AgriPoliS 2.1 â Model documentation - IAMO

LEIBNIZ INSTITUTE OF AGRICULTURAL DEVELOPMENT IN CENTRAL AND EASTERN EURO PE

AgriPoliS 2.1 – Model documentation

Konrad Kellermann1) Kathrin Happe1) Christoph Sahrbacher1) Alfons Balmann1) Mark Brady2) Hauke Schnicke1) Amanda Osuch1)

October 2008

1) Leibniz Institute of Agricultural Development in Central and Eastern Europe; Halle (Germany) 2) Swedish Institute for Food and Agricultural Economics; Lund (Sweden)

Table of contents 1 2

Introduction .................................................................................................... 4 Conceptual framework ................................................................................... 6 2.1 2.2 2.3 2.4

3

Implementation of the conceptual model ..................................................... 11 3.1 3.2 3.3

4 5

Agents involved.......................................................................................................... 7 Farm agent (inter-)actions and behaviour .................................................................. 7 The spatial, technological and political environment................................................. 9 Central modelling assumptions ................................................................................ 10 Object-oriented structure and design........................................................................ 11 Model dynamics ....................................................................................................... 14 Data Structure........................................................................................................... 15

An abstract spatial model of the agricultural landscape .............................. 17 Factor inputs and production activities ........................................................ 22 5.1 Production factors .................................................................................................... 22 5.1.1 Land.................................................................................................................. 22 5.1.2 Labour .............................................................................................................. 22 5.1.3 Capital .............................................................................................................. 23 5.2 Production activities................................................................................................. 25 5.3 Factor and product markets ...................................................................................... 26 5.3.1 General market rule.......................................................................................... 26 5.3.2 Regional markets.............................................................................................. 27 5.3.3 Exogenous price projections ............................................................................ 28 5.3.4 Regional and spatially-organised markets........................................................ 28 5.3.5 Transaction costs on land markets ................................................................... 33

6

The farm agent.............................................................................................. 36 6.1 Behavioural foundation ............................................................................................ 38 6.1.1 Farm planning .................................................................................................. 38 6.1.2 General remarks about expectation formation ................................................. 38 6.1.3 Price expectations............................................................................................. 39 6.1.4 Cost expectations.............................................................................................. 39 6.1.5 Expected profit/household income and opportunity costs ............................... 41 6.1.6 Expectations about policy changes .................................................................. 43 6.1.7 Managerial ability ............................................................................................ 43 6.1.8 Retired farms .................................................................................................... 44 6.2 Farm actions ............................................................................................................. 46 6.2.1 Renting/releasing land...................................................................................... 47 6.2.2 Investment ........................................................................................................ 51 6.2.3 Production ........................................................................................................ 53 6.2.4 Farm accounting............................................................................................... 55 6.2.5 Farm exit decision ............................................................................................ 57

7 Policy framework ......................................................................................... 60 8 Data input, results preparation and data output............................................ 62 References ........................................................................................................... 64

List of tables Table 3-1: AgriPoliS data structure and variable names.......................................................... 16 Table 4-1: Table summarizing global landscape representation.............................................. 21 Table 5-1: Investment attributes............................................................................................... 24 Table 5-2: Product attributes .................................................................................................... 25 Table 5-3: Table summarizing global product- and factor market related settings ................. 36 Table 6-1: Calculation of expected income and opportunity costs .......................................... 42 Table 6-2: Calculation of the household income ..................................................................... 45 Table 6-3: Indicators calculated in the financial statement...................................................... 56 Table 6-4: Definition of variables used in financial statement (selection) .............................. 57 Table 6-5: Table summarizing global settings related to farms ............................................... 59 Table 7-1: Overview and description of policy parameters ..................................................... 60 Table 8-1: Data output at the farm and sector level (selection of key data) ............................ 62

List of figures Figure 2-1: A static conceptual model of a regional agricultural system................................... 6 Figure 3-1: UML-representation of a class consisting of attributes and methods ................... 12 Figure 3-2: Static class-diagram of AgriPoliS ......................................................................... 13 Figure 3-3: Model dynamics implemented in the Manager class.......................................... 15 Figure 4-1: Abstract landscape representation in AgriPoliS.................................................... 19 Figure 4-2: Initialisation of the abstract landscape in pseudo code ......................................... 21 Figure 5-1: Procedure of land allocation in pseudo code......................................................... 30 Figure 5-2: Rent adjustment of new and old rental contracts .................................................. 33 Figure 6-1: Expected cost savings for machinery investments, depending on farm size and the average number of adjacent plots..................................................................................... 41 Figure: 6-2: Exemplary scheme of a mixed-integer programme matrix.................................. 46 Figure 6-3: Course of events in one planning period for one farm agent ................................ 47 Figure 6-4: Modelling of labour savings in investments.......................................................... 53 Figure 6-5: Farm exit decision in AgriPoliS ............................................................................ 59

The Agricultural Policy Simulator – AgriPoliS

1

4

Introduction

This contribution presents the current version of the model documentation of the agent-based model AgriPoliS (Agricultural Policy Simulator), which is an agent-based model of regional agricultural structures. The origin of this model dates back to work by BALMANN (1997), who studied path-dependencies in agricultural structural change with an agent-based approach. The model itself was first published in HAPPE 2004 and HAPPE et al. 2006. Whereas Balmann’s model was based on a hypothetical farm structure, AgriPoliS can be calibrated to empirical data derived from farm accounting data and regional statistics. Accordingly, this makes the model applicable for policy analysis and empirically-based analysis of regional structural change. Since the version documented in HAPPE et al. (2006), AgriPoliS was calibrated to further case study regions. Both the necessity of covering diverse regional characteristics and the extended scope of the conducted analysis has led to an amended version of the model, which is presented here. Although the first version of the model is presented in detail in HAPPE (2004) and more recently in HAPPE et al. (2006), the improvements of the model are substantial enough that we present a revised version of the whole model’s documentation instead of merely presenting the extensions (an intermediate step is presented in KELLERMANN et al. 2007). However, some parts of the documentation are based on previous versions of the documentation. This report is structured as follows: Section 2 provides an overview of the conceptual framework of the model, a brief explanation of the main entities and how these entities are related to each other. Section 3 shows how this conceptual model is implemented by outlining the general model dynamics and its data structure. Sections 4 to 7 describe the individual entities of the model in detail. In AgriPoliS, the behavioural foundation and the strategy space of the farms, that is, the core entities of the model highly depend on the surrounding environment, which is given by the spatial representation of the landscape, commodities and factors, and the respective markets and the policy framework. Consequently, it is reasonable to start the documentation from the point of view of the environment surrounding the farms. To guide the reader through the documentation of the model, and especially those who are familiar with previous versions of AgriPoliS, we first highlight those parts of the model which have undergone major changes since the version documented in Happe et al. (2006).


5

When reviewing the enhancements of the current version, five major groups can be identified: •

Spatial environment: Initialisation of a hypothetical landscape that reflects statistical properties of a real landscape, generic number of soil types, and spatially explicit assignment of production activities.

•

Political environment: Flexible implementation of decoupling policies such as the different single payment schemes (EU-15), single area payment schemes (EU-10) and bond schemes. Possibility of variable coupling rates over time, mix of different schemes and stepwise phasing out/in of policies, possibility to implement progressive modulation schemes.

•

Behavioural foundation: Extended expectation formation in case of radical policy changes. Extended possibilities to model farm exit/entry decisions.

•

Markets: Model interface to aggregated sector models, improved regional markets.

•

Land market: Extension of land market representation to allow for an adaptation to the regional specific properties of land markets. Introduction of several auction protocols (first or second price auction), implementation of different assumptions about contract types. Implementation of a framework to consider transaction costs on land transactions. Consistent consideration of size effects during land allocation.


2

6

Conceptual framework

The core of AgriPoliS is the understanding of a regional agricultural structure as a complex, evolving system.1 This regional agricultural system is shown schematically in Figure 2-1. The figure shows the interactions between the three central components of agricultural structures: farms, markets, and land.

Farms

e.g. soil quality

farm1

farm4

farm2

renting land

Land/ space

farm3

depends on price signals

production

influences

Markets Input: labour, capital, quota, manure, land Output: products

Source:

technology

policy

Own figure.

Figure 2-1: A static conceptual model of a regional agricultural system The key entities in the model are a number of individual farms which evolve subject to their actual state and to changes in their environment. This environment consists of other farms, factor and product markets, and space, which are again all embedded within the technological and political environment. Farms, land, and markets either directly depend on each other or they exert influence on each other. Direct dependence implies that one component cannot exist without another. Mutual dependence between farms, land, and markets results from the fact that farms require land on which to produce. Farm management practices in return influence the state of the land, the quality of which is characterised, for example, by soil fertility. On the other hand, mutual in-

1

This section is partly based on HAPPE (2004) and HAPPE et al. (2006).


7

terdependence between farms and markets takes place because farms can purchase production inputs on factor markets and sell products to product markets. Before translating the conceptual model into a computer simulation programme in the following, the core contents will be sketched in more detail along the following questions: • Which agents are involved and what makes them heterogeneous? • How do agents behave and what actions drive the system? • Which factors comprise the individual agent's spatial, technical, and political environment? • How do interactions between agents, and between agents and the environment, take place in the model?

2.1 Agents involved For the purpose of AgriPoliS, an agent is defined as an entity that acts individually, senses parts of its environment and acts upon it.2 In the context of regional agricultural structures, the main group of agents we defined in the model are the farm agents. In the context of AgriPoliS, one farm agent corresponds to one farm or agricultural holding. In accordance with the above definition, a farm agent is an independently acting entity that decides autonomously on its organisation and production to pursue a defined goal (e.g. farm household income maximisation). Furthermore, a farm agent reacts to changes in its environment and its own state by adjusting its organisation in response to available factor endowments and observable actions of other farm agents.

2.2 Farm agent (inter-)actions and behaviour Farm agents are assumed to act autonomously and to maximise profits from their economic activities. For this, production and investment decisions are made simultaneously based on a recursive mixed-integer linear programme including integer activities (c.f. HAZELL and NORTON

1986). From the solution of the linear programme, investment activities as well as pro-

duction factor shadow prices can be derived. However, decision-making of a farm is bound2

See e.g. HAPPE (2004), TESFATSION (2002, 2001) for a more in-depth discussion of agents and agent-based systems.


8

edly rational (SIMON 1955) since decision-making is myopic and strategic aspects are only included in a rudimentary manner. Except for the price information on rents and product and input prices, individual farms in AgriPoliS do not know about other farms' production decisions, factor endowments, size, etc. Farm agents are also boundedly rational with respect to expectations; in the majority of cases, farm agents follow adaptive expectations. Only policy changes are anticipated one period in advance and included into the decision-making process. Farm agents can produce a selection of goods. In order to produce, farm agents utilise buildings, machinery, and facilities of varying type and capacity. With respect to this, AgriPoliS implements economies of size; with an increasing size of production, unit investments costs decrease. Moreover, labour is assumed to be more effectively used with increasing size. AgriPoliS also aims to mimic the effect of technological progress. More specifically, it is assumed that with every new investment, unit costs of the product produced with this investment decrease by a certain percentage. Farms can engage in rental activities for land, production quotas, and manure disposal rights. Labour can be hired on a fixed or per-hour basis, and vice versa, farm family labour can be offered for off-farm employment. To finance farm activities and to balance short-term liquidity shortages, farm agents can take up long-term and/or short-term credits. Liquid assets not used within the farm can be invested with a bank. Farm agents quit production and withdraw from the sector if equity capital is zero, the farm is illiquid, or if opportunity costs of farmowned production factors are not covered.3 New investments affect production capacities for the operating lifetime of said investment. This implies investment costs to be sunk. A farm agent is handed over to the next generation after a given number of periods. In case of such a generation change, opportunity costs of labour increase. Accordingly, continuation of farming can be interpreted as an investment into either agricultural or non-agricultural training. Moreover we assume that opportunity costs of labour can decrease with increasing age of the farmer. Finally, farm agents differ not only with respect to their specialisation, farm size, factor endowment and production technology, but also with respect to person and managerial ability. The concept of interaction between agents is central to agent-based systems. Interaction takes place when two or more agents are brought into a dynamic relationship through a set of recip3

As investment costs are assumed to be sunk, only opportunity costs for land and labour are considered.


9

rocal actions. Interactions develop out of a series of agents’ actions, the consequences of which effect the future behaviour of the agents (FERBER 1999). At this development stage, agents in AgriPoliS interact indirectly by competing on factor and product markets. Interaction is organised by markets that explicitly coordinate the allocation of scarce resources such as land or the transaction of products. The markets we consider range from local, spatially-organised markets such as markets for land or manure, to supra-regional markets like markets for crops, etc. In this respect, the land market is the central interactional institution between agents. In reality, the land market is of particular relevance, as farms very often cannot develop independent of land.

2.3 The spatial, technological and political environment Land is an essential input for most kinds of agricultural production activities, be it for plant production, as fodder ground, or as manure disposal area. Hence, space is a factor that cannot be neglected if agriculture is concerned. One way of considering the spatial environment of a farm in the context of an agent-based model is to use Geographic Information Systems (GIS) as a representation of landscape, as some recent examples show (e.g., BERGER 2001; PARKER et al. 2002). GIS provide a way for organising spatial data and assigning certain properties to space. A common way to organise space in a GIS is to define a grid of cells. A grid, or layer, categorises land with respect to attributes of the cells. For example, this could be the soil type, ownership, or ecological parameters such as the nitrogen load. However, due to data availability, using maps with the exact location of farms and land plots is often impractical, at least when the model should cover multiple regions. Instead, in AgriPoliS a more basic approach was chosen to represent a region in a stylised way, where space is represented by a set of cells/plots assembled into a two-dimensional grid. In the model presented in HAPPE (2004) this stylised landscape was limited to two different soil types and a representation of only some, though not explicit, spatial relationships. In the current version of the model, this approach was extended. First of all, we allow for an unlimited number of soil types, where we also explicitly consider non-agricultural land to represent some kind of natural borders around the fields. Although the landscape is still based on grid cells of equal size, plots can now be aggregated to form larger areas of contiguous land. At different layers, the contiguous plot (land) area is either used to reflect the ownership structure, the physical landscape or the us-


10

age structure in the region. Therefore, we introduced a procedure which facilitates the representation of a set of statistical landscape indicators taken from the real landscapes (such as average block size and size distributions). Furthermore, we introduced spatially-explicit mapping between farm production and the landscape. In this manner, the model can be used for environmental impact assessments (cf. BRADY and KELLERMANN, 2005). The basic approach we followed was to incorporate environmental impacts within the production economic framework of AgriPoliS. Here we view environmental impacts as outputs or inputs of a multioutput and multi-input agricultural production process. From that, we derive a series of indicators to characterise e.g. biodiversity in terms of species diversity and species/area relationships, landscape diversity, nutrient balances and also the risk of soil erosion, all of which are dependent on regional characteristics. The farm’s technological environment is given by technologies of different vintages and technological standards. Over time, technology is assumed to underlie a constant technological progress created in the up-stream sector, but not on the farms themselves. Farm agents are assumed to benefit from technological progress by realising additional cost savings when adapting new technologies. The political environment represents the third building block of a farm agent's external environment; the first two are space and technology, respectively. Agricultural (and environmental) policies affect the farm at different instances such as prices, stocking density, direct payments, or interest rates.

