The future of agriculture. Prospective scenarios and modelling approaches for policy analysis

Land Use Policy 31 (2013) 102–113

Contents lists available at SciVerse ScienceDirect

Land Use Policy journal homepage: www.elsevier.com/locate/landusepol

The future of agriculture. Prospective scenarios and modelling approaches for policy analysis Sergio Gomez y Paloma a , Pavel Ciaian a,∗ , Adriana Cristoiu b , Frank Sammeth a a b

European Commission, Joint Research Centre, Institute for Prospective Technological Studies, C/ Inca Garcilaso 3, 41092 Seville, Spain European Research Council Executive Agency, Place Charles Rogier 16, BE-1210 Brussels, Belgium

a r t i c l e

i n f o

Article history: Received 25 April 2011 Received in revised form 12 December 2011 Accepted 19 December 2011 Keywords: Modelling Agricultural policy Scenarios Drivers CAP

a b s t r a c t The objective of this paper is (i) to compare and discuss literature related to global and European outlooks in relation to the farming sector and rural areas and (ii) to provide an overview of policy modelling methodologies, especially but not only those used in assessing the impact of the Common Agricultural Policy (CAP). There is significant variation in terms of both the policies and external drivers that are taken into account in global and European outlooks, driven predominantly by the heterogeneity in focus of studies, the approach applied and/or external pressures. An increasing number of studies take on board the new CAP challenges. However, an area where improvements are needed is in the understanding of the sensitivity of policy effects to assumptions on external drivers. Two key modelling approaches applied for policy impact analysis include structural models and econometric models, with the former dominating the latter mainly due to its better adaptability to the needs of policy makers. However, with the CAP evolving towards ever more complex instruments, the relevance and predictive accuracy of structural models will possibly improve only as long as methodological and data issues are addressed. © 2012 Elsevier Ltd. All rights reserved.

Introduction and objectives The Common Agricultural Policy (CAP) is an important player in rural areas in the EU and has a profound impact on agricultural and rural markets spanning from output and factor markets to income and environment. The changing focus of the CAP generated different pressures on the rural economy over time. The emphasis of the early CAP was on encouraging agricultural productivity, ensuring a stable supply of affordable food to consumers and ensuring a viable agricultural sector. The support to farmers was implemented predominantly through market price support. Market and environmental pressures of the early CAP led to a substantial overhaul of the policy instruments in the 1990s. The 1992 MacSharry reform and the Agenda 2000 reduced the importance of price support and introduced coupled direct payments. At the same time, the rural development issue was strengthened and environmental concerns were integrated into the CAP. The 2003 CAP reform decoupled most direct payments and made environmental aspects of agriculture a compulsory requirement (i.e. cross-compliance requirements) for receiving support. The recent

∗ Corresponding author. Tel.: +34 95 448 8429; fax: +34 95 448 8274. E-mail addresses: [email protected] (S.G.y. Paloma), [email protected], [email protected] (P. Ciaian), [email protected] (A. Cristoiu), [email protected] (F. Sammeth). 0264-8377/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.landusepol.2011.12.005

European Commission proposals on the future CAP (for the CAP post-2013) aim to further strengthen and enhance these environmental objectives by improving the targeting of direct payments (European Commission, 2010, 2011). In recent years important structural developments have taken place in the global markets. Of particular importance for the agricultural sector are the energy price rise and the expansion in bioenergy production, greater commodity price volatility, the shift in consumption patterns in developing countries and climatic changes. All these factors have important implications for the agricultural sector because they may lead to structural changes in production and farming systems thus creating pressures on rural markets, the environment and sustainability of food production. These global developments are also important from a policy perspective because they likely affect policy needs besides policy effects and thus may induce future adjustments of the CAP. The purpose of this paper is twofold: firstly it reviews some of the major recent studies addressing the medium to long-run projections of agriculture and rural areas at world and EU levels with the aim of identifying prospective scenarios and main external drivers, and secondly it assesses the main methodologies used to analyse the ex ante and ex post impacts of agricultural and rural policies on the farming sector and rural areas. Prospective global and EU level scenarios and national level survey results concerning the future development of the CAP and external drivers are presented and discussed. They are derived from the non-exhaustive literature review. External drivers identified to

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

be the most important factors affecting the future evolution of agriculture and rural areas at world and EU levels are also presented in greater detail. They include demography, climate change, economic growth, energy prices and agricultural world markets. Concerning methodologies, an attempt is made to identify the key challenges of modelling agricultural policies. The focus is mainly on the CAP but for illustration purposes a global overview is, in some cases, also provided. The aim here is not to provide an exhaustive list of models applied for policy impact analyses. The main objective is to identify the main practical and theoretical challenges to model policies in the framework of the CAP. The paper is structured as follows. Drivers, trends and prospective scenarios are presented in the next section, including “Drivers and prospective scenarios” section and “Trends and projections”. Section “Modelling Approaches for policy analysis” outlines the key challenges of modelling agricultural policies. It includes a first subsection on the “Classification of policy modelling approaches”, and a second one on “Modelling first versus second pillar CAP policies”. The last section concludes.

Drivers, trends and prospective scenarios Drivers and prospective scenarios Focusing on the most recent studies addressing the evolution of world and EU agriculture, this section summarises the key underlying assumptions on prospective scenarios, drivers and projections in the medium- to long-term. The results of a recent survey among EU Member States on the opinion at national level concerning the future of the CAP are also included. The studies consulted belong to two broad groups. The first group includes projections on the most likely evolution of the agricultural sector (either at global or regional level, e.g. OECD Members) subject to a set of drivers identified by the studies as being essential for the development of agriculture (FAO, 2006; OECD-FAO, 2011; USDA, 2009). These studies do not include scenarios on alternative future developments. They provide projections for macro-indicators and development of agricultural markets. In FAO (2006), projections for world agriculture for the 2030/2050 time horizon are conducted with four key drivers considered to be the most instrumental: demography, energy markets, food consumption and world commodity markets. The OECD-FAO (2011) and the USDA (2009) studies provide projections for shorter time horizons; for 2020 and 2018, respectively. However, both studies include similar drivers and assumptions, together with a more explicit representation of agricultural policies and agricultural market developments (Table 1). The projections of these studies (particularly macro-indicators) form the basis of the underlying macro assumptions on which most other studies build their futureorientated scenario exercises usually conducted at country level or for a group of countries (e.g. the EU). The second group of studies focuses on the European agricultural sector and its key drivers. Additionally from proposing a baseline scenario considered to be the most likely development, these studies also propose alternative scenarios about the evolution of the sector and the CAP, including some on rural economies (e.g. SCENAR2020, EURURALIS, SENSOR, MEASCOPE, etc.). The alternative scenarios are justified by the focus these studies take (e.g. EU agricultural sector and rural areas in SCENAR2020, land use in SENSOR, etc.). The additional scenarios are used to explore the potential implications for the evolution of the topic of interest under an alternative assumption of drivers and/or policies (Table 1). SCENAR2020 (SCENAR2020 – Scenario study on agriculture and the rural world) takes an in-depth approach to identify future trends and drivers of change in European agriculture and rural areas

103

in the 2020 time horizon (SCENAR2020, 2006). Drivers are grouped into two broad categories based on whether they are influenced by policy intervention: (i) exogenous to the EU policymaking or (ii) endogenous to the EU-policymaking system. The exogenous drivers include demography, macro-economic growth, agri-technology and world agricultural markets, whereas the endogenous drivers include trade and agricultural policies, environmental policies, enlargement and international agreements. SCENAR2020 (2006) simulates future effects of three scenarios. The baseline (reference) scenario assumes the observed developments in exogenous macro drivers to continue into the future and unchanged policy environment. The other two scenarios consider an alteration of EU policies: (i) the regionalisation scenario assumes sustained policy preference to promote regional economic strength and social welfare, as well as considering the maximum degree of agricultural support allowed under the WTO framework; and (ii) the liberalisation scenario assumes the reduction of policy intervention to a socially acceptable minimum. Five macro-drivers are considered in the SENSOR1 project (Kuhlman et al., 2006; Bakker and Verburg, 2009): oil prices, world GDP excluding the EU-25, population, labour force and R&D expenditure (all sectors). Quantitative assumptions are made about their evolution under three scenarios (Baseline, High-growth and Lowgrowth scenarios) using information from various international statistical sources (e.g. under Baseline population and R&D expenditure projections are from Eurostat, World GDP projections are based on PROMETHEUS, a stochastic model of the world energy system, etc.). The primary objective of the SENSOR project is to asses the impact of CAP policies on land use in the EU. For this purpose the bio-diversity scenario (extension of area protected under Natura 2000) and CAP reform scenario (gradual abolition of CAP) are simulated. EURURALIS (2008) includes a wider set of drivers to develop scenarios about EU rural areas at the 2030 time horizon. EURURALIS is an integrated impact assessment tool for exploring the future development of rural areas in the EU within a global context. Four contrasting world visions (scenarios) are developed along two dimensions: the first dimension representing a range from a global to a more regional integration, and the second one representing a range from market orientation to a higher level of governmental intervention to ensure specific social, economic and environmental objectives. The macro-drivers’ set includes demographic developments, consumer preferences, macro-economic growth, agro-technology, border support, income support, Less Favourable Areas (LFA), nature, spatial, erosion and energy policies. Further assumptions are made in EURURALIS about the CAP settings under each of the four scenarios. For example, the Global Economy scenario assumes full liberalisation and phasing out of income support (with CAP market support still maintained in 2010 but its total abolition after 2020, all CAP income support abolished after 2010, no support to biofuels combined with 0% blending obligations, no taxes or subsidies and from 2010 no special LFA policy and no designated Natura2000 areas). On the other hand, the Continental Markets scenario assumes no change in border support and domestic support. The other two scenarios are intermediary cases where the Global Co-operation scenario considers more trade integration and removal of trade protection, whereas the Regional Communities scenario assumes more environmental support. The same four-quadrant-model approach to building scenarios used in EURURALIS is also used in MEASCOPE (Micro-economic instruments for impact assessment of multifunctional agriculture to implement the Model of European Agriculture) (MEASCOPE,

