Bio-based and Applied Economics BAE © 2024 Author(s). Open access article published, except where otherwise noted, by Firenze University Press under CC-BY-4.0 License for content and CC0 1.0 Universal for metadata. Firenze University Press | www.fupress.com/bae Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Citation: Baldi, L., Calzolai, S., Arfi- ni, F., & Donati, M. (2024). Predicting the effect of the Common Agricultural Pol- icy post-2020 using an agent-based model based on PMP methodology. Bio-based and Applied Economics 13(4): 333-351. doi: 10.36253/bae-14592 Received: April 6, 2023 Accepted: March 27, 2024 Published: December 31, 2024 Data Availability Statement: All rel- evant data are within the paper and its Supporting Information files. Competing Interests: The Author(s) declare(s) no conflict of interest. Editor: Simone Cerroni ORCID LB: 0000-0002-7505-1174 SC: 0000-0001-9719-2392 FA: 0000-0002-5179-2541 MD: 0000-0002-3957-842X Predicting the effect of the Common Agricultural Policy post-2020 using an agent- based model based on PMP methodology Lisa Baldi1, Sara Calzolai2, Filippo Arfini2,*, Michele Donati1 1 Department Department of Chemistry, Life Sciences and Environmental Sustainability, University of Parma, Italy 2 Department of Economics and Management, University of Parma, Italy *Corrisponding author. E-mail: filippo.arfini@unipr.it Abstract. The objective of this study is to perform an ex-ante assessment of the poten- tial impacts of agro-environmental measures included in the post-2020 Common Agri- cultural Policy (CAP), by estimating farmers’ responsiveness in adopting organic agri- cultural practices and an eco-scheme that incentivises extensive forage systems. This research is conducted by means of an Agent-Based Model (ABM), based on Positive Mathematical Programming (PMP), implemented in GAMS. The ABM facilitates the simulation of interaction among farmers, allowing for an analysis of farm heterogene- ity. The PMP methodology adds a non-rational dimention to the farmers’ economic drivers. The model is calibrated using 2019 Farm Accountancy Data Network (FADN) data specific to the Emilia Romagna region in Italy. Our findings reveal significant impacts on land use, with a notable decrease in cereal cultivation in favour of protein and fodder crops. Moreover, structural shifts are observed, notably a decrease in the number of small-scale farms. We also assess environmental and economic implications, observing a modest reduction in CO2 equivalent emissions per hectare, an increase in water demand, and an overall economic stability among farms, as indicated by changes in gross margin per hectare. Keyword: CAP Reform, Agent Base Model, Land use, Structural change, CO2 Emis- sion. JEL Codes: C61, Q15, Q18, Q52. 1. INTRODUCTION Since its first implementation in the early 1960s, the Common Agricul- tural Policy (CAP) has greatly impacted European Union agriculture, driv- ing farm behavior through subsidies, direct and indirect payments, produc- tion constraints, and trade regulations. The CAP objectives have gradually moved from strengthening agricultural production to providing public goods through different reforms. However, despite the environmental principles embedded in the CAP regulations, as from the Fischler reform in 2003, the intensification of agricultural practices has progressively eroded several criti- https://creativecommons.org/licenses/by/4.0/legalcode https://creativecommons.org/publicdomain/zero/1.0/legalcode http://www.fupress.com/bae https://doi.org/10.36253/bae-14592 https://doi.org/10.36253/bae-14592 https://orcid.org/0000-0002-7505-1174 https://orcid.org/0000-0001-9719-2392 https://orcid.org/0000-0002-5179-2541 https://orcid.org/0000-0002-3957-842X 334 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati cal environmental components such as climate, water quality, pollination, biodiversity, physical and psycho- logical well-being, as well as cultural heritage (European Environmental Agency 2019; Nègre, 2022). These devel- opments have had significant repercussions on the pro- visioning of ecosystem services. The “greening” meas- ures introduced in the 2014-2020 CAP reform proved inadequate to meet social demand for an EU agriculture more aware of its role in enhancing regulatory and cul- tural services (Cortignani & Dono, 2019; Alons, 2017). The CAP post-2020 reform aimed to redress past failures in meeting EU Green Deal objectives and following tar- gets established by the Farm-to-Fork and Biodiversity strategies. The new green CAP architecture is based on eco-schemes, one of the most important innovations introduced by the CAP post-2020 reform, which obliges Member States to allocate at least 25% of first pillar pay- ments to measures beneficial for the environment and the climate. Strategic Plan regulations limit eco-schemes to active farmers, which can apply voluntarily (European Commission 2020). During the last decade, several economic models have been developed to help policymakers and stake- holders to evaluate CAP greening mechanisms from an ex-ante perspective. The main results provided by CAPRI, PASMA, and IFM-CAP models suggested that the CAP measures generating environmental benefits are not as effective as expected (Solazzo et al., 2016). A recent ex-post analysis confirmed these results (Bertoni et al., 2021). This empirical evidence supports the idea that economic modelling is a useful decision tool for designing more effective agricultural policies, increasing researcher and policymaker interest in in-depth impact assessment of agricultural policies at the farm scale (Kremmydas et. al 2018). The aim of this paper is to present an ABM, based on Positive Mathematical Programming, for conduct- ing an ex-ante impact evaluation of the agri-environ- mental measures incorporated into the post-2020 Com- mon Agricultural Policy (CAP). This evaluation entails a comparative static exercise, whereas we equate the baseline scenario with two simulated scenarios wherein farmers receive the organic payments or the payment for extensive forage systems, if economically viable. The baseline scenario also represents the counterfac- tual scenario, enabling the evaluation of the impacts of a particular policy (where farmers receive basic cou- pled and decoupled payments) against alternative poli- cies. The model employed is static in the sense that it evaluates the initial sample at a particular moment in time and compares it to the same sample where farm- ers have altered their behaviour to maximize their util- ity function due to different payment conditions. The quadratic functions, commonly used in dynamic mod- els to capture temporal dynamics, are introduced, in this study, with the PMP to represent nonlinear rela- tionships between variables at a specific point in time. Positive Mathematical Programming is widely used in agricultural policy assessment (Howitt, 1995; Britz et al., 2012; Solazzo et al., 2014; Reidsma et al., 2018; Mat- thews, 2022). A distinctive feature of PMP is its ability to recover important entrepreneurial decision variables, such as hidden costs related to past farming experience, risk attitude, and production expectations, useful for simulating more realistic behaviours, not solely driven by economic rationale. In this research, the PMP model is an agent-based model (ABM) which can capture inter- actions between farms in the use of scarce resources. ABMs are better suited to fulfilling important disaggre- gated specifications, to capturing farm heterogeneity at the regional level, and considering interaction between farmers in the use of scarce resources. They bring sub- stantial innovations to mathematical programming models (Reidsma et al. 2018; Berger & Troost 2014). Integrating positive mathematical programming (PMP) techniques within ABMs provides a rigorous framework for modelling agents’ decision-making pro- cesses, particularly with respect to optimising their behavior subject to constraints and policy incentives. PMP helps in simulating how agents respond to policy changes based on economic principles represented by explicit and implicit variable cost. Moreover, the integrat- ed methodology of ABMs and PMP enables the assess- ment of ex-ante agricultural policies by examining their potential effects on farmers’ behaviour related to agri- cultural production choices, land use, structural adjust- ments, as well as their environmental and economic impacts, supporting policymakers in making informed decisions while considering farms heterogeneity. That said, Implementing ABMs with PMP requires detailed data on agent characteristics, preferences, deci- sion rules, and interactions, which can be challenging to obtain, especially at fine spatial scales. Limited or inaccurate data may lead to uncertainty and biases in model outcomes. ABMs can become highly complex, particularly when modelling large-scale agricultural sys- tems with numerous interacting agents and processes. Calibrating such models to real-world data and ensuring their validity and reliability can be time-consuming and computationally intensive. With over one million hectares of UAA (8.6% of national UAA), in 2016 Emilia Romagna accounts for respectively 10.9% (€3,221.91 million) and 15.17% (€2,292.83 million) of Italian crop production and ani- 335Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 mal value, making this region one of the most produc- tive agricultural areas in Italy. Moreover, for the same reference year, 55% of agricultural land is under high intensity input agricultural practices, 37% under medi- um intensity and 8% under low intensity input practices. Agricultural activities have a strong climate-change impact, accounting in Europe for 10% of total Green- house Gases emission (Eurostat 2022). Italy is the fifth largest contributor, after France, Germany, Spain and Poland, emitting 8% of total agricultural GHGs. Not surprisingly, the high level of agricultural pro- ductivity and related impacts, as well as the consolidated presence of industrial and logistic infrastructures, heavy urbanization, and the peculiar geographical conforma- tion of the Po Valley, make Emilia Romagna, together with the other three regions of the Valley – Lombardia, Piemonte and Veneto, the most polluted and impacted areas in Italy (Raffaelli et al. 2020). This study is organised as follows. The materials and methods section presents the characteristics of the farm sample and discusses how PMP is particularly suitable for developing ABM models. The policy scenario section describes the main agricultural policy instruments used in the simulation, and the results are discussed in the last section. 