Københavns Universitet. A regional econometric sector model for Danish agriculture Jensen, Jørgen Dejgård; Andersen, Martin; Christensen, Knud

Size: px
Start display at page:

Download "Københavns Universitet. A regional econometric sector model for Danish agriculture Jensen, Jørgen Dejgård; Andersen, Martin; Christensen, Knud"

Transcription

1 university of copenagen Købenavns Universitet A regional econometric sector model for Danis agriculture Jensen, Jørgen Dejgård; Andersen, Martin; Cristensen, Knud Publication date: 2001 Document Version Publiser's PDF, also known as Version of record Citation for publised version (APA): Jensen, J. D., Andersen, M., & Cristensen, K. (2001). A regional econometric sector model for Danis agriculture: a documentation of te regionalized ESMERALDA model. Købenavn: Statens Jordbrugs- og Fiskeriøkonomiske Institut. Rapport / Statens Jordbrugs- og Fiskeriøkonomiske Institut, No. 129 Download date: 28. Jan. 2019

2 Statens Jordbrugs- og iskeriøkonomiske Institut Rapport nr. 129 A Regional Econometric Sector Model for Danis Agriculture A Documentation of te Regionalized ESMERALDA model Jørgen Dejgård Jensen, Martin Andersen, Knud Kristensen Købenavn 2001

3 2 SJFI A Regional Econometric Sector Model for Danis Agriculture

4 Contents Preface... 5 Executive summary Introduction Background and objectives of te model Status of agricultural sector modelling New developments compared to an earlier version of ESMERALDA Overview of te model General features of te model Structure of te model Outline of te econometric estimation procedure Application and limitations of te model Data Agricultural data Farm level accounts data Economic data for individual lines of agricultural production Using te data for econometric estimation Estimation of conditional input demand equations Deriving demand equations on te basis of duality teory Econometric estimation of input demand equations Estimation results Interpreting te estimation results in terms of elasticities Equations for profit maximising crop yield levels Equations for profit maximising yield levels Econometric estimation of te derived yield equations Econometric estimation results Interpreting te estimation results in terms of elasticities Equations for profit maximising land use, animal stocks and capital input Teoretical framework A Regional Econometric Sector Model for Danis Agriculture SJFI 3

5 6.2. Empirical model Dynamic specification Data Estimation procedure Ad 1. Identify order of integration in data series Ad 2. Identify and estimate possible cointegrating relations Ad 3. Calculate deviations from long-run equilibrium Ad 4. Estimate error correction equation Ad 5. evaluate te model Discussion Aggregation from farm level to national and municipality level Aggregation to te national level Distribution of te national aggregation to te municipality level Evaluation of te adjustment Discussion of te aggregation approac Discussion and perspectives Ad 1. Economic analysis at te national and regional level Ad 2. Integration wit oter economic models Ad 3. Integrated economic and environmental analysis Perspectives for furter development References Appendix A. Notational conventions Appendix B. Estimation results for input demand equations Appendix C. Estimated coefficients in yield equations (capter 5) Appendix D. Output-compensated elasticities of input demand Appendix E. Estimated yield response parameters to canges for eigt farm types Appendix F. Yield-endogenous elasticities for variable input demands (Marsall elasticities) SJFI A Regional Econometric Sector Model for Danis Agriculture

6 Preface Tis report describes and documents te latest version an econometric sector model of te Danis agricultural sector. Particular attention is given to te description of te metodology applied, te data used and te description of estimated beavioural parameters of te Danis agricultural sector. A fortcoming report from te institute will supplement te present report by giving a more verbal introduction to te model, including analyses of a few illustrative scenarios. Te development and te use of te model is an integral part of te researc project entitled Agriculture in rural districts: economy and development financed by te Directorate for Food, Fiseries and Agribusiness under te Danis Ministry of Food, Agriculture and Fiseries. Te overall objective of te researc project is to provide a basis for evaluating te development of rural areas and te possibilities for influencing tis development troug coice of political actions and promotions of alternative activities complementary to existing activities. Te empasis is on a olistic model based approac were te agricultural sector s economic importance is seen as an integral part of te local communities and teir economy, and environmental factors and matters relating to te natural landscape are taken into account. Te project is conducted in collaboration wit te Institute of Local Government Studies. Te report as been prepared by Senior researcer Jørgen Dejgård Jensen and researc assistants Martin Andersen and Knud Kristensen. Furtermore, te model development as drawn on data work and statistical analyses by Stine Hjarnø Jørgensen and Malene Krag Petersen. Researc Director Søren E. Frandsen as participated in te editing process. Danis Institute of Agricultural and Fiseries Economics, December 2001 Ole P. Kristensen A Regional Econometric Sector Model for Danis Agriculture SJFI 5

