja, Y. (Eds.), Bioenergy and Food Security: The BEFS Analysis for Tanzania. Food and Agriculture Organization of the United Nations, Rome, Italy. Headey, D., Fan, S., 2008. Anatomy of a crisis: the causes and consequences of surging food prices. Agric. Econ. 39, 375–391 (Supplement). Kapinga, R., Mafuru, J., Jeremiah, S., Rwiza, E., Kamala, R., Mashamba, F., Mlingi, N., 2009. Status of Cassava Production in Tanzania: Implications for Future Research and Development. Report Prepared for the International Fund for Agricultural Development, Rome, Italy. Kretschmer, B., Peterson, S., 2010. Integrating bioenergy into computable general equilibrium models. Energy Econ. 32 (3), 673–686. Melillo, J.M., Reilly, J.M., Kicklighter, D.W., Gurgel, A.C., Cronin, T.W., Paltsev, S., Felzer, B.S., Wang, X., Sokolov, A.P., Schlosser, C.A., 2009. Indirect emissions from biofuels: how important? Science 326, 1397–1399. MINAG, 2004. National Sample Census of Agriculture, 2002/2003. Ministry of Agriculture, Food Security and Cooperatives, Dar es Salaam, Tanzania. NBS, 2002. Household Budget Survey 2000/01. National Bureau of Statistics, Dar es Salaam, Tanzania. Peters, J., Thielmann, S., 2008. Promoting biofuels: implications for developing countries. Energy Policy 36, 1538–1544. Rosegrant, M., 2008. Biofuels and Grain Prices: Impacts and Policy Responses. Testimony for the U.S. Senate Committee on Homeland Security and Governmental Affairs, May 7, 2008. International Food Policy Research Institute, Washington, DC. Rothe, A., Görg, K., Zimmer, Y., 2007. The future competitiveness of sugar beet production in the EU in comparison to sugar cane production in developing countries. Report prepared by the Federal Agricultural Research Centre, Institute of Farm Economics, Braunschweig, Germany. Salvatore, M., Johnston, M., Kassam, A., Bloise, M., Marinelli, M., 2010. Biomass Potential. In: Maltsoglou, I., Khwaja, Y. (Eds.), Bioenergy and Food Security: The BEFS Analysis for Tanzania. Food and Agriculture Organization of the United Nations, Rome, Italy. Searchinger, T., Heimlich, R., Houghton, R.A., Dong, F., Elobeid, A., Fabiosa, J., Tokgoz, S., Hayes, D., Yu, T., 2008. Use of U.S. Croplands for Biofuels Increases Greenhouse Gases Through Emissions from Land-Use Change. Science 319, 1238–1240.