1 SENSOR stands for “Sustainable Impact Assessment: Tools for Environmental, Social and Economic Effects of Multifunctional Land Use in European Regions.”

104

Table 1 Summary of main drivers and study characteristics. Study

Area of concern

Main drivers

Scenarios

Region

Time horizon

FAO (2006)

Prospective developments in food, nutrition, agriculture and major commodity groups Medium-term development of global commodity markets

Demography; energy markets; food consumption; world commodity markets Economic growth; population; exchange rate; oil prices; agricultural policies Economic growth; population; exchange rate; oil prices; U.S. agricultural policy - Exogenous drivers to the EU policy-making system: demography; macro-economic growth; world agricultural markets; consumer preferences; quality of life and social well-being; human and animal health concerns; agri-technology; environmental trends; - Endogenous drivers to the EU policy-making system: trade and agricultural policies; environmental policies; enlargement; WTO and other international agreements Oil price; world GDP excluding EU-25; population; labour force; R&D expenditure; CAP policies

Baseline

World

2030 and 2050

Baseline

OECD and Non-OECD countries U.S. and World

2020

- Baseline - Regionalisation scenario: support of regional development; maximum agricultural support allowed under the WTO framework. - Liberalisation scenario: policy intervention in the economy reduced to minimum.

EU

2020

- Baseline - High and low growth scenarios Bio-diversity scenario: extension of current Natura2000 areas by 20%. - CAP reform scenario: gradual abolition of CAP. - Four baseline scenarios: - Global economy: global integration and low governmental intervention. - Continental markets: regional integration and low governmental intervention. - Global co-operation: global integration and high governmental intervention. - Regional communities: regional integration and high governmental intervention. - Baseline “Competitiveness” scenario: focus on economic functions; - “Rural Viability” scenario: focus on social functions - “Environment” scenario: focus on environmental functions - Baseline - Climate Shock: climate change and the acceleration of related environmental impacts as the driving disruption factor - Energy Crisis: energy supply vulnerability of Europe - Food crisis: food connected to health and society as a source of disruption - Cooperation with nature: science and technology effectively deployed to ensure sustainable development

EU

2025

EU and World

2030

EU

Agenda 2000 + 9 years

EU and World

20 year horizon

OECD-FAO (2011)

Long-run projections for the agricultural sector

SCENAR2020 (2006)

Identification of future trends and driving forces for the European agricultural and rural economy

SENSOR

Policy impacts on land use

EURURALIS (2008)

Integrated impact assessment tool for exploring the future development of rural areas in EU within a global context

GDP growth; demographic developments; consumer preferences; agro-technology; border support; income support; Less Favourable Areas (LFA); nature, spatial, erosion and energy policies.

MEASCOPE (2007)

The impact of the CAP reform on multifunctionality of agriculture and rural development

SCAR (2007)

Identification of long-term research priorities to support a European Knowledge-based Biosociety

CAP drivers: international trade liberalisation talks; costs of the CAP; society’s changing needs (e.g. food safety, environment, biofuels); rural economy/development Economy and trade; climate change; environment; energy; science and technology; societal and demographic changes; rural economy and regional development; health; policy; globalisation

Baseline

2018

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

USDA (2009)

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

2007). MEASCOPE develops alternative scenarios targeted at analysing the impact of the CAP reform on multifunctionality of agriculture and rural development. The main drivers of the future development of the CAP are identified through the participatory approach of regional stakeholders, EU officials, scientific experts and computer modellers. The baseline scenario assumes the continuation of the CAP Agenda 2000 to the selected time horizon (i.e. Agenda 2000 + nine years). Each of the other three scenarios focuses on different dimension of sustainability of agriculture and rural development: “Competitiveness” scenario (resulting from production intensification and general policy trend combination) assumes a stronger focus on the economic functions, “Rural Viability” scenario (productivity intensification and territorial/individual trends) focuses on the social functions and “Environment” scenario (production extensification and territorial/individual focus) centres on the environmental functions of agriculture and rural development. A foresight exercise regarding food, rural and agricultural future for the 20–30-year time horizon is investigated by the Standing Committee on Agricultural Research (SCAR, 2007, 2011).2 The SCAR advises the European Commission on the coordination of research in agriculture by identifying research and innovation needs in the medium to long term. The set of drivers the SCAR considers includes economy and trade, climate change, environment, energy, science and technology, societal and demographic changes, rural economy and regional development, health, policy and globalisation. Scenarios are built reflecting different pathways in the development of key drivers. For example, in the SCAR (2007) four scenarios were investigated. A baseline scenario is developed which assumes a continuation of existing trends in drivers. The other four scenarios (Climate Shock, Energy Crisis, Food crisis, Cooperation with nature) were built by considering a major change or discontinuity in one or more of the drivers, which in some cases may result in crisis (Climate Shock, Energy Crisis, Food crisis) and in another in utopian development (Cooperation with nature). Regarding the policy future prospects at national level, a survey conducted among representatives from EU Member States reveals the diversity of views on the future of the CAP and illustrates the difficulty of drawing policy scenarios that would achieve a broad common agreement (Council for the Rural Area, 2009). The respondents from Denmark, Finland, Sweden, the UK and Latvia indicate the need for further liberalisation and more market orientation of the CAP (including elimination of price support, export refunds, production quotas etc.), together with full decoupling of the Single Farm Payment. However, the respondents from Austria, France, Portugal and Spain seem to be more resistant to the idea of further liberalisation. The role of the future CAP is also seen differently by the respondents. Those from Austria, Spain and Portugal see the CAP as an instrument supporting the functions and structure of European agriculture, while a more broad-based territorial approach of the CAP is advocated in countries such as Italy, the Netherlands and Sweden. Moreover, the structure and organisation of the first and second pillars are seen differently. The debate in the UK, the Netherlands, Finland and Italy seems to argue for shrinking the role of the first pillar and increasing the size of the second; results from Austria seem to favour the maintenance of the current two pillar structure. Integration of the second pillar into regional development policy is argued for in Finland, Italy and Latvia. Furthermore, the debate in the Netherlands and the UK emphasises the need to give more room for manoeuvre to individual Member States in determining their agri-environmental priorities and other rural development needs. Finally, a re-nationalisation of the CAP

2 The SCAR was established by Regulation (EEC) No. 1728/74 of the Council of 27 June 1974 on the coordination of agricultural research.