2.4 Central modelling assumptions As with every model, AgriPoliS rests on a number of assumptions. Two kinds of assumptions can be differentiated. On the one hand, there are assumptions that represent central characteristics of an agricultural system. These form the cornerstones of the model. BALMANN (1995) has listed the central characteristics of agricultural systems and structures, which shall be mentioned here again: • The evolution of agricultural structures follows a dynamic process. • Agricultural structures are path dependent, i.e., the history of the system significantly determines its present state and certain events are irreversible. • For the most part, decision-making follows goal-oriented economic considerations.


11

• Certain activities, decisions and actions are indivisible. • There are feedback mechanisms, particularly on the local scale, between the actions of individuals and between the results of individual actions. On the other hand, there are assumptions that are model-specific and are necessary to make the model operational and to keep it tractable and clear. Particularly, these assumptions concern farm behaviour, expectation formation, the definition of the planning period, the representation of markets and interaction with other sectors.

3

Implementation of the conceptual model

3.1 Object-oriented structure and design A natural way of transferring the conceptual framework into a computer programme is to use an object-oriented programming language such as C++, Java, or Smalltalk.4 Objectorientation provides a way to break a problem into its component parts. In brief, objectorientation describes a system of entities in terms of elements called objects. Objects consist of data (or attributes) and actions (or methods). The data represent the state of the object. The actions operate on an object's data and change it. For example, a farm agent's investment activity (action) changes the agent's capital endowment (data). In other words, an object provides functionality in terms of data and actions. A programme built using an object-oriented design usually contains a large number of objects, of which many are the same. For example, in an agricultural structure all objects representing farms will be treated in the same way. When designing a computer programme such as AgriPoliS, which uses objects, it is therefore sufficient to describe the behaviour of sets of similar farms as a whole. A group of objects with the same data and actions is called a class. Because of this, it is actually more common and useful to define the functionality of classes instead of individual objects in the design of object-oriented computer programmes. To summarise, object-oriented programmes thus consist of a set of classes, the data associated with these classes, and the set of actions the classes can be asked to undertake.

4

This section on object-oriented design is largely based on REISS (1999) who gives an intuitive introduction to object-oriented programming and design.


12

One key to understanding object-oriented design is to consider the objects to be living, intelligent entities of various types (REISS 1999); they are living in the sense that their properties change over time. Objects are intelligent in that they can undertake actions and know how to perform them. To visualise and document the design of an object-oriented computer programme, it is convenient to use a standardised language such as the 'Unified Modelling Language' UML (BOOCH et al. 1999). UML simplifies the representation of complex software design. Accordingly, a representation of a class based on UML is given in Figure 3-1. The upper part of the class representation shows the class name. The middle and lower parts list the attributes and the methods that the class can be asked to undertake.

Class Attributes Methods Source:

Own figure.

Figure 3-1: UML-representation of a class consisting of attributes and methods When building an object-oriented programme, one is first concerned with identifying the individual classes, then with defining the data and actions of these classes, and finally with describing the connection between classes. Figure 3-2 shows the object-oriented class design of AgriPoliS. Class names, as used in the C++ programme are in parentheses. The grey shaded classes are agent classes.

The Agricultural Policy Simulator – AgriPoliS one-way association

Input data (DataInput)

Output data (DataOutput)

aggregation

Programme Manager (Manager)

Results Sector (SectorResults)

Region (Region)

Plot of land (Plot)

Investment list (InvestList)

Investment object (InvestObject)

List of products (ProductList)

Product (Product)

Labour (Labour)

Farm agent (Farm)

Linear programme algorithm (LP/MIP) Auctioneer (Auctioneer)

Note: Source:

13

Product market (ProductMarket)

Names in brackets denote the class names used in AgriPoliS' C++ programme code. For clarity, the figure does not show attributes and methods. The complete model code can be provided by the authors upon request. HAPPE (2004).

Figure 3-2: Static class-diagram of AgriPoliS For the model to perform its task, it is not necessary that all classes are related to each other or can invoke each other's methods. In the figure, lines are used to express different kinds of relationships between classes. In general, a line between two classes denotes an association relationship. Properties of this line, such as the arrowhead, are used to further specify the character of the association. For example, the relationship between the classes Farm and LP/MIP is implemented as a one-way association by using an arrow. This indicates that a Farm object can invoke the methods of the LP/MIP object, but not the other way around. Likewise, a Farm object determines its location by querying the Region object to return the position of the farm in the region, but the reverse is not possible. Another type of association is aggregation, denoted by a diamond. For example, the line from Region to Plot starts with a diamond, which denotes an aggregation. In this case, the region contains a set of plots. Similarly, each list of production contains a set of products. From the classes shown in Figure 3-2, four kinds of objects can be derived: objects representing agents (Farm, Auctioneer, ProductMarket), objects representing production


14

inputs and outputs (Product, ProductList, Labour, InvestObject, InvestList, Plot, Region), results and data management objects (SectorResults, DataOutput, DataInput, LP/MIP) and the Manager, which controls the programme’s flow. Accordingly, agent objects use the functionality embodied in input and output objects to achieve their respective goals. Results and data management objects offer some auxiliary functionality in that they provide optimisation methods on the one hand, and functions to summarise farm data on the other.

3.2 Model dynamics Whereas Figure 3-1 presents the static structure of the AgriPoliS model, Figure 3-3 illustrates the dynamics which are implemented and controlled by the Manager objects. As can be seen, the Manager essentially includes two model phases: the initialisation phase and the simulation phase. In the initialisation phase, the model structure is created. This includes the creation of objects based on class definitions, and assigning values to the respective attributes of the various objects. The initialisation phase ends by further individualising farms with respect to attributes for which empirical data is not available or difficult to obtain. Following the initialisation phase, the simulation phase starts with setting the political framework conditions that are valid during the subsequent simulation period. Following this, the Manager invokes the Auctioneer agent to carry out the land auction to allocate unused land to farm agents. After the land auction has finished, farm agents have the possibility of investing in new machinery, buildings, or equipment, and following this, to produce using the available production factors. After production, the Manager invokes the Market agent to bring together production of all farm agents in the respective region and to determine a price for each type of product produced by the farm agents.


15

Simulation Phase

Initialisation phase Initialise objects create objects assign data to objects

Set policy

Land auction Individualise agents managerial ability investment vintage farm and plot location

Investment Production Update product markets

Output to file farm data sector data

repeat

Period results farm assessment sector results Continue farming

End

Source:

Yes

stop ?

No

Happe, 2004.

Figure 3-3: Model dynamics implemented in the Manager class At the end of each simulation period, farm agents assess their economic performance during that particular period. Based on this assessment and the given prospective policy changes, the farm agents form expectations about the next production period and decide on whether to continue or stop farming. For this decision, farm agents take into account all possible adjustment options such as off-farm labour opportunities, selling excess quota, and terminating land rental contracts. Fixed assets cannot be disinvested due to the mentioned sunk cost assumption. Results for each individual farm agent and the sector as a whole are written to an output file. The simulation terminates when the number of specified simulation periods is reached.

3.3 Data Structure Table 3-1 lists the data structure and variable names used in AgriPoliS and the following documentation.


16

Table 3-1: AgriPoliS data structure and variable names Farm agent f(k=1,…,K) Z Utilised agricultural area of farm a) LU Stocking density per farm MP Manpower hours Managerial ability factor M A (l=1,...,L) Fixed assets Aec(l=1,...,L) Equity financed share of assets Vintage of asset nc LA Land assets L Liquidity EC Equity capital BC Borrowed capital MR Minimum equity capital reserve Y Farm household income Ye Expected farm household income GMA

Gross margin agriculture

IR RE S

Interest on working capital Rent paid Support payments

MC D OV Capital CRF Iec

Current upkeep (maintenance) Depreciation Farming overheads

Ibc Ibcs

Capital return factor Interest on equity capital Interest on borrowed capital Interest on short-term borrowed capital

Farm agent (continued) TC Transport costs IC Interest paid HW Wages paid W Off-farm income Withdrawal for consumpWD tion ε Additional consumption Investment I (h=1,…,H) to produce i d (d=1,…,D) Investment type Investment costs A AC Average annual costs N Useful life MC Maintenance costs p.a. l Technical change factor machinery F Technical change factor buildings and equipment LS Labour substitution Production activities x (i=1,...,I) Production activity Variable prod. Cost c (i=1,...,I) ce (i=1,...,I) Expected variable costs pe (i=1,...,I) Expected product price γ Price trend b (j=1,...,J) Factor capacities q (j=1,…,J) Shadow price of b r (j=1,...,J)

Factor demands

Land

u yj , x

Plot of owner j at grid position y,x

sb l

Sub block (set of single plots of one type with adjacent borders) managed by farm l Field (set of single plots of one type with adjacent borders) managed by farm l which is used for a single activity i Transaction Costs for farm i of plot u yj , x

flil

TAC yi , x

β

Bid adjustment

Ry,x

Rent paid for plot uy,x Average rent in region Transport costs between farmstead and plot Distance between plots

R TCy,z DIy,z

Notes: Bold letters denote vectors; a) 1 LU corresponds to approximately 500 kg live weight.


4

17

An abstract spatial model of the agricultural landscape

Land is an essential input to most agricultural production, and most livestock production depends on the availability of land. The availability of suitable land is therefore an important determinant of the potential for farm growth. Yet the more fragmented and dispersed fields are in a region, the lower the potential for profitable farm expansion. The size distribution of fields is also a defining characteristic of an agricultural landscape and its mosaic. Furthermore, land-use or the availability of suitable agricultural or semi-natural habitat is important for the conservation of biodiversity. Hence, considering space is important to obtain a realistic model of structural development and its impacts on the agricultural landscape. Space in AgriPoliS is represented by a 2-dimensional grid of equally-sized cells or plots. Each plot represents a standardised spatial entity of a specific size that can take different states such as soil type (e.g., arable or grassland). In this idealised representation, all aspects of the landscape not directly related to agriculture such as roads, rivers, lakes, etc., are subsumed as a single plot type: non-agricultural land. This abstraction is based on the assumption that only agricultural land use is affected by changes in agricultural policy, and thus all other features of the landscape remain unchanged. In addition, in this version, AgriPoliS does not model the location of farms and land as found in the real region, such as in a GIS that is used to obtain an explicit5 spatial representation of an area6. Instead, using the landscape calibration procedure developed in IDEMA, AgriPoliS is initialised with a abstract landscape based on the size distribution of agricultural blocks and non-agricultural land in the region. Thus, rather than producing an exact replication of the real landscape, a model of the real landscape is created. This approach captures some important characteristics of the actual landscape (field size distribution and fragmentation) while other characteristics are ignored (field shape). We refer to this as an abstract landscape modelling approach because we abstract from the complexities of reality, which is also consistent with the economic modelling approach adopted. The practice of using abstract or synthetic landscapes to study landscapes is fairly common (O'NEILL et al. 1992; WITH 1997). Modelling landscapes explicitly requires enormous 5

However, as we will describe later, the representation of the landscape is spatially explicit with respect to the mapping between farm production and land plots. 6 The functionality to initialise AgriPoliS with a spatially explicit landscape and an exact allocation of farms in the landscape is provided in the EU-funded project MEA-Scope (SSPE-5016)


18

amounts of computing power and data generated from landscape images. In many cases, the availability of both is restricted. Abstract landscapes are, however, quite capable of reproducing the statistical characteristics of real landscapes. Therefore, they are considered a valid proxy for landscape analysis (SAURA and MARTINEZ-MILLAN 2000; LI et al. 2004). If a synthetic landscape has statistical properties similar to a real landscape, then it represents an abstraction of the real landscape and can thus be used to answer questions related to the statistical properties of the landscape (e.g., mean field size and variation). In the EU, these landscape properties are available without major problems, as government authorities have detailed information on field sizes and the type of agricultural crops grown on fields from farmers’ applications for support (through e.g. the Integrated Administration and Control System, IACS). This information can then be used by AgriPoliS to initialise an abstract agricultural landscape that is a statistical, and hence abstract, representation of the real landscape. Five different landscape layers are used in AgriPoliS to represent the structure of agriculture and the landscape in a region; ownership, soil distribution, block, allocation and usage/field layers. The ownership layer denotes the ownership of a specific plot. The second layer reflects the distribution of different soils over the landscape. The block layer then replicates the distribution of contiguous areas of a soil type (e.g., arable land, grassland, etc.) reflecting the geophysical conditions in a region. Such a block is defined as a group of contiguous plots7 of the same soil type that are separated from other plots of the same type by either non-agricultural land or another soil type. The allocation layer represents the allocation of the geophysicallydefined blocks to farms. As it is also possible to allocate parts of a block to different farms at the allocation layer, we talk about sub-blocks, i.e., a subset of a block which is itself again a number of contiguous plots; these sub-blocks can either be rented or owned by the farm. The fifth layer reflects a farm’s land use, i.e., a field comprising a number of contiguous plots used by a particular farm for a particular land use (e.g., wheat). An example of such a landscape is given in Panel 1), which shows the basic representation of land in AgriPoliS as a two dimensional grid with cells of equal size - in this example as a 10x10 grid8. A cell defines the basic unit of land in AgriPoliS. Furthermore, this layer of the abstract landscape represents the ownership structure, where each plot has an owner. The second layer of the landscape as shown in Panel 2) reflects soil distribution in the landscape, where we allow for an arbitrary number of soil types, plus an additional soil type, “non7 8

Two plots are contiguous if they have an adjacent border. Note that in AgriPoliS, the landscape is represented as a torus to avoid border effects.