7 6 SJFI A Regional Econometric Sector Model for Danis Agriculture

8 Executive summary Troug te last decade, agricultural policy as canged significantly in te European Union. Tis is partly a result of te pressure for removing support measures, wic distort international trade (e.g. support in terms of export subsidies, import duties and market intervention), but also a result of an increased awareness of te needs for differentiating te regulations and interventions between farms and between regions. As an example, tere is an increased focus on te economic role of agriculture in rural areas lacking beind in attracting oter industries. Tere is also an increasing focus on te environmental and landscape impacts of agricultural activities and te variation in suc impacts from different farm types. Tus, te range of agricultural policy instruments moves in te direction from orizontal measures to more targeted and differentiated instruments. Tis development is expected to continue in te future. In order to perform quantitative assessments of suc more differentiated instruments in agricultural policy, tere is a need for quantitative economic models, wic take into account te regional diversities witin te agricultural sector, and te possibility for representing suc policy instruments properly. Togeter wit anoter report (Jensen, 2002), tis report describes and documents an econometric sector model ES- MERALDA wic enables economic analyses of canges in te Danis agricultural sector at te national as well as te local (municipality) level, integrates agricultural sector analysis wit oter model tools, e.g. macroeconomic models at te national and local level, and tus enables more olistic analysis of te economic development in Danis rural districts and in te country as a wole, and improves te possibilities for integrated economic and environmental analysis related to agriculture, e.g. environmental problems wit a non-point source nature and policy instruments addressing te farm level. Te present report provides a formal documentation of te teoretical principles and econometric metodologies applied for constructing te model, wereas Jensen (2002) gives a more verbal introduction to te model, including a number of illustrative examples of model applications. A Regional Econometric Sector Model for Danis Agriculture SJFI 7

9 ESMERALDA describes te agricultural production, demand for inputs, land allocation, livestock density and various economic and environmentally relevant variables on representative Danis farms, and subsequently in te Danis agricultural sector at relevant levels of aggregation. Te model covers 14 lines of agricultural production, of wic 11 yield a marketed output: (7 cas crops, 2 cattle sectors, pigs and poultry), wereas te output from te remaining 3 rougage crop sectors is mainly consumed on-farm. Furtermore, te model describes te demand for 7 variable inputs (energy, labour, fertilisers, pesticides, contract operations, purcased rougage and concentrate feeds). A basic assumption underlying te model s description of tese variables is tat farmers exibit economic optimisation beaviour. Te model simulates economic beaviour on a large sample of representative Danis farms as a tree-stage procedure, comprising te determination of: cost minimising composition of variable inputs for given input s, profit maximising yield levels in respective lines of agricultural production for given input and commodity s and profit maximising land allocation and livestock densities for given s and quantitative restrictions. Te representative farms are subsequently aggregated into national, regional/local or typology aggregates using a set of aggregation factors, wic aim at ensuring te maximum feasible consistency wit existing structural population variables at te relevant levels of aggregation. Tus, key parameters in te model are: a) parameters describing te tree stages of economic beaviour at te farm level, and b) te set of aggregation factors for aggregating farm data to iger levels. Empirical estimation of tese key parameters is te main issue in te present report. Econometric estimation of te key beavioural parameters as been carried out using panel data econometric metods on a large set of farm accounts data. In order to allow for major tecnological differences between farm types and soil types, distinction as been made between four farm types: part-time farms and full-time crop, cattle and pig farms, and between farms on two different soil types: and in total eigt different groups of farms. 8 SJFI A Regional Econometric Sector Model for Danis Agriculture

