land use change impacts of biofuels: near-var evidence from the us

Ecological Economics 84 (2012) 98–109

Contents lists available at SciVerse ScienceDirect

Ecological Economics

j ourna l homepage: www.e lsev ie r .com/ locate /eco lecon

Analysis

Land use change impacts of biofuels: Near-VAR evidence from the US☆

Giuseppe Piroli a, Pavel Ciaian a,b,c,d, d'Artis Kancs a,c,d,⁎a European Commission (DG Joint Research Centre), Belgiumb Slovenská Polnohospodárska Univerzita (SAU), Slovakiac Economics and Econometrics Research Institute (EERI), Belgiumd Catholic University of Leuven (LICOS), Belgium

☆ The authors acknowledge helpful comments from Cornconference participants in Zürich, GTAP conference in GenEuropean Economic Association and the European MeetinOslo and Malaga. Financial support from the FP7 project ‘Nand Bioenergy Trade’ (AgFoodTrade), and the Federal MAPVV-0706-07, VEGA 10714/09 and KEGA K-08-004-00 isthors are solely responsible for the content of the paper.those of the authors andmay not in any circumstances be rsition of the European Commission.⁎ Corresponding author.

E-mail addresses: [email protected] (G. [email protected] (P. Ciaian), d'[email protected]

0921-8009/$ – see front matter © 2012 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.ecolecon.2012.09.007

a b s t r a c t
a r t i c l e i n f o
Article history:Received 29 August 2011Received in revised form 17 September 2012Accepted 23 September 2012Available online 29 October 2012

JEL classification:C14C22C51C61D58Q11Q13Q42

Keywords:Near-VAREnergyBioenergyPricesLand useBiofuel policiesClimate change

The present paper studies the land use change impacts of fuels and biofuels. We test the theoretical hypoth-esis, which says that changes in fuel prices cause changes in land use both directly and indirectly and that,because of price inter-dependencies, biofuels reinforce the land use change impacts. We apply time-seriesanalytical mechanisms to five major traded agricultural commodities, the cultivated area of agriculturalland and crude oil price. Our data consists of yearly observations extending from 1950 to 2007 for the US.The empirical findings confirm that markets for crude oil and cultivated agricultural land are interdependent:an increase in oil price by 1 dollar/barrel increases land use between 54,000 and 68,000 ha. We also find thatthe increase of bioenergy sector accelerates land use change in the US, i.e. food commodities are beingsubstituted for bioenergy crops.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

Biofuel production is heavily supported in many countries aroundthe world, particularly in developed economies. The main objectives ofbiofuel support policies are to reduce the dependence on fossil fuels, todecrease the greenhouse gas emissions, and to ensure a demand for

elis Gardebroek as well as EAAEeva, the Annual Congress of theg of the Econometric Society inew Issues in Agricultural, Foodinistry's of Education projectsgreatly acknowledged. The au-

The views expressed are purelyegarded as stating an official po-

iroli),uropa.eu ( Kancs).

rights reserved.

surplus agricultural production (OECD, 2008). However, as shown inrecent studies of Fischer (2008); Gardebroek (2010); Kancs andWohlgemuth (2008); andCiaian andKancs (2011), due to price interde-pendencies between the energy, bioenergy, and agricultural markets,biofuel support policies may have far-reaching side-effects on othermarkets. For example, they may directly or indirectly increase foodprices (e.g. Ciaian and Kancs, 2011; Gardebroek, 2010), cause nega-tive environmental impacts on farmland cultivated with biofuel crops(Kancs and Wohlgemuth, 2008), or induce land use changes (Ciaianet al., 2012; Fischer, 2008).

Theoretical models provide two types of explanations for biofuelsupport policies impact on land use: a direct land use change impactand an indirect land use change impact (Ciaian and Kancs, 2011). Thedirect impact on land use change captures the agricultural land con-version to producing bioenergy crops, i.e. biofuel support policiescause substitution in land use between different types of agriculturalcrops. The indirect impact on land use change captures expansion inthe total agricultural area, implying that new land, which previously

http://dx.doi.org/10.1016/j.ecolecon.2012.09.007

mailto:[email protected]

mailto:[email protected]

mailto:d'[email protected]

http://dx.doi.org/10.1016/j.ecolecon.2012.09.007

http://www.sciencedirect.com/science/journal/09218009

1 See Appendix for formal description of the model.2 We use the term ‘channel’ to describe the mechanism of adjustments in prices,

production, land use, etc.3 The term ‘input’ refers to inputs used in the agricultural production process, e.g. fuel,

land, seeds, and fertilisers.

99G. Piroli et al. / Ecological Economics 84 (2012) 98–109

was not used for agricultural production (such as idle land and forestland), is converted into farmland.

The empirical literature on land use change impacts of biofuelsmostly relies on partial and general equilibrium simulation models(CGE) (e.g. Birur et al., 2010; Kancs, 2002, 2006, 2007; Kancs andWohlgemuth, 2008; OECD, 2008; Rosegrant, 2008; Searchingeret al., 2008; Tyner et al., 2010). Searchinger et al. (2008) use a world-wide model to analyse the land use change impact of ethanolincrease in the US. They estimate that ethanol increase of 56 billion lbrings 10.8 million ha of additional land into cultivation worldwide:2.8 million ha in Brazil, 2.3 million ha in China and India, and2.2 million ha in the US. Their results also show that new cropsdo not have to necessarily replace all corn diverted to ethanol(12.8 million ha), because the ethanol by-product replaces roughlyone third of animal feed, which otherwise would be diverted awayin the absence of feed by-product. Tyner et al. (2010) apply a CGEmodel to estimate land use change impact of the US corn ethanolproduction increase from 6.7 billion l in 2001 to 56.8 billion l, asimplied by the Energy Independence and Security Act of 2007.According to their study, the estimated land use changes heavilydepend on model assumptions, such as yields, population growthand base year. Their simulation results imply that producing50.0 billion l of ethanol requires between 1.72 and 2.96 million haof additional land, of which between 0.42 and 1.01 million ha (25%and 34%, respectively) occur in the US. In contrast, Darlington(2009) finds that the expansion of corn ethanol production to56.8 billion l per year by 2015 will unlikely result in conversion ofnon-agricultural land, arguing that yield improvements will offsetthe global demand for cropland to meet the corn ethanol productiongrowth.

Most of the simulation studies find that biofuel support policiesaffect land use positively. However, there is wide variation in themagnitude of the simulated impacts between studies (Kancs andWohlgemuth, 2008). An important source of this variation is due tocalibrated parameters, particularly those that directly or indirectly af-fect biofuels, such as price transmission elasticities. This issue isamong the key advantages of econometric models, as they havemuch stronger emphasis on the use of observed data in the past. Onthe other hand, CGE models have richer theoretical structure andare able to capture induced general (partial) equilibrium feedback ef-fects. Hence, the methodological approach followed in the presentstudy is complementary to the partial and general equilibrium simu-lation models. While the econometric approach employed in thepresent study allows to estimate the land use change elasticitieswith respect to fuel and biofuel price changes, the CGE models aresuperior to simulate long-run land use changes using the estimatedparameters.