19

agricultural lands”, to subsume natural borders, roads, rivers, etc. In this example, there are two arable land types, grassland (green) and arable land (brown), and non-agricultural land (grey). 1) Ownership structure Single Plot u yj , x of an owner j as the basic unit of land with the size a

Non-agricultural land defining the natural borders of physical land blocks Land plots of distinct quality, e.g. arable land or grassland

2) Soil distribution

Distribution of physical land blocks reflecting key landscape statistics of the real landscape Sub-block sb kl ∈ SB l of a farm 3) Geophysically given land blocks

l consisting of one single or a number of contiguous plots reflecting the land allocation in a region

Usage structure resulting from a farm’s land use decisions. A field flikl ⊆ sb kl defines a contiguous area of land used for a particular activity (shaded areas) 4) Land allocation

5)Land use

Farm B

Farm A

Farm B Farm C

Farm C

Farm B

Farm B Farm A

Farm A Farm C

Source:

Farm A

Farm C

Own figure.

Figure 4-1: Abstract landscape representation in AgriPoliS In Panel 3), a thick red line delineates land blocks. The third layer represents the geophysical conditions of the abstract landscape. In this case, there are two grassland blocks and two arable land blocks. Block borders represent permanent boundaries and cannot change during


20

simulations (which is equivalent to assuming that soil type is defined by the geophysical characteristics of the region and cannot be altered by farmers, at least in the short run). The fourth layer as shown in Panel 4) represents the allocation of land (sub blocks) to farms in the region. All agricultural plots belong to one of the farms A, B or C. Note that plots comprising a single block can have different owners and that the allocation of a plot to a farm can change during simulations via the land market. Finally, Panel 5) illustrates the fifth layer of the chosen landscape representation with the usage structure of fields resulting from the farms’ land use decision. In the example, the usage structure of the arable (sub-) land blocks managed by Farm C result in C’s specific field structure. In this case, the farm endowment consists of three separate fields comprising 5, 6 and 9 contiguous plots, respectively. As the field structure is a result of a farm’s decision–making, it can change over time. The distribution of block sizes in AgriPoliS can be controlled by varying the landscape initialisation parameters (see BRADY and KELLERMANN 2005 for a more detailed description). We introduce two parameters, NON_AG_LAND and OVERSIZE which facilitate, in a twostage process, steering the share of non-agricultural land plots until the size distribution of contiguous plots generated by AgriPoliS has statistical characteristics similar to that of the real landscape. The basic procedure for initialising the landscape is as follows. Based on the selected farm sample, the necessary amount of land for each soil type is calculated. This is referred to as the “calculated amount of plots” for each soil type. The amount of nonagricultural land is given exogenously by the variable NON_AG_LAND. To influence the allocation of a specific type of plot to the farms, we define a buffer for each soil type, i.e., we create more plots of one type than is actually needed to initialise the landscape. This buffer is defined by the variable OVERSIZE, which is the ratio of initialised plots to the calculated amount of plots. In a first step, all defined soil-types, including non-agricultural land, are distributed over the grid according to the calculated amount of plots. In a second step, the farmstead of the defined farms are distributed on the grid at random. In an iterative process, each farm selects a plot that is most valuable to the farm9. This is repeated until every farm is equipped with its initial endowment of land. Afterwards, every plot which is still unused is set to non-agricultural land. The algorithm for this is summarized in Figure 4-2:

9

To define which plot is most valuable for a farm, the same decision rule applies as for the selection of plots during the land auction described in Section 5.3.4. If a farm just wants to minimise transportation costs, this is the plot closest to the farmstead.


21

Data : fi : i ∈ {1, . . . , NO FARMS} (set of all farms) NO OF SOILS, NON AG LAND, OVERSIZE grid (2-dimensional gird of cells) Result : grid (initialized abstract landscape) begin total land ← 0 land of typei ← 0, ∀i ∈ {1, . . . , NO SOIL TYPES} foreach i : i ∈ {0, . . . , NO SOIL TYPES} do foreach j : j ∈ {0, . . . , NO FAMRS} do land of typei ← total landi + fj .getInitialLandOfType(j) · OVERSIZEi foreach i : i ∈ {0, . . . , NO SOIL TYPES} do total land ← total land + land of typei total land ← total land + NON AG LAND grid.initialize(totall and) repeat plot ← grid.getRandomFreePlot(NON AG Land) plot.setSoilType(NON AG LAND) until grid.getLandOfType(NON AG LAND) ≤ NON AG LAND foreach i : i ∈ {0, . . . , NO SOIL TYPES} do repeat plot = grid.getRandomFreePlot(i) plot.setSoilType(i) until grid.getLandOfType(i) ≤ land of typei repeat until fi .getLandOfType(j) ≤ fi .getInitialLandOfType(j), ∀i, j foreach i : i ∈ {1, . . . , NO FARMS} do foreach j : j ∈ {1, . . . , NO SOIL TYPES} do fi .occupyFreePlotOfType(j) grid.setUnusedPlotsToNonAgLand() end Source:

Own figure.

Figure 4-2: Initialisation of the abstract landscape in pseudo code In Table 4-1 the main parameters to steer the landscape initialisation are summerized. Table 4-1: Table summarizing global landscape representation Name of the parameter NO_OF_SOILS OVERSIZEi NON_AG_LAND

Description Number of different soil types defined in a region Share of additional land initialised in the region Share of non agricultural land

Type

Range [1..∞]

Default 2

ℝ

[0..∞]

1.1

ℝ

[0..∞]

0.15

ℝ


5

22

Factor inputs and production activities

5.1 Production factors Production factors in AgriPoliS primarily concern the classical production factors land, labour and capital, whereby the factor capital includes both money and assets for production. 5.1.1

Land

As described above, spatial representation in AgriPoliS is organised by way of equal-sized cells called plots (class Plot). Taken together, the plots make up the entire region (class Region). Plots differ on three aspects: land quality, usage structure and ownership. Land of either quality is assumed to be homogeneous. Regarding the usage structure, area utilised for agriculture is classified as either managed land or abandoned land. And finally, at the outset of the model, area utilised for agricultural is either owned by farm agents or rented. All land not owned by farm agents is assumed to belong to external land owners, which, in the current version of AgriPoliS, are explicitly modelled, albeit in a rudimentary manner (cf. section 5.3.5 on transaction costs). Individual plots in AgriPoliS are characterised by a number of attributes defining the plot's state, its location on the grid of plots, and its location relative to the location of the farm interested in renting the plot or the farm agent already managing the plot. A plot of either land quality can take different states: no agricultural use, abandoned land currently not managed, the respective soil type, plot rented by farm agent k, plot is farmstead, plot is owned by farm agent k. 5.1.2

Labour

Labour is supplied in three forms (class Labour). The first is labour supplied by the farm family. The amount of farm family labour is derived from accountancy data and is expressed in labour units.10 Furthermore, farms can hire additional workers either on a fixed contract or an hourly basis. Hiring fixed labour is treated as an investment for a period of one year. The total labour capacity is determined in the mixed-integer programme, where variable labour and fixed labour are activities. In addition to hiring labour, farms can also offer their own farm family labour on the labour market. This offers the possibility of non-professional farming, on the one hand, and reducing the overall amount of farm labour, if necessary, on the other. Corresponding to the hiring of 10

One labour unit corresponds to the annual labour input per hour provided by one worker.


23

labour, fixed and variable off-farm labour activities are introduced as activities in the mixedinteger programme. 5.1.3

Capital

To produce, a farm agent needs capital, both in the form of liquid funds to pay running costs, and in the form of fixed-asset capital (investments), which determines a farm agent's productive capacity. Investments are introduced into AgriPoliS through an investment catalogue (class InvestList), which depicts a list of investment objects containing investment possibilities and production technologies typical for the region under investigation. The investment catalogue is available to all farm agents and provides the basis for farm agents’ investment decisions. The individual objects in the catalogue differ with respect to the type of investment (e.g. dairy, fattening pigs, machinery), as well as the size of the investment, which is reflected in the production capacity. For each type of investment, the catalogue contains a variety of sizes. Differently-sized objects affect a farm agent in three ways: first, the effect of a larger scale of production is reflected in lower average annual unit costs when compared to an object of the same type, but of smaller size. Second, larger investments are also considered to have lower labour requirements relative to smaller investments. Third, the technology underlying investment objects is assumed to improve over time, whereby larger investment objects are assumed to be technologically more advanced. Although technological change is not modelled explicitly by changing the technical coefficients of production, AgriPoliS nevertheless aims to mimic two effects of technically more advanced production technologies. On the one hand, AgriPoliS assumes that with every new investment, unit production costs of the product produced with this investment decrease. The extent of this cost-saving effect depends on the technical standard of the investment (see Section 6.1.4 on cost expectations). Stated more formally, each investment object I h ,i ,d (h=1,…,H) for producing product i (i=1,…,I) is defined by the set of attributes in Table 5-1. In particular, this is the investment's type d (d=1,…,D), investment costs, production capacity, maximum useful life, labour sub-


24

stitution in hours, maintenance costs, and a factor representing the impact of technological change.11 Maintenance costs are expressed as a percentage of total investment costs. Table 5-1: Investment attributes Investment attributes ID-number Type of investment (d) Investment costs (€) Production capacity (heads or hectares) Maximum useful life (periods) Labour substitution (hours) Maintenance cost (% of investment costs) Technological change factor (%)

-

The maximum time that an investment can be used in production is given by its useful life. Also, no investment object can be sold before it has reached its maximum useful life. Accordingly, an object's salvage value at the end of its useful life is zero, such that it is non-tradable. This particular assumption has important consequences for the decision-making of farm agents because it implies that investment costs are fully sunk once an investment is made. Because of this, depreciations are not variable and are treated as fixed costs in every case. Capital required for production and investments is considered in three forms: short-term credit, long-term credit, and liquid equity capital.12 Short-term credit is taken up by farms in the case of short-term liquidity shortages. The amount of short-term credit is not explicitly limited, but interest is higher than for long-term credit, which therefore sets a kind of natural limit for borrowing in the short-term. Long-term borrowed capital can be used to partially finance investments. It is assumed that a maximum share (1 − v) of investment costs is partially-financed with borrowed capital, with

the remaining share v representing the equity-financed share. Borrowed capital for investment is supplied by an annuity credit that runs for the entire useful life of the investment. The maximum amount of borrowed capital is also not directly restricted. Nevertheless, it is assumed that a farm only invests if the equity-financed share of total investment costs does not exceed a minimum equity reserve threshold value MR given by:

11 12

For greater clarity, subscripts i, and d will be omitted in the following. Liquid equity capital is defined as total equity, minus land assets, minus equity bound in asset capital.


L

L

∑ (v ⋅ A ) ≤ L + 0.7 ⋅ LA + 0.3 ⋅ ∑ A l =1

l

l =1

ec ,l

with L = ECt −1 − LA −

25

L

∑A l =1

ec ,l

.

(1)

That is, there is a limit on the maximum equity capital that can be used for investment. The limit is introduced to prevent putting the substance of the farm at risk.13

5.2 Production activities Production activities in AgriPoliS are distinguished into livestock production (e.g. fattening pigs, turkeys), plant production activities (e.g. crops, sugar beets, grassland), short-term capital activities (e.g. short-term borrowing), short-term labour activities (e.g. short-term hiring), and 'additional' activities. Most livestock and plant production activities consist of the production of marketable products. Exceptions are grassland production activities and silage maize, which serve as intermediate products for livestock production. Additional activities relate to those activities besides capital and labour which are needed to balance capacities in the shortrun. This includes, for example, manure disposal, machinery contracting, or milk quota lease. Similar to investment objects, each individual production activity is characterised by a set of attributes (Table 5-2). Table 5-2: Product attributes Product attributes

ID-number Production branch (e.g. sows for breeding, dairy production) Product produced with investment Io of type d Price (€/unit) Variable unit production costs (€/unit) Price flexibility Price trend (% change per period) Support payment (direct payment) (€/unit) In the simulation, products are managed by the farm agents in a product list that keeps track of the total units produced as well as the gross margins associated with each product. Product prices change in response to developments on product markets (see Section 0). Variable unit production costs are affected by both technological change, and by the individual managerial ability of a farm agent. 13

This means that 70% of land assets LA and 30% of the total equity-financed fixed assets have to be covered by total equity capital EC t −1 at all times.


26

5.3 Factor and product markets Markets for production factors and production outputs are the main instance of interaction between farms. In AgriPoliS, there are three groups of markets: national and EU-wide, regional, regional and spatially organized. The former subsumes markets for products like crops, oil seeds, sugar, capital, labour and services. Products traded on regional markets differ from region to region and depend on the specific characteristic of the region. At the regional scale, these are, e.g. a market for manure (Germany, France), a market for calves (Sweden, Västerbotton and Jönköping), or a market for milk quota (all regions). Furthermore, we consider regional and spatially-organised markets where we do not assume that goods are traded on spot markets. In our case, this applies to the land market. 5.3.1

General market rule

Generally, prices on markets for inputs and outputs are derived from the market rule given in equation (2). The price function comprises two main elements. The first is a general price trend reflecting changes in macroeconomic conditions such as increasing costs of labour. The second element is the price formation itself. Here, we assume neither a fully elastic nor a fully static demand or supply. Hence, similar to the function for gross margins developed by Balmann (1995), it is assumed that for most products i the price pi in period t is the function ⎛ ∑ xik,t −( t +1) ⎜ k ⋅⎜ p i ,t = p i , 0 ⋅ γ i k ⎜ ∑ Zt ⎝ k

⎞ ⎟ ⎟ ⎟ ⎠

− bi

with k = 1,K, K ,

(2)

where pi ,0 denotes the initial price of product i at the outset of the simulation (period t=0), the coefficient γ i controls for a price trend over time, xik,t denotes the activity level of a farm k in period t and Z tk is the land input of farm k in period t, so that the last term allows for price variation depending on the cumulative quantities produced by K farm agents. The parameter

b i ,t represents price flexibility, which is equivalent to the inverse demand/supply elasticity (cf. BALMANN 1995). As we are predominantly modelling small regions, the expected effect on prices should be marginal. From this it follows that in almost all cases, we assume the price flexibility to be zero.