105

financing is seen as undesirable by the respondents from Austria, Denmark, Portugal and Spain, while the debate in the UK, the Netherlands and Sweden points to the need to assess whether agricultural support should be financed at EU or national level. In general, the diversity of interest among Member States indicates that the definition of the policy scenarios is a challenging task as the future policy developments are subject to uncertainties related to political decision process. Studies analysing future development of agriculture need to consider a set of alternative policy scenarios in order to encompass most probable trajectories of policy changes. As the recent policy developments show, the role of environmental concerns within the CAP is increasing, which is also reflected, as discussed above, in the scenario analysis of the possible (near) future development of farming and rural areas. The importance of integrating environmental aspects has been reiterated in the current proposal of the European Commission on the CAP post 2013. From the policy making side, climate change is seen as a cause exacerbating market instability while active management of natural resources by farming is considered a contributor to the mitigation of climate change. Yet, both environment and climate change are further identified as future challenges facing European agriculture and the provision of environmental public goods is highlighted as an objective of the future CAP along with green growth through innovation and climate change mitigation. Environmentally friendly farming and rural areas are seen to stabilise local employment and to reduce the negative effects of economic crises on rural areas, producing benefits in competitiveness and innovation (European Commission, 2010). The identification of prospective scenarios which correctly reflect both policy priorities and external drivers is a first step for conducting future projections and policy impact analyses in agriculture and rural areas. Ultimately, the actual identification and quantification of effects of and interactions between policies, drivers, environment, climate change and sustainable development requires a robust methodological approach and is subject to future research in this field (see further). Trends and projections This section presents trends and projections of drivers indentified as the most important factors affecting the future of agriculture at world and EU levels. They are derived from the above nonexhaustive literature review and include demography, climate change, economic growth, energy prices and agricultural world markets. Demography Population growth is one of the key drivers identified by most of the studies to affect the future development and structure of the agricultural sector. According to UN population data and projections, world population growth will continue and will reach 7.7 billion inhabitants in 2020. Most of this growth is expected to occur in developing countries, especially China and India whose populations will represent 19% and 18% of 2020 total world population, respectively (UN, 2008a, 2008b). For the EU-27, the rising population trend observed over the 1999–2008 period is expected to continue. Eurostat projects that the EU population will peak in 2035 at about 521 million persons, thereafter falling to about 506 million inhabitants by 2060 (Eurostat, 2009). Migration is expected to contribute to structural changes in EU demography. The migration pattern in the EU changed after 1999 (SCENAR2020 II, 2009). Increased migration to intermediate rural areas suggests that suburbanisation and counter-urbanisation remain very strong drivers for population development in EU rural regions. Still, consistently low birth rates and increasing life expectancy will mark the

106

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

transition towards a much older EU-27 population and the old-age dependency ratio will rise to 31% in 2020 and to 53.5% by 2060.3 The rising world population is expected to create strong upward pressure on food demand, having effects on both price and the way natural resources are utilised (e.g. intensive production). Although the largest population growth is expected to occur outside the EU, European agriculture is expected to be affected indirectly through world price rises causing adjustment in production technology and the intensity of exploitation of natural resources. Climate change Climate change is a permanent driver of agricultural sectors and its importance has been growing in recent decades, not only in terms of its direct impact on the sector but also through its influence on policies. For agricultural production, one of the main concerns relates to the effects of climate change on agricultural yields. Changes in temperature and rainfall patterns are important but difficult to accurately predict. According to the PESETA (2011) study which simulates crop yield changes by the 2080s relative to 1961–1990, under a high CO2 emission scenario (IPCC A2) and a climate model, a 10% yield decrease may occur in the South and West of Europe.4 In Central and North-Eastern Europe, crop reduction could be lower (0–5%), while a 15% yield improvement is projected for Nordic countries. For the 2020 time horizon, the annual average surface air temperature change is projected to be under 1 ◦ C. Warmer and wetter climate conditions in Northern Europe would see the area under certain crops (such as maize) expanded while the increased average annual temperatures and recurrent droughts would result in lower yields for others (such as wheat) (SCENAR2020, 2006). The climate change impacts could be mitigated, but probably not totally offset, by adaptation of production practices (e.g. fertilizer and pesticide use, new varieties, crop rotation), alteration in crop and animal production mix and/or adjustment in farming systems. Economic growth Economic growth has a profound effect on agriculture by affecting both the demand side (e.g. consumer preferences) and the supply side (e.g. technology, input use) of the agricultural sector. The GDP has experienced a volatile period over the last few years mainly due to the economic crisis. However, when looking at development over a longer time horizon world GDP grew by 3.6% per year over the 2001–2010 period. In advanced economies GDP grew at a 1.6% rate over the same period while rates were higher for emerging and developing economies (6.3%). The EU-27 GDP growth rate over the same period was about 1.5% per year, lower than the growth in other high income countries. The overall trend in the last three decades is an acceleration of GDP growth in the emerging and developing economies and deceleration in the advanced economies (IMF, 2011). Continued economic growth is expected over the next decade in almost all regions of the world, especially in transitional and developing countries (Brazil, China, India and new EU Member States). By 2020, the OECD and World Bank expect world economic growth to increase from 2.4% annually to 3.1% annually. World and EU economic growth prospects depend on the amount of investments in education and research, technological opportunities and the liberalisation of world commodity and factor markets (SCENAR2020, 2006). GDP growth will also trigger higher income and food consumption. By 2020, the daily

3 Old-age dependency ratio is the projected number of persons aged 65 and over expressed as a percentage of the projected number of persons aged between 15 and 64 (Eurostat). 4 See Ciscar et al. (2009). The impact of climate change on agricultural production in Africa is the focus of Barrios et al. (2008).

per capita calorie consumption is expected to reach 3600 kcalories/capita/day in China; all the other regions are also expected to follow the same trend, although with only marginal improvement of calorie consumption in the developed world (Rosegrant et al., 2001). Energy Recently an important structural change took place in the energy sector. The rapid growth in demand for fossil fuels in several developing countries, as well as increased concerns over the security of supplies in the Middle East, increased the pressure on energy prices. The rise in oil prices since 2002 transmitted its effects to the whole EU economy, resulting in higher prices for energy and consumers, including electricity (since electricity production depends to a large extent on oil and gas as main fuels) (Eurostat, 2008a). The uncertainty of prospects for future world oil resources makes it difficult to project the associated prices. EIA projects world oil prices to rise to approximately USD 130 per barrel in 2035 (in 2009 dollars) as the most likely scenario and a ranging path from USD 50 to USD 210 per barrel (EIA, 2010). The evolution of energy prices is crucial for the viability and development of the agricultural sector. Energy prices directly affect the agricultural sector through altering production costs and thus determining its profitability. In 2002, of the total on-farm energy consumption in the EU-15, petroleum products represented 61% of the total consumption, followed by gas (19%), electricity (14%) and the remaining 4% included renewable and other sources (derived heat and solid fuel) (Eurostat, 2008b). Another channel through which the energy sector influences agricultural production and prices is through the bioenergy demand for agricultural commodities. Bioenergy, particularly energy crops, has a direct effect on the agricultural sector, because it uses biomass as input which, together with agricultural food commodities, is produced on a limited area of agricultural land. Competition between energy and food crops for this limited area of agricultural land tends to result in increasing food prices (FAO, 2008b). Policies targeted at increasing the share of renewable energies also impact on agricultural commodity prices. For example, the implementation of the 10% biofuel directive in the EU is expected to induce important future changes in world agricultural commodity prices, thus having consequences on agricultural production, farm practices and environment. However, overall price effects of bioenergy depend on the availability of fallow land5 and induced land productivity increases. Agricultural commodity markets and trade Over the last few years an important structural change has taken place in the agricultural commodity markets. Several factors such as expansion of biofuels, the shift towards animal-based protein sources in developing countries, and climatic changes led to upward adjustment and higher volatility of agricultural commodity prices. Over the medium-term structural changes are expected to maintain agricultural commodity prices at a higher level than observed before this period. These price adjustments may be reduced by two factors. Firstly, new technological developments may improve yields and lead to an offsetting effect in the supply of agricultural commodities. Secondly, with rising agricultural profitability, unused fallow land may be brought into cultivation. However, because technological improvement is costly and the

5 FAO (2008a) reports a substantial amount of additional land – up to 2 billion ha – potentially suitable for crop production. Fischer (2008) estimates that between 250 and 800 million hectares are potentially available for expanded crop production after excluding forest land, protected areas and land needed to meet increased demand for food crops and livestock.

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

fallow land brought into cultivation is usually less productive, these factors may not fully offset the upward adjustment of prices. A key development that may affect world and EU trade patterns is the WTO trade liberalisation. As for the agricultural trade, the WTO negotiations are still ongoing. The main policy areas under negotiation and subject to liberalisation are market access, export subsidies and domestic support. The improvement of market access occurs via the reduction of import tariffs and non-tariff barriers to trade but the speed and period of their reduction is seen as partially dependent on the political agreement. This liberalisation of EU trade combined with abolition of export subsidies may affect the competitiveness and structure of the EU agricultural sector. Increased competition may, among other things, lead to stronger regional specialisation and abandonment of agricultural production in non-competitive regions. However, due to the continuous rising world demand for agricultural commodities (e.g. due to world GDP growth, higher population, bio-energy) the trade induced effects may be partially offset, particularly if the increase in world demand for agricultural commodities will be reflected in higher prices. Concerning the domestic EU support, most of the CAP direct payments are decoupled from production. The decoupling of direct payments has reduced the pressure on commodity markets. However, several studies show that the decoupled direct payments are not completely delinked from producers’ behaviour and they exert certain impacts on agricultural markets particularly through wealth, risk and credit channels. Thus the possible future removal of direct payments may induce structural as well as supply adjustments in EU agriculture (Goodwin and Mishra, 2006; Sckokai and Moro, 2006). Modelling approaches for policy analysis This section attempts to identify the key challenges of modelling the CAP instruments. The focus is mainly on the CAP but for illustration purposes a global overview is provided where necessary. However, we do not aim to provide an exhaustive list of the models applied for CAP impact analyses. For more detailed assessments of applied agricultural models the reader is referred to studies by van Tongeren et al. (2001), Conforti (2001), de Muro and Salvatici (2001), Sckokai (2001), Britz and Heckelei (2008) and Ahumada and Villalobos (2009) among others. Our main objective here is to identify the main practical and theoretical challenges to model the common agricultural policies. Classification of policy modelling approaches An important distinction between models used for policy impact analysis is between econometric models and structural models6 (de Muro and Salvatici, 2001; Sckokai, 2001; van Tongeren et al., 2001; Keane, 2010). The econometric estimations methodology is well suited to explain the impact of already implemented policies. The most prominent areas where this approach has been extensively used are the income distributional effect of subsidies (e.g. Goodwin and Ortalo-Magné, 1992; Lence and Mishra, 2003; Roberts et al., 2003; Patton et al., 2008; Kirwan, 2009), production quota analyses (e.g. Helming et al., 1993; Guyomard and Mahé, 1994), trade policies (e.g. Haniotis, 1990; Duffy et al., 1990; Wilson, 1994;