10 Selected estimated beavioural parameters for te eigt farm groups are given in te table below. Te parameters include tree types of parameters: yield-endogenous own- elasticities, output-compensated own- elasticities and elasticities of transformation between different production activities. Te parameters are described in more detail in capters 4-6, and in appendix D and F. Selected estimated beavioural parameters for all eigt farm types Crop, Crop, Cattle, Cattle, Pigs, Pigs, Parttime, Parttime, Yield-endogenous ( Marsallian ) own- elasticities Fertiliser Pesticide Labour Output-compensated ( Hicksian ) own- elasticities Fertiliser Pesticide Labour Elasticities of transformation between production activities Land-non-land Cereals-pulses/rape Spring barley-weat Yield-endogenous (or Marsallian) elasticities reflect te combined effects of yield adjustments (cange in crop yield per ectare or yield per animal) and input substitution for given land allocation and livestock activity. Yield adjustments are due to canges in te input-output relationsips, wereas te input substitution effects are due to canges in relations between different inputs. Tere seem to be similarities in te own- elasticities for many of te inputs across most farm types, but also some systematic patterns concerning farm and soil types. For example, te elasticities for purcased fertilisers are around 0.6 for most farm types on y soil and 0.5 for most farm types on y soil. Pig farms, owever, seem to ave significantly lower elasticities for fertilisers (around 0.1). Tis may reflect a ig self-sufficiency wit animal manure on tese farms, but it may also reflect tat pig farmers are less concerned about partial optimisation of fertilisation tan e.g. crop producers, because te cost of fertiliser is a minor sare of pig farmers total costs. A pattern similar to tat for fertilisers is observed wit respect to labour, wereas no systematic pattern across farm and soil types seems to be present in te case of pesticides. It sould be noted tat cross elasticities are less omogenous, because tey to a iger extent also reflect differences in input composition on te eigt farm types. A Regional Econometric Sector Model for Danis Agriculture SJFI 9

11 As indicated above, te yield-endogenous effect may be decomposed into a yield effect and an input substitution effect. Output-compensated (or Hicksian) own elasticities represent te latter component, i.e. te pure sort-run input substitution effect triggered by a cange in te considered, e.g. te of fertilisers, assuming tat yield levels (per ectare or ead) as well as activity levels remain constant. Tese elasticities are in general lower (in absolute value) tan te yieldendogenous elasticities, reflecting te fact tat te yield effect of an input increase in general is non-positive. Across farm and soil types, te output-compensated elasticities sow patterns similar to tose of te yield-endogenous elasticities above. Tus, for fertilisers and labour, te elasticities appear to be larger (in absolute value) on y soil tan on y soil, wereas te pattern is less obvious for pesticides. Te difference between te Marsallian and Hicksian elasticities represent te yield effect of an input cange. For example, te own- effect of a fertiliser cange due to te pure crop yield effect can be calculated as for crop farms on y soil. Tus, te yield effect constitutes a relatively small sare of te total effect of input canges on input use. Elasticities of transformation represent canges in te composition of agricultural activities due to canges in te economic returns in te respective activities. Specifically, te elasticities describe te percentage cange in te ratio of activity levels due to a percentage cange in te ratio between economic returns of te two activities. For example, if te economic returns to non-land based (pig and poultry) production increases by one per cent wile keeping te economic return to land-based activity (crop and cattle production) uncanged, te number of pigs/poultry per ectare increases by per cent on crop farms on y soil. Or if te economic returns to spring barley area increases by one per cent relative to tat of weat, te ratio between areas for spring barley and weat will increase by per cent for crop farms on y soil. Due to data sortages, it as not been possible to estimate all tese elasticities. However, te results in te table seem to indicate tat te difference between farm and soil types is relatively limited concerning te transformation between landbased and non-landbased activities. On te oter and, farms on y soils seem to be more eager to switc between cereals and pulses/rape tan farms on y soils, wereas te opposite is te case for te switc between spring barley and weat, as te elasticities are iger for y soil farm types. 10 SJFI A Regional Econometric Sector Model for Danis Agriculture