The objective of this paper is to assess the land use change impactof increasing oil prices which, together with bioenergy support poli-cies, make the production of bioenergy more profitable and hence re-quire more farmland. We identify two channels through whichchanges in oil prices are transmitted to agricultural input and outputmarkets: an indirect input channel and a direct output channel. Theindirect input channel affects farm production costs in agriculture,whereas the direct output channel interacts through changes in de-mand for agricultural commodities (which can also be used for biofu-el production). Theoretically, the total effect is ambiguous, becausethe two channels affect land use in opposite directions. The relativestrengths of the two channels, which depend on the relative impor-tance of energy-based inputs in agricultural production and on thesize of biofuel production, respectively, determine the long-run equi-librium on energy, bioenergy and agricultural markets.

The theoretical model offers testable hypothesis, which says thatdue to the expansion of biofuel production, the output channel willlikely offset the input channel leading to net positive impact on thetotal agricultural land use, and causing important substitution in

land use between different agricultural crops. This may be particularthe case of the last 15 years, when the US production of ethanolincreased continuously 1100 million gal in 1996 to 6500 million galin 2007 (RFA, 2012).

Empirically, we estimate the net effect (direct output net ofindirect input channel) of changes in oil prices on agricultural landuse, because the two effects cannot be identified separately in ourdata. Our estimation approach is based on the near Vector AutoRegression (VAR) model. The main advantage of the near VAR ap-proach is that it allows to control for structural effects, such as GDP,yield, consumption, production, that may impact land use changes.To examine the role of biofuels, we take into account structuralbreaks for the period of biofuel expansion. From the obtained esti-mates, we calculate long-run land use change elasticities. The esti-mated elasticities of land use change reveal the interdependenciesbetween energy, bioenergy and agricultural markets. These resultshave high policy relevance, because knowing the potential impact ofmacroeconomic and policy shocks on all inter-dependent marketsallows to increase policy efficiency, and to avoid negative/offsettingside effects.

2. Theoretical Framework

2.1. The Model1

Building on theoretical models of Gardner (2007); Gorter and Just(2009); Ciaian and Kancs (2011), we develop a stylised theoreticalmodel in order to identify the channels transmitting price signals be-tween energy, bioenergy and agricultural markets.2 While acknowl-edging that the energy, bioenergy and agricultural markets areaffected by a multitude of other factors in reality, developing a theo-retical model with all the linkages would unnecessary increase thecomplexity of the analysis, without adding significant insights foridentifying the key linkages between the energy, bioenergy and agri-cultural markets. However, most of the relevant covariates, such asmacroeconomic development, are introduced in the empirical analy-sis (e.g. GDP growth, exchange rate, population growth, and infla-tion). Compared to theoretical models in the literature, the mainadvantage of our approach is that it explicitly models the agriculturalsector and explicitly identifies the key transmission channels be-tween the energy, bioenergy and agricultural markets: a direct outputchannel and an indirect input channel.3

Following Ciaian and Kancs (2011), we model five mutuallyinterdependent food-energy markets: agricultural, biofuel, fossil fuel,by-product, and input. The representative farm can substitute its out-puts between biomass and food, according to a constant returns toscale production function of two substitutable inputs: fuel and other in-puts (referred to as ‘land’). Biomass output is supplied to food and bio-fuel markets, whereas food commodity is supplied solely to the foodmarket. Biofuel sector produces biofuels and by-product from biomass.The aggregate fuel supply is a sum of biofuel and fossil fuel.

Price signals between the fuel and the agricultural markets aretransmitted through two channels: an indirect input channel and adirect biofuel channel. The indirect input channel affects farm pro-duction costs on the agricultural market, whereas the direct biofuelchannel interacts through biofuels' demand for agricultural commod-ities on the agricultural markets.

100 G. Piroli et al. / Ecological Economics 84 (2012) 98–109

2.2. Total Land Use Change

Fuel price affects agricultural production costs and hence the prof-itability of land through the indirect input channel by causing, forexample, a decrease in land demand as a result of an increase infuel price. More precisely, due to higher input (fuel) costs, the pro-duction and hence land use decreases, because agricultural land andfuel are imperfect substitutes (see Ciaian and Kancs, 2011 for a moredetailed exposition). Second, the direct biofuel channel has the opposite(increasing) effect on the total land demand. Higher fuel price stimu-lates biofuel demand, leading to an upward adjustment of agriculturalprices, thus improving land profitability. Higher agricultural land de-mand stimulates conversion of idle and forest land into agricultural land.

The overall effect depends on the relative strengths of the twochannels. If biofuels play an important role in agricultural markets,then the direct output (biofuel) channel will offset the indirectinput channel resulting in higher land use. Otherwise, the total landuse will decline. In the US, the output channel will likely be strongerthan the input channel, particularly in the period of biofuel produc-tion expansion starting from 1996. For example, the ethanol produc-tion increased from 1100 million gal in 1996 to 6500 million gal in2007 (RFA, 2012).4

According to the underlying theoretical framework, the indirect landuse change is caused only by the biofuel channel. In the empirical anal-ysis, however, wewill be able to estimate only the combined (total) im-pact of both channels on the indirect land use change, because theobserved indirect land use change is a result of the total land use re-sponse net of the land use change caused by the input cost adjustments(indirect input channel). Hence, our estimates will represent a lowerbound of the impact of biofuels on land use. As noted above, in orderto account for the expansion of biofuels after 1996, we include adummy variable controlling for structural break in the data series.

2.3. Land Use Substitution between Commodities

As a result of biofuel expansion, land substitution between com-modities may take place. The indirect input channel has the sameimpact on both biomass and food commodities. An increase in fuelprice causes higher production costs, leading to lower land cultivationof both commodities. The exact impact depends on the relative fuelintensity of agricultural commodities. With fuel price increase, thefuel intensive commodities will reduce land demand relatively morethan the fuel extensive commodities.

The direct biofuel channel will affect biomass and food commoditiesdifferently. The demand for biomass in the biofuel production due tohigher fuel prices will stimulate land cultivation. On the other hand,the food commodity's land demand will decline due to rising biofuelprice, as farmers will substitute production to more profitable biomass.Depending on output substitutability and input energy's intensity, theoverall effectwill be different for the two types of commodities. For bio-mass, the land use change depends on the relative strengths of the twochannels (the same effect as in the case of indirect land use change). Forthe food commodity, both channels have the same direction, land usedecreases in fuel prices, i.e. due to higher input (fuel) costs, the produc-tion and hence land use decreases.

3. Empirical Implementation

3.1. Econometric Specification

Theoretical findings from the previous section suggest that ener-gy, bioenergy and agricultural markets are interdependent. Energy

4 Biodiesel production is considerably less important, consisting only of a small sharein the total biofuel production in the US.

affects agricultural markets through altering agricultural productioncosts, and through biomass demand for biofuel production. Reversely,agricultural markets impact energy markets through agricultural fueldemand and biofuel supply on the fuel market (Ciaian and Kancs,2011). In order to account for the induced effects on land use changes,the econometric approach needs to account for these inter-sectorallinkages.