2. METHODOLOGY AND DATA 2.1. Agent-based models and PMP A key feature of ABMs is their capacity to evaluate the interactions between agents (farms) and to describe the impact on land use and structural change according to the structure, productivity, efficiency, and spatial het- erogeneity of the agents in their territory (Reidsma et al., 2018). Agents can represent different individual farms, entrepreneurs, or aggregated entities, such as farm types. The ability of ABMs to capture the interactions between farmers can be leveraged under the assumption of non-full rationality in production preferences. This can be done because farmers tend to maximize their utility function, rather than their profit function (Nolan et al., 2009; Kremmydas et. al 2018). This is plausible only if agents represent individual farm-households, in which family structure and other individual char- acteristics are particularly important in determining transaction costs affecting the economic objective to be maximized. Decisions are based on production factor endowment and level of technological knowledge, as well as the perception of economic and technical risks.The literature provides some attempts to measure the effect of CAP provisions through ABM-type models, such as AgriPoliS (Happe et. al 2004), MP-MAS (Schrein- emachers and Berger, 2011), and RegMAS (Lobianco & Esposti, 2010), however none of them is associated with the PMP. Linking Agent-Based Models (ABMs) with Mathematical Programming (MP) models offers the advantage of creating micro-level models that can depict technological variations based on the structural char- acteristics of farms. For more insights into the different types of ABMs, Kremmydas et al. (2018) have conducted a systematic literature review on ABMs for evaluating agricultural policies. The integration between ABM and PMP models enables the optimization of the cost func- tion for each farm within the sample. This optimiza- tion takes into consideration the unique characteristics and behaviors of individual farmers, starting from the observed optimal scenario. The cost function is hypoth- esized to be a quadratic functional form in output quan- tities: C(x) = x’Qx/2, where the Q matrix is symmetric and positive semidefinite. Additionally, this integration allows for the simulation of structural and technologi- cal changes, such as changes in farm size or the poten- tial abandonment of farm activities. An ABM based on PMP can estimate these choices by simulating land exchange, the introduction of new activities and chang- es in agricultural management practices. Aggregating these results can provide a useful and solid insight into the general trend of the agricultural sector at regional, national, and international levels. PMP is generally used as a straightforward calibration technique as seen in the CAPRI model, where specific technical coefficients are applied. In this study, the PMP methodology employed for calibration is based on farm marginal costs, which consider accounting costs c and the marginal implicit cost λ, intended as “transaction costs”, or socio-economic costs (Anderson et al., 1985), perceived by the farmers. These costs are estimated under economic constraints using the dual property of a profit maximisa- tion problem implicit in the model. This results in shadow prices linked to production activities that precisely equate to the combined total of the estimated accounting cost and the estimated differential marginal costs. The esti- mated accounting cost corresponds to the farm account- ing values, whereas the estimated differential marginal costs can be viewed as the opportunity cost linked to each activity. The estimated differential marginal cost, usually referred as hidden cost, represents the portion of the esti- mated total marginal cost not documented in the farm accounting sheet but taken into account by farmers when formulating production plans (Cesaro and Marongiu, 2013). The hidden costs refer to the specific and individual opportunity costs that each farmer considers when decid- ing whether to introduce a given crop in the production 336 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati plan. These hidden costs incorporate the specific and indi- vidual opportunity costs that each farmer weighs when determining whether to incorporate a particular crop into the production plan. These costs are important not only for the marginal cost calculation but also for the calibra- tion. It is for this reason that the PMP guarantees that the cost estimates obtained can be used for reproducing the basic production situation, enabling the assessment of each farm’s response within the sample to the policy measures implemented. Although there is no theoretical rationale requir- ing a specific functional form for farmers’ reactions, the quadratic form is employed in this study because it is widely used in Agricultural Economics and inherently represents the cost function. Additionally, the Cholesky decomposition ensures to obtain a symmetric and posi- tive semidefinite matrix. 2.2. The model structure AGRISP (Agricultural Regional Integrated Simu- lation Package), the model described in this paper, is a supply ABM, based on the PMP approach, which models farm-holders as agents and analyzes the impact of new CAP measures on agents’ behaviours related to land use, gross margin, carbon emission, and water consump- tion. AGRISP is implemented in GAMS (GAMS 2023) and articulated in a calibration module and a simulation module, depicted in Figure 1. The exact production level for each farm is estimat- ed with the “self-selection”. A detailed explanation of self-selection rules and a comparison between the farm and frontier cost functions can be found in Paris and Arfini (Paris & Arfini, 2000). Leveraging on the self-selection process, in AGRISP, agents belonging to a specific regional farm sample can exchange production techniques or adopt new agricul- tural practices, if experimental research makes technical information available. This is accomplished through the use of a common frontier-cost function, shared at the regional level, esti- mated using the PMP and which incorporates the costs associated with all crops and cultivation techniques, and the deviation of each individual farm from this func- tion (Arfini and Donati 2012). The common frontier-cost function serves as a link among the farms in the sample. The deviation from the common cost function is regard- ed as a basis for comparing costs and profitability among the farms included in the sample. The introduction of a subsidy or a tax, which trig- gers changes in output prices of variable costs, leads farms to different cost-efficiency crop or techniques combination, as result of the optimization run in the simulation phase. This can be viewed as a form of “social learning process” or, more accurately, as an exchange of technical and economic information made available, because observed, in the sample. The interconnected- ness stems from the fact that all farms are aware of the potential techniques available. The latent technologies or crops are those options that agents could potentially adopt but remain “unused” by a farm due to their lack of economic viability within a particular simulated sce- nario.. Supports coupled to a specific technique or tied to the acreage can alter the economic ratios among various production plans. As a result, farm holders may choose to adopt a new crop or technology from the array of agronomic techniques practiced by the farms in the sample, originally latent in their production plan, and their decision is influenced not only by the accounting cost but also by the utility cost unique to each farm. Following calibration, the simulation module assesses the repercussions of alternative policy scenarios by lev- eraging the positive information embedded in the non- linear cost function and employing a set of hypothetical behavioural rules. These agent-based rules offer a more realistic representation of the