The Agricultural Policy Simulator – AgriPoliS 5.3.2

27

Regional markets

Depending on specific regional characteristics and conditions, demand and supply for some products is explicitly modelled. We assume that these products are traded at a regional scale, for example, due to high transportation costs. Accordingly, we have to ensure that there is a balance between demand and supply at the regional level. Examples for this type of product are e.g. markets for calves (Sweden), manure (Germany, France) or milk quota (all regions). Furthermore, we consider cross-price effects between some products, such as for piglet production and fattening pigs (all regions). In general, the price adjustment on regional markets follows a simple Tâtonnement process, where we assume the price pi ,t +1 to be a function of excess demand z i ,t = x d ,t − x s ,t in period

t , whereas s and d reflect the supply and demand activities of a product i, respectively. We assume that prices change proportional to the ratio of excess demand to the total supply in the market with a proportionality factor β i : ⎛ z i ,t ⎞ ⎟. pi ,t +1 = pi ,t ⋅ ⎜1 + β i ⎟ ⎜ x s ,t ⎠ ⎝

(3)

For the proportionality factor, we choose β = 1 , assuming a nearly elastic reaction14. Milk quota: For milk quota, this rule is modified. Since the year 2000, in many regions (e.g. in Germany), prices for milk quota have been determined by regional quota exchange markets, where offers and bids are collected and on a periodic basis, market clearing prices are determined. In many cases, these markets are limited to closed regions. Accordingly, we assume a total reference quota ( REFQUOTA ) for each region. Therefore, excess demand is calculated by z milk ,t = REFQUOTA − x smilk ,t . In contrast to the real quota market, the quota market as implemented in the model resembles a quota leasing market.

14

However, we did not derive this assumption from empirical data; instead, it is based on expert knowledge.


28

Piglet production: Piglets are assumed to be used as intermediate inputs in fattening pig production. Hence, the total quantity of piglets produced in a region is reduced by the quantity of piglets used for fattening pig production.15 5.3.3

Exogenous price projections

As we follow a regional modelling approach, we assume that adjustments in farms’ production patterns do not necessarily lead to changes in commodity prices. This is why most commodity prices are treated exogenously in the model. But on the EU-level, adjustment reactions to reform packages such as the 2003 CAP reform are expected to result in potentially substantial price reactions. Neglecting these price reactions would lead to biased model results. Moreover, it would hinder a common discussion of results with models at higher levels of aggregation (e.g. with ESIM in the IDEMA project). Therefore, we created an interface to introduce price projections from the partial equilibrium model ESIM. To minimise problems due to different calibrations, we consider relative prices changes for every product and year ( cti ,t ). 5.3.4

Regional and spatially-organised markets

Land markets play a crucial role in agricultural structural change. Because the dynamics on land markets are mainly determined by the interactions between individual farms, an agentbased approach offers many advantages for creating an endogenous land market. The first version of the land market presented here dates back to the work of BALMANN (1997), HAPPE (2004) and HAPPE et al. (2006). A literature review reveals few similar attempts. For example, the model by BERGER (2001) includes a land market which is of a similar structure to the approach presented here, and is also based on BALMANN (1997). Other models such as FEARLUS (cf. POLHILL et al. 2001) also include a land market, although the price formation for land is not endogenous. Rather, the model uses fixed exogenously-given prices for land plots. The land market used in AgriPoliS serves as the institutional framework for exchanging land between individual farms. Although farms are initialised with both owned and rented land, transactions on the land market take place exclusively via renting activities. Basically, there are two reasons for this decision: first, we assume that the capitalised rent equals the sales price of a unit of land. Even though this equivalence could not be observed in reality, the difference in capitalised rents and sales prices is often explained by determinants such as reli15

At the current developmental stage, there is no interdependence between the price of piglets and the gross margin of pig fattening.


29

ability, stability, eligibility for use as collateral or taxation reasons. As all of the mentioned determinants are beyond the scope of the model, this simplifying assumption seems acceptable. The second reason for the decision to focus on rental markets is that in many countries, land transactions take place on land rental markets.. This is especially true for EU-10 countries (cf. SWINNEN and VRANKEN. 2006) such as the CZ, where in 2003, 89 % of the land was rented (MZE 2004) and the majority of land transactions still take place on the rental market; but it also holds for the other case study regions (LATRUFFE and MOUEL, 2006). AgriPoliS uses an auction mechanism to allocate land. To support the specific auction mechanism we have chosen, we would first like to provide a brief overview of possible mechanisms for the land allocation process (for an introduction into auction theory, see KLEMPERER 1999). The four basic types of auction protocols are the English and the Dutch auctions as oral auctions, and the first-price sealed-bid and second-price sealed-bid (Vickery) auctions that represent closed auctions. Open auctions take place at a common marketplace and a certain point in time. The amount of a bid is public. For land transactions, these characteristics do not hold. In land transactions, we assume a formal (e.g. through advertisement or real estate agent) or informal (direct/indirect conversation) announcement. Each actor interested in the land plot communicates a bid to the land owner which is not public. To mimic this situation, we restrict the possible options to closed auction protocols. Within the group of closed auctions there is the choice between a first-price and a Vickery auction. In a Vickery auction, the bidder with the highest bid (the winner) pays the price of the second-highest bid, whereas in the first-price auction the winner has to pay its bid. From the perspective of a bidder, there is a strategic difference between the two forms of auction. In a Vickery auction, the decision for a bid is independent of the other bidders and bidding the true valuation is the dominant strategy. In a firstprice auction, this is not true. There, a bidder has to form an expectation about the valuation of the other bidders. Based on these expectations, the bid for a good has to be lower than one’s own valuation. From a modelling perspective, this can be used, given a fixed bidding strategy, to assume different strategic capabilities of the agents. In the case of a Vikery auction, this implicitly allows the bidder to anticipate the valuation of the other bidders. The model’s supply of land is given by three sources: 1) Land which was previously unused. 2) Land which is free because either the rental contract has ended, or a farm manager terminated the contract, depending on the used contract type. 3) Land which is available because a farm has quit.


30

Data : fi , i ∈ {1, . . . , NO FARMS} (set of all farms) NO SOILS TYPES (number of soil types) grid (2-dimensional gird of cells) Result : grid, fi (resulting land allocation) begin repeat foreach i : i ∈ {1, . . . , NO SOIL TYPES} do if grid.getFreeLandOfType(i) > 0 then maxoffer ← 0 (highest bid for a plot) secondoffer ← 0 (second highest bid for a plot) equalbidder ← ∅ (set of bidders with equal bids) foreach j : j ∈ {1, . . . , N O F ARM S} do fj .searchPlot() offer ← fj .formulateBid() if offer > maxoffer then equalbidder ← ∅ equalbidder ← equalbidder ∪ {j} secondoffer ← maxoffer maxoffer ← offer else if offer = maxoffer then equalbidder ← equalbidder ∪ {j} foreach k : k ∈ equalbidder do fk .occupyWantedPlot() if AUCTION TYPE = 0 then fk .setPriceOfWantedPlot(maxoffer) else fk .setPriceOfWantedPlot(secondoffer) until ∃i ∈ {1, . . . , NO SOIL TYPES} : bidi > 0 end Source: Own figure.

Figure 5-1: Procedure of land allocation in pseudo code

To allocate these set of plots, the land market is designed as a sequential auction, i.e., it is possible to bid for only one plot at a time. This means that the auctions for a single plot are repeated until all plots are allocated or there are no further positive bids. To avoid a first mover advantage, during each auction turn, a farm first selects a plot which is most valuable to the farm16 and then calculates the corresponding bid.17 That is, the winning farm receives the plot with the highest valuation for the farm. As described in Section 2.3, in the current 16

In the basic setting of the model, the plot with the highest valuation for a farm is the plot with minimum transport costs, i.e., the free plot closest to the farm stead. However, as we describe in Section 5.3.5, we may also consider plot-specific transaction costs or scale effects resulting from the merging of neighbouring plots. This can alter the valuation of plots compared to a situation where we only consider transport costs. 17 Organising the auction in a way that each farm is asked to bid for a specific plot would create a first mover advantage, resulting from the random selection of free plots to be offered.


31

version of AgriPoliS, it is possible to consider an arbitrary number of soil types. Because of the indivisibility of investments, the shadow price function for land is discontinuous and most notably non-monotone. Scale and size effects, together with the fact that some production activities rely on the presence of different soil types (e.g. for dairy production, both arable land and grassland is necessary for fodder production) cause complementarities between plots, and especially between plots of different soil types. To reduce these complementarities, the auction process alternates between the defined soil types. The procedure is summarised in Figure 5-1. To complete the picture of the land market model used in AgriPoliS, we describe the two types of rental contracts considered in AgriPoliS. The first is a rental contract with a fixed duration. After a plot is allocated to a farm, the contract length is randomly drawn from a uniform distribution with a minimum and maximum contract length. The contract length is binding for the duration of the contract, and neither the land owner nor the land manager can terminate or renegotiate the rental contract during the contractual period. In contrast, in the second type of contract, we give more contractual power to the land manager. Instead of having a fixed contract length, the plots can be “renegotiated” at the end of each production period. This is implemented in the following way: at the end of a production period for each rented plot, a farm decides to either keep the plot or to release it. The decision rule itself is based on the expected revenue of a plot in the next production period and is described in more detail in Section 6.2.1. In contrast to the farmer, the landowner cannot terminate the rental contract. Although the first type of contract seems more natural, we introduced this second type of contract to take into account the specific situation in some case study regions, especially those in which rental contracts are less specified and short-term contracts prevail (HURRELMANN, 2005) or when there are special regulations (e.g. in France the SAFER (Société d’Aménagement Foncier et d’Etablissement Rural), a private body with public service missions, sets each year a maximum level for rental prices). Moreover, one may argue that the second type of contract provides some market power to the land manager. Yet this argument is put into perspective by the fact that in the case of the second contractual type, rental prices for new and already exiting contracts are adjusted from one period to the next, allowing some kind of contractual “renegotiation” which otherwise cannot be terminated by the land owner.


32

This “renegotiation” is carried out in a two-stage process. In the first stage, the rental prices for new rental contracts are adjusted towards the average rent paid in the region, and in the second stage, the price for already existing contracts is adjusted towards the regional average. For adjusted rent R ynew , z we use a weighted geometric average of the original paid rent R y , z and the regional average R as it is given in (4). λ (1− λ ) R ynew ,z = R ⋅ R y,z

(4)

In case of new rental contracts, λ is specified as 0.5. Prices for new rental contracts are adjusted because of shadow prices, and thus the bids for a plot of land can vary significantly between farms. By “pushing” the prices – especially those of extreme outliers – towards the regional average, we introduce some kind of market transparency, as it would be unrealistic in a real market setting to enforce prices which are very far away from the market price. For already existing rental contracts, we assume that they are adjusted, for example in the case of strong product price changes, policy changes, or changes in the regional rental reference. Such provisions are common in real contractual rental agreements, and hence mimicked in AgriPoliS. Again, we use the weighted geometric average of the average rent in the region and the previous rent of the plot according to (4) to adjust the rent for old contracts, whereby the weight λ is given by the share of newly-rented land in the entire region. In Figure 5-2, the relationship between the deviation of a rental price from the regional average, the parameter λ and the resulting adjustment is displayed.


33

25 20 15

Adjustment %

10 5 0 -30

-25

-20

-15

-10

-5

0

5

10

15

20

25

30

-5 -10 -15 Deviation from regional average (%)

Source:

Own figure.

Figure 5-2: Rent adjustment of new and old rental contracts 5.3.5

Transaction costs on land markets

In the version of AgriPoliS presented in HAPPE (2004), the model was calibrated to an intensive livestock region in the south-east of Germany. With respect to the conditions given in that region, it was reasonable to assume well-functioning markets for products and production factors. However, in the current version of AgriPoliS, the model is also calibrated to three regions in the EU-10, where even after some years of transition, markets, especially markets for land, are still far from being smooth and formal. There is much evidence that market imperfections on land markets have a major influence on structural development in these countries (DALE and BALDWIN 1999; LERMAN, CSAKI and FEDER 2004). Imperfections are mainly due to high transaction costs (TAC), and the market power of large corporations given by an often prevailing dual farm structure or fragmented ownership structure. In order to explicitly take the particular situation in these countries into account, we introduce a transaction cost framework into the model. The framework itself is derived from a theoretical analysis by CIAIAN and SWINNEN (2006). Like CIAIAN and SWINNEN, we basically differentiate three types of transaction costs connected with the land transactions: those which are specific i) to the plot, ii) to the owner, and iii) to the user.


34

The costs specific to the plot (i) mainly result from unclear or missing boundaries in terrain, co-ownership of land, and the related problems of unknown ownership and inaccessibility of plots. Transaction costs related to the owner (ii) occur due to asymmetric information between the land owner and the manager of the farm who uses the land. The managers have better knowledge about the economic situation of the farm and about regulations concerning land transactions than the land owner. Transaction costs of this type are especially relevant if land owners do not live where their land is located or if they have no knowledge about agriculture (CIAIAN and SWINNEn 2006). Costs related to the user (iii) are a result of land fragmentation. For example, large fields cultivated by corporate farms often comprise numerous (tens or hundreds) individual land plots, and the withdrawal of one plot would split up the entire field. To avoid this, it has been common practice that operators give back substitute plots to landlords instead of the plot with the actual ownership title. However, these substitute plots could be of lower soil quality or further away from the farmstead than the original one, which creates additional costs for the new tenant. Furthermore, for the implementation of the TAC framework, we assume that the plot-specific portion is dependent on the field size distribution in the analysed region, as it is very likely that TAC increases with the size of the field in which a plot is embedded, e.g. due to unclear boundary conditions. Therefore, we utilise the synthetic landscape introduced in Section 4, where we mimic the real landscape in a way that some statistical properties of the real landscape are reflected in the landscape generated by AgriPoliS. As described in Section 4, we distinguish five different layers to represent a region’s agricultural structure and landscape. With these prerequisites it is possible to incorporate owner-specific and plot-specific transaction costs. At the same time, plot-specific costs could be modelled dependent on the size of the land block in which a plot is embedded. Let u yj, x ∈ PLi be a specific plot of a farm i in the previous production period. Following Ciaian and Swinnen (2006), a farm’s TAC yi , x if it wants to rent an additional plot u yj , x of owner j are a function of plot and owner-specific transaction costs.