6 Here we make the distinction between the two types of models for expositional purposes to facilitate the presentation of the applied agricultural models. Note that the use of features from both model types in one modelling framework results in a structural econometric model (see further). Most of the applied agricultural models contain elements of both types of models as either part of the full model is econometrically estimated and/or structural model elasticities are in many cases extracted from econometric studies.

107

Sparks and Ward, 1992; Williams and Shumway, 2000) and production and market effects of decoupled payments (e.g. Sckokai and Moro, 2006; Goodwin and Mishra, 2006; Bougherara and Latruffe, 2010; Weber and Key, 2011). The advantage of this approach is that it is less driven by assumptions regarding model parameters and behavioural effects. Rather the effects are calculated based on the observed behaviour of market agents. Additionally, the estimated econometric model can be statistically tested and validated, which permits the assessment of the quality of econometric models in all phases of the modelling work: specification of the model, estimation, hypotheses testing and simulation. However, a key requirement for obtaining reliable results is the availability of data in sufficient detail and quality. This is not often satisfied in reality as either data are not available to study the impact of a policy of interest or the quality of data is not appropriate to identify the policy impacts. An additional important shortcoming of the econometric approach is that it is not always able to capture the structure of the analysed sector sufficiently well. Commonly applied econometrically estimated models adopt a simplified representation of the analysed markets (e.g. due to data constraints) or have a reduced form structure which makes it difficult to identify potential nonlinearities, sectoral inter-linkages and various channels of policy transmission to agricultural markets. These disadvantages of econometric models make them less useful for policy impact analyses and in particular for new policies. Structural models can address several shortcomings of the econometric approach. However, the econometric approach may play a supplementary role. Among others, it may be instrumental for the estimation of behavioural parameters used in structural models hence resulting in a combined structural econometric model. Next we focus on structural applied models as they are extensively used for agricultural policy analyses.7 A key distinction of applied structural models for agricultural policy analysis is between partial equilibrium (PE) and general equilibrium (GE) models.8 Each has its advantages and disadvantages. Some of the examples of applied multi-regional PE models include AGLINK, AGMEMOD, Aropaj, CAPRI, ESIM, FAPRI, FSSIM and WFM (van Tongeren et al., 2001). The main advantage of PE models, compared to GE, is that they are able to represent agricultural markets and the policy instruments in greater detail. This allows the structure of the analysed markets and the intervention logic of policies to be modelled more accurately. The error that PE models are prone to is that they cannot capture policy-induced changes on the rest of the economy and their feedback on agriculture. This limits the application of these models to small sectors which represent only a small portion of the economy whereby the rest of the economy can be treated as an exogenous factor. In general, the agricultural sector fulfils this requirement, particularly for developed economies where agriculture represents a small share of the overall economy.9 On the one hand, the GE models address this PE model problem but on the other they are less flexible when capturing the specificities of agricultural markets and implementation details of agricultural policies.10 Using meta-analysis of partial and general equilibrium results on the Doha Development Round liberalisation scenarios, Hess and von Cramon-Taubadel (2006) find that, ceteris

7 Note that most of the structural models included in this section partially include elements of econometric models (e.g. by using econometrically estimated parameters). 8 For example, PE models include farm-level mathematical programming modules/models, supply demand equilibrium models, agent-based models, etc. They have a sectoral focus (e.g. agricultural sector, farming sector, crop sector) as opposed to GE models which encompass in their structure the whole economy. 9 However, in the EU context this may be problematic in certain regions. In several Member States agriculture represents an important share of the overall economy. 10 Examples of applied GE models include GTAP, GLOBE, RUNS.

108

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

paribus, PE models produce significantly larger estimates of welfare gains from trade (by US$ 2.3 billion) than GE models. In the context of the CAP, GE models may be more attractive particularly for modelling Rural Development Policies (RDP). Several RDP measures go beyond agriculture and impact other sectors in the economy (e.g. tourist industry, construction sector). The GE models can better capture these effects because of their inter-sectoral coverage. However, the implementation complexity and the regional differentiation of rural development measures can also be an obstacle to incorporating them properly into a GE modelling framework (Psaltopoulos et al., 2011). There are three key areas where applied structural models need to make progress to be able to react to current policy changes: (i) improvement of modelling CAP instruments, (ii) accounting for regional variation in the CAP implementation, (iii) accounting for farm level variation in the CAP implementation, and (iv) accounting for non-market environmental goods. Modelling CAP instruments The CAP instruments have had a tendency to increase their implementation complexity over time. The CAP tended to get more complicated in two directions: (i) in terms of combining several policy instruments in one programme and (ii) in terms of targeting. The early CAP instruments (e.g. price support) are relatively straightforward to model. The market price intervention dominated the early CAP as a means to provide income support to farmers and to ensure a stable supply of affordable food to consumers (mainly before the 1992 reform). The theoretical and empirical implementation of the price intervention policy in the PE or GE models is relatively simple because of the unambiguous identification of the incentive mechanism that it creates in the agricultural sector. The model variables targeted by the price support policy can be rather easily identified from both the theoretical and the empirical viewpoint. They involve key model variables such as domestic prices and market support instruments (e.g. intervention prices, import tariffs and/or export subsidies) which are straightforward to theoretically model, and data availability for model calibration purposes is usually not a constraining factor. The CAP reforms introduced from the early 1990s onwards have increased the complexity of policy instruments. Firstly, often CAP support policies are combined in policy programmes involving multiple instruments working at the same time, none of which can be considered isolated from the others. For example, this is the case for decoupled SPS payments (but not only for them)11 which are combined with cross-compliance requirements and modulation. The combined intervention logic of policies complicates their theoretical and empirical implementation in the model. In an ideal situation, modellers need to theoretically model the joint implementation of the combined instruments. However, this increases model complexity and requires a greater number of model parameters. With the aim of reducing the model complexity and parameter needs, the approach commonly applied by current models is to model only the key instrument of the whole scheme (e.g. SPS payments) or selected accompanied instruments (e.g. SPS payments and modulation). To the best of our knowledge, none of the available models are able to completely model the joint implementation logic of all SPS accompanying instruments (SPS payment, crosscompliance and modulation). Furthermore, the policy complexity increases the demand for more data because policy complexity often implies more policy variables. Of particular importance are

11

Other examples include the compensatory payments and set-aside introduced as part of the Agenda 2000 and many RDP measures.

the data needs regarding all combined policy instruments, as well as data concerning the related effects they may induce on the agricultural sector (e.g. compliance costs). Secondly, the CAP has refined its focus several times since its inception. Initially the main objective was to support farm income through the price intervention mechanism. In the last two decades the emphasis has shifted to supporting environmentally friendly agricultural production. This shift in the CAP logic has led to the redesign of most policy instruments to comply with the new objectives. The targeting of policy instruments has become more complex, addressing multiple objectives dispersed over multiple farm activities12 and often linked to farm activities that are difficult to quantify and measure (e.g. friendly farm management practices). From a modelling perspective, this complicates the modelling exercise because it is not straightforward to identify the link between policies and model variables. Due to this identification problem, modellers often choose a simplified way of modelling the new policies by imposing ad hoc behavioural assumptions. For example, consider agri-environmental payments. This measure supports farm practices designed to promote environmentally friendly agricultural production. Their implementation however is very diverse and can affect any farm activity altering variable inputs, fixed inputs, management practices, etc. This is in sharp contrast to the ‘old’ CAP measures such as market price support. Modellers need to identify theoretically and empirically their intervention logic which is often not possible due to measurement problems or due to lack of data and/or modelling constraints. An approach often applied to overcome data and modelling constraints is to supplement the modelling exercise with an extensive outside model analysis (e.g. through econometric/statistical analysis) in order to avoid problems related to lack of data, and theoretical and empirical complexities, as well as to identify channels through which the policy instrument in question affects agricultural markets. However, the weakness of this approach is that it implicitly introduces additional error to the modelling exercise because, in general, additional assumptions are imposed when ambiguities in data and/or in theoretical and empirical limitations are addressed, particularly if parameter values and behavioural relations estimated/calculated outside the model are based on hypotheses not consistent with the structure of the underlying model. Regional variation in the CAP implementation There is an increasing trend in the CAP development to give more freedom to MS to choose their specific implementation of certain measures. This concerns both first and second pillar policies. For example, in the case of decoupled SPS, implementation models (e.g. historical, regional, dynamic hybrid, static hybrid) as well as cross-compliance requirements vary by MS. An even stronger variation in implementation is evident for RDP. The accepted flexibility of RDP is due to many factors which can be altered at MS level, such as sectoral eligibility, type of farm activities supported, eligibility thresholds, co-financing rate, etc. Moreover the variation of CAP implementation also exists at lower resolutions than MS (e.g. regional). This policy variation across MS and regions is the main challenge in terms of model structure and data needs. The model structure needs to be extended to deal with the different implementation possibilities of policies, including a flexible theoretical structure which would reflect the regional variation of policies. Furthermore, this leads to the need to collect more disaggregated data and to extend the model to lower resolution levels (e.g. MS,