interactions among farms, encompassing resource exchange, as well as the choices made by the farmers regarding different agricultural prac- tices, taking into account the specific social and family characteristics. More specifically, as argued by Möhring et al., farm dynamism is correlated with the farm holder’s age and successors’ presence (Möhring et al. 2016). The authors of this study make the assumption that once farmholders reach the age of 65, they are more inclined to reduce farm activity rather than expand it. Likewise, it is assumed that farms with holders aged 65 or older, without successors, are unlikely to lease addi- tional hectares or opt for the conversion of farms from conventional to organic practices. Farmers over 65 are more likely to rent out their land, totally or partially. In the model, the complete rental of land is regarded as equivalent to abandoning the farming activity. On the other hand, if the holder is younger than 65 or the pos- sibility of a generational renewal exists, they may con- sider expanding the farming activity by leasing addi- tional land from neighboring farms or transitioning from conventional to organic practices.. It is important to highlight that all these decisions are contingent upon cost-effectiveness. Therefore, the economic cost function that needs to be optimized incorporates factors such as the cost of land rental and the supplementary expenses associated with converting and sustaining organic crops. The equations associated with the key characteris- tics of the model are outlined below, and more details 337Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 on the implementation of the policy instruments can be found on the Appendix 1. The interactions between agents (1-3), related to the adoption of a specific pro- duction plan, are given by sharing the same frontier- cost function (Q) plus a deviation (u) and the adop- tion of the self-selection rule (4-5) by the nth farm. The self-selection allows for the replication of the observed production plan through a comparison between the marginal cost of the current activity (or technology) and the average cost of a new activity or technology, which is defined within the Q matrix as latent activity or technology. Figure 1. Model Structure. Source: authors’ own elaboration. 338 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati (p’nxn – 1/2x’nQ̂nxn – unxn) (1) Anxn ≤ bn (2) xn ≥ 0 (3) To simulate the fact that not all farms in the sample cultivate all the crops encountered in the region two sets of constraints are postulated. The first set deals with the crops, which are pro- duced, and, thus, the marginal cost relation is an equa- tion: mcnk | xRk > 0 λnk + cnk = QkxRn + unk if the k-th activity is produced, k=1, …, Jn (4) where mcnk is the marginal cost for the n-th farm associ- ated with the k-th activity. The second set of constraints deals with the activi- ties which are not produced by the n-th farm, in which case the marginal cost relation is a weak inequality with respect to the level of the frontier-cost function: mcnk | xRk = 0 : λ_nk + c_nk ≤ QkxRn + unk if the k-th activity is not produced, k = 1,…,Jn (5) R is the level of production observed for activity k and the vector unk assumes the role of indexing the cost function with the farm n specific characteristics. λ represents the implicit component of the marginal cost associated to the production of the activity k by the farm n. Restrictions (4) and (5) enable farmers also to select possible production activities from all activities present in the region among the activities observed in the first phase of the PMP (Paris and Arfini, 2000). In the case of conversion to organic farming, equa- tions (1-3) are replaced by Equations (6-8). p’cxc + p’gxg – 1/2 [xc xg] Qcg [xc xg] (6) S.t. Acxc + Agxg ≤ b (7) Ancxnc ∙ Angxng ≤ 0 (8) Any farm using conventional technology (c) can convert to organic technology (g) if it is more profitable. In the Italian FADN, information regarding the agronomic management practice (organic or conven- tional) is provided for each farm. From this information the average costs, yield and output prices of the organic production are extrapolated. When a farm converts to organic farming those values are applied for the crops included in its production plan. Appendix 1 explains the operational implementation of the conversion from con- ventional to organic agriculture in the simulation phase. The objective function, with the non-linear cost component, takes advantage of the self-selection prop- erty, allowing the substitution of technology or crops based on the cost information provided in the Qcg matrix.Consequently, farms that decide to convert to organic farming change their production plan and cost structure. Equations (9-14) represent and rules related to the exchange of the land factor between agents. Setting j activities, n and m farm holdings exchange land between each other. Equation 9 indicates that the available uti- lised area is equal to the available area plus the rented-in land minus the rented land. Equations (9 - 14) indicate that a farmer can either rent or rent out land, and that the total amount of rented land must be equal to the total rented-out land at the agrarian region level.. More precisely constraint (9) requires that the total land allo- cated to the different crops j (j = 1, … ,J), must be less than or equal to the observed total available land at the j farm level, bn, plus the land rented (Zn) minus the land rented out (Vn). Anjxn ≤ bn + Zn – Vn (9) The land rented is represented as: Zn = ∑ m ZZnm (10) and the land rented out is represented as: Vm = ∑ n VVnm (11) where ZZnm and VVnm are the matrix tracing the trans- fer of land for each pair of farms for renting and renting out, respectively. Furthermore, for each pair of farms, the land rented by one farm must be equal to the land out by the other, as follows: ZZnm – VVnm = 0 ∀n≠m (12) To avoid a given farm renting and renting out land at the same time, a specific constraint has been added: Zn ∙ Vn = 0 (13) Finally, to ensure that the exchange of land is con- sistent with the total available land at regional level, we establish that the total land rented must be equal to the total land rented out: 339Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 ∑ n Zn = ∑ n Vn (14) Therefore, we assume that the exchange of land is limited to the farms located in the same agrarian region. Each farm has a marginal cost level, estimated with the PMP, beyond which acquiring additional land provides no further advantage. Introducing a price shock or a policy incentive can lead to a change in the shadow price of land for a specific farm. However, the land rental price remains constant, as it is treated as exogenous to the model and is assumed to be uniform throughout the Emilia-Romagna region. Agents’ interactions are regulated by the behaviour- al rules already mentioned in the previous section and here summuarised: i) Conventional farmers older than 65 and without successors cannot move to organic prac- tices; ii) Farms are only allowed to exchange land within the agrarian regions where they operate; iii) Farmers older than 65 and without successors cannot rent land. The input level is calculated based on the spending on purchased inputs, both for crops and livestock, per hectare of UAA. The inputs are purchased fertilizers and soil improvers, plant protection products, other means for protection, bird scarers, anti-hail shells, frost protec- tion and purchased feed. To provide environmental impact assessment, we integrated the Italian FADN data with environmental information on greenhouse gas (GHG) emission factors and water consumption for the different crops. GHG emissions from agricultural activities were estimated by applying the ICAAI methodology (Impronta Carbonica dell’Azienda Agricola Italiana), developed by CREA-PB, following the guidelines provided by the IPCC for estab- lishing a national inventory of greenhouse gas emis- sions (Coderoni and Vanino, 2022; IPCC, 2008). This procedure, already implemented by Solazzo et al. (2016) assumes that the amount of atmospheric emissions is linearly related to the level of economic activity, and the emission factors considered for the agricultural sector are carbon dioxide, methane and nitrous oxide, expressed in ton CO2eq per hectare or head of livestock. The con- version factors referred to the 100-year Global Warming Potential and are provided by the Fourth Assessment Report of the IPCC (2007), following Equation (15): CO2eq = CO2 + 298 ∙ N2O + 25 ∙ CH4 (15) More in detail, carbon dioxide emissions comprise emissions due to mechanical cropping operations (Rib- audo 2011) and soil organic carbon (SOC) estimation; methane emissions are due to livestock enteric fermenta- tion and rice cultivation; nitrogen emissions include ani- mal manure management, synthetic fertilizer application and atmospheric deposition (Solazzo et al. 2016). The water consumption measurement uses the Water Footprint Network, based on the extensive work of Mekkonnen and Hoekstra (Mekonnen and Hoekstra 2010) that estimates the water footprint of 147 crops and over 200 products, and which also calculates the water footprint at national and sub-national level of each crop worldwide. The concept of Water Footprint was previ- ously introduced by Hoekstra in 2002 in order to assess the direct and indirect use of freshwater resources along a production chain (Hoekstra and Hung 2002), as a sum of i) Blue water, surface water or groundwater for irriga- tion; ii) Green water, the water naturally embedded in the rhizosphere and available for plant assimilation; iii) Grey water, the volume of water necessary to dilute eco- toxic compounds (mainly used in crop protection) to restore specific quality standards. Results are analised using the aggregation depicted in Table 1. 