TAC yi , x

⎧0, p yj , x ∈ PLi ∧ ( LT i = LT ∨ TAC _ ONLY _ LT ) ⎪ = ⎨ g yj , x G ij ⎪ a j + A j , else ⎩ y,x

35

(5)

G ij is defined as the TAC specific to the relationship between the owner j and a farm manager i . A j is the total area owned by owner j . The variable g yj , x are the transaction costs specific to plot p yj , x of owner j and a y , x is the size of this plot. In case a plot was already rented by a farm, no transaction costs occur. In order to consider the situation that transaction costs only occur when a land owner wants to withdraw land from a farm with a specific legal type, LT (e.g. a cooperate farm) we introduce a variable TAC _ ONLY _ LT ∈ {true, false} . If this variable is true, transaction costs are only taken into account if the plot was previously rented by a farm with the legal type LT . The legal type of a farm is given by LT i ∈ {CF , IF }, whereas in the current version of AgriPoliS, a farm can either be a cooperate farm ( CF ) or an individual farm ( IF ). This framework allows us to model a wide range of typical settings. KELLERMANN et al. (2006) apply this framework when modelling the restructuring process between cooperate and individual farms in Czech Republic. For this application we assumed that when an individual farm wants to withdraw land from a cooperate farm, it at least has to pay the shadow price of the cooperate farm, plus the TAC TAC yj, x associated with a plot in order to be able to withdraw the plot. Therefore, we argued that G ij , A j and a yj , x is constant, i.e., G ij = G , A j = A and a yj , x = a . This was done to meet the assumption we made above for the first layer of our landscape, i.e., that plot size distribution is the same for all landowners and land ownership is distributed equally among land owners (in fact, every owner holds exactly one plot of the same size). Because G is equal for all landowners, the owner-specific part of the TAC is constant for all plots. Furthermore, we assume that g yj , x is dependent on the size of the sub-block

sb i of the farm i where the plot p yj , x of owner j is embedded. First of all we introduce the possibility that there is a linear relationship between g yj , x of plot p yj , x and the size of sb i . In this case, g yj , x is given by g yj , x = Sizeof ( sb i ) ⋅ tcv , where tc v is a constant used to adjust the variable part of the transaction cost. For analysing the farm restructuring process between co-


36

operate and individual farms, we furthermore assumed that TAC only occurs if the plot was previously rented by a cooperate farm. At the end of this section Table 5-3 summarises important global parameters related to products and factor markets. Table 5-3: Table summarizing global product- and factor market related settings Name of the parameter Capital v

Description

Type

Equity finance share

ℝ

Variable which selects between first- and second-price auction

ℕ

CONTRACT_TYPE

Variable which selects between different contract types

ℝ

λ

Adjustment weight for new rental contracts Duration of the rental contracts integer

ℝ

Consider transaction costs dependent on legal form of a farm Legal for to be considered for transaction costs Constant which defines the variable part of transaction cost in dependent of the size of a subblock

Land AUCTION_TYPE

MIN_CONTRACT_LENGTH, MAX_CONTRACT_LENGTH Transaction costs TAC_ONLY_LT LT tcv

6

Range

Default

0: first-price auction 1: secondprice auction 0: fixed contract length 1: renegotiation [0..1]

0

0

ℕ

[1..∞]

9, 18

boolean

true, false

true

enum

CF, IF

CF

ℝ

[0..∞]

0.5

The farm agent18

To characterise the farm agent, it is useful to first describe why farm agents do what they do and based on what. That is, this section will first describe a farm agent's behaviour and the goal of its actions before describing the farm agent's actions themselves. Farm agents can produce a selection of goods. In order to produce, farm agents utilise buildings, machinery, and facilities of various types and capacities. Accordingly, AgriPoliS implements economies of size: as size of production increases, unit investment costs decrease. Moreover, labour is assumed to be used more effectively with increasing size. AgriPoliS also aims to mimic the effect of technological progress. More specifically, it is assumed that with 18

In this section, subscript k is omitted to increase clarity. All formulae concern one farm agent only.


37

every new investment, unit costs of the product produced with this investment decrease by a certain percentage. Farms can engage in land rental activities, production quotas, and manure disposal rights. Labour can be hired on a fixed or a per-hour basis, and vice versa, farm family labour can be offered for off-farm employment. To finance farm activities and to balance short-term liquidity shortages, farm agents can take up long-term and/or short-term credit. Liquid assets not used within the farm can be kept in a bank. Farm agents quit production and withdraw from the sector if equity capital is zero, the farm is illiquid, or if opportunity costs of farm-owned production factors are not covered.19 Farm agents are assumed to act autonomously and to maximise farm household income. For this, production and investment decisions are made simultaneously based on a recursive linear programme including integer activities (c.f. HAZELL and NORTON 1986). From the solution of the linear programme, production factors’ shadow prices can be derived. Farm decisionmaking is myopic or boundedly rational, that is, agents make decisions based on the information available to them, which can possibly even be wrong (SIMON 1955). Because of this, the decision problem of the model farms is highly simplified compared to that of real farmers in that strategic aspects are not included. Except for the price information on rents, as well as product and input prices, individual farms in AgriPoliS do not know about other farms' production decisions, factor endowments, size, etc. On the contrary, unbounded rationality would imply that farms take account of all interactions between farms, plus the technical and political framework conditions present now and in future periods, and would include these into the individual decision problem. Farm agents are also boundedly rational with respect to expectations. In the majority of cases, farm agents follow adaptive expectations. Policy changes alone are anticipated one period in advance and included into the decision making process. New investments affect production capacities for the operating lifetime of the investment. This implies that investment costs are sunk. A farm agent is handed over to the next generation after a given number of periods. In case of such a generational change, opportunity costs of labour increase. Accordingly, the continuation of farming can be interpreted as an investment into either agricultural or non-agricultural training. Finally, farm agents differ not only with respect to their specialisation, farm size, factor endowment and production technology, but also with respect to the identity of the farmer, and with respect to managerial ability. 19

As investment costs are assumed to be sunk, only opportunity costs for land and labour are considered.


38

6.1 Behavioural foundation 6.1.1

Farm planning

To model farm behaviour, it is necessary to make assumptions about goals, expectations, managerial ability, and the variety of actions that a farm agent can pursue. As for objectives, we assume that farms in AgriPoliS follow a profit maximisation calculus.20 The action space given to farm family members is defined by: on-farm factor endowments (land, labour, fixed assets, liquidity); the market situation for production factors and products; the vintage of existing fixed assets; technical production conditions; overall economic framework conditions (work opportunities outside the farm, interest rate levels, access to credit); and the political framework conditions. According to this action space, the farm’s objectives and the expectations, a farm performs an optimisation. In more formal terms, the objective function and restrictions can be expressed as (abbreviations are given in 3.3)

max Y e (x, p e , c, A, I, r, MP, D, RE, L, BC , IC ,K) with Y e = x′(p e − c) + IR + S + W − RE − MC − D − OV − TC − IC − HW s.t.

b ≥ x ′r

with

r = (r1 , K , rI , K , rH , K , rJ )

x≥0

6.1.2

(6)

.

General remarks about expectation formation

Production planning, investment, and the decision to continue or quit farming is based on expectations about future developments of prices, costs, technologies, investment possibilities, and policies. In AgriPoliS, farm agents can form short-term expectations about the next planning period. However, farm agents are not capable of forming long-term expectations. With respect to all other future periods, they expect prices and costs to remain constant.21 Thus, dy20

Depending on the legal type of the farm, this could either be to maximise the household income (in case of individual farms) or profit maximisation. Other goals, e.g. to retain employment, which are perhaps relevant for other organisational forms, are not considered here. 21 This assumption has some implications, in particular for investment activity, because farm agents make long-term investment decisions based on short-term expectations. If farm agents would be able to articulate medium- or long-term expectations, some investments probably would not be made. The introduction of long-term expectations may be desirable, but is currently limited by practical considerations. Indeed, it appears to be particularly difficult to consider short-term and long-term expectations simultaneously. The problem would be even


39

namic effects resulting from expectations about the development of markets and demand developments are neglected. Farm agents also follow the same pattern of expectation formation, i.e., there is no differentiation between optimists and pessimists. 6.1.3

Price expectations

Regarding prices, a farm agent follows adaptive expectations defined in terms of the weighted geometric average of actual and expected prices.22 A farm agent bases all planning decisions on expected prices, because actual prices are only determined at the end of a production period as a result of farm activity. The expected price of production activity i in period t + 1 is determined as α

pie,t +1 = ( pi ,t ⋅ pie,t

(1−α )

−1

) ⋅ γ i with 0 ≤ α ≤ 1 and p > 0 for i = 1,K, I

(7)

The coefficient γ controls the price trend of production activity i , whereby prices increase (decrease) if γ < 1 ( γ > 1 ). In AgriPoliS, the actual price and the expected price in period t are equally weighted, i.e., α = 0.5 . 6.1.4

Cost expectations

A farm agent also forms expectations about production costs. With regard to cost expectations, livestock and plant production activities are treated differently from additional production activities. For the group of additional production activities, a farm agent forms cost expectations in the same manner of price expectations, but without the price trend. Accordingly, expected costs of additional production activities are calculated as the weighted geometric average with equal weights: α

cie,t +1 = ci ,t ⋅ cie,t

(1−α )

with 0 ≤ α ≤ 1

for i = 1,K, I .

(8)

Cost expectations for livestock and plant production activities are determined in a different way in order to introduce the cost-saving impact of technologically more advanced production

more complex if expectations were also made with respect to the behaviour of other farm agents. 22 Unlike the more common definition as the weighted arithmetic mean, the chosen definition tones down expectations for period t+1 if expected prices and actual prices in period t differ (cf. BALMANN 1995).


40

technologies (see discussion on technological change). With respect to this, it is necessary to distinguish between plant production activities and livestock production activities. As mentioned above, it is assumed that the technological standard of production technology improves over time. Thus, with every new investment into livestock production, the expected production costs cie,t +1 of livestock production activity i produced with investment object I are computed as cie,t +1 = ci ,t − f o ,i ⋅ ci ,t with 0 ≤ l < 1

for i = 1, K , I ,

(9)

whereby factor f represents the size of the investment, and is higher for larger investments. On the subject of plant production activities, cost savings can only be realised as a combination of larger machinery together with larger field sizes.23 Expected costs of plant production activities cie,t +1 are thus a function cie,t +1 = ci ,0 − l ⋅ ci ,0 with 0 ≤ l < 1

for i = 1,K, I

(10)

of costs at the outset of the simulation in period t=0, adjusted by factor l , which is a function of the average number of adjacent plots and the size of the farm. The factor l thus captures the effect of larger field sizes and is defined as ⎤ 0.15 ⎤ ⎡ 0.45 ⎡ ⋅ ⎢1 − lt = ⋅⎢1 − ⎥. ⎥ ⎣ 1 + 100 / Z ⎦ ⎣ 1 + 100 /(10 ⋅ APt + 1) ⎦

(11)

Figure 6-1 shows values of l for different farm sizes and average numbers of adjacent plots. Accordingly, a farm agent with an initially small amount of scattered land can realise large cost savings if it considerably increases its acreage. The potential cost effect is much lower if a farm agent's acreage is already high and if the plots are nearby.

23

KUHLMANN and BERG (2002) quantify the cost difference between a 1 ha plot and one of 60 ha at 250 €/ha, which corresponds to about a third of the current revenue for wheat.


41

0.3

8 7 6 5 4 3

l-factor

0.25 0.2

2 1

0.15

0

0.1

number of adjacent plots

0.35

0.05 0 0

50

100

150

200

250

300

350

400

Farm size [ha] Source:

Own figure.

Figure 6-1: Expected cost savings for machinery investments, depending on farm size and the average number of adjacent plots 6.1.5

Expected profit/household income and opportunity costs

After each period, every farm decides whether it stays in agriculture or not. This decision is based on the calculation of the expected farm-household income/profit and the calculation of the opportunity costs for owned factors. The expected farm-household income (or profit in case of corporate farms) is calculated for the next period by calculating the expected total gross margin from agriculture with a mixed integer programming model (MIP) based on the farm's current factor endowment. Thereby, increasing costs for hired labour and price changes for agricultural outputs caused by policy changes and possible investments in new assets are considered. The total gross margin is then reduced according to Table 6-1.


42

Table 6-1: Calculation of expected income and opportunity costs Expected farm-household income

total gross margin agriculture + interests on liquid capital at bank + off-farm income + subsidies - interests on long-term debts - depreciation of fixed assets - wages paid - transport costs to plots - rent expenditures

Opportunity costs Input factor family labour + liquid capital + owned land + milk quota - interests on long-term debts - depreciation of fixed assets

valued at off-farm income long-term savings rate average regional rent quota leasing price

For total opportunity costs of owned factors we consider costs for family labour, working capital, owned land and owned milk quota. Costs for fixed assets are not considered as we assume these costs to be sunk. Family labour is valued at the costs for hired labour, liquid capital at the long-term savings rate, which is assumed to be equal to the long-term interest costs. Milk quota is valued at the quota leasing price and owned land at the expected average rental price REs = Rs ⋅ λ s of the simulated region, which is endogenously determined. We assume that this expected rental price for a soil type s comprises of the average rental price in a region Rs and a factor λ s which expresses the expected rental price change for the next period. This leads to opportunity costs for land OCOL given by: OCOL = ∑ [OLs ⋅ R s ⋅ λs ] , S

(12)

s =1

whereas OLs is the owned land of a farm of soil type s To consider rental price changes also in the in the calculation of the expected household income the expected rent expenditures REE are calculated accordingly:

REE =

∑ [RL ⋅ ( R E + R ) / 2 ⋅ λ ] , S

s

s

s

s

s =1

whereas RLs is the rented land of a farm of soil type s

(13)


43

As we in general assume that farms act myopic this means that for farms the expected rental price equals the current rental price in the region, i.e. we assume λ s equals 1.Only in case of major policy changes we deviate from that rule as explained in 6.1.6. 6.1.6

Expectations about policy changes

When forming expectations about the next planning period, policy changes have to be taken into account as well, particularly if changes are expected to be striking. It is assumed that a farm agent knows about major policy changes one period before the policy becomes effective. This influences decision-making primarily when it comes to evaluating the farm agent's profitability at the end of a planning period. In the previous paragraph we showed how both the expected household income/profit and opportunity costs are influenced by changes in rental prices. However, our basic rule is that farms assume that the rental price in the next period equals the current rental price. Only in case of major policy changes we deviate from that rule. Therefore, the year of the policy change (tp) is simulated in advance and the average rental prices of the year in which the policy change takes place ( Rs , tp ) is divided by the average rental price in the year before the policy change ( Rs , tp − 1) . This gives the adjustment factor λ for the rental prices in the expectation formation (see equation (12), (13)). Then the same period is simulated again under consideration of the resulting adjustment factors. λs = R s , tp / R s , tp − 1 ,

6.1.7

(14)

Managerial ability

In real world agriculture, farmers’ economic performance can differ substantially, even if they operate under more or less the same production conditions using the same production technologies. These differences in farmers’ economic performance are often attributed to differences in their managerial ability (Nuthall 2001; Rougoor et al. 1998). Managerial ability can be understood as a farm agent’s ability to use its technology to realise all potential cost savings. Accordingly, production costs are lower if managerial ability is higher. In AgriPoliS, a farm agent’s managerial ability is introduced by a factor m , which is drawn randomly from a uniform distribution at start-up. The factor affects the production costs of all products in the initial period according to


cinew , 0 = m ⋅ ci , 0 .