In standard regression models by placing particular variables on theright hand side, the endogeneity of all variables sharply violates theexogeneity assumption of a regression equation (Kielyte, 2008). Besidesthe energy–bioenergy–agriculturalmarket interdependencies shown inthe theoretical model, one needs to control for other factors that mayaffect the three markets and linkages between them. The agriculturaland energy markets depend on macroeconomic developments, suchas GDP growth, consumption development, and exchange rate. Forexample, a favourable macroeconomic development may induce ad-justments in both energy and agricultural markets through stimulatingproduction and hence causing land use changes and fuel price rise.These structural adjustments may confound the estimations, likelycausing upward bias of the estimated land use effects. In order to takeinto account these factors, we consider a “structural equilibrium frame-work”, by including a number of macroeconomic and policy controlvariables, such as GDP growth, population, exchange rate, index ofshare prices, agricultural subsidies, share of agriculture in GDP/valueadded, oil production and price indices. The inclusion of these structuralvariables allows us to take into account a number of effects that other-wise would bias the results. Many econometric approaches applied inthe literature for biofuel market analysis relay on simple cointegrationand autoregressive model. This is an important shortcoming as theydo not take into account structural effects which may significantlybias their estimates.

The problem of endogeneity can be circumvented by specifying aVector Autoregressive (VAR) model on a system of variables, becauseno such conditional factorisation is made a priori in VAR models. In-stead, variables can subsequently be tested for exogeneity, and re-stricted to be exogenous then. Given these advantages, a generalapproach in the literature to analyse the causality between variablesin the presence of endogenous variables is VAR model (Lütkepohland Krätzig, 2004). VAR is a statistical model used to capture the lin-ear interdependencies among multiple time series, which generalisesthe univariate autoregression (AR) models. All the variables in a VARare treated symmetrically, each variable has an equation explainingits evolution based on its own lags and the lags of all the other vari-ables in the model. VAR models can be seen as a ‘theory-free’ ap-proach to estimate economic relationships. Given that in VAR allvariables are modelled together as endogenous, it serves as an alter-native to the “incredible identification restrictions” in structuralmodels (Kielyte, 2008).

In the present paper, we employ a near-VAR model, which pro-vides an extension of the structural VAR approach, as it does not im-pose the same lag lengths in all the equations of the system.5

Compared to the standard VAR, a near-VAR offers the possibility to in-corporate economic constraints, which affect the parameter of laggedregressors and, in thisway, avoiding double-counting the effect of inter-action effects between the variables. This becomes very important if themodel relies on data at different levels, for example micro/macro vari-ables, or, as in our case, sectoral and aggregate data (Imai et al., 2008).In the context of the present paper, another important advantage of

5 A near-VAR is a special case between the VAR models, which allows for linear re-strictions in the system of equations. This means that it is possible to have differentnumber of regressors in each equation. Adjusted-VAR, quasi-VAR and restricted-VARare other terms used by different authors in the literature.

Table 1Data sources and variable description.

Variable Average Std. dev. Unit Source

Value added inagriculture

2.63 1.46 % of GDP USDA

World population 4,434,079 1,274,294 Thousands OECDExchange rate 2.79 1.07 Mark (Euro)/USD PACIFIC ERSShare price 28.29 35.01 Index, 2005=100 IMFDirect governmentpayments

10.86 6.98 Index, 2005=100 USDA

Oil price 15.68 14.46 USD/barrel IOGAConsumer Price Index 3.84 2.87 Index, 1982=100 BLSProducer PriceIndex in agriculture

78.89 31.42 Index, 1982=100 BLS

Oil production 7.41 1.29 Million barrels/day DOEGPD growth 3.48 2.31 % change of GDP IMFCropland harvested 314.05 17.79 Million acres USDABarley price 1.83 0.75 USD/100 pounds USDABarley harvested 8901.64 3058.46 Thousand acres USDABarley yield 46.32 11.84 Pounds USDACorn field price 1.96 0.72 USD/100 pounds USDACorn field harvested 66,991.36 6983.48 Thousand acres USDACorn field yield 94.64 34.36 Pounds USDARice price 6.92 2.36 USD/100 pounds USDARice harvested 2481.34 634.03 Thousand acres USDARice yield 47.71 12.7370 Pounds USDASoybeans price 4.68 1.99 USD/100 pounds USDASoybeans harvested 49,524.43 19,639.37 Thousand acres USDASoybeans yield 29.88 6.76 Pounds USDAWheat price 2.75 1.05 USD/100 pounds USDAWheat harvested 57,152.38 8879.16 Thousand acres USDAWheat yield 31.80 7.69 Pounds USDA

Notes: USDA = US Department of Agriculture; IOGA = Illinois Oil & Gas Association;PACIFIC ERS = PACIFIC Exchange Rate Service, a service for academic research andteaching provided by UBC's Sauder School of Business; DOE = US Department of Ener-gy; OECD= Organisation for Economic Cooperation and Development; IMF = Interna-tional Monetary Fund; BLS = Bureau of Labor Statistics; USD = United States dollar.


the near-VAR model is that it offers a robust statistical framework withthe ability of integrating alternative economic constraints and structuraleffects that may impact land use change. Important structural effectsrepresent e.g.: (i) inter-sectoral effects as derived in the theoretical sec-tion, and (ii) macroeconomic effects on the energy and agriculturalmarkets, as shown for example in Ciaian and Kancs (2011).

In order to estimate the land use change impacts of biofuels, wespecify the following near-VAR model:

YtXct

Xbt

Xrt

Xst

Xwt

26666664

37777775¼

A Lð Þ 0 0 0 0 0B Lð Þ C Lð Þ 0 0 0 0D Lð Þ 0 E Lð Þ 0 0 0F Lð Þ 0 0 G Lð Þ 0 0H Lð Þ 0 0 0 I Lð Þ 0J Lð Þ 0 0 0 0 K Lð Þ

26666664

37777775

Yt−1

Xbt−1

Xct−1

Xrt−1

Xst−1

Xwt−1

26666664

37777775

þ

MNOPQR

26666664

37777775Zt þ

M Lð ÞN Lð ÞO Lð ÞP Lð ÞQ Lð ÞR Lð Þ

26666664

37777775Zt−1 þ S

εytεctεbtεrtεstεwt

26666664

37777775

ð1Þ

where set Zt contains exogenous variables, set Yt contains endogenousmacroeconomic variables, and set Xt

i contains sector-specific endoge-nous variables. All equations include a constant term and, in order toaccount for adjustments over time, equations in set Yt include also atrend variable and a dummy variable controlling for structural breakin the data. The blocks from A(L) to R(L) contain the parameters ofthe lagged variables that are estimated by the model. For example,variables in set Yt are not affected by lagged variables in set Xt

i, butare affected by their own (and exogenous) variables. The only param-eters to be estimated, therefore, are those contained in the block A(L),while the others are bounded to 0.

Matrix (1) describes a general representation of the underlyingstructural model. More precisely, long-run relationships between en-ergy, bioenergy and agricultural markets are identified, and macro-economic variables affecting these markets are determined. Thevariables included in matrix (1) are detailed in the next section.

3.2. Data and Variable Construction

Our data consists of yearly observations for crude oil, land use forfivemajor traded agricultural commodities (barley, corn, rice, soybeans,andwheat), 6 the total land, and several structural and control variables(e.g. GDP growth, world population, exchange rate, and agriculturalsupport). The data covers US agricultural and energy markets for theperiod 1950–2007. Crude oil prices are represented by historical freemarket (stripper) prices. In empirical analysis, prices are normalisedto the year 1982. Also agricultural data, such output prices, land useand subsidies are all collected on yearly basis. Descriptive statistics ofkey variables and data sources are detailed in Table 1.