2.3. Data The economic agents in the model are the individual farms included in the “Rete di Informazione Contabile Agricola” (RICA or FADN) database, which has been operational in Italy since 1968. This database is managed by CREA and provides data for the year 2019. The ini- tial sample is specific to the Emilia-Romagna (NUTS2) Region and comprises 739 farms out of the nearly 11,000 sampled farms across Italy. Since RICA assigns a sample weight to each farm to ensure it is representative of the entire population, the weighted sample corresponds to a total of 40,753 farms. Table 2 illustrates the distribution of farms based on their size class (measured in hectares) and their management practices, which can be either conventional or organic. The set of farm data includes information on geo- graphical location (region, province, altitude, agrar- Table 1. Crop aggregation. Macrocategory Aggregated Crops Cereals wheat, barley, rice, sorghum, other cereals Forages alfalfa, forage maize, other forages Proteic/Oilseeds sunflower, soja, protein crops, other oilseeds Maize maize Meadows Pastures meadows and pastures Indutrial tomato industrial tomato Other industrial crops beetroot, potato 340 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati ian zone), agricultural practices (conventional, organic), household characteristics (age and gender of the farm holder, number of potential farm holder’s successors), land use, specific production costs per crop (cost of seeds, fertilizers, pesticides, energy, water), gross total product, and CAP payments. Table 3 depicts the hetero- geneity of the sample based on class of age, per farm size and percentage of organic farms, that represent almost 15% of the farms population in Emilia Romagna. Within the sample, the average age of the land- holders is 61 for conventional farms and 54 for organic farms. The “agrarian region” spatial definition is a pecu- liarity of the FADN and it further segments Italian prov- inces (NUTS3) based on geographical location and alti- tude range. Although similar to the European sampling, the Italian FADN is notably more comprehensive, con- sidering over 2,500 variables for each sampled farm, in contrast to the European FADN, which only takes into account approximately 1,000 variables (CREA-PB 2021). Table 4 detailed the observed land use in Emilia Romagna region in the year 2019. The prevailing land use relates to cereals (33.26% of the total Utilized Agricultural Area (UAA)) followed by forage (33.17%); meadows and pastures count for 10.54% of the regional UAA. 2.4. Policy scenarios To model how farmers respond to the adoption of organic agricultural practices and eco-scheme, two scenarios are implemented in AGRISP and evaluated through a comparative static analysis. More specifically, we compare the baseline scenario, represented by the calibrated FADN data for the year 2019, wherein farmers receive the basic coupled and decoupled payments, with the simulated scenario. Greening measures of the previ- ous CAP reform: crop diversification, maintenance of permanent grassland, and the establishment of Ecologi- cal Focus Areas are simulated (European Commission 2017) are also included in the baseline. The two CAP post-2020 scenarios implemented in the simulation module of AGRISP are: 1. the “Organic” scenario, where payments are made to encourage farm holders to adhere to organic agricultural practices in order to increase the area Table 2. Number of Farms according to size class (ha) and management practices. Size (ha) Conventional Farms Organic Farms Total Initial Sample Weighted Sample Initial Sample Weighted Sample Initial Sample Weighted Sample < 10 246 17,312 23 1,397 269 18,710 10-20 120 7,714 17 1,950 137 9,664 20-50 152 5,975 34 1,610 186 7,585 50-100 68 2,197 25 964 93 3,160 100-300 47 1,249 3 92 50 1,342 > 300 1 51 3 61 4 112 total 634 34,499 105 6,074 739 40,573 Table 3. Farms per age and size class, based on management practices. Holder’s Age Conventional Farms Organic Farms Total % Organic Farms ≤40 41–64 ≥65 ≤40 41–64 ≥65 - Size class % Age class/Farm type 6.81 46.55 46.64 15.02 64.31 20.67 - - < 10 1,122 7,470 8,721 159 1,034 205 18,710 3.44% 10-20 230 3,350 4,134 283 1,293 375 9,664 4.81% 20-50 525 3,235 2,215 130 1,001 478 7,585 3.97% 50-100 439 1,112 645 340 481 142 3,160 2.37% 100-300 34 842 374 - 52 41 1,342 0.23% > 300 - 51 - - 46 15 112 0.15% Total 2,350 16,059 16,089 913 3,906 1,255 40,573 14.97% 341Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 under organic agriculture to 25%, according to the Farm to Fork strategy target (Appendix 1). Regional payments for organic crops are listed in the RDP of Emilia Romagna (DG AGRI 2021). In this scenario farmers will opt for organic farming if economical- ly convenient, considering transition costs, organic yield and prices for organic products. 2. the “Eco-Scheme” scenario simulates the 4th eco- scheme in the Italian National Strategic Plan (MAS- AF 2022). It envisages incentives in the form of additional payment of 110 €/ha added to the basic payment, for an extensive forage system. In our model, we consider the crop category “Meadows and Pastures” as eligible for this payment. The “Eco- Scheme” scenario is added to the subsidy foreseen to support the conversion to the organic agronomic management practice. The ABM rules and the PMP methodology integrat- ed in the AGRISP model trigger farm owners’ decisions on farm organisation, including factors such as land endowment and utilisation. This is achieved by optimis- ing the individual utility functions of each farm, which subsequently influence the environmental impact and the overall regional gross margin. Other models can be used to perform similar com- parative analysis, such as partial equilibrium models based on farm types (e.g. CAPRI), providing a mac- roeconomic perspective by analysing the interactions between supply and demand in agriculture. However, these models can offer insights into how policies affect market equilibrium, prices and production but do not consider the farms heterogeneity. As noted above (Equations 9 - 14), in both sce- narios farmers can exchange land according to specific agent-based constraints that trace a one-to-one rela- tionship between all the farms included in the sample, in the sense that each farm has the option to rent or rent out arable land with the other farms located in the same agrarian region. Farmers exchange land as a way of making optimal use of their resources. Farmers can adopt different structural strategies, such as leasing out their land and exiting the market entirely, or alternative- ly, they may choose to lease out only a portion of their land while continuing their farming activities. The rental price for land is not resulting from a land market equilibrium but is assumed to remain fixed at 690€/ha. This price is derived from the “Survey on the Land Market” conducted by CREA-PB (2019) in Emilia Romagna. 3. DISCUSSION OF RESULTS The “Organic” scenario and the “Eco-Scheme” sce- nario are executed using the calibrated 2019 Italian FADN data, and subsequently compared to the baseline, which does not incorporate any agent-based or policy constraints. The main emerging phenomena are: (i) The impact on land use, including technological changes for conversion to organic farming; (ii) The structural chang- es recorded in total number of farms per sized-class and in terms of farm holder age ; (iii) The environmental impact related to the carbon emissions and water con- sumption; (iv) The impacts on farmers’ gross margin. 