44

(15)

In the current version of AgriPoliS, farm agents cannot learn to improve managerial ability. 6.1.8

Retired farms

To show the distributional effects of e.g. policy reforms it would be desirable to also take into account already retired farms. For this we extended the model in a way that retired farms are not removed from the simulation, instead we introduced an accounting system in analogy to the accounting for active farms (c.f. paragraph 6.2.4) also for retired farms. This means that these farms consider payments for land, labour, liquid capital and borrowed capital. For their owned land, leaving farmers receive the rent paid by the leaseholder. Wages for labour are based on the off-farm wages of surviving farms with the respective change rates. In case a farm closes during the generation change the possible higher opportunity costs (cf. paragraph 6.2.5) of a successor are considered. For liquid capital we assume that the leaving farms can invest their money long-term. These payments are reduced by depreciations and interest costs for still existing agricultural buildings and machinery. The remaining income is reduced by minimum expenditures covering the costs for basic requirements of each family member. This minimum level is equal to the minimum withdrawal (WDmin) of the surviving farms. Further details about the calculation of the household income of surviving and leaving farms can be found in Table 6-2.

The Agricultural Policy Simulator – AgriPoliS Table 6-2: Calculation of the household income Indicator

Calculation for surviving farms

Calculation for leaving farms

household income =

total gross margin agriculture + off-farm income + interest on working capital + subsidies - rent expenditures - depreciation - interest costs - current upkeeping of machinery and equipment - farming overheads - transport costs - wages paid

total gross margin agriculture + off-farm income + interest on working capital + subsidies + rent received - depreciation - interest costs

equity capital change =

household income (Y) - withdrawal (WD)

household income (Y) - expenditures (WD)

withdrawal =

WDmin + (Y − WDmin ) ⋅ ε with 0 < ε ≤ 1 and Y > WDmin

Source: Own

45


46

Buy/sell variable labour

Hire contractor

Plant production

Livestock production

Set-aside land

Buy/sell manure

Buy/sell milk quota

Investment activities

Buy/sell fixed labour

Unused labour savings

Mixed-integer programme

Short-term loans/saving

6.2 Farm actions

c

c

c

c

c

c

c

c

i

i

c

Other restrictions

Factor capacities

Objective function Liquidity (€) Min. equity capital reserve (€) Labour (h) Utilised agricultural area (ha) Winter fodder (ha) Livestock capacities (places) Machinery (ha) Organic N-balance (kg N/ha) Rape seed max. (% of UAA) Sugar beet max. (% of UAA) Set aside (% of UAA) Milk quota (litres) Direct payments (€) Stocking density (LU/ha) Normalised Labour savings

Notes: Source:

Gross margin x

x x

x x x x

x x x

x x x

x x

x x x

x x x x x x x

x x x

x x x

x

x x

x

x

x x x x x x

x x x x

c = continuous activities, i = integer activities Own figure.24

Figure: 6-2: Exemplary scheme of a mixed-integer programme matrix

For the maximisation problem described in (6), production and investment possibilities have to be optimised simultaneously given the farm’s factor endowments and other restrictions. A suitable setting for this is a mixed-integer optimisation problem, the solution to which gives the optimal combination of action possibilities subject to the given framework conditions. Figure: 6-2 schematically shows the main elements of a matrix of the mixed-integer optimisation problem. In this scheme, investments and fixed labour are considered as non-divisible and are introduced as integer activities. The set of constraints consists of on-farm production capacities, but some constraints also reflect political framework conditions such as the setaside requirement, the limit on livestock density, or the nutrient balance. To be able to use this 24

Compared to highly-differentiated and detailed farm-based linear programming models, the optimisation model in AgriPoliS is aggregated. Yet, with respect to the objective of AgriPoliS, it is not the specific farming system which is of interest in this study, but rather a basic representation of central organisational characteristics, as well as financial/economic considerations.


47

model for the farms’ decision-making, the decision-making process must be systematised to take into account the interdependencies between farms, and also within single parts of the decision process. Therefore, it was necessary to conceptualise a decision hierarchy for a farm’s decision during one period of the simulation. The decision hierarchy is displayed in Figure 6-3. Based on this figure, the most important actions undertaken by a farm agent are renting land (renting additional land and disposing of unprofitable land), investment, production, farm accounting, and the decision of whether to quit farming or stay in the sector. Start

Rent land Farm planning

Investment decision

rent

Production decision Pool of investment objects No

Production results

Land market Product markets / Expectation formation Expectation formation

Farm accounting

Let land / disinvest quit farming?

Farm accounting

Farm planning let

Yes

Stop

Source:

step

influence

Own figure based on BALMANN (1995).

Figure 6-3: Course of events in one planning period for one farm agent 6.2.1

Renting/releasing land

Initially, each farm is endowed with a certain amount of land consisting of owned and rented land of different types. Depending on the type of rental contract chosen for the simulation (see section 5.3.4), rental contracts either have a fixed duration or farms have the option of renegotiation. In the current version of the model, one type of contract is globally valid for both the landowner and the farmer. The type of the contract is exogenously given in order to best reflect the situation on the land market in the respective case study region. A detailed discussion of the implications resulting from the different contractual types is given in section 5.3.4.


48

In AgriPoliS, all farmland is categorised as plots of the same size. Plots are not divisible, and their size is fixed during one simulation run. Accordingly, the size of a plot defines the smallest unit by which farm acreage can change. Assuming that a plot is the basic land unit, the shadow price for land q Land represents the marginal utilisation of the plot. That is, the costs for a plot, which consist of the paid rent R yBasic , x , the transportation costs TC y , x between the plot and the farmstead, and additional costs25 C y , x associated with this plot, must not be higher than the shadow price. As we use a mixed integer approach, it is not possible to derive the shadow price directly from the dual solution of the optimisation. Instead, we solve the optimisation problem given in (6) twice. We first assume the current factor endowment of the farm for the optimisation, and then increase the acreage by the size of N additional plots. The shadow price of a plot of land in then calculated as shown below (16):

q

N Land

=

max Y e (..., bLand + N ⋅ sizeof ( p y , x ),...) − max Y e (..., bLand ,...) N

.

(16)

Accordingly, the maximum amount a farm is willing to pay for a plot is given by the shadow price, minus the transportation costs26 and additional costs for the plot. R yBasic = q 1Land − TC y , x − C y , x , with ,z

(17)

max R yBasic ⇔ min (TC y , x + C y , x ) . ,x y,x

y,x

That is, the farm will search for a free plot in the region, which minimises the sum of transportation costs and additional costs (e.g. transactions cost) or benefits (e.g. due to scale effects of having a bigger plot). However, in the basic setting of the model, additional costs are not considered. Consequently, a farm searches for a plot closest to its farmstead.

25

Under the “additional” costs C y , x of a plot, we subsume different types of costs dependent on the kind of

model setting we choose, e.g. transaction costs associated with the transfer of land or cost savings due to realising economies of size. 26 The transportation costs TC y , x for a plot are given by the distance between the farmstead in km, the plot size and a constant TC defining the transportation costs per ha and km.


49

The relationship given in (17) is the most basic way in which a farm derives a bid in the land market specified in AgriPoliS and is based upon economic reasoning. However, in order to adopt the land market model to the specific situation in the case study regions, we derived a number of further decision rules which are based on (17). First, with the decision rule given in (17), the full rent associated with the utilisation of the plot is transmitted to the land owner. This seems to be unrealistic, as other costs such as taxes, administrative costs, labour costs or fees associated with leasing land are not covered. Furthermore, one can expect that a farmer wants to keep a part of the rent as a security mark-up because of the risk associated with the utilisation of the plot. Moreover, a certain degree of market power could lead a farmer to keep a portion of the rent. This is especially plausible in regions with large enterprises in terms of acreage and a highly fragmented land ownership structure such as in the Czech Republic. To consider this, we assume a parameter β which defines the share of the shadow price of an additional plot remaining with the farm agent. A farm’s bid is then defined as R yadj, x = β ⋅ R yBasic , x . Hence, the higher β is, the higher the share of the rent transmitted to the land owner. Second, there is a problem with the basic bidding rule which arises from the farms’ specified production functions. Because of the indivisibility of investments, one will expect size effects associated with farm growth, i.e., plots can be complementary to each other in a way that the valuation for a set of plots is higher than the sum of the valuations for single plots. In order to consider this, we allow a farm to “anticipate” its production function. Hence, in addition to the calculation of the shadow price for one additional plot, we compute the average shadow price for a fixed number PLOTN of plots. The maximum of both values is chosen to derive the bid as shown in (18). PLOTN R yadj, z = β ⋅ (max(q1Land , q Land ) − TC y , x − C y , x )

(18)

Third, the additional cost component C y , x can play a decisive role depending on regional conditions. The additional cost component C y , x is either used to a) account for economies of size in cases where plots are put together to form a larger contiguous plot ( EOS y , x ) or b) as part of the transaction cost framework to satisfy the properties of less developed land markets that can be found in some EU-10 countries ( TAC yi , x ). Accordingly, the additional costs per


50

plot are composed by representing the transaction costs associated with the transfer of a plot, as well as cost savings due to the size of the (sub-) land block which the plot is embedded in. C y , x = TAC yi , x − EOS y , x

(19)

In detail, the two parts of the additional cost component are based upon the following reasoning. First, in AgriPoliS, a single plot is the basic land unit. However, a farm can exploit economies of size if it puts together adjacent plots in order to farm a larger contiguous plot. We assume that with these contiguous plots, a farm can use its machinery more efficiently, which results in lower costs of production. The procedure of these cost adjustments was introduced in detail in section 6.1. To take full advantage of these cost savings, a farm must include expectations about cost savings into its decision-making, e.g. the decision to rent additional land. The expected cost savings realised for one plot are assumed to be a linear function of the number of adjacent plots. With n adj being the number of adjacent plots, we adjust the resulting cost savings to EOS y , x = −δ ⋅ nadj , where δ is a constant reflecting the expected cost savings per additional land unit. As the cost savings are a part of the additional cost component, the approach ensures that the cost savings are taken into account both for searching for the most preferable plot in the region and for formulating the bid. Second, in the proposed transaction cost framework, we consider three different types of transactions costs specific to the plot, to the owner, and to the user (see section 5.3.5). Here, we assume that transaction costs a farm i has to pay can be compiled into a single figure TAC yi , x based on s single plot. Given that the transaction costs TAC yi , x are a part of the addi-

tional cost component, it is again ensured that they are properly considered in the rental decision-making of a farm. However, as with the decision rule given above the bid for a plot is reduced by the transaction costs, the farm nevertheless has to compensate these costs resulting in the final rental price of: R y , x = R yadj, x + TAC yi , x

The decision rule for releasing a land plot is again dependent on the type of contract used. If contracts are renegotiated after every production period, the following decision rule applies:


27

q Land < max ( R y , z + TC y , z + C y , x ) . y,z

51

(20)

That is, a farm would give up a rented plot p y , z if the shadow price does not cover the plot's costs, which consist of the rent R y , z , the transport costs TC y , z , and the additional cost component C y ,c . After giving up a plot, the farm recalculates the shadow price of land. The procedure is repeated until the shadow price of land is at least equal to the costs of a plot. In case of contracts with a fixed contract length, the contract is terminated when the end of the contract is reached. 6.2.2

Investment

Farm investment activity is typically concerned with the purchase of machinery, buildings, facilities, and equipment. As investment and production are mutually interdependent, they are simultaneously considered in the mixed-integer planning programme. In general, the possibility of further investment is taken into account every time the solution to the farm planning problem described above is determined. This is especially the case during farm growth/shrinking, i.e., on the land market or during a farm’s exit decision, according to the decision-making hierarchy of the farm. However, during all planning calculations, the agent does not invest in real terms, but plans 'as if' he invested, i.e., production capacities are not actually changed. The possibility of actually investing is given only once per production period, between the land allocation and the production decision. The number, kind, and combination of investments are not restricted. In principle, a farm agent will invest in one object or a combination of objects if the expected average return on the investment, determined in the farm-planning problem, is positive, i.e., if profit increases. For investment-planning purposes, all expenditures and payments related to an investment are distributed equally over the investment's useful life and considered in the optimisation. Accordingly, the average annual costs AC h of investment I h ,i considered in the objective function of the farm-planning problem are calculated as

27

Adjacent (contiguous) plots are not considered when rental contracts are terminated.


⎡ ⎤ v ACh = Ah ⎢(1 − v) ⋅ CRFibc , N h + + MCh ⎥ . Nh ⎣ ⎦

52

(21)

Maintenance costs MCh are expressed as a percentage of total investment costs. The average annual opportunity costs of bound equity capital is determined as

Ah ⋅ v ⋅ f ,

with

f =

(1 + iec ) N h 1 − . Nh (1 + iec ) − 1 N h ⋅ iec

(22)

As described above, for each type of investments (e.g. a hog stable) we introduced various gradations in the size of the investment to allow for economies of size. Economies of size arise from decreasing costs per unit of investment and lower labour requirements of bigger investments. Lower costs per unit of investment are simply introduced by the ratio of acquisition costs of an investment to the capacity of this investment. For modelling the labour requirements dependent on the size of investments, the attribute “labour substitution” of the investment defines how much additional labour is needed/can be saved compared to the labour demand of the associated production activity. After investment takes place, the labour substitution of the investment is added to the labour capacity of the farm. In general, this approach is unproblematic. However when a farm’s capacities are unused, this could lead to a situation where either the farm’s labour capacity is fixed or there a farm can realize a labour saving out of an investment , even if no production is taking place. To circumvent this problem, we assume that the labour substitution must be less than or equal to zero and decrease with the size of the investment, i.e., only a situation with fixed labour capacities can occur. Given that assumption, we can then introduce an additional activity which allows us to “free” the fixed labour capacities if there are unused investments. To illustrate this approach, an extract of the mixed integer program is shown in Figure 6-4 for a hypothetical dairy investment. Accordingly, the biggest stable (Dairy I) has no additional labour requirements; the medium-sized stable (Dairy II) has an additional requirement of 3 hrs. per place; and the smallest-sized stable (Dairy III) requires an additional 5 hrs. per place. Let us assume that a arm is equipped with one stable of the type Dairy II. Thus, the labour capacity of this farm is reduced by 600 hrs. To keep track of the additional labour requirements, they are summed up and normalised to the difference in labour requirements between the smallest and largest stable (normalised


53

labour substitution=120= 600/5). If the capacity of this stable is not used, the fixed labour

Unused Dairy

Dairy III

Dairy II

Dairy I

Dairy production

could be freed through a corresponding amount of the activity “Unused Dairy”.