In line with the underlying near-VAR model, all variables areregrouped into three sets: exogenous variables, Zt, endogenous mac-roeconomic variables, Yt, and sector-specific endogenous variables, Xt

i.Set Zt contains exogenous variables that control for change in the

total world demand and structural characteristics of economic systemand the US agricultural policy: share of agricultural value added inGDP, agr _sharet, world population, poptw, US dollar exchange rate,

6 These commodities represent almost 75% of the total utilised agricultural area inthe US.

exc _markt, index of US share prices, share _pricet, and agriculturalsubsidies, subsidiest:

Z′t ¼ agr

Psharet popwt exc

Pmarkt share

Ppricet subsidiest

h ið2Þ

Set Yt includes agricultural aggregates and macroeconomics vari-ables that, according to the underlying theoretical framework, are en-dogenous: oil price, oil _pt, oil production, oil _prodt, consumer priceindex, cpit, farm producer price index, farm _pt, GDP growth, and totalagricultural land, tot _ landt:

Y ′t ¼ oil

Ppt cpit f pit oil

Pprodt Δgdpt tot

Plandt

� �: ð3Þ

Equations in setY ′t include also a trend variable and a dummy var-

iable, which controls for structural break in the series. In the empiricalanalysis we estimate four models: a basic model, and three augment-ed models. The basic model contains only trend variable. To ensurerobustness of basic model's results and to take into account structuralbreak due to biofuel expansion in the US,7 the three augmented

7 Note that the dummy variable controlling for structural break may control also forother effects besides the biofuel expansion. For example, it may control for thedecoupling of agricultural subsidies in US in 1996, or support measures for biofuel pro-duction in US (e.g. blender's tax credit and consumption mandate). Under theblender's tax credit, the blender received a subsidy per gallon of biofuel blended witha fossil fuel (i.e., gasoline or diesel). The mandate established that a fixed amount ofbiofuel be blended with a fossil fuel.

8 In our identification strategy, we assume that K is an identity matrix.9 Consumer price index also affects the producer price index of farm products.


models include also a dummy variable capturing the structural breakin 1996, when the biofuel production (ethanol) started to expand inthe US (RFA, 2012). According to the underlying theoretical model,expansion in biofuel sector would imply an increase in the relative im-portance of the direct biofuel channel. The first augmented model con-siders a structural break in 1996. Dummy variable, d _biofuel _ 1996,takes value 0 in the first period (1950–1995), and value 1 in the secondperiod (1996–2007). According to RFA (2012), ethanol productionincreased by 27% in 1998, and 48% in 2000 relative to 1996. This impliesthat in the following years the relative importance of the direct biofuelchannel might have increased. To ensure the robustness of our results,we estimate two additional augmented models, which include dummyvariables, d _biofuel _ 1998 and d _biofuel _ 2000, for the years 1998and 2000.

Finally, set Xti contains variables, which disaggregate the total ag-

ricultural sector into sub-sectors. We consider five agricultural(sub-)sectors: corn, wheat, rice, barley and soybeans. For each agri-cultural sector we include three variables in set Xt

i: price, land useand yield:

Xi′

t ¼ priceit landit yieldit

h ið4Þ

where i=corn, barley, rice, soybeans and wheat.Total land, tot _ landt, commodity-specific land use, landti, and oil

price, oil _pt, are the key variables of our study. They interlink thethree related markets: energy, bioenergy and agricultural. Accordingto the underlying theoretical model, the estimated impact of oil priceson land use will be a result of both channels: the direct biofuel andthe indirect input channels. If the biofuel channel is stronger thanthe input channel, then the overall estimated effect on the indirectland use, landti, will be positive. In the reverse case, the estimatedimpact will be negative. This implies that in the empirical analysisthe biofuel link to the land use can be tested for only indirectly.

The underlying theoretical model also suggests that when ac-counting for structural break in biofuel expansion by including adummy variable, the relative importance of biofuel channel woulddecline compared to the basic model. In addition to biofuel expan-sion, the structural dummy variable controls also for other structuraleffects, such as policy impacts, and thus may reduce the responsive-ness of land use to oil prices. For example, if oil price induces a pos-itive indirect land use change in the basic model, then we expectthat in the augmented models with dummy variable the overall esti-mated effect on the indirect land use would be reduced.

The key difference between sets Yt and Xti is that variables of Yt im-

pact variables in set Xti but, according to the underlying theoretical

framework, there is no causality in the reverse direction. The causalitybetween the agricultural aggregates and macroeconomic variables isaccounted for in set Yt. To avoid duplication of the effects, set Xt

i isassumed not to affect set Yt. Implicitly, this setup implies thatcross-effects between the five agricultural sectors are captured byvariables in set Yt. This specification ensures that the specifiednear-VAR model has sufficient degrees of freedom, while avoidingthe duplication of causal effects between variables. In the sametime, it allows to estimate commodity-specific effects.

3.3. Market Interdependencies

Next, we analyse the contemporaneous interactions between en-dogenous variables, for which we set up an identification scheme.Denote θt as residuals obtained from the reduced-form estimationwhich, according to the underlying structural model, are related tothe structural shocks:

Wθt ¼ Kεt : ð5Þ

The underlying theoretical framework suggests the followingidentification scheme for the impulse response functions8:

1 0 0 0 a15 0 0 0 0a21 1 0 0 a25 0 0 0 0a31 a32 1 0 0 0 0 0 00 0 0 1 0 0 0 0 00 0 0 0 1 0 0 0 00 0 0 0 0 1 0 0 0ai71 ai72 ai73 0 0 0 1 0 00 0 0 0 0 0 0 1 00 0 0 0 0 0 0 0 1

26666666666664

37777777777775

θoilptθcpitθf pit

θoilprodtθΔgdptθtotlandtθipricetθilandtθiyieldt

2666666666666664

3777777777777775

¼

εoilptεcpitεf pit

εoilprodtεΔgdptεtotlandtεpriceitεlanditεyieldit

26666666666664

37777777777775

:

ð6Þ

Matrix (6) represents the identification of contemporaneous rela-tionships between model variables and their responsiveness to exog-enous shocks. In line with the underlying theoretical framework, weassume that prices are more responsive to shocks than other vari-ables. For example, for agricultural production and land use it takeslonger time to adjust due to sunk costs, costly adaptation of technol-ogy and natural constraints. Imposing these constraints on therelationships between the variables allows for a contemporaneousimpact of oil price changes on all other prices. Similarly, cpi and fpiare assumed to affect agricultural prices contemporaneously.9 Rowseven in matrix (Gardebroek, 2010) can be interpreted as a costbased inflation of each agricultural sector. Further, we assume a con-temporaneous oil price (and consumer price index) response toshocks in GDP growth.

4. Results

4.1. Specification Tests

As usual, we start with determining the lag length. Three standardtests are performed to determine the optimal lag length: Akaike In-formation Criterion, Schwarz Criterion and Hannan–Quinn Criterion.According to all three tests, the optimal lag order is one. Next, wetest for the existence of cointegrating vectors. Both the likelihoodratio and the trace tests confirm that our data are co-integrated inlevels, which justifies the specified near-VAR model in levels.