3.1. Impacts on land use The impact of the two scenarios on land use has been analysed both in total hectares allocated and as a percentage (Table 5). Cereals, the less profitable crops, decrease overall by 13.74% in the Organic scenario and by 13.90% in the Eco-scheme one respectively. Meadows and pastures experience a modest decrease in the organic scenario, but the eco-scheme subsidy helps bring produc- tion back up slightly. All other crop categories show an increase. Among them, protein/oleaginous crops reveale the highest rise, with an increase of 8.58% for the Organ- ic scenario and 8.59% for the Eco-scheme scenario. The greening requirement leads to land set-aside of 0.28% on “Organic” and 0.27% in “Eco-scheme” farms. Additional elaboration is provided for each class of dimension concerning the four crops that exhibit higher Table 4. Land Use in thousands of hectars. Land Use (1000 ha) Cereals Forages Proteic/ Oilseeds Maize Meadows Pastures Industrial Tomato Other Industrial Crops Total Conventional 263 230 52 81 45 23 35 729 Organic 46 78 10 5 53 3 4 199 Total 309 308 62 87 98 26 39 928 % 33.26% 33.17% 6.64% 9.33% 10.54% 2.82% 4.24% 342 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati variation: cereals, forage, protein/oleaginous crops, and industrial tomatoes (Figure 2). Delving into the results we notice that the decrease in cereal production is mostly accentuated in the small medium-sized farms (under 100 hectares), whereas the decrease is of lower intensity in farms between 100 and 300 ha and almost not relevant in farms over 300 ha. This could be explained with the fact that cereals are typically grown on large plots of land, and they tend to require less labor and inputs per unit of land compared to other crops. Large-scale cereal farms may have specialised equipment and processes optimised for extensive agriculture, making it less practical or eco- nomical to switch to different crops or practices and they may have more stable market contracts or subsidies that incentivise the continuation of existing cereal production methods. In smaller to medium-sized farms, the decrease in cereal production could be more pronounced when switching to organic or eco-schemes due to the relative increase in labor and management required for these practices. Smaller farms might not benefit from econo- mies of scale in the same way larger operations do and may feel the shifts in practice more acutely. For farms under 50 hectares there is no incentive to increase the production of forage. This is probably due to the relatively low amount of the subsidy for conversion to organic (only 120€/ha for alfalfa and other forage) that the Eco-scheme scenario is not able to counterbal- ance. However, for larger farms (50 hectares and above), the trend reverses, with the forage under Organic and Eco-scheme scenarios having more allocated land than in Baseline, with the largest increases seen in the 100- 300 hectares size class. This could also be driven by the concentration of the dairy farms in class 3-5 (86.33%), which may have further interest in forage. Strong positive shift towards protein/oleaginous crops production is reported in both the Organic and the Eco-scheme scenarios consistently across all farm sizes, suggesting that farmers find agroecological prac- tices economically viable for these products, notably more profitable. This might be due to more favorable subsidies for these crops (351€/ha) or higher market price for organic products. The percentage increase in land allocation is higher in larger farms, especially in those over 300 hectares. This could be due to the great- er financial resilience of larger farms, allowing them to take on the risk of transition and the associated costs more readily than smaller farms. Also for these crops, data suggests significant economies of scale for larg- er farms, more likely to distribute the costs and labor required for organic farming more efficiently. The total increase of around 129% for both Organic and Eco- scheme scenarios is particularly notable. It underscores a widespread and significant adoption of these practices across the sector. A remarkable increase is depicted for smaller farms (<10 hectares) in the Organic and Eco-scheme sce- narios for industrial tomatoes, which might be due to the high subsidy of 427€/ha. This makes it financially attractive for smaller operations to switch to these practices. Medium-sized farms (10-50 hectares) also show substantial increases for both scenarios. This could suggest that the subsidy is sufficient to cover the additional costs of transitioning and that the market for organic or eco-friendly tomatoes is strong. There’s a notable decrease in land allocation for farms larg- er than 300 hectares, where there’s no activity in the Organic and Eco-scheme scenarios. This stark contrast to other size classes might be influenced by several fac- tors, including the possibility that farms producing industrial tomatos practice more intensive farming and consequently have relatively smaller size. Tomato cul- tivation typically involves higher costs for seeds, ferti- lizers, pesticides, and water. They also require careful management and more labor for tasks like pruning, Table 5. Impact on land use, per crop in hectares and in %. Crops Land allocation in hectares per crop Land allocation in % per crop Baseline Organic Eco-scheme Baseline Organic Eco-scheme Cereals 308,691.60 181,205.60 179,665.30 33.27 19.53 19.36 Forages 307,796.00 331,914.00 333,666.00 33.17 35.77 35.96 Protein/oleaginous 61,604.60 141,340.40 141,267.40 6.64 15.23 15.22 Maize 86,561.00 90,632.00 88,917.00 9.33 9.77 9.58 Meadows Pastures 97,817.00 93,449.00 97,454.00 10.54 10.07 10.50 Industrial tomato 26,184.00 32,444.00 33,143.00 2.82 3.50 3.57 Other industrial crops 39,318.80 54,361.80 51,379.50 4.24 5.86 5.54 Greening - 2,620.70 2,475.40 0.00 0.28 0.27 Total 927,973.00 927,967.50 927,967.60 - - - 343Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 trellising, and harvesting. Intensive crops like tomatoes are often grown in smaller areas with a higher yield per hectare and are more labor-intensive than exten- sive crops. The increase in Organic and Eco-scheme scenarios for the 50-100 and 100-300 hectares classes is lower compared to the smaller farms, and this could be because these larger operations might already be pro- ducing at scale, and the relative benefit of the subsidy is lower compared to their overall operations. Despite the differences in subsidies, the overall trend shows that there is a significant move towards Organic and Eco-scheme practices across most crop types and farm sizes. The data for the Industrial Tomato crop, especial- ly the impressive increases in the smaller size classes, shows that when subsidies are perceived as significant and worthwhile, they can be a powerful motivator for changing farming practices. However, for larger farms, especially those over 300 hectares, the current subsidy rates and perhaps other factors related to scale, market dynamics, or the specifics of tomato cultivation may not provide enough incentive for a shift to Organic or Eco-scheme practices. Overeall organic land increases significantly in the Organic scenario (+43% at aggregated level) but is lower at (+35% at aggregated level) for the Eco-scheme. Look- ing at the impact of the two payments schems per class of dimension (Figure 3) we notice that the more reac- tive are the medium size farms, particularly those in the class 100-300 ha. It’s worth mentioning that the sig- nificant rise in organic surface area within this category might be attributed to the absence of a cap on the subsi- dies that farms can request. 3.2. Structural changes The impact of the scenarios on the number of farms, in terms of farm size, is illustrated in Figure 4. - 20.000 40.000 60.000 80.000 100.000 <10 10-20 20-50 50-100 100-300 >300 Ha c ul tiv at ed w ith c er ea ls Farms size class Cereal (Ha) Baseline Organic Eco-scheme - 20.000 40.000 60.000 80.000 100.000 <10 10-20 20-50 50-100 100-300 >300 Ha c ul tiv at ed w ith fo ra ge Farms size class Forage (Ha) Baseline Organic Eco-scheme - 10.000 20.000 30.000 40.000 50.000 <10 10-20 20-50 50-100 100-300 >300 Ha c ul tiv at ed w ith p ro te ic /o ils ee d Farms size class Proteic/Oilseed (Ha) Baseline Organic Eco-scheme - 5.000 10.000 15.000 20.000 <10 10-20 20-50 50-100 100-300 >300 Ha c ul tiv at ed w ith in du st ria l to m at os Farms size class Industrial Tomatos (Ha) Baseline Organic Eco-scheme Figure 2. Land use per crop and per scenario. 344 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati Overall, there is a noticeable decline in the weighted fig- ures, showing a drop of 5,381 units for the Organic sce- nario and a drop of 5,325 units for the Eco-scheme sce- nario. (Table 6). The farms appearing to be the most affected are the smaller ones, with a decrease of 18% in farms smaller than 10 hectares, a 13% decrease in the class with a UAA of 10-20 hectares, and a 10% decrease for farms smaller than 50 hectares altogether (Figure 4). The activation of the land exchange constraints, allowing for land rental, as highlighted in Figure 5, emerges as the primary trigger for this structural trans- formation in the scenarios. However, there is an excep- tion with very small farms (less than 10 hectares), where the incentives for organic conversion and eco-scheme 4 do not seem adequate to support them. These phenomena might be explained with the fact that small farm holders are more likely to leave the mar- ket, while big farms tend to consolidate. For small farms, with shadow prices lower than market prices, it becomes more economically efficient to lease out their land rather than continue farming. We can make the assumption that larger farms exhibit greater resilience, as they can capitalize on their economies of scale, as well as on the subsidies tied to their larger land holdings. From an age-based analysis, and considering the initial agronomic practices of the sample, results reveal (Figure 6) that young farm holders (aged below 40), who represent only a small portion, experience a slight increase in the size class of less than 10 hectares, in conventional farming, due to the impact of the land exchange rules. However, their overall decrease remains relatively stable. Within the organic com- part the decline is perceivable in the smaller size class - 20.000 40.000 60.000 80.000 100.000 120.000 140.000 160.000 180.000 <10 10-20 20-50 50-100 100-300 >300 Ha c ul tiv at ed w ith o rg an ic c ro ps Farm size class Organic (Ha) Baseline Organic Eco-scheme Figure 3. Changes in hectares cultivated under organic farming per scenario and size class. Table 6. Impact of scenarios on number of farms (weighted). Farm size class Baseline Organic Eco-scheme < 10 18710 15368 15297 10-20 9664 8465 8387 20-50 7585 6852 6852 50-100 3160 3053 3159 100-300 1342 1342 1342 > 300 112 112 112 0 10000 20000 30000 40000 50000 Baseline Organic Eco-scheme N um be r o f f ar m s Structural changes per scenario < 10 10-20 20-50 50-100 100-300 > 300 Figure 4. Structural changes according to farm size in ha. -4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 500 < 10 10-20 20-50 50-100 100-300 > 300 N um be r o f f ar m s Farm size class Impact of Land Exchange Land Organic Eco-scheme Figure 5. Effect of land exchange on number of farms per size class (in ha) compared to Baseline. 345Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 (less than 10 hectares) and in the 10-20 hectare range, primarily influenced by land exchange. In the 50-100 hectare class instead, incentives have a minor but still positive effect. In the age range of 41-64, the land exchange rules contribute to a decrease in the number of very small farms, while subsidies help retain some of the 10-20 hec- tare farms in the market. For organic farms in this same age range, subsidies appear to be beneficial in the 20-50 hectare size class, although the impact of land exchange still remains a significant driver in reducing the number of small farms. Farmers aged 65 or older, constituting 43% of the initial sample, appear to be the less responsive to change triggered by subsidies, with a slight exception for conven- tional farms in the 10-50 hectare range. The primary fac- tor leading to the decrease in the number of very small conventional farms is the opportunity to exchange land. If the total Utilized Agricultural Area (UAA) is assumed constant, the average farm size increases from the 31 hectares in the “Baseline” to the 41 and 40 hec- tares in “Organic” and “Eco-scheme” scenarios. This result is consistent with the ongoing trend according to the 7th General Census of Agriculture (ISTAT 2022). 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 < 10 10-20 20-50 50-100 100-300 > 300 N um be r o f f ar m s Farm size class Age under 40 - Conventional Farms Baseline Land Organic Eco-scheme 0 200 400 600 800 1000 1200 1400 < 10 10-20 20-50 50-100 100-300 > 300 N um be r o f f ar m s Farm size class Age under 40 - Organic Farms Baseline Land Organic Eco-scheme 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 < 10 10-20 20-50 50-100 100-300 > 300 N um be r o f f ar m s Farm size class Age 41-64 - Conventional Farms Baseline Land Organic Eco-scheme 0 200 400 600 800 1000 1200 1400 < 10 10-20 20-50 50-100 100-300 > 300 N um be r o f f ar m s Farm size class Age 41-64 - Organic Farms Baseline Land Organic Eco-scheme 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 < 10 10-20 20-50 50-100 100-300 > 300 N um be r o f f ar m s Farm size class Age above 65 - Conventional Farms Baseline Land Organic Eco-scheme 0 200 400 600 800 1000 1200 1400 < 10 10-20 20-50 50-100 100-300 > 300 N um be r o f f ar m s Farm size class Age above 65 - Organic Farms Baseline Land Organic Eco-scheme Figure 6. Variation in number of farms per size class and age range. 346 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati Census results depict an overall decrease in the number of farms, while across all regions of Italy, farm sizes are increasing, which confirms that incentives to counter the disappearance of small farms need to be well-planned. 3.3. Environmental impacts The environmental impact of the CAP post-2020 reform on climate change can be evaluated in terms of GHG emissions per agricultural activity. GHG emissions are measured in CO2 equivalent. Implementing subsidies to support organic agriculture, in this research, leads to a total reduction of almost 6% of tons of CO2 equivalent emitted at the regional level, resulting in a total reduc- tion of 1,294 thousand and 1,297 thousand respectively for scenario Organic and Eco-scheme at the regional level (Table 7), confirming that organic practices impact less on the climate than conventional ones (Holka et. al 2022). In line with these results is the average carbon emis- sion per hectare (Figure 7). Carbon footprint aggregated per crop shows that the reduction in emissions is mainly due to the reduction of cereal cultivation (-11%), while there is a slight increase in emissions related to forage, protein crops and oilseeds. The per farms-size analysis of the evolution of the GHG emissions across scenarios depicts (Figure 8) how the implemented policies generally lead to a significant reduction in CO2 equivalent emissions across most farm sizes, with the exception of the largest farm size cat- egory (above 100 hectares), where emissions actually increase. This suggests that while subsidy-driven policies can effectively reduce GHG emissions in smaller to mid- sized farms, their impact on larger farms may require additional considerations or tailored approaches. The results underline the importance of carefully design- ing agricultural subsidies to ensure they achieve desired environmental outcomes across all farm sizes. It may also point towards the need for diversified strategies that cater specifically to the operational and environmental conditions of different farm sizes. Unlike carbon emission, water resources are in gen- eral strongly affected by the transition to organic pro- duction. Water consumption in the Organic scenario increases by 9,4% (Figure 9), which is mainly due to the decrease in cereal production, offset by an increase in oilseeds and protein crops. Forage cultivation consumes the most water of all crops, accounting for over 60% of the regional water footprint. The result is coherent with the fact that alfalfa is one of the most widespread crops in Emilia Romagna (Solazzo et al. 2016). However, if we delve further in the results per farm size, we note that for farms smaller than 20 hectares, both subsidy scenarios lead to a reduction in water con- sumption, suggesting that the adoption of organic and eco-friendly practices can effectively decrease water usage in smaller scale operations. For farm sizes larger than 20 hectares, both subsidy scenarios result in an Table 7. Carbon Emission in 1,000 tCO2 equivalent aggregated per crop. Baseline Organic Eco-scheme Cereals 493.22 290.93 290.67 Forages 184.15 196.02 198.88 Proteic/oilseeds 54.05 126.34 126.31 Maize 305.00 319.34 313.30 Meadows Pastures 219.08 209.29 218.26 Industrial tomato 55.34 68.57 70.04 Other industrial crops 62.90 83.81 79.58 Total 1,373.75 1,294.30 1,297.04 1,48 1,39 1,4 0,00 0,20 0,40 0,60 0,80 1,00 1,20 1,40 1,60 Baseline Organic Eco-scheme tC O 2e q/ ha Carbon Emission per hectare Figure 7. Average carbon emission (tCO2eq) per hectare. 0 50.000 100.000 150.000 200.000 250.000 300.000 350.000 400.000 <10 10-20 20-50 50-100 100-300 >300 1, 00 0 tC O 2 eq . Farm size class GHG Emissions Baseline Organic Eco-scheme Figure 8. GHG Emissions (ton of CO2 equivalent) per class of farm size. 347Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 increase in water consumption. This trend is especially pronounced in the largest farm size category (100-300 hectares), which could reflect the more water-intensive nature of some organic and eco-friendly practices, or possibly the increased water requirements for these prac- tices to be effective at a larger scale. The results indicate that while subsidy-driven poli- cies can support water conservation in smaller farms, they may exacerbate water use in larger operations. This could have significant implications for water resources management, especially in regions facing water scarcity. These findings underscore the importance of designing agricultural subsidies and practices that are tailored to farm size and local water availability conditions. Policies should consider the varying impacts of organic and eco- friendly practices on water consumption across different farm sizes to ensure sustainable water use. The increased water consumption under both sce- narios for larger farms highlights the need for compre- hensive environmental assessments of subsidy programs. Ensuring that efforts to reduce one form of environ- mental impact do not inadvertently increase another is crucial for the overall sustainability of agricultural prac- tices. 3.4. Economic results Gross margin per hectare increases in both “Organ- ic” and “Eco-scheme” scenarios. The increase of 8.8% in the “Organic” scenario, corresponding to 81€/ha, can be attributed to the implementation of subsidies for organic farming conversion. Adding to these subsidies the pay- ment for extensive forages leads to an overall increase in gross margin of 9.2% (85€/ha) (Figure 10). Looking at gross margin relative variation according to size class (Figure 11), less economically efficient farms are those with an UAA over 300 Ha, followed by those between 50 and 100 Ha. All the other classes show an increase in the gross margin per hectare. 