12 For example, the SPS provides income support and enforces environmental rules through cross-compliance thus, among other things, it may affect land use and farm management practices.

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

NUTS2, NUTS3). In this respect there is no uniform disaggregation of policies in the available models. Some models have relatively good regional coverage (e.g. AGMEMOD, CAPRI, van Delden et al., 2010) whereas others group MS/regions in one or several trading blocks (e.g. AGLINK) (Britz and Witzke, 2008; Gocht, 2009, 2010; Louhichi et al., 2010; OECD, 2006). Further improvements need to be conducted regarding the flexibility of theoretical structure to account for policy variation. Models often apply a standardised model structure for all resolutions (MS, regions). This standardisation is suitable for keeping models manageable from a maintenance and application point of view but render them less able to capture variation in the CAP implementation and policy impacts across regions. Farm level variation in the CAP implementation The ongoing policy shift towards support of environmental protection requires model disaggregation at the level of policy implementation to maintain their usefulness for policy-making support. The implementation of many environmental measures is strongly linked to the specificity of management practices at farm level (e.g. stocking density, cross-compliance) and their activation is often dependent on the intensity of certain farm activities and processes (e.g. heard size, crop rotation, manure handling) (Durandeau et al., 2010). The 2003 CAP reform already led to variation in decoupled SPS payments among individual farms within MS implementing the historical and hybrid models. The aggregate PE and GE models were inappropriate for modelling the impact of policy proposals attempting to alter the variation in SPS among farms (e.g. redistribution of payments between farmers and regions) (Gocht et al., 2011). The recent European Commission proposal on the CAP post-2013 goes further in this direction by proposing to introduce a mandatory “greening” component to direct payments with the aim of enhancing the environmental performance of agriculture (e.g. permanent pasture, crop rotation and ecological set-aside) (European Commission, 2010, 2011). These “greening” measures are introduced at farm level and the induced behavioural adjustments are farm-specific. The shifting policy paradigm requires the disaggregation of models beyond MS/regions to farm level downscaling. The aggregated models at regional level (e.g. MS or NUTS2) will tend to underestimate the policy effects because they average out the heterogeneity across farms. The farm heterogeneity may be relevant in activating certain policies when a given threshold is reached. This particularly concerns policy measures which impose constraints at farm level (e.g. stocking density, crop rotation) which may be less binding if modelled at aggregated level. Furthermore, the policy effects are often farm specific (e.g. environmental policies) and depend on farm decisions on production related activities and processes (e.g. specialisation, technology; structural change, soil preparation, fertilisation, feeding practise etc.). The farm activities and processes interact between themselves and with other factors such as biophysical components (e.g. climate, soil, geology, hydrology, topography) creating an interdependent socio-economic and environmental system. The interdependency strongly varies across farm types and location and is subject to change, which could be a result of policies themselves or external drivers. Sufficiently detailed farm level models are able to take into account these interdependencies giving them an advantage over regionally aggregated models when modelling new CAP instruments (Schils et al., 2007; Viaggi et al., 2010, 2011). There is growing body of research on farm level modelling of agricultural policies. Two lines of research can be identified in the field: (i) individual farm models (e.g. Buysse et al., 2007; Hennessy et al., 2009; Sckokai and Moro, 2009; Viaggi et al., 2010) and (ii) farm-type models (e.g. Annetts and Audsley, 2002; AROPAj;

109

CAPRI-FARM; FSSIM; FARMIS; Jones et al., 1995; Shrestha et al., 2007).13 The main distinction between the two modelling approaches is that the former models single farm units (usually built from survey data), whereas the latter approach models aggregated or representative farm units often grouped by production specialisation, farm size, production technology or other characteristics. The key advantages of the individual farm modelling approach are that it allows the modeller to better account for variation in farm characteristics (e.g. specialisation, technology) and it can cover farm activities and processes in greater detail. Its main disadvantage is its high computational requirement particularly if the geographical coverage is large (e.g. EU) combined with high model complexity. The farm-type modelling allows better interlinkage of farm types with market models and thus it can generate market feedback at farm-type level. However, the aggregated/representative farm-type models average out a significant share of farm heterogeneity (similar to but at a smaller scale than models aggregated at regional or country level) which can curtail or bias the policy impact assessments.

Accounting for non-market environmental goods Over the last two decades there has been an increasing trend to target agricultural policies at improving the supply of nonmarket environmental goods (e.g. agricultural public goods and externalities) linked to agricultural production. In the EU context, since the 1990s the policy reforms were focused on integrating environmental aspects into the agricultural support programmes. Different measures have been introduced (e.g. cross-compliance, agri-environmental schemes; Natura 2000) in order to give farmers incentives to adjust their management practices and thus reduce the harmful effects of their activities on nature and the environment. The recent European Commission proposal for the CAP post-2013 aims to further strengthen and enhance these environmental objectives (European Commission, 2010, 2011). The modelling of non-market environmental goods requires a different approach to the modelling of market goods. At farm level, non-market environmental goods do not affect the behavioural parameters of models as usually they are not taken into account in the decision making of farmers due to the costly enforcement of property rights to them (OECD, 2001b; Meister, 2001). They tend not to affect the incentive structure within the farming sector because prices do not normally reflect the costs or benefits of producing non-market goods (as long as they are not enforced by policies). An appropriate way of modelling them in a farm level non-equilibrium or PE modelling exercise is by introducing an accounting equation which specifies the relation between agricultural activities on the one side and the non-market environmental effects on the other. This model extension allows quantification of non-market environmental goods and their response to policy changes. In a wider context (i.e. in models going beyond the farm level), non-market environmental goods may affect the behaviour of market agents as agricultural activities may inflict costs (negative externality) or benefits (positive externalities and public goods) on them. This may complicate construction of GE models and PE models (which go beyond farm sector) as modellers need to parameterise the models in this respect. However, when policies target the improvement of non-market environmental goods, models (including farm level models) need to consider both an accounting equation for their production level and their behavioural impacts on the market agents. The policy affects

13 Note that classification of farm models can be done along many dimensions such as from the perspective of modelling, time dimension, uncertainty, methodological application, etc.

110

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

the behaviour of market agents when it imposes a cost or a gain on the production/consumption of public goods and externalities. An important constraint faced by modellers in modelling policy impacts on non-market environmental goods is a measurement problem and the identification of relationships between farm activities and non-market environmental goods. Both are strongly linked to data availability. However, a more fundamental problem is value measurement as opposed to quantity measurement. The quantity measurement is less problematic because data on various environmental indicators can be retrieved from physical, biological and chemical measurements. Value measurement is more complex because it is set in exchange transactions. Because nonmarket environmental goods are non-traded their values cannot be observed, thus they are not available from traditional statistical sources. Modellers have to resort to non-traditional sources (e.g. surveys, developing indicators) with the disadvantage that they often reflect only approximate values of non-market goods. Addressing this data and measurement problem may improve the modelling of this aspect of the CAP and may strengthen the usefulness of models as a tool for policy impact analysis. Modelling first versus second pillar CAP policies There is a significant difference in the present state of modelling of the first and second pillar CAP measures. Generally, the modelling of the first pillar CAP measures is noticeably more advanced than the modelling of the second pillar measures. This is due in particular to the fact that the former policies were implemented earlier than the latter ones hence considerably more theoretical and empirical work was conducted. It is important to note that the intervention logic of first pillar measures is simpler hence their inclusion in the model requires a less complex model structure and less data and parameter requirements compared to second pillar measures. Although important advancements in modelling the first pillar type of measures were made in the relevant literature, there are several areas where improvements are needed. The most important unsolved issue is the modelling of decoupling and cross-compliance. Modellers rarely follow a theoretically rigorous approach to introduce decoupled payments in the model. Some models (e.g. CAPRI) do not consider other impacts besides the land market effects of decoupled CAP payments. Often models apply a highly synthetic representation of the interaction between decoupled payments and farm activities (e.g. usually through an exogenous coupling factor). An explicit theoretically consistent modelling of decoupling in accordance with the effects identified in the literature (e.g. risk-related effects, wealth effects, dynamic and credit effects, etc.) is not considered in almost any structural model (Gohin, 2006; OECD, 2001a). Even more problematic is the modelling of cross-compliance. With few exceptions, a consistent incorporation of cross-compliance in the model is not done in any of the available models. The lack of its modelling is mainly due to missing empirical evidence and the complexity of its implementation. The current available theoretical and empirical knowledge on how cross-compliance affects farmers’ behaviour is not sufficient to render it appropriate for modelling purposes. The state of modelling of RDP (second pillar) is less advanced but a growing body of research is emerging in the field. Various types of models are applied to analyse the effects of RDP, e.g. econometric models, partial equilibrium models, general equilibrium models (based on regional SAMs), integrated assessment models, etc. (Copus et al., 2008). However, these models show several limitations. Most of them analyse a selected set of RDP measures. To the best of our knowledge a study which would provide a comprehensive modelling framework for all RDP measures is not available,