4.2. Indirect Land Use Change Impact

After identification of the structural model, in order to determinewhether the specified model is reasonable, we perform innovationaccounting. The impulse response functions for the total land usechange for the basic model are reported in Fig. 1, which shows the in-novation (oil price shock of one standard deviation) response (area oftotal land use change) in a horizon up to 20 years. In other words, weincrease oil price, and predict the response of land use change bytaking into account the structural relationships between energy,bioenergy and agricultural markets, and by controlling for macroeco-nomic effects.

The impulse response results suggest important indirect land usechanges: the total cultivated area increases significantly in the analysedperiod as a result of an increase in oil price (Fig. 1). The largest impactcan be observed 8 years after the oil price shock of one standard devia-tion (14.46 USD/barrel), when the total agricultural area increases byalmost 1 millionha. In the following years the increase in the cultivatedland is slightly lower (778,000 ha in 20 years). The small negative de-crease in the first year after the oil price shock can be explained bythe fact that farmers are not able to expand their crop land area imme-diately, although higher input costs are effective instantly. As a result, if

Fig. 1. Impulse response function of total land use (basic model). Notes: Y axis: area inmillion acres; X axis: years. Impulse: positive oil price shock (1 STD) in USD; response:changes in land use.


everything else would stay constant, then higher input costs wouldimply reduction in input use, also land.

Next, we examine whether structural changes, such as the expan-sion of biofuel incentives and increase in biofuel production, alongwith other effects, such as the decoupling of agricultural subsidiesin 1996, might have affected indirect land use change. As explainedin the previous section, to control for structural break in the secondhalf of nineties, in addition to the basic model we also estimatethree augmented models, which include a dummy variable (for1996, 1998 and 2000). We expect that the inclusion of the dummyvariable would decelerate the positive effect of oil price on the totalcultivated area, because part of the indirect land use change inducedby the biofuel channel may be absorbed by the structural dummy var-iable. The results are reported in Fig. 2.

The results suggest that, when controlling for structural break in1996, the significant increase in the total cultivated area is confirmed,though the magnitude is lower. In the peak (5–6 years after the oilprice shock), the increase in the total cultivated area is 813,000 ha(Fig. 2), which is around 25% less compared to the basic model (with-out controlling for structural break) (Fig. 1). Also in the followingyears the response to oil price shock is lower, when accounting forstructural break.

Fig. 2. Impulse response function of total land use (controlling for structural break in1996). Notes: Y axis: area in million acres; X axis: years. Impulse: positive oil priceshock (1 STD) in USD; response: changes in land use.

As in the basic model, the impact is larger in the medium-run(5–6 years) compared to the long-run (20 years), when the increasein the total cultivated area is 680,000 ha (Fig. 2). The differences be-tween the medium- and long-run elasticities in the indirect landuse change can be explained by the output (price) channel (increasein the bioenergy supply exerts downward pressure on oil price), andinput (cost) channel, because farming itself uses crude oil products(e.g. diesel), the price of which would increase due to higher oilprices. 10

A similar pattern is observed also for the other two augmentedmodels with structural breaks in 1998 and 2000 (see Figs. 3 and 4).The indirect land use change is lower compared to basic model and rel-ative to the model with structural break 1996. Compared to the basicmodel, the indirect land use response in the augmented model withstructural break in 1998 is lower by 35% after 10 years and by 56%after 20 years. In the model with structural break in 2000, the respec-tive differences are by 52% after 10 years and by 62% after 20 years.

Generally, these findings are in line with the underlying theoreti-cal model and with our expectations: both the magnitude of increasein the total crop land area, and the responsiveness of the cultivatedarea to changes in oil price are lower, when basing our calculationson the augmented model's elasticity estimates. This can be explainedby the fact that in the second period (1996–2007) the size ofbioenergy sector was considerably higher than in the first period(1950–1995) which, according to the underlying theoretical frame-work, would marginally decrease the indirect land use change impactof biofuels.

The magnitude of our indirect land use change estimates are com-parable to previous studies (Searchinger et al., 2008; Swinton et al.,2011; Tyner et al., 2010). Searchinger et al. (2008) find an increasein agricultural land use due to biofuels of 2.2 million ha. Swintonet al. (2011), report that the 2006–2009 leap in field crop prices andthe attendant 64% gain in typical profitability led to a 2% increase incrop planted area. In order to compare our estimates with those ofSwinton et al. (2011), we need to convert the estimated elasticitiesand marginal effects into total (cumulative) changes in crop landarea. Taking the average cultivated area of 127.09 million ha as thebase, 2% would imply an increase of 2.54 million ha. Our results sug-gest that an increase in oil price by 1 dollar/barrel increases land usebetween 54,000 and 68,000 ha. Given that between 2006 and 2009the crude oil price increased by around 50 dollars/barrel, this impliesthat the cultivated area increased by 2.70–3.40 million ha, which isslightly higher compared to Searchinger et al. (2008) and Swintonet al. (2011). However, the results are of the same order of magni-tude, which underlines the general finding that the three marketsare interlinked and changes in energy/bioenergy markets does affectagricultural markets, e.g. through changes in the demand for agricul-tural land.

4.3. Direct Land Use Change Impact

The impulse response functions for the direct land use change of indi-vidual agricultural commodities are reported in Fig. 5 (basic model). Theresults suggest that all agricultural commodities' land use is affected byincreasing energy prices, including those commodities that are not di-rectly used for bioenergy production (e.g. rice and barley). As expected,

10 The cost effect hypothesis is confirmed by the coefficient of lagged oil price. In thebasic model, the lagged oil price has a negative direct impact (−0.533) on the total ag-riculture land, which is likely caused by increasing cost of production. In the augment-ed model with structural break, the estimated parameter is still negative (−0.519) andstatistically significant (p value 0.004), which reveals the depressing impact on totalagricultural land expansion of an increase in oil price.

image of Fig.�2



the land use of corn increases significantly (233,000 ha after 10 years)in response to an increase in oil price by one standard deviation(14.46 USD/barrel), which is due to the fact that corn is the maincommodity used for biofuel production in the US. Similarly, alsothe land use for wheat and barley production increases (130,000and 223,000 ha, respectively). In contrast, the area of soybean isnegatively affected by a positive fuel price shock (−292,000 ha)(Fig. 5). This could be due to lower importance of soybeans in biofuelproduction, which is more than offset by the indirect input channel.Additionally, farmers may substitute from non-biofuel crops to bio-fuel crops (e.g. corn). Further, soybeans represent an importantinput for animal production and the indirect input channel may im-pact the land use of soybean also through adjustments of input costsin animal sector. The smallest response in land use change is estimat-ed for rice (−46,000 ha).

These results are in line with the underlying conceptual frame-work and confirm that, in addition to the indirect land use change im-pact, also important direct land use changes take place. The resultsalso suggest an important role of biofuels in land use change. Thoseagricultural commodities, which are used also for the production ofbioenergy, e.g. corn and wheat, are more sensitive in terms of changes


in the cultivated area, than those commodities, which are not directlyused for the production of bioenergy, e.g. rice.

As above,we examinewhether structural changes, such as the expan-sion of biofuel incentives and increase in biofuel production, along withdecoupling of agricultural subsidies, might have affected indirect landuse changes. We employ exactly the same augmented models withdummy variable specification for structural break. Commodity-specificresults with structural break in 1996 are reported in Fig. 6.