4. DISCUSSION AND CONCLUSION In this study, the application of the agent-based methodology within the AGRISP model has proven to be an effective tool for quantifying the supply-side impacts of CAP measures. Methodologically, AGRISP introduces unique features to capture the diverse charac- teristics of farms, their decisions, and interactions with- in their economic and social contexts. It facilitates pre- dictions of the effects of CAP reform at a granular level, including individual farms, and enables analysis at both territorial and sectoral levels. The social variables, such as family structure and farmers’ age, are taken in con- sideration in the model through the definition of specific rules, to characterise the behaviour of the entrepreneur. The choice of the social variables and the socio-structur- 4.248 4.646 4.673 4.000 4.100 4.200 4.300 4.400 4.500 4.600 4.700 4.800 Baseline Organic Eco-scheme M ill io ns o f m 3 Water Consumption Figure 9. Water consumption (m3) per hectare. 918 999 1003 0 200 400 600 800 1000 1200 Baseline Organic Eco-scheme Gr os s M ar gi n (€ /h a) Gross Margin per hectare Figure 10. Gross margin variation. 100 11 6 -15 9 -25 9 100 14 8 -16 8 -26 9 -40 -20 0 20 40 60 80 100 <10 10-20 20-50 50-100 100-300 >300 Total GM /H a % v ar ia tio n Scenario Gross Margin variation per hectare Organic Eco-scheme Figure 11. Gross margin variation per ha according to size class compared to baseline scenario. 348 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati al rules in this paper was made to assess how the CAP strategies may benefit young farmers, however, other socio-structural rules linked to the characteristics of the agricultural family business can be included. Another innovative feature is the capabilities of simulating the farmers’ attitude to change their produc- tion plans or their production factors endowment. In order to model farmers’ willingness to make changes, the PMP methodology was employed to calculate the marginal cost of individual agricultural productions and the constraining factor, represented by the availability of land. Comapring costs with alternative options acts as a benchmark for farmers when considering the adoption of new technologies and adjustments to their farm struc- ture. Furthermore, the PMP methodology coupled with the self-selection process, enables agents to adapt their production plans by broadening their decision-making options, incorporating production methods and technol- ogies employed by other farms in the sample, as well as considering new production technologies that may emerge due to policy interventions. Consequently, farmers can introduce new processes or modify production intensity, when these choices prove to be more advantageous. Using this approach, AGRISP enabled the simulation of the transition to organic farming in response to the introduc- tion of additional payments and the Eco-scheme 4. The analysis of the model results may highlight which farm categories are advantaged and which are penalised when policy measures are implemented, whether they are designed for specific production cat- egories or are applicable to all farms across the agricul- tural region. Micro-based farm models, capable of simulating farmers’ behaviour and their aptitude to change produc- tion plans under economic, market, technological and environmental scenarios, are becoming increasingly important, however supply-side farm models, while accu- rately simulating the entrepreneur’s strategies, have the limitation of assuming the farm as a “closed” produc- tion system whose decisions consider only the production resources available. Nontheless, the exchange of produc- tion factors between farmers, particularly land, in order to adjust to fluctuations in their marginal value, allows the sample’s dynamics to be brought closer to reality. The results illustrated in this paper showing how less efficient farmers rent out land to more productive ones, enabling the latter to expand their operations and lev- erage economies of scale and scope, well reproduce the decline in number of farms depicted in the most recent Italian agricultural census. Furthemore, our preliminary results show that the ambitious objectives of the new CAP reform would have significant impacts on land use as well as non-negligible effects on climate change mitigation and water resource consumption. The complexity of the new CAP, due to potential contradicting objectives such as competitiveness and environment sustainability, requires careful ex-ante eval- uation of the possible outcome. This study reveals that the subsidies allocated to organic farming conversion and the Italian Eco-scheme 4, applied to the Emilia-Romagna FADN sample (2019), may lead to: 1. a considerable decrease in the number of small farms, 2. a shift from cereal cultivation towards protein and feed crops, 3. a substantial economic stability among farms, meas- ured by changes in gross margin per hectare, 4. a modest reduction in CO2 equivalent emissions per hectare, and 5. an increased demand for water resources. Overall, the effect appears to be positive in terms of CO2 reduction. However, concerns are raised by the fur- ther increase of capital-intensive agriculture at detriment of small farms. This work presents some results aggregated at the regional level, but further analysis could be done to high- light findings at the sub-regional level, to suggest more targeted actions able to consider the individual character- istics of different rural areas, allowing, for instance, dif- ferent payment scheme better calibrated to the territorial conditions and specific regional policy objectives. To conclude, it is noteworthy that like any modeling approach, ABMs with PMP involve simplifications and assumptions about agents’ behavior, market dynamics, policy implementation, and other factors. These assump- tions may not always hold true in practice, leading to potential limitations in the model’s predictive accuracy and generalizability across different contexts. Integrat- ing variuos modeling approaches could provide a com- prehensive assessment of agricultural policies, taking into account farm heterogeneity, farmers’ cost and risk perceptions, and the dynamic nature of production deci- sions and techniques. Collaboration between interdisci- plinary teams of researchers and stakeholders is essential to develop and apply these models effectively in policy analysis and decision-making processes. ACKOWLEDGMENTS This research was funded by the European Union’s Horizon 2020 research and innovation programme 349Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 under grant agreement No 816078 for the AGRICORE Project REFERENCES Alons, G. 2017. «Environmental Policy Integration in the EU’s Common Agricultural Policy: Greening or Greenwashing?» Journal of European Public Policy 24(11): 1604–22. https://doi.org/10.1080/13501763.2 017.1334085. Anderson, J.R., Dillon, J.L., Hardaker, J.B., 1985. Socioec- onomic modelling of farming systems. In: Hardaker, J.B. (Ed.), Agricultural systems research for develop- ing countries: proceedings of an international work- shop. ACIAR-Proceedings-Series,- Australian-Cen- tre-for-International-Agricultural-Research., Hawkes- bury Agricultural College, Richmond, N.S.W., Aus- tralia, 12-15 May 1985, pp. 77-88. Arfini, F., Donati M. 2012. «Organic Production and the Capacity to Respond to Market Signals and Policies: An Empirical Analysis of a Sam- ple of FADN Farms.» Agroecology and Sustain- able Food Systems. 37: 149–71. https://doi.org/. doi: 10.1080/10440046.2012.695328. Arfini, F., Donati M., Grossi L., Paris Q. 2008. «Revenue and Cost Functions in PMP: a Methodological Inte- gration for a Territorial Analysis of CAP», 107th Seminar, January 30-February 1, 2008, Sevilla, Spain 6636, European Association of Agricultural Econo- mists. Berger, T., Christian T. 2014. «Agent-Based Modelling of Climate Adaptation and Mitigation Options in Agri- culture». Journal of Agricultural Economics 65(2): 323–48. https://doi.org/10.1111/1477-9552.12045. Bertoni, D., Aletti G., Cavicchioli D., Micheletti A., Pre- tolani R. 2021. «Estimating the CAP Greening Effect by Machine Learning Techniques: A Big Data Ex Post Analysis». Environmental Science & Policy 119 (maggio): 44–53. https://doi.org/10.1016/j.envs- ci.2021.01.008. Britz, W., van Ittersum, M., Oude Lansink, A., & Heck- elei, T. (2012). «Tools for Integrated Assessment in Agriculture. State of the Art and Challenges ». Bio- Based and Applied Economics, 1(2), 125–150. htt- ps://doi.org/10.13128/BAE-11232 Cesaro, L., Marongiu, S., «The use of RICA to estimate the cost of production in agriculture». Application of the PMP Model to estimate specific cost in Italy. 101- 127. INEA 2013. Coderoni S., Vanino S. (2022), The farm-by-farm rela- tionship among carbon productivity and eco- nomic performance of agriculture, Science of the Total Environment 819 (2022) 153103, https://doi. org/10.1016/j.scitotenv.2022.153103 Cortignani, R., Dono G. 2019. «CAP’s Environmental Policy and Land Use in Arable Farms: An Impacts Assessment of Greening Practices Changes in Italy». Science of The Total Environment 647 (gennaio): 516– 24. https://doi.org/10.1016/j.scitotenv.2018.07.443. CREA-PB. 2019. «Land Market Research». 2019. https:// www.crea.gov.it/en/web/politiche-e-bioeconomia/-/ indagine-mercato-fondiario. CREA-PB. 2021. «RICA». 2021. https://rica.crea.gov.it/ index.php. DG AGRI. 2021. «Rural Development Programme - Emilia-Romagna. 2014-2022». DG AGRI. European Commission. 