12 Te above results demonstrate some similarities, but also systematic differences between te different farm groups. Tus, compared wit a more uniform approac to agricultural sector modelling, te current framework must be expected to yield a more precise description of economic beaviour in te agricultural sector, because te variation in beavioural parameters are explicitly taken into account. Te combination of te farm-based approac and an aggregation sceme ensuring consistency wit aggregate figures as proven useful for several types of economic analysis related to te agricultural sector. Some illustrative applications of te model are described in Jensen (2002), were economic impacts of various canges in te economic conditions are assessed for te agricultural sector as a wole and for different groupings of te agricultural sector. Te model as also proven useful for spatial analyses (Rygnestad et al., 2000), were te model as been combined wit GIS-data in order to investigate te economic potentials for aforestation and drinking water protection witin a specific study area in Denmark. Te combination of a large number of diverse farm types and a consistent aggregation sceme also enables linkage wit oter analytical tools. One example at te aggregate level is a set of scenario analyses, were te model is linked to macro-economic models at te national as well as te municipality level in order to assess te impacts of canges in agricultural and agri-environmental policies on te economy of Danis rural municipalities (Hasler et al., 2002). Anoter example is te use of te model in environmental assessment of macroeconomic projections (Andersen et al., 2001). Te farm-level basis as also proven useful for analysing environmental policy instruments, because it takes into account te diversity of farms, tus enabling analysis of instruments, wic address te farm level. For example, Jensen et al. (2001) supplement te model wit detailed data for nitrogen balances and pesticide use, and te expanded framework as been used for economic comparison of transferable versus non-transferable reduction requirements on te use of pesticides. Te ig level of disaggregation is expected to make te link wit specific environmental satellite models (e.g. a model for nitrate leacing) relatively easy. As mentioned at te outset, te main part of tis report provides a tecnical documentation of te specification and estimation of ESMERALDA. Capter 1 provides an introduction to te problems addressed by ESMERALDA, as well as a discussion of te similarities and differences wit a previous version of te model, wereas capter 2 gives a structural overview of te model. In capter 3, te data underlying te A Regional Econometric Sector Model for Danis Agriculture SJFI 11

13 model are presented, and capters 4-6 describe te teoretical specification and econometric estimation of model s tree types of beavioural equations at te farm level: input composition equations (cap. 4), yield equations (cap. 5), and activity level equations (cap. 6). Capter 7 describes and demonstrates te aggregation procedure used for aggregating variables from te farm level to te national or local level. Finally, capter 8 draws some perspectives from te work and discusses some of te strengts and limitations to te approac. Readers wit main interest in model potential applications may tus skip te relatively tecnical capters 3-7, and peraps supplement wit illustrative applications from Jensen (2002). 12 SJFI A Regional Econometric Sector Model for Danis Agriculture

14 1. Introduction Te present capter provides an introduction to te contents of te report, including a discussion of te aims of te model, a brief overview of existing approaces to agricultural sector modelling, and a sort discussion of te main similarities and differences between te current and a previous version of te model Background and objectives of te model Te roles of te agricultural sector in rural development as gained attention during te last decade, in te European Union and oter regions among te industrialised countries. Tis increased attention can be seen in te ligt of an ongoing structural development in most industrialised countries. On te one and, te structural development implies growt in manufacturing and service industries, wereas primary industries are stagnating. Consequently, te significance of agriculture and oter primary sectors for income and employment is on te decrease. On te oter and, agricultural activities are concentrated on still fewer production units, in terms of a rapid development in farm structure and a relatively strong growt in labour productivity in te farming sector, wit consequences for agricultural employment, rural landscapes etc. At te same time, parttime farms constitute an increasing sare of te total number of farms. Te focus on rural development can owever also be seen in te ligt of efforts to liberalise international trade in agricultural products, as well as ongoing or expected canges in environmental policies related to agriculture. A major concern in tis respect is, to wic extent te economy in (agriculture-based) rural areas will be affected by suc canges in te economic conditions surrounding te agricultural sector. Te increasing focus on rural development and environmental aspects, as well as te pressure to remove trade-distorting policy instruments, as implied a movement in te range of agricultural policy instruments in te direction from orizontal measures to more targeted and differentiated instruments. Tis development is expected to continue in te future. In order to perform quantitative assessments of suc questions, concerns and policy trends, tere is a need for quantitative economic models, wic take into account te A Regional Econometric Sector Model for Danis Agriculture SJFI 13

15 regional diversities in te agricultural sector and in te economic structures in general, as well as te interactions between te agricultural and oter economic sectors. Based on tese considerations, te objective of te present model is treefold: 1. to enable economic analyses of canges in te Danis agricultural sector at te national as well as te local (municipality) level. Te scope for economic analysis includes comparative analysis of different situations as well as different scenarios for development of te agricultural sector. 2. to integrate te agricultural sector analysis wit oter model tools, e.g. macro-economic models at te national and local level, and ence enable more olistic analytis of te economic development in Danis rural districts, and Denmark as a wole. 3. to improve te possibilities for integrated economic and environmental analysis related to agriculture specifically environmental problems wit a non-point source nature. Agricultural sector economic analysis at te national and local level Te model aims at representing te Danis agricultural sector at te national level as well as for eac of te 275 municipalities in Denmark, were results from te individual municipalities add up to te national aggregate. Te modelling approac is based on farm data from a large sample of Danis farms, and te representation of individual municipalities is based on an aggregation sceme, taking into account caracteristics of te farm structure in te respective municipalities. Integration of agricultural sector analysis wit oter model tools Te development of te model as aimed at facilitating te link to macro-economic models, in order to enable consistent economic analyses, wic take into account te interrelations between agriculture and oter economic sectors including te feedback of responses and te consistency of relevant beavioural parameters. 14 SJFI A Regional Econometric Sector Model for Danis Agriculture