probably due to the considerable heterogeneity of RDP measures and their implementation complexity. One of the key shortcomings of the available models is the weak theoretical accuracy of modelling RDP. Often the studies combine several RDP measures in one policy group (e.g. SCENAR, 2020), or group them either by axis (e.g. Bergmann and Thomson, 2008) or by the area of RDP support (e.g. Psaltopoulos and Balamou, 2006; Vollet, 1998). To reduce the complexity of the model, the grouped measures are normally treated as one policy variable and are assumed to affect farm incentives equally. This is an important weakness since different RDP measures are expected to create different incentives in agricultural and rural markets. Theoretical consistency is an additional weak point of current models. This is probably mainly due to a combination of several factors, such as complexity of RDP implementation and unavailability of sufficiently detailed data. Often the same RDP measures are modelled differently in different studies, implying that the behavioural effects of the RDP differ by study. For example, Psaltopoulos and Balamou (2006) model the agri-environmental measures equivalent to an income transfer. On the other hand, Oglethorpe and Sanderson (1999) are more explicit. They assume that agri-environmental policies lead to an adjustment in farm management practices (e.g. restricting stocking density, fertilizer use, etc.). In some cases the modelling of RDP appears to be ad hoc and simplification of their intervention logic is often preferred. For example, Psaltopoulos and Balamou (2006) assume that RDP payments exogenously increase the output demand of the construction sector in the analysed region. In most cases, the choice of behavioural assumptions is a compromise solution between the various constraints faced by modellers such as the focus of the study, type of model used, model structure, data availability, or/and differences in regional implementation of the RDP measures. Rarely a consistent theoretical approach is followed. Many of the applied models are case- or region-specific (e.g. Haile and Slangen, 2009; Oglethorpe and Sanderson, 1999; Psaltopoulos et al., 2012) and therefore there is limited possibility for their direct use in other economic context and/or regions. Often this is caused by the fact that the modelling of RDP requires a large set of region/farm/situation-specific data which are usually not available and by the fact that the RDP impacts tend to be caseor region-specific. This is particularly the case for models focusing on agri-environmental measures (van Ittersum et al., 2008).

Conclusions The development of prospective policy scenarios and conducting quantitative policy modelling in the presence of the changing nature of CAP is similar exercise as trying to hit a multi-directional ‘moving target’. First, the CAP instruments are changing due to gradual reform of the CAP which requires constant adaptation of models and scenario design in order to capture new and/or adjusted instruments, i.e. the ‘target’ maintains its distance but its location changes. Second, the CAP instruments tend to be more complex with each CAP reform. This requires theoretical and empirical improvement of models and approaches for scenario design in order to capture the new/adjusted intervention logic of instruments, i.e. the ‘target’ tends to move further away. Overall, the CAP changes require prospective scenario development exercises and underlying applied agricultural models to constantly adjust their focus, theoretical and empirical approach, structure and disaggregation level. Since the 1990s there has been a significant shift in the emphasis of the CAP. The CAP reforms implemented over the last decades have moved away from the price intervention policy, towards promoting environmental support programmes. The ongoing political discussions on the future of the CAP (the CAP

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

post-2013) indicate a further strengthening and enhancement of the environmental objectives of EU agricultural policy (European Commission, 2010, 2011). Alongside policy changes, important structural developments have taken place in the global markets, such as the energy price rise and the expansion of bioenergy production, greater commodity price volatility, shift in consumption patterns in developing countries and climatic changes. These external drivers put new pressures on the agricultural sector but there is also a reaction from the policy side leading to an adjustment in policy objectives and a redesign of policy instruments. All these changing factors complicate the development of prospective scenarios for analysing the future development of the agricultural sector as they pose problems to correctly identify future policy changes, as well as external drivers affecting the agricultural sector. In this paper we have attempted to summarise global and European outlooks which focus on medium to longterm scenario projections and policy impact analysis. There is a big variation in terms of both policies and external drivers taken into account in these analyses. This variation can be explained by the heterogeneity in focus among studies and/or the approach applied. Often assumptions on policies and drivers also depend on the time period when they were undertaken, reflecting the main policy and external concerns valid at that period. Consistent with the ongoing CAP changes, an increasing number of prospective scenario outlooks attempt to take on board new policy challenges (e.g. environmental pressures, bioenergy, energy market pressure, climate change). However, an area where improvements are desired to make better use of prospective scenario analyses’ results in policy making is the role of external drivers in shaping the agricultural sector and rural areas, and the sensitivity of policy effects to assumptions on external drivers. In the context of increased dynamism and volatility in global markets, such an improvement is desired because the pattern of agricultural and rural development is a combined effect of policies and external forces. The identification of policy scenarios and drivers is just a first step in conducting robust policy impact analysis. A crucial factor in putting scenarios into practice is the development of appropriate modelling methodologies. This paper has reviewed methodologies applied in policy impact studies with the aim of identifying the main practical and theoretical challenges in conducting quantitative modelling in the framework of the CAP. Two types of approaches dominate much of this field: structural models and econometric models. However, structural models represent a key modelling tool applied extensively in the literature for providing quantitative support to policy impact assessment. The main reason is their higher adaptability to the needs of policy makers although they may suffer from providing less reliable results and offer lower possibility for conducting statistical and validation tests relative to econometric models. Although structural models play a prominent role in this field, with the CAP evolving towards ever more complex instruments, their relevance and predictive accuracy will decline as long as methodological and data issues are not addressed. Over the years the CAP has introduced increasingly complex policy measures and has altered the philosophy of the support. We have identified three key areas where improvement of structural applied models is desirable to tackle these CAP developments (particularly the CAP post-2013 proposals and beyond). The increasing complexity of policy measures, regional and farm level variation in implementation and accounting for non-market environmental goods are factors which increase model complexity in terms of theoretical structure, policy intervention logic, data and parameter requirements. The current state of models is far behind addressing these new CAP developments. Modellers are often forced to impose many ad hoc assumptions to countervail constraints associated with modelling new policy issues. A simplified theoretical foundation behind modelling the intervention logic of the CAP is

111

one of the key underdeveloped areas. The standard considerations that tend to prevail across majority of currently applied models is perfect competition and over-simplified representation of policies. In most cases, models do not go beyond this paradigm, for example by considering a limited range of market imperfections in output and factor markets, or with respect to farm decision making, which is desirable for modelling some few-key CAP instruments. Improper modelling of these new CAP developments may lead to the fact that models may fall short of being able to provide robust quantitative support to policy makers. As modelling new CAP measures is more challenging and combined with the unavailability of sufficiently detailed data and model parameters, modellers often resort to simplified estimation procedures and expert judgments to address these shortcomings. The consequence is that modelling errors cumulate, leading to less theoretically robust results and often driven by the assumptions and expert judgement imposed in the model. This is in sharp contrast to the ‘old’ type of CAP instruments which were much simpler in terms of modelling as well as with regards to data and parameter requirements. Overall, the use of current structural models for scenario analyses of new CAP measures is associated with significant prediction error due to outlined empirical and theoretical shortcomings, which reduces their ability to provide accurate quantitative support to policy makers This is in sharp contrast to their application for modelling the ‘old’ CAP instruments. Addressing these shortcomings is a promising area for future research. A number of ongoing research initiatives co-financed by the European Commission’s Technological Development and Demonstration Framework Programme (e.g. FADNTOOL, CAPRI-RD, etc.) are trying to overcome the existing gap between the increasing complexity of CAP instruments and the capacity to predict/test their effects via modelling tools. Whether or not, and to what degree, the research community will succeed in these challenging tasks is an open issue.