RHS

Objective function Labour (h) 23.4 Capacity (places) -400 Normalised labour substitution

600 600 -200 -120

-5 1 1

cap-600 200 120

…

Figure 6-4: Modelling of labour savings in investments 6.2.3

Production

Each farm agent is assumed to optimise production in any one planning period subject to available production capacities using the planning approach described above. All production activities enter the optimisation as continuous activities. That is to say, products are assumed to be fully divisible. In addition to fixed assets (buildings, machinery, equipment), production requires liquidity to cover operating costs in the short-run. Products produced continuously throughout the year (mostly livestock production) have a constant demand of working capital, which in AgriPoliS is defined as liquid assets. Other products such as crops are seasonal, and therefore require working capital only during parts of the year. To overcome short-term liquidity shortages, farm agents can take up loans to finance working capital. Two additional properties of agricultural production are uniquely captured in the revised version of AgriPoliS; spatially-explicit organisation and the development of production patterns. Their addition was necessary in order to conduct an environmental impact assessment of the analysed policies. A more detailed description of the procedure of environmental impact assessment, as well as the extensions of the AgriPoliS model, can be found in BRADY and KELLERMANN

(2005). An outline is given here, where we focus on the modifications of the model

necessary for a spatially-explicit assessment of land use. One prerequisite for analysing the


54

spatial organisation of production patterns is an adequate model of the landscape upon which production is taking place. In AgriPoliS, we therefore introduced an abstract spatial model of landscapes, which is described in Section 4. The spatially-explicit organisation of production patterns refers to layer 5 of the landscape, which reflects land use in a region. Thereafter, the land use layer reflects the field structure of a farm, where a field is a contiguous area of land (i.e., one or more contiguous plots) used for a particular activity. As described above, as a result of the farm planning problems, optimal activity levels could be derived. However, there is no information on the spatial distribution of activities at this stage. This can be done under the assumption that farmers aim to maximise farm profits, and the duality of the associated costminimisation problem. In Section 5.3.4, it was shown that farmers take into account transportation costs and field size when they determine their valuation of a plot. As larger field sizes imply lower unit costs of production, minimising the costs of production for a crop means maximising the field size of a particular activity. Given this knowledge of how farmers frame their crop allocation decisions, we should be able model the allocation of crops to fields, and hence spread optimal levels of each production activity over the landscape in a fashion which is both similar to that of farmers in reality, and which is consistent with decision-making in AgriPoliS. The duality of the farmers’ profit maximisation problem is associated with minimising the costs of producing the optimal level of output. Since activity levels generated by AgriPoliS represent profit maximising levels of production given any set of farm and policy parameters, then a dual of this problem is to allocate optimal cropping levels across plots of agricultural land in the synthetic landscape such that the costs of producing crops is minimised. Therefore, farmers will minimise the costs of crop production by growing their various crops, as far as is possible, on contiguous areas of land. Given that farmers have maximised profits, the following mathematical programme (23) will allocate the optimal activity levels of a farm across the plots of land managed by a farm in the synthetic landscape such that the costs of production are minimised (cf. BRADY and KELLERMANN

2005):

max ∑ ∑ (q i a ikl ) 2 i∈I k ∈K

s.t.

(23)


∑a ∑a k

≤ x il ,∀i

kl i

kl i i

≤ y kl ,∀sb kl ∈ SB l

a ikl ≥ 0

55

Optimal activity levels Contiguous plots of land

No negative field sizes,

where aikl is the size of a field flikl and represents a contiguous area within block sb kl allocated to cropping activity i by farm l, xil is the optimal level of activity i provided by the solution to farm l’s profit maximisation problem, and y kl is the size of the (sub-) block sb kl managed by the farm (i.e., contiguous plots of agricultural land). In the problem statement above, qi reflects the diagonal value of the Q-matrix of the optimisation problem. Setting qi allows us to assign specific weights to production activities, as a farmer probably would allocate the most intensive crops to the largest fields (high q-value), but would possibly treat set-aside as a residual ( q set − aside = 0 ) and fill up area not used for agricultural production. The data required to solve this problem are optimal activity levels and the (sub-)blocks managed by each farm. The solution to this problem is the set of fields { flikl } that maximise the sum of the square of field size subject to the constraints that total activity levels cannot exceed optimal levels, and that the sum of field sizes cannot exceed the area of contiguous plots of land. The nonlinear form of the objective function implies that larger fields are better than smaller. That is, if fields are chosen that are smaller than permitted by the constraints, then the objective function will be lower than is possible. The quadratic form was chosen because it is straightforward to solve. One limit of the approach described above is that we do not take crop rotation restrictions into account; a refinement of the problem would allow us to consider crop rotation restrictions as further constraints. However, as crop rotation is considered on the farm level, this would be beyond the scope of this analysis. 6.2.4

Farm accounting

A farm agent’s financial year ends with an annual financial statement. This statement produces indicators on incomes and profits, the stability and financial situation of the farm agent,


56

and the remuneration of fixed factors. Table 6-3 lists central indicators and how they were derived; Table 6-4 shows a list of selected variables in the financial statement. Table 6-3: Indicators calculated in the financial statement Indicator (end of period t)

Calculation

Profit (farm income) (t) =

Gross margin + Interest on working capital + Subsidies - Rent paid - Current upkeep of machinery and equipment - Depreciation - Farming overheads - Transport costs - Interest paid - Wages paid

Household income (t) =

Profit + Off-farm income

Farm net value added (t) =

Profit + Rent paid + Interest paid + Wages paid

Equity capital (t) =

Equity capital (t-1) + (Household income - Withdrawal)

Change in equity capital is an indicator of a farm agent's economic stability. The higher is the equity-debt ratio of the farms, the more economically stable a farm is, i.e., the higher the share of equity capital of total capital. Consequently, it would be reasonable for a farm to stop farming if equity capital were less than zero. In this case, all own resources which could be used up, for example credit security, would be. The accumulation of equity capital is the result of balancing total farm income with living expenses. In AgriPoliS, equity capital stock increases because total household income is greater than withdrawals. Regarding withdrawals, it is assumed that each family labour unit working on the farm consumes at least WDmin per year. A share ε of the remaining farm household income after deducting WDmin is consumed in addition to the minimum withdrawal. The remaining share (1 − ε ) ⋅ (Y − WDmin ) is then charged to the farm agent's equity capital. Table 6-4 illustrates this.


57

Table 6-4: Definition of variables used in financial statement (selection) Variable (at end of period t)

Definition

Equity capital Withdrawal Gross margin Interest on borrowed capital

EC = ECt −1 + Y − WD WDmin ≤ WD ≤ (Y − WDmin ) ⋅ ε + WDmin with 0 < ε ≤ 1 GMA = x′(p − c) IC = f( BC , iBC )

Repayment

RP = (1 − v)∑ Ac ⋅ (1 + ibc ) ( nc −1) ⋅ (CRFibc , N c − ibc )

S

c =1

Long-term loans

[

BC = BCt −1 − RP + BC new

∑ [A ⋅ v ⋅ (1 + i S

D=

c

bc

c =1

Depreciation

)( nc −1) ⋅ (CRFibc , N c − ibc )

∑ [A ⋅ (1 − v) ⋅ (1 + i S

+

c

ec

c =1

]

)( nc −1) ⋅ (CRFiec , N c − iec )

Farming overheads

OV = γ ⋅ GMA with γ ≤ 1

Current upkeep (maintenance)

MC = ∑ MCc

Rent paid

RE = ∑∑ R y , z

Transport costs

TC = f( DI y ,z )

Liquiditya) Interest on working capital

L = ECt −1 − LA − Aec IR = iec ⋅ L

Notes:

]

]

S

c =1

y

z

a) Liquidity is updated throughout the accounting year whenever the total equity capital stock changes due to investment or disinvestment.

Lasting farm profitability requires that all farm-owned production factors (own land, family labour, liquid equity capital, and quota) receive an adequate payment when used on-farm. To assess farm profitability, all on-farm production factors have to be valued at their opportunity costs. Since costs of fixed assets are assumed sunk, they are not considered in this calculation. 6.2.5

Farm exit decision

In AgriPoliS the basic decision rule for a farm to quit production and withdraw from the sector is if the farm is illiquid, or if opportunity costs of farm-owned production factors are not covered by expected household income (individual farms) or expected profit (corporate farms). Especially with respect to the possibility of generating off-farm income for family members of individual farm, the decision rule described above is quite restrictive because it assumes that the possibility of a farmer working off-farm is independent of age and educational level. This is in contrast to findings in BAUM et al. (2006), RIZOV and SWINNEN (2004) and BOJNEC


58

et al. (2003), which claim that especially in CEECs, middle-aged and older farmers have little or no off-farm employment possibilities. Furthermore, other studies show that the decrease in farms is not mainly caused by the exit of farm operators, but is rather due to the non-entry of young farmers (cf. GALE 2003). This can be explained by the fact that it is mostly the younger, better-educated family members who are able to work off-farm. In order to reflect these empirical observations in our model and to allow a better representation of these macroeconomic conditions in the case study regions, we introduced a set of refinements for the above mentioned decision rule according to Figure: 6-2. The basic decision rule described above is shown on the left-hand side of the graph. Both individual farms and corporate farms can leave the sector in this way. Since corporate farms are operating with hired labour only, opportunity costs are calculated for land and capital. The first refinement in this respect is that we allow varying the opportunity costs for labour in dependence of the age of the farm operator of individual farms. For this we divide the time span between a generation change is taking place into three periods and define for each period separately the opportunity costs for labour as a share of off-farm income at that time. For individual farms, a farm exit decision is also related to the question of succession at the end of the previous operator’s operating life cycle (right-hand side of graph). In AgriPoliS we assume that a potential successor can be a farm family member but also an external person willing to enter and work on the farm. Moreover, exit can occur if it is not profitable for a willing successor to enter the farming business. To reflect that this is a life-time decision and requires some investment into agricultural training we add a mark-up to opportunity costs for potential successors. In contrast we allow also for the possibility that farm exit takes place because there is no successor or no willing successor to take over the farm even if it were profitable to do so. For this we define a probability that a farm successor is willing to take over the farm. Closely related to the problem of entry and exit decision is the empirically observed age distribution of farm operators of individual farms. In many countries especially in CEEC we are facing the situation of overaged farm operators. To reflect this we allow in alternative to a uniform initial age distribution to draw the operators ages from a triangular distribution during initialisation. In that way we can model the typically observed, left skewed age distribution.


59

Figure 6-5: Farm exit decision in AgriPoliS

Exit?

consequence

at

at

After each simulation period due to

Illiquidity

End of operator lifecycle

due to

farm has

farm has

No successor/ No willing sucessor

Opportunity costs (labour, land, capital)

due to

Potential successor

No

Entry?

Yes

due to

Non-economic reasons no interest in agriculture, personal characteristics, ...

farm operation closes land is released to the land market and rented by other farms assets are no longer used

Economic reasons (opportunity costs)

successor takes over farm operation land and assets are kept on the farm

Source: Own figure

As in the previous section we conclude with a table summarising the main global settings which apply for the decision making of farm agents. Table 6-5: Table summarizing global settings related to farms Name of the parameter Renting/releasing land

β

TC PLOTN

Production/ investments qi [1..N]

v Farm accounting WDmin

Description

Type

Range

Default

Share of the shadow price of an additional plot remaining with the farm agent (bid adjustment) Parameter for the transportation costs per km and ha Number of additional plots used to calculate the average valuation for a plot.

ℝ

[0..1]

0.5

ℝ

[0..∞]

50 € km ⋅ ha

ℕ

[1.. ∞]

8

Relative importance of a production activity for the distribution of activities over the fields of a farm (q-value of the corresponding QP-matrix).

ℝ

[0.. ∞]

1

Equity-finance share

ℝ

[0..1]

0..3

Minimum withdrawal for consumption

ℝ

[0..∞]

15400 € (e.g.


60 Hohenlohe)

Farm exit decision GENERATION_CHANGE OPP_C1, OPP_C2, OPP_C3

OPP_SUCC

PROB_WILLING

TRIANGULAR_AGE_DIST

TRI_MIN,TRI_MAX,TRI_M

7

Time period between farm being handed over to a successor Opportunity costs for labour defined as a share of the offfarm income for the first, second and third period of a generation Opportunity costs for labour of farm successor in the year of the generation change, defined as share of off-farm income Probability of farm successors to take over a farm if it is profitable. Initial age distribution is drawn from a triangular instead of a uniform probability function. Minimum , maximum and mode of the triangular distribution

ℕ

[0..∞]

25

ℝ

[0..∞]

1,1,1

ℝ

[0..∞]

1.25

ℝ

[0..1]

1

Boolean

true, false

false

ℕ

[0..∞]

0,25,18

Policy framework

Within the AgriPoliS framework, we are able to generate projections under various policy assumptions. Looking more closely to a single reform package, e.g. the 2003 CAP reform, one can see that even within a reform, the possible options are various and the finally chosen options differ from country to country. Moreover, the 2003 CAP reform, as well as other reforms, e.g. the accession policy of the NMS, are not static events that take place at a single point in time, but rather develop over time. To cover this whole range of policy options, beginning with the Agenda 2000 reform, moving to the 2003 CAP reform policies, and on to the pre-accession and accession policies for the new member states, it is necessary to have a flexible framework which allows for a varied and dynamic setting of different policy measures. Therefore, we provide a set of parameters which allow us to steer, over time, the policy framework that the farms find themselves embedded in. Table 7-1: Overview and description of policy parameters Direct payments PREMIUMi,t PREM_ELIGIBILITYi Decoupled payments REF_PERIOD REF_PREMIUMi

ℝ boolean

ℕ ℝ

Direct payment for product i in period t Value which specifies if a product is eligible for premium payments Period in which the reference premium is calculated Premium value used for calculating the reference premium for product i

The Agricultural Policy Simulator – AgriPoliS REF_PERIODi

ℕ

COUPLING_RATEi,t

ℝ

MODULATIONt

ℝ

DEGRESSION_Lt, DEGRESSION_Mt, DEGRESSION_Ht

ℝ

UPPER_LIMIT_Lt, UPPER_LIMIT_Mt, UPPER_LIMIT_Ht DECOUPLING_SCHEMEt

ℝ

Price policies MIN_PRICEi PRICE_ELIGIBILITYi Structural policies1 RET_PAY

61

Time period assumed for calculating the reference premium Share of reference payments used for the decoupling scheme Share of decoupled payments modulated in a period t Degression of decoupled payments in a period t, whereas we allow different degression rates dependent on the amount of payments. Therefore we divide the payments into three quantiles. Upper limit of the low, middle and high degression quantiles. Decoupling scheme at period t, whereas we allow for a single payment scheme as a) a farm-specific payment based on premium entitlements per ha or as b) a region-specific premium entitlement per ha or c) a bond scheme.