16 Integrated economic-environmental analyses Due to te farm-based structure of te ESMERALDA-model 1, it is relatively wellsuited for analyses of fairly detailed environmental policy regulations, and te municipality focus of te model also enables analysis at a fairly geograpically disaggregated level, a feature wic is useful for analyses addressing site-specific environmental issues Status of agricultural sector modelling ESMERALDA can be caracterised as an agricultural sector model, altoug tis term as been used for various analytical concepts and metodologies during te last 2-3 decades. Depending on te issue to be analysed various approaces ave been applied, spanning te range from te igly detailed specific farm level (see e.g. Jacobsen et al., 1999) over te agricultural sector level to te global general economic level (for a review of international trade models, see e.g. van Tongeren et al., 2000). In te following, we use te term agricultural sector model for a model focusing on te agricultural sector at te national or regional level. Internationally, te dominating approac to agricultural sector modelling as been tat of matematical programming 2, wereas te use of econometric metods for establising large agricultural sectoral models as been less extensive in te most recent years 3. Te matematical programming approac builds on explicit optimisation of an objective function (e.g. minimisation of costs, maximisation of agricultural profits or some equilibrium/welfare measure), subject to a number of restrictions (e.g. resource constraints, crop rotation restrictions etc.). Te explicit formulation of tecnical restrictions etc. provides an opportunity for model specification at a relatively detailed level, including for instance te switc between different manure andling tecnologies, different feeding strategies etc., but also requires proper specification of all details. Te use of matematical programming for agricultural sector analysis as been exten- 1 Econometric Sector Model for Evalutating Resource Application and Land use in Danis Agriculture. 2 Examples are Stryg et al., (1995), Wiborg (1999), Day (1963), Eurostat (1995), Heckelei & Britz, (2000), Letonen (1999), Bauer & Kasnakoglu (1989), Osterburg et al. (2000), Flury et al. (2000), Malitius et al. (2000), Helming et al. (2000), Jacobsen et al. (1999). 3 See owever Oude Lansink & Peerlings, 2000, for a survey. A Regional Econometric Sector Model for Danis Agriculture SJFI 15

17 sive since te work of Day (1963). One fundamental problem in matematical programming models as been te calibration problem, i.e. te ability of te model to reproduce empirical observations. Data availability often prescribes a linear specification, but suc linear programming models tend to produce corner solutions, wic typically deviate from empirically observed situations. Te use of matematical programming metods as owever undergone a renaissance since te launc of socalled Positive Matematical Programming (Howitt, 1995), wic as furter been developed by te use of Maximum Entropy metods in te calibration procedures, despite te lack of empirical quantitative foundation for te smoot response patterns implied by te quadratic calibration. Normally, an econometric model is not formulated as an explicit optimisation problem, altoug te econometrically estimated beavioural relations implicitly reflect economic optimisation. Compared wit te matematical programming approac, te econometric approac is less sensitive to aving full knowledge of all tecnical details in te production processes (altoug more knowledge is better tan less knowledge also in econometric models). Econometric models are empirically founded on istorical variations in data. Hence, te validity of te beavioural parameters in an econometric model is in principle limited to te data ranges, for wic tese parameters are estimated New developments compared to an earlier version of ESMERALDA Te model described in tis paper can be seen as a furter development of an earlier version of ESMERALDA, altoug te new version differs significantly from te previous version in several respects. Te following gives a brief overview of te main similarities and differences between te two model versions. For more details on te old model version, see Jensen (1996). Te current model version describes production in 14 (wic can be disaggregated into 21) of te most important agricultural sub-sectors: weat, winter barley, oter cereals (spring barley, rye, oats, oter), pulses, rape, potatoes, sugarbeets, fodder beets, green fodder in rotation (grass and silage cereals), permanent grass, dairy cattle (dairy cows and eifers), beef cattle (slaugtering calves and nurse cows), pigs (sows and baconers) and poultry in a relatively simple dynamic setting. Te earlier model version was purely static comparative and described production beaviour in 19 subsectors. Among te variables modelled in bot model versions are activity levels 16 SJFI A Regional Econometric Sector Model for Danis Agriculture

