Global Land Use Impacts of U.S. Ethanol: Revised Analysis Using GDyn-BIO Framework

Abstract

This paper describes dynamic extension of the comparative static computable general equilibrium (CGE) GTAP-BIO model—framework employed in assessments of biofuel policies. In the dynamic extension, called GDyn-BIO, several structural components of the static model, including food demand responses to higher incomes and intensification options in land-based sectors and food processing, were revised to better capture changes in derived demand for land under pressure of growing population and per capita incomes. The impact of 15-billion gallon biofuel mandate on land use, analyzed with the GDyn-BIO model, evolves significantly over time. In particular, net global cropland brought into production due to the mandate declines over time, which is in sharp contrast to the results of static analysis where policy impacts are pictured as fixed for the next 30 years. Despite the fact that land use change impacts of this policy are transitory, environmental impacts and the global warming implications of such policies should not be underestimated. The policy causes earlier conversion of forest and pasture lands to cropland, resulting in earlier GHG emissions and lost carbon sequestration that contribute to global warming.

The research was funded by Electric Power Research Institute.

Appendices

Appendix 1: Calibration of the Crop Production Functions

Following GTAP-BIO, in this modeling the functional form of a crop sector (coarse grains, wheat…) is represented by a nested CES. All non-energy inputs (including land) and capital-energy composite are in the same nest. For the CES production structure, the own Allen-Uzawa elasticity of substitution (AUES) of input i entering the top of the production structure can be calculated using this formula

$$ \sigma_{ii} = \left( {1 - \frac{1}{{S_{i} }}} \right)*\alpha $$

(1)

where σ _ii is own-AUES elasticity of input i, S _i is cost share of input i in total cost of the sector, and α is substitution parameter in the CES production function. Using (1), parameter α can be expressed as a function of the cost share and own-AUES

$$ \alpha = \sigma_{ii } S_{i} /\left( {S_{i} - 1} \right) $$

(2)

For a given estimate of own-AUES, parameter α can be calibrated using (2). Keeney and Hertel (2009) develop a simple theoretical model of supply and producer demand for production factors. Assuming factor market equilibrium, they solve the model for own-price elasticity of sector output. The elasticity depends on the AUES, factor supply elasticities and cost shares of production factors. Next, those authors define a long run as a period over which prices of non-land factors are determined outside of the agriculture, and these factors are available to crop sectors at infinitely elastic supply. Then, land remains the only quasi-fixed factor of crop production. Given these assumptions

$$ \varepsilon_{i} = - \sigma_{\text{land, land}} $$

(3)

where ε _i is supply elasticity of crop output (e.g., coarse grains), and σ _{land, land} is land own-AUES.

If land is fixed, then the only way producers can increase the coarse grains output is through the increase in yield. Thus, using Eqs. (2) and (3) and land cost shares, crop production function can be calibrated to a given own-price yield elasticity.

Appendix 2

In this model, land supply is represented with three-level nested CET frontier (Fig. 2). The values of the CET parameters are calibrated to estimates of own-return land supply elasticities (Lubowski 2002; Ahmed et al. 2008). The relationship between CET parameter in the top of the nested structure and own-return land supply elasticity to forest use is

$$ \sigma_{1} = \frac{{\varepsilon_{\text{f}} }}{{\theta_{\text{f}} - 1}} $$

(4)

where σ ₁ is the CET parameter in the top nest of the nested structure, ε _f is own-return land supply elasticity to forest use, and θ _f is forest land rents share in total land rents. In U.S., for 10 year time span, Ahmed et al. (2008) report ε _f = 0.01 (Fig. 2). Share of forest land rents in total U.S. land rents is 0.13 (GTAP 7 Data Base). Using (1), the calibrated elasticity of transformation in the upper nest of the land supply tree in U.S. is −0.01.

Moving down the tree structure to the agricultural land nest, the relationship between CET parameter and own-return elasticity of land supply to cropping activity is

$$ \sigma_{2} = \frac{{\varepsilon_{\text{c}} - \sigma_{1} \left( {\theta_{a} - 1} \right)\theta_{\text{c}} }}{{\theta_{\text{c}} - 1}} $$

(5)

where σ ₂ is the CET parameter within agricultural land nest, ε _c is own-return land supply elasticity to cropping activity, and θ _c is crop land rents share in total land rents. In U.S., ε _c is about 0.1 (Fig. 2 in Ahmed et al. 2008). Share of crop land rents in total U.S. land rents is 0.82 (GTAP 7 Data Base). Using (2), the calibrated elasticity of transformation in the agricultural land nest of the land supply tree in U.S. is −0.55. The elasticity of transformation among crops and the elasticity of transformation in ruminant livestock nest are left the same as in the earlier analysis with GTAP-BIO, at −0.75 and −10, respectively (Tyner et al. 2010).

To calibrate the elasticities of land transformation in other regions, region specific estimates of own-return elasticities of land supply are needed. For many regions, however, such estimates do not exist. In this analysis we assume that elasticity of transformation in a given level of the CET structure is uniform across AEZs and regions.

Appendix 3

Table 5 Aggregation of GTAP regions

Table 6 Aggregation of standard GTAP sectors and new biofuel-specific sectors