Acknowledgements Financial support from the FP7 project ‘Assessing the multiple Impacts of the Common Agricultural Policies (CAP) on Rural Economies (CAP-IRE)’ is greatly acknowledged. The authors are solely responsible for the content of the paper. The views expressed are purely those of the authors and may not in any circumstances be regarded as stating an official position of the European Commission or the European Research Council Executive Agency.

References Ahumada, O., Villalobos, J.R., 2009. Application of planning models in the agri-food supply chain: a review. Eur. J. Oper. Res. 195, 1–20. Annetts, J.E., Audsley, E., 2002. Multiple objective linear programming for environmental farm planning. J. Oper. Res. Soc. 53, 933–943. Bakker, M., Verburg, P., 2009. Scenario based forecasts of land use, land management at 1 km2 grid, NUTS-X level, based on CLUE-S. In: Helming, K., Wiggering, H. (Eds.), SENSOR Report Series 2008/09. ZALF, Germany, p. 22, www.sensor-ip.eu. Barrios, S., Ouattara, O., Strobl, E., 2008. The impact of climatic change on agricultural production: is it different for Africa? Food Policy 33 (4), 287–298. Bergmann, H., Thomson, K., 2008. Modelling policies for multifunctional agriculture and rural development in a remote EU region (Caithness & Sutherland, Scotland, UK). In: Proceedings of the 107th EAAE Seminar “Modelling of Agricultural and Rural Development Policies”, Seville, Spain, 29 January–1 February. Bougherara, D., Latruffe, L., 2010. Potential impact of the EU 2003 CAP reform on land idling decisions of French landowners: results from a survey of intentions. Land Use Policy 7 (4), 1153–1159. Britz, W., Heckelei, T., 2008. Recent developments in EU policies – challenges for partial equilibrium models. In: Proceedings of the 107th EAAE Seminar Modeling of Agricultural and Rural Development Policies, Seville, Spain, 29 January–1 February. Britz, W., Witzke, P., 2008. CAPRI Model Documentation 2008: Version 1, 181 pp. http://www.capri-model.org/docs/capri documentation.pdf.

112

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113

Buysse, J., Fernagut, B., Harmignie, O., De Frahan, B.H., Lauwers, L., 2007. Farm-based modelling of the EU sugar reform: impact on Belgian sugar beet suppliers. Eur. Rev. Agric. Econ. 34 (1), 21–52. Ciscar, J.C., Iglesias, A., Feyen, L., Goodess, C.M., Szabó, L., Christensen, O.B., Nicholls, R., Amelung, B., Watkiss, P., Bosello, F., Dankers, R., Garrote, L., Hunt, A., Horrocks, L., Moneo, M., Moreno, A., Pye, S., Quiroga, S., van Regemorter, D., Richards, J., Roson, R., Soria, A., 2009. Climate change impacts in Europe. Final report of the PESETA research project. EUR Number: 24093 EN. JRC-Scientific and Technical Report (EUR24093EN). European Commission, IPTS, Seville, 115 pp. Conforti, P., 2001. The Common Agricultural Policy in Main Partial Equilibrium Models. Working Paper No. 14806, National Institute of Agricultural Economics, INEA, Italy, p.40. Copus, A., Psaltopoulos, D., Skuras, D., Terluin, I., Weingarten, P., 2008. Approaches to rural typology in the European Union. In: Handan Giray, F., Ratinger, T. (Eds.), JRC-Scientific and Technical Report (EUR23634EN). European Commission, IPTS, Seville, p. 134. Council for the Rural Area, 2009. Implementation and vision of Common Agricultural Policy. RAAD, Utrecht, http://www.rlg.nl/cap/index.html. de Muro, P., Salvatici, L., 2001. The Common Agricultural Policy in main multisectoral models. Working Paper No. 11. National Institute of Agricultural Economics, INEA, Italy, 54 pp. Duffy, P.A., Wohlgenant, M.K., Richardson, J.W., 1990. The elasticity of export demand for U.S. cotton. Am. J. Agric. Econ. 72, 468–474. Durandeau, S., Gabrielle, B., Godard, C., Jayet, P.A., Le Bas, C., 2010. Coupling biophysical and micro-economic models to assess the effect of mitigation measures on greenhouse gas emissions from agriculture. Climatic Change 98 (1-2), 51–73. EIA, 2010. Annual Energy Outlook 2010. U.S. Energy Information Administration, Washington, DC, 231 pp. European Commission, 2010. Communication from the Commission to the Council, the European Parliament, the European Economic and Social Committee and the Committee of the Regions. The CAP Towards 2020: Meeting the Food Natural Resources and Territorial Challenges of the Future. European Commission, Brussels, 16 pp. European Commission, 2011. Legal Proposals for the CAP After 2013, European Commission, Brussels. http://ec.europa.eu/agriculture/cap-post-2013/legalproposals/index en.htm. Eurostat, 2008a. Energy. Yearly Statistics 2006. Statistical Books. Eurostat, Luxembourg, 490 pp. Eurostat, 2008b. Key Figures on Europe. 2007/08. Eurostat, Luxembourg, 236 pp. Eurostat, 2009. Europe in Figures. Eurostat Yearbook 2009. Eurostat, Luxembourg, 567 pp. EURURALIS, 2008. The Future of Rural Europe – An Anthology Based on the Results of the Eururalis 2.0 Scenario Study Wageningen University, Wageningen, 56 pp. http://www.eururalis.eu/current/background/documents/The%20future%20of %20rural%20Europe.pdf. FAO, 2006. World Agriculture: Towards 2030/2050. Food and Agriculture Organization of the United Nations, Rome, 78 pp. FAO, 2008a. Growing demand on agriculture and rising prices of commodities; an opportunity for smallholders in low-income, agricultural-based countries? In: Proceeding of the 31st session of IFAD’s Governing Council, Rome, 13–14 February. FAO, 2008b. The State of Food and Agriculture. Food and Agriculture Organization of the United Nations, Rome, 138 pp. Fischer, G., 2008. Implications for land use change. In: Expert Meeting on Global Perspectives on Fuel and Food Security, FAO, Rome, 18–20 February. Gocht, A., 2009. Methods in Economic Farm Modelling. Doctoral Dissertation. Institut für Lebensmittel- und Ressourcenökonomik der Rheinischen Friedrich-Wilhelms-Universität zu Bonn, 127 pp. http://hss.ulb.uni-bonn.de/ 2010/2007/2007.pdf. Gocht, A., 2010. Update of a quantitative tool for farm systems level analysis of agricultural policies (EU FARMS). In: Dominguez, I.P., Cristoiu, A. (Eds.), JRC-Scientific and Technical Report (EUR24321EN). European Commission, IPTS, Seville, p. 92. Gocht, A., Britz, W., Adenäuer, M., 2011. Farm level policy scenario analysis. In: Ciaian, P., Gomez, Paloma, S. (Eds.), JRC-Scientific and Technical Report (EUR24787 EN). European Commission, IPTS, Seville, p. 83. Gohin, A., 2006. Assessing CAP reform: sensitivity of modelling decoupled policies. J. Agric. Econ. 57 (3), 415–440. Goodwin, B.K., Mishra, A.K., 2006. Are ‘decoupled’ farm program payments really decoupled? An empirical evaluation. Am. J. Agric. Econ. 88, 73–89. Goodwin, B.K., Ortalo-Magné, F.N., 1992. The capitalisation of wheat subsidies into agricultural land value. Can. J. Agric. Econ. 40, 37.54. Guyomard, H., Mahé, L., 1994. Measures of distorting support in the context of production quotas. Eur. Rev. Agric. Econ. 21, 5–30. Haile, D., Slangen, L., 2009. Estimating the willingness to pay for the benefit of AES using contingent valuation method. J. Nat. Resources Policy Res. 1 (2), 139–152. Haniotis, T., 1990. European Community enlargement. Impact on U.S. corn and soybean exports. Am. J. Agric. Econ. 72, 289–297. Helming, J., Oskam, A., Thijssen, G., 1993. A micro-economic analysis of dairy farming in the Netherlands. Eur. Rev. Agric. Econ. 20, 343–363. Hennessy, T., Shrestha, S., Laurence, L., Wallace, M., 2009. The inefficiencies of regionalised milk quota trade. J. Agric. Econ. 60, 334–347. Hess, S., von Cramon-Taubadel, S., 2006. Meta-Analysis of General and Partial Equilibrium Simulations of Doha Round Outcomes , IAAE Conference, Gold Coast, Australia, 12–18 August.