The results for the augmented models (controlling for structuralbreak in the series) suggest that the shape of the impulse response func-tion is similar to the basic model but, as above, the magnitude of landuse change is lower (except for wheat). For example, the response inland use in the augmentedmodel with structural break in 1996 relativeto the basic model is 4%–68% lower (depending on commodity) after5 years. For wheat, the land response is 80% higher after 5 years com-pared to the basic model. This can be explained by the fact that in thesecond period (1996–2007) the size of bioenergy's production wasconsiderably higher than in the first period (1950–1995) which,according to the underlying theoretical framework, would increasethe direct land use change impact of biofuels. As above, most of thecommodities show larger response in medium-run (5–6 years), andfaster absorption of oil price shock on indirect land use change.

4.4. Elasticities of Land Use Change Impacts

The estimated near-VAR model allows us to assess the impact ofboth types of land use change impacts of biofuels: land use changecaused by the indirect input channel, and land use change causedby the direct biofuel channel. In addition, the results from the impulseresponse analysis allow us to calculate medium- and long-run (5, 10and 20 years) land use change elasticities. The calculated elasticitiesare reported in Table 2.

The magnitude of the estimated land use change elasticities rangesbetween −32,500 ha and 67,700 ha per 1 dollar/barrel increases infuel price in the basic model. Wheat, barley and corn have the elasticitybetween 5100 and 17,900 ha per 1 dollar/barrel increases in fuel price.The elasticity of rice and soybean is between −32.5 and −1.4 ha per1 dollar/barrel increases in fuel price. The largest effect is estimated forthe total land use, which is equal up to 67,7000 ha per 1 dollar/barrel.

The land use elasticities are lower in the augmented models com-pared to the basic model. Analogously, the elasticities are lower, thelater the structural break ismodelled (1996, 1998 and2000). The only ex-ception is wheat, for which the elasticity increases. The magnitude of theestimated land use change elasticities ranges between −32,600 ha and55,200 ha per 1 dollar/barrel in the augmented models.

From these results we can further imply that biofuels indeed playan important role, i.e. they affect land use, because the land usechange impact is larger for food/bioenergy commodities (e.g. corn)than for food only commodities (e.g. rice). At the same time, whencontrolling for structural break, there is a reduction in the magnitudeof elasticities, implying that part of the biofuel effect is absorbed bythe structural break dummy. These findings are consistent bothwith the theoretical hypothesis and with previous studies (Birur etal., 2010; Searchinger et al., 2008; Swinton et al., 2011; Tyner et al.,2010). For example, Birur et al. (2010)) find that the bioenergy poli-cies in the US and EU would substantially increase the share of agri-cultural products (e.g., corn in the US, oilseeds in the EU, and sugarin Brazil) utilised by the biofuels sector. In the US, this share couldmore than double from 2006 levels, while the share of oilseedsgoing to biodiesel in the EU could triple.

5. Limitations and Outline for Future Research

The magnitudes of the results reported in Table 2 should be con-sidered cautiously, however, due to several caveats of our model.First, we cannot separately identify the direct and indirect land use

image of Fig.�3

image of Fig.�4

Fig. 5. Impulse response functions of direct land use change (basic model). Notes: Y axis: area in thousand acres; X axis: years. Impulse: positive oil price shock (1 STD) in USD;response: changes in land use. Commodities from the top: barley, corn, rice, soybeans and wheat.


change impacts of biofuels. This implies that our empirical estimatesrepresent a lower bound of the indirect biofuel effects. Given thatthe estimated land use change is net of land use change caused bythe fuel cost adjustments (indirect input channel) and, given that

Fig. 6. Impulse response functions of direct land use change (controlling for structural breakshock (1 STD) in USD; response: changes in land use. Commodities from the top: barley, co

we find a substantial response of land use to changes in oil price,the actual size of the indirect land use change impact of biofuelsmay be even higher in reality than our results suggest. However,Ciaian and Kancs (2011) show that the indirect input channel is

in 1996). Notes: Y axis: area in thousand acres; X axis: years. Impulse: positive oil pricern, rice, soybeans and wheat.

image of Fig.�5

image of Fig.�6

Table 2Estimated long-run land use elasticities.

1000 ha 1000 ha/1 USD/barrel

5 years 10 years 20 years 5 years 10 years 20 years

BarleyBasic modelStructuralbreak

216.1 223.6 164.5 14.9 15.5 11.41996 165.0 131.8 97.6 11.4 9.1 6.71998 198.5 176.5 91.4 13.7 12.2 6.32000 110.1 87.2 50.3 7.6 6.0 3.5Corn

Basic modelStructuralbreak

258.6 233.7 165.0 17.9 16.2 11.41996 248.3 181.1 135.7 17.2 12.5 9.41998 226.2 166.9 83.5 15.6 11.5 5.82000 132.1 98.8 57.7 9.1 6.8 4.0Rice


−63.8 −45.7 −20.2 −4.4 −3.2 −1.41996 −48.0 −25.7 −3.9 −3.3 −1.8 −0.31998 −65.5 −49.1 −20.9 −4.5 −3.4 −1.42000 −15.3 −13.2 −3.4 −1.1 −0.9 −0.2Soybeans


−470.2 −291.8 −58.7 −32.5 −20.2 −4.11996 −152.4 35.6 156.6 −10.5 2.5 10.81998 −470.8 −325.9 −118.7 −32.6 −22.5 −8.22000 −215.5 −39.8 35.1 −14.9 −2.7 2.4Wheat


137.1 129.9 74.1 9.5 9.0 5.11996 246.1 259.4 212.2 17.0 17.9 14.71998 136.0 108.4 49.9 9.4 7.5 3.52000 364.4 224.9 94.1 25.2 15.6 6.5Totalland


877.0 979.7 777.6 60.6 67.7 53.81996 798.4 746.3 680.1 55.2 51.6 47.01998 698.6 640.6 343.1 48.3 44.3 23.72000 546.2 467.0 297.4 37.8 32.3 20.6

Source: Authors' estimations based on US data for 1950–2007. Notes: Land use changeelasticities are calculated based on the impulse response analysis.


small and statistically not significant, implying that the downwardbias is likely to be small.

Second,we do not explicitly estimate the impact of changes in biofuelpolices on land use changes. The impact of policy changes is capturedonly partially through the dummy variable, and through changes in bio-fuel production (and hence in land use) due to changes in fuel prices. Inreality, however, fuel price variationmay affect biofuels differently underdifferent policy regimes, e.g. tax credit versus mandate (Gorter and Just,2009). Given that fuel price determines the profitability of biofuels joint-ly with biofuel polices (e.g. tax credit), the fuel price impact on biofuelsmay be strongerwith thanwithout biofuel support policies. For example,a fuel price increase may not be sufficient to make biofuel productionprofitable (and hence cause land use changes) in the absence of biofuelsupport policies. In such a case, the impact of biofuels on land use willalso reflect the impact of biofuel support policies. However, if a higherfuel price causes the same change in biofuel production both with andwithout policies, then the policy effect will not be identified in ournear-VAR model. In this case, we capture only market induced effect ofbiofuels. Further, some policesmay isolate biofuels from the fuel market.For example, with a binding mandate, the biofuel prices tend to bedecoupled from the fossil fuel market, implying that biofuels (andhence land use through the direct biofuel channel) does not respond tofuel prices changes. This effect of biofuels is not identified in our estima-tion, and hence we do not capture this policy effect, as a result of whichthe induced land use adjustments would further bias our resultsdownward.