2017. Commission Delegated Regulation (EU) 2017/ 1155 Am. DG AGRI. 2020. “Recommendations to the Member States as regards their strategic plan for the Common Agricultural Policy”. DG AGRI, European Commis- sion COM(2020) 846 final, 2020. European Environmental Agency. 2019. «The Europe- an Environment: State and Outlook 2020 : Knowl- edge for Transition to a Sustainable Europe». LU: Publications Office. https://data.europa.eu/ doi/10.2800/96749. Eurostat. 2022. «Eurostat». Eurostat. 2022. https:// ec.europa.eu/eurostat/web/agriculture/data/database. GAMS. 2016. «McCarl GAMS User Guide». McCarl Expanded GAMS User Guide Version 24.6, 962. Happe, K., Balmann A., Kellermann K. 2004. «The Agri- cultural Policy Simulator (AgriPoliS): An Agent- Based Model to Study Structural Change in Agri- culture (Version 1.0)». Discussion Paper, Institute of Agricultural Development in Central and Eastern Europe, 50. Hoekstra, A.Y, Hung P.Q. 2002. «Virutal Water Trade. A Quantification of Virtual Water Floes between Nations in Relation to International Crop Trade.» 11. Value of Water Research Report Series. Netherlands: IHE Delft. https://www.waterfootprint.org/media/ downloads/Report11.pdf. Holka, M., Kowalska J., e Jakubowska M. 2022. «Reduc- ing Carbon Footprint of Agriculture—Can Organic Farming Help to Mitigate Climate Change?» Agri- culture 12 (9): 1383. https://doi.org/10.3390/agricul- ture12091383. Howitt, R.E. 1995. «Positive Mathematical Program- ming». American Journal of Agricultural Economics 77 (2): 329–42. https://doi.org/10.2307/1243543. IPCC. 2007. «Climate Change 2007: The Physical Sci- ence Basis: Contribution of Working Group I to the https://doi.org/10.1080/13501763.2017.1334085 https://doi.org/10.1080/13501763.2017.1334085 https://doi.org/10.13128/BAE-11232 https://doi.org/10.13128/BAE-11232 http://dx.doi.org/10.1016/j.scitotenv.2022.153103 http://dx.doi.org/10.1016/j.scitotenv.2022.153103 350 Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 Lisa Baldi, Sara Calzolai, Filippo Arfini, Michele Donati Fourth Assessment Report of the Intergovernmental Panel on Climate Change». Cambridge; New York: Cambridge University Press. IPCC. 2008. «2006 IPCC Guidelines for for ational Greenhouse Gas inventory». IGES, Japan. ISTAT. 2022. «7°Censimento generale dell’agricoltura». Kremmydas, D., Athanasiadis I.N, Rozakis S. 2018. «A Review of Agent Based Modeling for Agricultural Policy Evaluation». Agricultural Systems 164: 95–106. https://doi.org/10.1016/j.agsy.2018.03.010. Le, Q.B., Soo Jin Park, Vlek P.L.G., Armin B.C. 2008. «Land-Use Dynamic Simulator (LUDAS): A Multi- Agent System Model for Simulating Spatio-Temporal Dynamics of Coupled Human–Landscape System. I. Structure and Theoretical Specification». Ecological Informatics 3(2): 135–53. https://doi.org/10.1016/j. ecoinf.2008.04.003. Lobianco, A., Esposti R.. 2010. «The Regional Multi- Agent Simulator (RegMAS): An Open-Source Spa- tially Explicit Model to Assess the Impact of Agri- cultural Policies». Computers and Electronics in Agri- culture 72(1): 14–26. https://doi.org/10.1016/j.com- pag.2010.02.006. MASAF. 2022. «Italy CAP Strategic Plan. Relazione 2021 sul piano strategico della PAC». MASAF. https:// www.reterurale.it/downloads/PSP_Italia_15112022. pdf. Matthews, A. 2022. «The Contribution of Research to Agricultural Policy in Europe». Bio-Based and Applied Economics 10(3): 185–205. https://doi. org/10.36253/bae-12322. Mekonnen, M.M, Hoekstra A.Y. 2010. «The Green, Blue and Grey Water Footprint of Crops and Derived Crop Products». 47. Value of Water Research Report Series. Delft, the Netherlands: UNESCO-IHE. Möhring, A., Mack G., Zimmermann A., Ferjani A., Schmidt A., Mann S.. 2016. «Agent-Based Modeling on a National Scale – Experiences from SWISSland», fasc. 30: 56. Nègre, F. 2022. «Gli strumenti della PAC e le loro riforme», “Fact Sheets on the European Union”, 04/2022, European Parliament, 5. Nolan, J, Parker D., van Kooten G.C., Berger T. 2009. «An Overview of Computational Modeling in Agricultural and Resource Economics». Canadian Journal of Agri- cultural Economics/Revue Canadienne d’agroeconomie 57(4): 417–29. https://doi.org/10.1111/j.1744- 7976.2009.01163.x. Paris, Q., Arfini F.. 2000. «Frontier cost functions, self- selection, price risk, PMP and Agenda 2000». Rivista Di Economia Agraria, gennaio, 211–42. Paris, Q., Howitt R.E. 1998. «An Analysis of Ill‐Posed Production Problems Using Maximum Entropy». American Journal of Agricultural Economics 80(1): 124–38. https://doi.org/10.2307/3180275. Raffaelli, K., Deserti M., Stortini M., Amorati R., Vasconi M., Giovannini G. 2020. «Improving Air Quality in the Po Valley, Italy: Some Results by the LIFE-IP- PREPAIR Project». Atmosphere 11(4): 429. https:// doi.org/10.3390/atmos11040429. Reidsma, P., Janssen S., Jansen J., van Ittersum M.K. 2018. «On the Development and Use of Farm Models for Policy Impact Assessment in the European Union – A Review». Agricultural Systems 159 (gennaio): 111– 25. https://doi.org/10.1016/j.agsy.2017.10.012. Ribaudo, F. 2011. Prontuario di Agricoltura. MIlano: Ulri- co Hoepli Milano. Schreinemachers, P., Berger T. 2011. «An Agent-Based Simulation Model of Human–Environment Inter- actions in Agricultural Systems». Environmental Modelling & Software 26(7): 845–59. https://doi. org/10.1016/j.envsoft.2011.02.004. Shannon, C E. 1948. «A Mathematical Theory of Com- munication». The Bell System Technical Journal 27 (ottobre): 379–423, 623–56. Solazzo, R., Donati M., Tomasi L., Arfini F. 2016. «How Effective Is Greening Policy in Reducing GHG Emis- sions from Agriculture? Evidence from Italy». Science of The Total Environment 573 (dicembre): 1115–24. https://doi.org/10.1016/j.scitotenv.2016.08.066. APPENDIX 1 – CONVERSION TO ORGANIC PRACTICE SCENARIO List of indexes, parameters, and variables: Indexes n = (1,2,…,N): index of farm j = (1,2,…,J): index of crop k = (1,2,…,K); k = j: index of crop Parameters pcnj: output prices for conventional crops pbnj: output prices for organic crops shnj: specific crop payment (€/ha) shbnj: specific payment for organic crops (€/ha) SFPn: single farm payment including basic and greening payments r: rent price for land (€/ha) Qjk: matrix Q unj: farm deviations ABnj: technical coefficients for organic crops Acnj: technical coefficients for conventional crops Variables GMn: gross margin 351Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Bio-based and Applied Economics 13(4): 333-351, 2024 | e-ISSN 2280-6172 | DOI: 10.36253/bae-14592 xhnj: land use xhcnj: land use for convention crops xhbnj: land use for organic crops xcnj: production for conventional crops xbnj: production for organic crops Vn: land rented Zn: land leased List of relevant equations: 1) Constraint linking land allocation to conventional and organic practices xhcnj + xhbnj = xhnj ∀n [conventional AND ((with farm owner ≤ 65 years) OR (with farm owner > 65 years AND with successor))]: Δj 2) Constraint ensuring the total conversion by crop xhcnj · xhbnj = 0 ∀n [conventional AND ((with farm owner ≤ 65 years) OR (with farm owner > 65 years AND with successor))]: Δj 3) Constraint linking organic land allocation and organ- ic production Abnj · xbnj = xhbnj ∀n [conventional AND ((with farm owner ≤ 65 years) OR (with farm owner > 65 years AND with successor))]: Δj 4) Constraint linking conventional land allocation and conventional production Acnj · xcnj = xhcnj ∀n [conventional AND ((with farm owner ≤ 65 years) OR (with farm owner > 65 years AND with successor))]: Δj 5) Objective function at the farm level ∀n [conventional AND ((with farm owner ≤ 65 years) OR (with farm owner > 65 years AND with successor))]: Δj _Ref114578152 _Ref114489287 _Ref114578337 _Ref115429676 _Hlk124753738 _Hlk124753769 _Ref114578159 _Hlk125619779 _Hlk125620100 _Hlk125620116 _Hlk168057609 _Hlk125617961 _Hlk125617978 _heading=h.1fob9te _heading=h.1y810tw _heading=h.3whwml4 _Hlk105387355 _Hlk162970176 _Hlk120503612 _Hlk94825625 _Hlk101307043 _Hlk112165281 _Hlk110795819 _Hlk163789378 _Hlk101572865 _Hlk101574469 _Hlk162424605 _Hlk163657897 _Hlk163658986 _Hlk163657719 _Hlk163657691 _Hlk163658456 _Hlk163658513 _Hlk163658646 _Hlk162446713 _Hlk125919866 _Hlk107164817 _Hlk162973538 _Hlk162447045 _Hlk107174352 _Hlk162447083 _Hlk124882770 _Hlk142394334 _Hlk142395728 _Hlk142430461 _Hlk142337917 _Hlk142342729 _Hlk142337969 _Hlk142337867 _Hlk142342461 _Hlk142342928 _Hlk142338077 _Hlk142348528 _Hlk164779807 _Hlk125467744 _Hlk141790649 Assessing the bioeconomy’s contribution to evidence-based policy: A comparative analysis of value added measurements Tévécia Ronzon1,2,*, Patricia Gurria2, Michael Carus3, Kutay Cingiz2, Andrea El-Meligi1, Nicolas Hark3, Susanne Iost4, Robert M’Barek1, George Philippidis5, Myrna van Leeuwen6, Justus Wesseler2 Predicting the effect of the Common Agricultural Policy post-2020 using an agent-based model based on PMP methodology Lisa Baldi1, Sara Calzolai2, Filippo Arfini2,*, Michele Donati1 Simulating farm structural change dynamics in Thessaly (Greece) using a recursive programming model Stamatis Mantziaris1,*, Stelios Rozakis2, Pavlos Karanikolas1, Athanasios Petsakos3, Konstantinos Tsiboukas1 Analyzing the impact of government subsidies on household welfare during economic shocks: A case study of Iran Mohammad Dehghan1,*, Seyyed Nematollah Moosavi2*, Ebrahim Zare3 Crop production, the pollinator deficit and land use management: UK farm level survey results Iain Fraser1,*, Michelle T. Fountain2, John M. Holland3