ℝ

ℝ

Minimum price for product i. Parameter which specifies if a product is eligible for price support.

Yearly payment if a farm withdraws from a sector, whereas we assume a planning horizon of 20 years. 1 In addition to the parameters presented here, we allow for the direct manipulation of the farms’ production programmes to allow for country-specific measures, e.g. cross-compliance regulation, the removal of minimum set aside restrictions, or the activation of production activities like basic land management. ℝ

The set of parameters provided for the policy framework is displayed in Table 7-1. As we are using an agent-based approach, we are also able to cover a set of structural policies like environmental payments or specific aspects of the decoupling schemes such as cross compliance regulations. These policies are introduced by directly changing the farm programmes as specified in Figure: 6-2 in terms of changes of the technical coefficients (e.g. for minimum setaside restriction), the upper bounds for activities, or the right-hand side constraints (e.g. for the introduction of minimum land management requirements). With the policy framework described above, we are now able to implement a whole range of policies in a quite detailed manner. This can be seen as a comparative advantage of the agentbased approach over other approaches such as equilibrium models, which have to rely on much more abstract representations of policy measures. This is especially valid if we consider the shift from policies toward structural and more regional measures. In the context of the IDEMA project, the focus is on analysing the decoupling policies. At the current stage of model development, we used the policy framework to implement four different decoupling schemes that reflect the basic options of the EU-10 and EU 15 countries, as well as one more “hypothetical” scheme. The four different decoupling schemes are a single payment scheme as implemented in the EU-10 in: a) a regional version, b) a farm-specific version based on the historical payments of a farm during a reference period; this results in


62

farm-specific payment entitlements per ha, c) a single area payment scheme (SAPS) as introduced in the EU-15 and, d) a bond scheme where payments are independent of land management activities or the acreage of a farm. Especially for a) and b) it is possible to implement a combination of both schemes, either in a static version “static hybrid” or a dynamic version “dynamic hybrid”. Sweden, for example, opted for the former, and a mixture of both payment schemes will persist over the next year. The latter option was chosen by Germany, where historic farm-specific payments are dynamically converted into a regionally-based payment.

8

Data input, results preparation and data output

AgriPoliS has an interface to a spreadsheet file that includes data on the regional agricultural structure to be studied when initialising the model. On the input side, broadly speaking, data input consists of farm accountancy data, regional statistics, and stylised data on technical coefficients, prices and costs. On the output side, AgriPoliS compiles aggregate data at the sector level (class SectorResults) on the one hand, and individual farm data on the other. More specifically, data output at the sector level and the farm level (class DataOutput) include data listed in Table 8-1. Based on these indicators, it is possible to draw conclusions with respect to production, economic performance of farms, production intensity, income distribution, farm structure and environmental impacts. Table 8-1: Data output at the farm and sector level (selection of key data) Farm level Structure Farm size Economic size Farm type Main income source

Production Output in quantities Output in value Costs Overheads Maintenance Depreciation Wages paid Rent paid Interest paid Annualised average costs of fixed capital Variable costs Subsidies

Unit ha ESU Professional/ non-prof.

ha, LU € € € € € € € € €/unit

Sector level Production Region totals Inputs Total land input Total capital input Total labour Investment Investment expenditure Sector totals of farm level data Land Block size distribution per soil type (mean, std) Environmental indicators Species per habitat Farm level Financial situation Profit Equity capital Change in equity Net investments Income and labour

Unit ha, LU ha € h € various units ha 1/habitat Unit € € € €

The Agricultural Policy Simulator – AgriPoliS Direct payments Land Economic land rent Rent paid per soil type (Sub-)block size distribution (mean, std) Field size distribution per activity (mean, std) Owned land Rented land Balance sheet Total assets Total fixed assets Total land assets Liquidity Borrowed capital Short-term borrowed capital Environmental indicators N,P,K usage Chemical consumption (fungicides, herbicides, insecticides) Water usage Soil loss

€ €/ha €/ha ha ha ha ha € € € € € € 1/ha 1/ha 1/ha 1/ha

Labour input Family labour Farm net value added Total household income Off-farm income

63

h h € €


64

References ALLB BIBERACH (2002): Marktentlastungs- und Kulturlandschafsausgleich. Retrieved from internet: August 2002, http://www.infodienst-mlr.bwl.de/allb/Biberach/dienstbezirk/Dienstbezirk.htm. BALMANN, A. (1995): Pfadabhängigkeiten in Agrarstrukturentwicklungen - Begriff, Ursachen und Konsequenzen. Berlin: Duncker und Humblot. BALMANN, A. (1997): Farm-based Modelling of Regional Structural Change: A Cellular Automata Approach, European Review of Agricultural Economics, 24(1), pp. 85-108. BALMANN, A., LOTZE, H., NOLEPPA, S. (1998): Agrarsektormodellierung auf der Basis 'typischer Betriebe'. Teil 1: Eine Modellkonzeption für die neuen Bundesländer, Agrarwirtschaft, Jg. 47, Nr. 5, pp. 222-230. BAUM, S., COOK, P., STANGE, H., WEINGARTEN, P. (2006). Agricultural employment trends in an enlarged European Union: does the CAP reform / introduction matter? Paper presented at the 46th Annual Conference of the Gesellschaft für Wirtschafts- und Sozialwissenschaften des Landbaues e.V., Giessen, 4-6 October, 2006. BERGER, T. (2001): Agent-based spatial models applied to agriculture. A simulation tool for technology diffusion, resource use changes, and policy analysis, Agricultural Economics, vol. 25, iss. 2-3, pp. 245-260. BOJNEC, S., DRIES, L., SWINNEN, J.F.M., 2003. Human Capital and Labor Flows out of the Agricultural Sector: Evidence from Slovenia. In: Reshaping agriculture's contribution to society. Proceedings of the 25th IAAE Conference, 16-22 August 2003, Durban, South Africa, 649-654. BOOCH, G., RUMBAUGH, J., JACOBSON, I. (1999): The Unified Modeling Language User Guide, Object Technology Series, Addison-Wesley, Redwood City. BRADY, M., KELLERMANN, K. (2005): Methodology for Assessing the Regional Environmental Impacts of Decoupling: A Focus on Landscape Values; SLI-Working Paper 2005:2, Swedish Institute for Food and Agricultural Economics. CIAIAN P., SWINNEN J., (2006): Land market imperfections and agricultural policy impacts in the new EU member states: A partial equilibrium analysis, American Journal of Agricultural Economics, vol. 88, no. 4, pp. 799 - 815. DALE, P., BALDWIN, R. (1999): Emerging Land Markets in Central and Eastern Europe; Paper presented at Second World Bank/FAO Workshop, Warsaw, June 27−29th. DEUTSCHE BUNDESBANK (2003): Monatsbericht März 2003. FERBER, J. (1999): Multi-Agent Systems: An Introduction to Distributed Artificial Intelligence, Addison-Wesley, Redwood City. GALE, H. F. (2003): Age-specific patterns of exit and entry in U.S. farming; 1978-1997, Review of Agricultural Economics Vol. 25: 168 - 186. HAPPE, K. (2004): Agricultural policies and farm structures - Agent-based modelling and application to EU-policy reform in: Studies on the Agricultural and Food Sector in Central and Eastern Europe, Vol. 30, IAMO. HAPPE, K., BALMANN, A., AND KELLERMANN, K. (2004). The Agricultural Policy Simulator AgriPoliS version 1.0. IAMO Discussion paper Nr. 71: IAMO. HAPPE, K., KELLERMANN, K., BALMANN, A. (2006): Agent-based Analysis of Agricul tural Policies: an Illustration of the Agricultural Policy Simulator AgriPoliS, its Adaptation and Behavior; Ecology and Society 11 (1): 49. [online] URL: http://www.ecologyandsociety.org/vol11/iss1/art49/. HAZELL, P.B.R., NORTON, R.D. (1986): Mathematical Programming for Economic Analysis in Agriculture, Macmillan, New York.


65

HEMME, T.; ISERMEYER, F.; DEBLITZ, C. (1997): TIPI-CAL Version 1.0: Ein Modell zur Politik- und Technikfolgenabschätzung für typische Betriebe im internationalen Vergleich. Arbeitsbericht 2/97 des Instituts für Betriebswirtschaft, FAL, Braunschweig. HURRELMANN, A. (2005): Agricultural Land Markets: Organisation, Institutions, Costs and Contracts in Poland, in: Institutional Change in Agriculture and Natural Resources. Vol. 24, Shaker, Aachen. KELLERMANN, K., SAHRBACHER, C., JELINEK, L., HAPPE, K. (2006): Impacts of an increasing land mobility in NMS – a case study of the Czech region Vysočina. Paper presented at the 93rd EAAE Seminar on the "Impacts of Decoupling and Cross-Compliance on Agriculture in the Enlarged EU", Prague, Czech Republic, September 20-22nd, 2006. KELLERMANN, K.; HAPPE, K.; SAHRBACHER, C.; BRADY, M. (2007): AgriPoliS 2.0 - Documentation of the extended model. Deliverable 20, IDEMA Project, Halle (Germany). KLEINGARN, A. (2002): Anpassungskosten von Agrarpolitikänderungen – Erstellung und Kalibrierung der Datengrundlage für ein räumlich dynamisches Multiagentenmodell der Region Hohenlohe. Diplomarbeit. Humboldt Universität zu Berlin. KLEMPERER, P. (1999) Auction theory: a guide to the literature. Journal of Economic Surveys, vol. 13(3), pp. 227-286. KTBL (2002): Taschenbuch Landwirtschaft 2002/03. 21. Auflage, Münster-Hiltrup: Landwirtschaftsverlag GmbH. KTBL (KURATORIUM TECHNIK UND BAUWESEN IN DER LANDWIRTSCHAF), various years. Betriebsplanung Landwirtschaft. Landwirtschaftsverlag, Münster. LERMAN, Z., CSAKI, C., FEDER, G. (2004): Agriculture in Transition: Land Policies and Evolving Farm Structures in Post-Soviet Countries; Lanham, MD: Lexington Books. LI, X., HE, H.S., WANG, X., BU, R., HU, Y., CHANG, Y. (2004). Evaluating the effectiveness of neutral landscape models to represent a real landscape, Landscape and Urban Planning, 69, pp.137-148. NUTHALL, P. L. (2001): Managerial ability – a review of its basis and potential improvement using psychological concepts. Agricultural Economics, 24, pp. 247-262. O'Neill, R.V., GARDENER, R., TURNER, M.G. (1992): A hierarchical neutral model for landscape analysis, Landscape Ecology, 7(1-2), pp. 55-66. PARKER, D. C., BERGER, T. MANSON, S. M., MCCONNELL, W.J. (eds.) (2002): Agent-Based Models of Land-Use / Land.-Cover Change, Report and Review of an International Workshop October 4-7, 2001, Irvine, California, USA. LUCC Report Series No. 6. REISS, S.P. (1999): A Practical Introduction to Software Design with C++, John Wiley & Sons, New York. RIZOV, M., SWINNEN, J.F.M. (2004). Human capital, market imperfections, and labor reallocation in transition. Journal of Comparative Economics 32: 745-774. ROUGOOR, C.W., TRIP, G., HUIRNE, R.B.M., and RENKEMA, J.A. (1998): How to define and study farmers' management capacity: theory and use in agricultural economics. Agricultural Economics, 18, pp. 261-272. SAHRBACHER (2003): Modellierung der Agrarstruktur Hohenlohes auf der Grundlage typischer Betriebe – Bereitstellung der Datengrundlage für das agentenbasierte Modell AgriPoliS, Diplomarbeit, Universitaet Hohenheim, Stuttgart. SAURA, S., MARTINEZ-MILLAN, J. (2000): Landscape patterns simulation with a modified random clusters method, Landscape Ecology, 15(7), pp. 661-678. SIMON, H.A. (1955): A Behavioral Model of Rational Choice, The Quarterly Journal of Economics, vol.69, pp. 99-118. SWINNEN, J., VRANKEN, L. (2006): The development of rural land markets in transition countries, ECSSD Report, The World Bank, Washington, DC


66

TESFATSION, L. (2001): Introduction to the special issue of agent-based computational economics, Journal of Economic Dynamics & Control, vol. 12, pp. 281-293. TESFATSION, L. (2002): Agent-Based Computational Economics, Artifical Life, vol. 8, iss. 1, pp. 55-82 WITH, A. (1997): The application of neutral landscape models in conservation biology. Conservation Biology, 11(52), pp.1069-1080.

AgriPoliS 2.1 â Model documentation - IAMO

AgriPoliS 2.1 â Model documentation - IAMO

Suggest Documents

Technical model documentation

Download Model documentation - wholeSEM

Policy Brief - IAMO

Policy Brief - IAMO

Policy Brief - IAMO

Policy Brief - IAMO

Policy Brief - IAMO

Policy Brief - IAMO

TIAM-UCL Global Model Documentation

agricultural producers in Uzbekistan - IAMO

Simulations results of AgriPoliS - AgEcon Search

Agricultural education in Uzbekistan - IAMO

Model 21 Manual - Chaparral Physics

Agenda 21: Information and Documentation â a Research Agenda - grif

Rethinking Agricultural Reform in Ukraine - IAMO

2003 DISCUSSION PAPER Institute of Agricultural ... - IAMO

Buch IAMO engl 2017.indb

Documentation of Hybrid Hydride Model for ...

Documentation and Verification of VST2D: A Model

Documentation of model components EXPAMOD ... - AgEcon Search

R DForm REPORT DOCUMENTATION PAGE APPLY MODEL ...

Servoy Model Tool documentation - Servoy Stuff

Solar Deployment System (SolarDS) Model: Documentation and ...

The national bioenergy investment model: Technical documentation

AgriPoliS 2.1 â Model documentation - IAMO