In summary, these limitations imply that we may capture the ef-fects of biofuel polices on land use change only partially. Hence, thepresented estimates can be considered as a lower bound, whereasthe true impact on land use change may be even larger.

A further potential drawback of our analysis is that by focusingon the US we do not account for price and land use effects ofbiofuels in other regions, such as Latin America, Africa. However,due to significant biofuel production potential in these regions,the land use impact may be even stronger, because of largernon-cultivated farmland reserves. Our findings and those of Ciaianet al. (2012) suggest that this question is a promising area for futureresearch.

6. Conclusions

The present paper studies the interdependencies between the ener-gy prices, bioenergy production, and land use changes. Our data coverthe US agricultural and energy markets for the period 1950–2007. Weapply near-VAR time-series analytical mechanisms to five major tradedagricultural commodities (barley, corn, rice, soybeans and wheat), andthe total agricultural land, alongwithweighted crude oil price and struc-tural variables. An important advantage of the employed near-VAR ap-proach is that it allows to set up a structural model and to control for anumber of macroeconomic variables, such as such as GDP, yield, con-sumption, and production. Estimating several near-VAR models, we in-directly examine the response of land use to changes in oil prices.

The underlying theoretical model suggests two types of impacts offuel and biofuel price increase on land use change: a direct land usechange impact and an indirect land use change impact. The direct impacton land use change refers to a situation, when the land already used foragriculture is switched to produce biofuel crops, causing substitution inland use between different types of crops. The indirect impact on landuse change reflects a case, when the total land cultivation expands, im-plying that new land, which before was not used for agricultural produc-tion (such as idle or forest land), is converted into farmland.

Our empirical findings confirm the theoretical hypothesis that en-ergy prices do affect land use in the US. The impulse response analysisresults suggest that land of all agricultural commodities is affected byenergy prices, including those that are not directly used for bioenergyproduction. However, the impact on land use change is higher for thefood/bioenergy commodities (e.g. corn) than for food only commod-ities (e.g. rice), suggesting the importance of biofuels in transmittingthe fuel price signals to land use changes.

Based on the innovation accounting results, we calculate medium-and long-run land use change elasticities. The magnitude of thelong-run price transmission elasticities varies between −32,000 and18,000 ha for individual crops and between 54,000 and 68,000 hafor the total land per 1 dollar/barrel increases in fuel price, dependingon the time horizon considered. When controlling for structural breakin biofuel production and public support policies, as expected, long-run price transmission elasticities are lower.

Our findings are highly important for policy makers, as they help tobetter understand the role of fuel prices, and suggest a potential role ofbiofuels (and biofuel policies) in determining the induced land usechanges. According to our results, biofuels may reinforce both the totalagricultural land use change, as well as land substitution among differ-ent crops. These results suggest that policies, which stimulate biofuelproduction (which is the case of many developed countries), may in-deed have undesirable environmental consequences and/or adverse cli-mate change impact leading to more carbon emissions due to land usechanges induced by the expansion of croplands for biofuel production.These findings contradict some recent statements made by the EU andUS policy executives, who attempt to play down the role of bioenergypolicy spillovers to the agricultural input and output markets.

The indirect land-use change effects are also very important to es-timate the impacts of biofuels on Greenhouse Gas (GHG) emissions.For example, Plevin et al. (2010); Sterner and Fritsche (2011) showthat most biofuels emit more GHG than fossil fuels when the indirectland-use change exceeds a certain threshold.


Appendix. Theoretical Model.

The present model builds on Gardner (2007); Gorter and Just(2009); and Ciaian and Kancs (2011). We introduce two important ex-tensions. First, we assume two agricultural commodities: one suitablefor biofuel production (referred to as “biomass”)11 and one not suitablefor biofuel production (referred to as “food”). Second, we consider thetransmission of prices also through the input channel. Both extensionsimprove the model predictions

The economy consists of vertically integrated agricultural, biofuel,fossil fuel, by-product, and input markets (Fig. 7). The representativefarm can substitute between producing two agricultural commodities(biomass and food) using constant return to scale production func-tions of two substitutable inputs: fuel and other inputs (referred toas “land”). Agricultural output can be supplied to food market andto biofuel market. We assume fixed Leontief coefficient transforma-tion function of biomass into biofuel and by-product from biofuel pro-duction. Further, we assume a downward sloping demand for foodand by-product, and an upward sloping supply curve for land. The ag-gregate fuel supply is a sum of biofuel and fossil fuel supplies. The ag-gregate fuel demand is a sum of agricultural and non-agricultural fueldemand. The fossil fuel supply (non-agricultural fuel demand) isgiven by upward (downward) sloping functions of prices.

The representative agricultural farmmaximises its profits accordingto:Π ¼ ∑

ipiQ i Ni; ;Ki

� �−wNi−rKi, which implies the following equi-

librium conditions:

pi∂Qi

∂Ni¼ w for i ¼ AB;AN ð7Þ

piQi

∂Ki¼ r for i ¼ AB;AN ð8Þ

where N is non-fuel input (land), K is fuel input, p is farm output price,w is land rental price, and r is fuel price. AB is index for biomass, and ANis index for food commodity. Eqs. (7) and (8) are marginal conditionsfor land and fuel input, respectively, and determine input demand andoutput supply of agricultural commodities.

In Fig. 8 the derived supplies of biomass, SAB, and food commodity,SAN, are shown as upward sloping curves (panels a and d, respective-ly). The aggregate world food demand for biomass and food commod-ity is denoted by DAB(pAB) and DAN(pAN), and are shown in panels aand d, respectively. The world supply of land is given by SN(w) (notshown).

We assume constant Leontief transformation technology in biofuelsectorwith constant extraction coefficient denoted by β. Each unit of bio-mass results in β units of biofuel. Additionally, biofuel production yieldsfeed by-product, γ, units quantity per one unit of biomass. To simplifythe analyses, we assume constant value of unit processing costs (adjustedfor mark-up), c, incurred to biofuel production from one unit of biomass.This implies that biofuel supply, SB(r), and by-product supply, SO(pO),represent the excess supply of biomass, SAB−DAB, adjusted by constantextraction coefficients: SB=β(SAB−DAB) and SO=γ(SAB−DAB), respec-tively, where # is the price for by-product. In Fig. 8 biofuel supply, SB, isshown in panel b.

The world supply of fossil fuel together with biofuel supply gener-ates the aggregate fuel supply curve, STF(r)=SF+SB, where SF(r) isthe world supply curve of fossil fuel (panel b in Fig. 8). The aggregatefuel demand, DTF(r), is a sum of agricultural fuel demand, KAB+KNB,and non-agricultural fuel demand, DNF(r) (panel b in Fig. 8). Inorder to simplify the analysis, we assume perfect substitutability be-tween biofuel and fossil fuel in consumption.

11 Note that we consider the case where the agricultural commodity suitable for bio-fuel production may be used for both food and biofuel production. We denote it as bio-mass to simplify the text.