18 (ectares or livestock units) in te sub-sectors, intensity variables (output and input use per activity unit) in te sub-sectors, and tus total production and input use. In contrast to te new version, were focus lies on optimisation at te farm level, te old model version covers te aggregate (national) level in eac agricultural subsector, and te model structure assumes separability between different agricultural sub-sectors, in tat optimization in te individual sub-sectors determines production intensity, and optimization at te aggregate level (for given production intensity) determines te activity levels. In bot versions, most product and input s are assumed to be exogenous, due to te fact tat te Danis agricultural sector accounts for only a small part of te entire EU agricultural production as well as te total Danis economy, and tus te market power on bot output and input markets is limited. Some of te s of internally produced inputs (rougage and some oter non-traded feeds and breeding animals) are endogenous, owever. Bot model versions are based on an econometric approac, were model equations are formulated on te basis of an assumption of economic optimisation beaviour using duality teory and te translog functional form. In te old version, econometric estimations were based on aggregate farm accounts data at te sub-sector level, provided by te Danis Institute of Agricultural and Fiseries Economics. In order to enable te econometric estimation, a number of separability assumptions were imposed, because suc assumptions reduce te number of parameters to be estimated econometrically, and tis was necessary, given te number of observations available for te estimation. Te new version is estimated on farm level panel data from a large sample of Danis farms, wic reduces te requirement for separability assumptions. Te old model version links te individual sub-sectors troug a set of model conditions, i.e. identities and pysical restrictions (e.g. land availability), economic equilibrium conditions, regulations, etc. Hence, te effects of specific regulations on e.g. land use can be assessed in te model, provided tat te model implementation of te regulation makes sense at an aggregate level. In te new model version, specific equations for land allocation, livestock density etc. ave been estimated directly. Tus, te new model version presented in tis report sares some caracteristics wit te old version, in tat it is based on an econometric approac, duality teory, translog functions, economic account data from Danis Institute og Agricultural and Fiseries Economics and an explicit modelling of te linkages between different lines of A Regional Econometric Sector Model for Danis Agriculture SJFI 17

19 agricultural activity. However, tere are also significant differences between te two model versions. 18 SJFI A Regional Econometric Sector Model for Danis Agriculture

20 2. Overview of te model Tis capter intends to provide an overview of te model, and te interrelations between te different components of te model. Altoug tere are parallels, it may be useful to distinguis between te structure of te model itself, and te structure of te econometric estimation procedure General features of te model Te purpose of te model is to describe production, input demands, land allocation, livestock density and various economic and environmentally relevant variables on representative Danis farms, and subsequently in te Danis agricultural sector at relevant levels of aggregation. Tese variables are assumed to be functions of te economic conditions facing te farms, including agricultural s, economic support scemes, quantitative regulations etc. A basic assumption underlying te model s beavioural description is tat farmers exibit economic optimisation beaviour. Te model covers 14 lines of agricultural production (plus fallow), of wic 11 yield a marketed output: - 7 cas crops: spring barley (covering spring barley, rye and oats), winter barley, weat, pulses, rape, potatoes and sugar beets - 3 rougage crops: fodder beets, green fodder in rotation (covering grass in rotation and silage cereals) and permanent grass - 2 cattle sectors: dairy cattle (covering dairy cows and rearing cattle) and beef cattle (covering nurse cows and slaugtering calves) - pigs (covering sows and baconers) - poulty - fallow Of tese, te outputs from rougage and fallow are not marketed in te model. Fallow is not supposed to yield an output at all, and te production of rougage is assumed to serve as on-farm input in cattle production. Along wit te 11 commercial outputs, te model determines demands for 7 variable inputs in te sort run (energy, labour, commercial fertilisers, pesticides, contract operations, purcased rougage and purcased concentrate feeds). In te longer run, te model determines canges in activity levels (land allocation and livestock density), input of capital and derived effects on outputs and demands for sort-run variable inputs. Based on canges in s, A Regional Econometric Sector Model for Danis Agriculture SJFI 19