IMF, 2011. World Economic Outlook Database. International Monetary Fund. http://www.imf.org/external/pubs/ft/weo/2011/01/weodata/index.aspx. Jones, P.J., Rehman, T., Harvey, D.R., Tranter, R.B., Marsh, J.S., Bunce, R.G.H., Howard, D.C., 1995. Developing LUAM (Land Use Allocation Model) and modelling CAP reforms. CAS Paper 32. The University of Reading, Centre for Agricultural Strategy. Keane, M.P., 2010. Structural vs. theoretic approaches to econometrics. J. Econ. 156 (1), 3–20. Kirwan, B., 2009. The incidence of U.S. agricultural subsidies on farmland rental rates. J. Polit. Econ. 117 (1), 138–164. Kuhlman, T., Le Mouël, P., Wilson, C., 2006. Baseline scenario storylines. SENSOR sustainability impact assessment: tools for environmental social and economic effects of multifunctional land use in European regions. In: Deliverable 2.1.1. LEI, The Hague, 36 pp. Lence, S.H., Mishra, A.K., 2003. The impacts of different farm programs on cash rents. Am. J. Agric. Econ. 85, 753–761. Louhichi, K., Kanellopoulos, A., Janssen, S., Flichman, G., Blanco, M., Hengsdijk, H., Heckelei, T., Berentsen, P., Lansink, A.O., van Ittersum, M., 2010. FSSIM, a bioeconomic farm model for simulating the response of EU farming systems to agricultural and environmental policies. Agric. Syst. 103 (8), 585–597. MEASCOPE, 2007. Analysis of Exemplary Policy Issues: Assessment of Multifunctional Impact. Deliverable D7.4, MEA-Scope Project. IAMO, Halle, 90 pp. Meister, A.D., 2001. Synthesis and Evaluation of the Evidence from the Country Case Studies Concerning Different Arrangements and Institutional Options for Providing Non-Commodity Outputs, OECD Workshop, Paris, 2–3 July. OECD, 2001a. Decoupling: A Conceptual Overview. OECD, Paris, 42 pp. OECD, 2001b. Multifunctionality Towards an Analytical Framework. OECD, Paris, 157 pp. OECD, 2006. Documentation of the AGLINK Model. Working Party on Agricultural Policies and Markets, AGR-CA-APM(2006)16/FINAL. Directorate for Food, Agriculture and Fisheries, Committee for Agriculture, OECD, Paris, 29 pp. OECD-FAO, 2011. Agric. Outlook 2011–2020. OECD and FAO, Paris and Rome, 194 pp. Oglethorpe, D.R., Sanderson, R.A., 1999. An ecological-economic model for agrienvironmental policy analysis. Ecol. Econ. 28 (2), 245–266. Patton, M., Kostov, P., McErlean, S., Moss, J., 2008. Assessing the influence of direct payments on the rental value of agricultural land. Food Policy 33, 397–405. PESETA, 2011. Projection of Economic Impacts of Climate change in Sectors of the European Union based on Bottom-up Analysis. http://peseta.jrc.ec.europa.eu/ docs/Agriculture.html. Psaltopoulos, D., Balamou, E., 2006. Modelling the effects of trade policy scenarios on multifunctionality in Greek agriculture: a social accounting matrix approach. ENARPRI Working Paper No. 14, ENARPRI, Brussels, 48 pp. Psaltopoulos, D., Balamou, E., Skuras, D., Ratinger, T., Sieber, S., 2011. Modelling the impacts of CAP Pillar 1 and 2 measures on local economies in Europe: testing a case study-based CGE-model approach. J. Policy Model. 33, 53–69. Psaltopoulos, D., Phimister, E., Ratinger, T., Roberts, D., Skuras, D., Balamou, E., Bednarikova, Z., Espinosa, M., Gomez y Paloma, S., Mary, S., Nohel, F., Santini, F., Sieber, S., 2012. Ex-ante Spatial Policy Impact Analysis of the Rural Development Policy in European Rural Areas (RURAL ECMOD). EUR Number: 25238 EN. JRC- Scientific and Technical Report (EUR25238EN), European Commission, IPTS, Seville, 96 pp. Roberts, M.J., Kirwan, B., Hopkins, J., 2003. The incidence of government program payments on land rents: the challenges of identification. Am. J. Agric. Econ. 85, 762–769. Rosegrant, M.W., Paisner, M.S., Meijer, S., Witcover, J., 2001. 2020 Global Food Outlook: Trends, Alternatives, and Choices. International Food Policy Research Institute, Washington, DC, 24 pp. SCAR, 2007. FFRAF Report: Foresighting Food, Rural and Agri-futures. European Commission Foresight Expert Group, 98 pp. http://ec.europa.eu/research/ agriculture/scar/pdf/foresighting food rural and agri futures.pdf. SCAR, 2011. European Commission – Standing Committee on Agricultural Research (SCAR). The 3rd SCAR Foresight Exercise, European Commission – Standing Committee on Agricultural Research, February, 150 pp. http://ec. europa.eu/research/agriculture/scar/pdf/scar feg3 final report 01 02 2011.pdf. SCENAR2020, 2006. Scenar 2020 – Scenario Study on Agriculture and the Rural World. European Commission, Directorate-General Agriculture and Rural Development, Brussels, 236 pp. SCENAR2020, I.I., 2009. Update of Analysis of Prospects in the Scenar 2020 Study. Final report prepared for the European Commission, Directorate-General for Agriculture and Rural Development, Brussels, 205 pp. Schils, R.L.M., Olesen, J.E., del Prado, A., Soussana, J.F., 2007. A review of farm level modelling approaches for mitigating greenhouse gas emissions from ruminant livestock systems. Livestock Sci. 112, 240–251. Sckokai, P., 2001. The Common Agricultural Policy in Econometric Models. Working Paper No. 1, National Institute of Agricultural Economics, INEA, Italy, 39 pp. Sckokai, P., Moro, D., 2006. Modeling the reforms of the Common Agricultural Policy for arable crops under uncertainty. Am. J. Agric. Econ. 88, 43–56. Sckokai, P., Moro, D., 2009. Modelling the impact of the CAP single farm payment on farm investment and output. Eur. Rev. Agric. Econ. 36 (3), 395–423. Shrestha, S., Hennessy, T., Hynes, S., 2007. The effect of decoupling on farming in Ireland: a regional analysis. Irish J. Agric. Food Res. 46, 1–13. Sparks, A.L., Ward, R.W., 1992. Simultaneous econometric model of world fresh vegetable trade, 1962–82: an application of nonlinear simultaneous equations. J. Agric. Econ. Res. USDA-ERS 44, 15–26.

S.G.y. Paloma et al. / Land Use Policy 31 (2013) 102–113 UN, 2008. World Population Prospects: The 2006 Revision. Department of Economic Social Affairs of the United Nations Secretariat. http://www.un.org/ esa/population. UN 2008. World Urbanization Prospects: The 2005 Revision. Department of Economic Social Affairs of the United Nations Secretariat. http://esa.un.org/unup/. USDA, 2009. USDA Agricultural Projections to 2018. Long-term Projections Report OCE-2009-1. US Department of Agriculture, Washington, D. C, 100 pp. van Delden, H., Stuczynski, T., Ciaian, P., Paracchini, M.L., Hurkens, J., Lopatka, A., Shi, Y., Prieto, O.G., Calvo, S., Vanhout, R., 2010. Integrated assessment of agricultural policies with dynamic land use change modelling. Ecol. Model. 221 (8), 2153–2166. van Ittersum, M.K., Ewert, F., Heckelei, T., Wery, J., Olsson, J.A., Andersen, E., Bezlepkina, I., Brouwer, F., Donatelli, M., Flichman, G., Olsson, L., Rizzoli, E.A., van der Wal, T., Wien, J.E., Wolf, J., 2008. Integrated assessment of agricultural systems – a component-based framework for the European Union (SEAMLESS). Agric. Syst. 96, 150–165. van Tongeren, F., van Meijl, H., Surry, Y., 2001. Global models applied to agricultural and trade policies: a review and assessment. Agric. Econ. 26, 149–172.

113

Viaggi, D., Raggi, M., Gomez, Paloma, S., 2010. An integer programming dynamic farm-household model to evaluate the impact of agricultural policy reforms on farm investment behaviour. Eur. J. Oper. Res. 207, 1130–1139. Viaggi, D., Raggi, M., Gomez, Paloma, S., 2011. Understanding the determinants of investment reactions to decoupling of the Common Agricultural Policy. Land Use Policy 28, 495–505. Vollet, D., 1998. Estimating the direct and indirect impact o residential and recreational functions on rural areas: an application to five small areas of France. Eur. Rev. Agric. Econ. 25, 528–549. Weber, J., Key, N., 2011. Do decoupled payments stimulate production? Proceedings of the 2011 AAEA & NAREA Joint Annual Meeting , Pittsburgh, Pennsylvania, 24–26 July. Williams, S.P., Shumway, C.R., 2000. Trade liberalization and agricultural chemical use: United States and Mexico. Am. J. Agric. Econ. 82, 183–199. Wilson, W.W., 1994. Demand for wheat classes by Pacific Rim countries. J. Agric. Resource Econ. 19, 197–209.