The equilibrium conditions for agricultural, input and fuel marketscan be summarised in five equilibrium conditions. Food productionfrom biomass with no biofuel production is given by:

if pABo ≥βr þ γpOo−c ⇒ SB ¼ SO ¼ 0 ⇒ DAB ¼ SAB ð9aÞ

where poAB is the equilibrium price for biomass in absence of biofuel

production, poO is by-product price in absence of production of

by-product from biomass. Food and biofuel production from biomassis given by:

if pABo bβr þ γpO−c ⇒ SB > 0; SO > 0 ⇒ SAB−DAB> 0

and pAB ¼ βr þ γpO−c:

ð9bÞ

Food commodity equilibrium:

SAN ¼ DAN: ð10Þ

Land market equilibrium:

NAB þ NAN ¼ SN : ð11Þ

By-product market equilibrium:

SO ¼ DO ð12Þ

where DO(PO) is by-product demand. Fuel market equilibrium:

STF ¼ DTF: ð13Þ

Eq. (3) determines the equilibrium condition for biomass. The unitreturn of biomass, if used to produce biofuels, is given by the adjustedfuel and by-product prices net of processing costs c: βr+γpO−c. Ifthe return from biofuel is smaller than the biomass equilibrium pricein the absence of biofuel production, poAB, then biofuel production isnot profitable in equilibrium. In this case, the equilibrium biomassprice is determined by intersection of biomass demand and supply,DAB=SAB (Eq. (3a)). Biofuel production is in equilibrium, SB>0 , whenthe unit return of biomass used for production of biofuels is higherthan the biomass price, poAB, on the food market: βr+γpO−c>po

AB. Inthis case, the equilibrium biomass price is determined by the price forfuel and by-product: pAB=βr+γpO−c (Eq. (3b)). Eq. (10) representsthe equilibriumcondition for food commodity. Eqs. (11) and (12) deter-mine the equilibrium on land and by-product markets, respectively.Eq. (13) is clearing condition for the aggregate fuel market equilibrium,where STF=SF+SB and DTF=DNF+KAB+KAN.

Fig. 8 shows the equilibrium prices and quantities for biomassmarket (panel a), fuel market (panel b) and food commodity market(panel d). The equilibrium price and quantity for biomass is poAB, QAB,and the equilibrium price and quantity for fuel is r, QF. There is no bio-fuel production in equilibrium, because the return from biomass forbiofuel production would be lower than the price obtained if soldon the food market. With fuel price r, the unit return of biomass forbiofuel production is given by pAB (=βr+γpoO−c), while the equilib-rium food price is po

AB. As shown in panel a, food price is lower thanthe equilibrium price for biomass, pABbpo

AB, implying no biofuel pro-duction in equilibrium. 12 Finally, the equilibrium price and quantityof food commodity is pAN, QAN.13

12 Note that the fuel price ro is equivalent to return from the biofuel production atprice of biomass poAB. Everything else equal, the biofuel production is profitable for fuelprices higher than ro.13 As noted above, we assume that this commodity is used only for food production.

Fig. 7. Structure of the model.

ro

0

p AN

DAN

(d) Food commodity market

pAN

Q AN

Q AB Q AB QF

DAB

0

p AB

p AB

S F

S2TF

0

r

SF

r

(a) Biomass Market (b) Fuel Market

poAB

SAB

SB

DTF

Fig. 8. Biomass, fuel and food commodity markets.


References

Birur, D.K., Hertel, T.W., Tyner, W.E., 2010. The global impacts of biofuel mandates.Energy Journal 75–100.

Ciaian, P., Kancs, D., 2011. Interdependencies in the energy–bioenergy–food price sys-tems: a cointegration analysis. Resource and Energy Economics 33, 326–348.

Ciaian, P., Kancs, D., Rajcaniova, M., 2012. Bioenergy and Land Use Change. EERIResearch Paper 2012/12. Economics and Econometrics Research Institute (EERI),Brussels.

Darlington, T.L., 2009. Land use effects of U.S. corn-based ethanol. Air ImprovementResource.

de Gorter, H., Just, D.R., 2009. The economics of a blend mandate for biofuels. AmericanJournal of Agricultural Economics 91 (3), 738–750.

Fischer, G., 2008. Implications for Land Use Change. Paper presented at the ExpertMeeting on Global Perspectives on Fuel and Food Security, 18–20 February 2008.FAO, Rome.

Gardebroek, C., 2010. Do Oil Prices Increase Corn Price Volatility? Paper presented atthe IATRC meeting, Berkeley, California. December 12 14, 2010.

Gardner, B., 2007. Fuel ethanol subsidies and farm price support. Journal of Agricultural& Food Industrial Organization 5 (2) (Article 4).

Imai, K., Gaiha, R., Thapa, G., 2008. Food and Oil Prices and their Implications for RuralPoverty. Working Paper.

Kancs, D., 2002. Integrated appraisal of renewable energy strategies: a CGE analysis.International Journal of Energy Technology and Policy 1, 59–90.

Kancs, D., 2006. Evaluation of renewable energy policies. Energy Studies Review 14,1–34.

Kancs, D., 2007. Applied general equilibrium analysis of renewable energy policies.International Journal of Sustainable Energy 27, 1–20.

Kancs, D., Wohlgemuth, N., 2008. Evaluation of renewable energy policies in an inte-grated economic–energy–environment model. Forest Policy and Economics 10,128–139.

Kielyte, J., 2008. Estimating panel data models in the presence of endogeneity andselection. Journal of Economics and Econometrics 51, 1–19.

Lütkepohl, H., Krätzig, M., 2004. Applied Time Series Econometrics. Cambridge UniversityPress.

OECD, 2008. Biofuel Support Policies: An economic Assessment. Organisation forEconomic Co-operation and Development (OECD), Paris.

Plevin, R.J., O'Hare, M., Jones, A.D., Torn, M.S., Gibbs, H.K., 2010. The greenhouse gasemissions from market-mediated land use change are uncertain, but potentiallymuch greater than previously estimated. Environmental Science and Technology44, 8015–8021.

image of Fig.�7

image of Fig.�8


RFA, 2012. Statistics — Historic U.S. Fuel Ethanol Production. Renewable Fuels Associa-tion. http://www.ethanolrfa.org/pages/statistics.

Rosegrant, M.W., 2008. Biofuels and Grain Prices: Impacts and Policy Responses. Testi-mony for the U.S. Senate Committee on Homeland Security and GovernmentalAffairs, May 7 2008. IFPRI.

Searchinger, T., Heimlich, R., Houghton, R.A., Dong, F., Elobeid, A., Fabiosa, J., Tokgoz,S., Hayes, D., Yu, T.H., 2008. Use of U.S. croplands for biofuels increasesgreenhouse gases through emissions from land use change. Science 319,1238–1240.

Sterner, M., Fritsche, U., 2011. Greenhouse gas balances and mitigation costs of 70 mod-ern Germany-focused and 4 traditional biomass pathways including land-use changeeffects. Biomass and Bioenergy 35, 4797–4814.

Swinton, S.M., Babcock, B.A., James, L.K., Bandaru, V., 2011. Higher US crop pricestrigger little area expansion so marginal land for biofuel crops is limited. EnergyPolicy 39, 5254–5258.

Tyner, W.E., Taheripour, F., Zhuang, Q., Birur, D., Baldos, U., 2010. Land Use Changes andConsequent CO2 Emissions due to US Corn Ethanol Production: A ComprehensiveAnalysis. Final Report Department of Agricultural Economics, Purdue University.

http://www.ethanolrfa.org/pages/statistics

land use change impacts of biofuels: near-var evidence from the us

Documents