21 quantities etc., a number of economic variables can be determined: output value, variable costs, gross margin etc. Te model is based on anonymous farm account data from Danis farms per year in te period 1973/74 to 1995/96. Tese data comprise land use, livestock erds, labour and capital input, output revenues from different agricultural products and variable input costs at te farm level Structure of te model Te main principle in te ESMERALDA model is to determine economic beaviour on a number of representative Danis farms, and subsequently aggregate tese farm level results to te relevant level or type of aggregation. Te economic beaviour at te farm level includes determination of input composition, production intensity in individual lines of production as well as activity levels (numbers of ectares or animals) in eac line of production. Examples of relevant aggregation scemes could be te construction of geograpic (national or regional/local) or typologic (main production, farm size etc.) aggregates. Te overall structure of te model is illustrated in figure 2.1. Model analysis is based on data from one year s sample in te agricultural accounts data base (approximately 2000 farms). Te contents of te dotted box in te figure represent beaviour at te farm level on eac of tese representative farms, as response to canges in economic conditions represented by te upper rigt module in te diagram. Farm level adjustment takes place in tree stages: 1) cost minimising adjustments in te composition of variable inputs, 2) sort-run profit maximising adjustments in te yield levels in individual subsectors (keeping land allocation and livestock numbers constant), and derived adjustments in te use of variable inputs 3) adjustments in activity levels (land allocation, livestock numbers) and capital, and derived canges in output and use of variable inputs 20 SJFI A Regional Econometric Sector Model for Danis Agriculture

22 Figure 2.1. Overview of ESMERALDA Economic and structural data from F sample farms v 1 v 2... v f... v F Canges in economic conditions (e.g. s, premia) Canges in composition of variable inputs in te sort run on F sample farms v 1 1 v 2 1 v f 1 v F 1 Canges in production intensities on F sample farms v 1 2 v 2 2 v f 2 v F 2 Canges in land use, livestock and capital, and derived canges in input and output on F sample farms v 1 3 v 2 3 v f 3 v 3 θ 11 θ θ 1f... θ 1F θ 21 θ 22.. θ 2F Aggregation of sample farm results : to aggregate results θ if V 1 Results V 2 at te : relevant V i level of : aggregation θ I1... θ IF V I In eac stage, te beavioural adjustments (e.g. to canges) are determined by econometrically estimated beavioural parameters (e.g. elasticities etc.). Specifically, 8 sets of beavioural parameters ave been estimated, representing 8 main farm types (part-time farms and full-time crop, cattle and pig farms on y and y soil, respectively). To eac farm in te model, te most relevant of tese 8 sets of parameters is attaced. A Regional Econometric Sector Model for Danis Agriculture SJFI 21

23 Te part below te dotted box in figure 2.1 represents te aggregation of individual model farms to te relevant level of aggregation. Te aggregation is carried out by means of an aggregation matrix, wic contains aggregation factors for eac model farm to eac of te relevant aggregates. Hence, te aggregation matrix represents te farm structure related to te considered grouping of farms. Te aggregation matrix is assumed to be independent of te economic conditions specified in te upper rigt module. Tis assumption migt be considered as a restrictive one. However, a study by Rasmussen & Wiborg (1996) indicates tat developments in te Danis farm structure seems to ave been fairly unaffected by observed canges in s and regulations. Te developed model structure can be considered as a refinement of te approac applied in Scou et al. (1998, 2000). Te model structure as similarities wit te approac used by Osterburg et al. (2000), altoug teir study is based on matematical programming simulation at te farm level Outline of te econometric estimation procedure Te econometric estimation procedure applied for providing estimates of te beavioural parameters in te farm level adjustments generally follows te stages outlined in te simulation model. Hence, te estimation procedure consists of tree stages wit a separate estimation model in eac stage: 1) model of sort run cost minimisation problem for given yield and activity levels. Te model determines te cost minimising composition of energy, labour, purcased fertilisers, pesticides, external services and purcased feeds for given s of tese components as well as given yield and activity levels in te respective agricultural sub-sectors (described in detail in capter 4). 2) extending sort run cost minimisation model to a sort run profit maximisation model, allowing yield levels to adjust, but maintaining activity levels in te respective sub-sectors (described in detail in capter 5). 3) model of medium- and long run profit maximisation in terms of adjustments in land allocation, livestock numbers and capital input (described in detail in capter 6). All tree stages of model formulation and estimation are based on te dual approac (see e.g. Cambers, 1988), were farmers economic optimisation problems are specified in terms of cost or profit functions. In all tree stages, te translog functional orm 22 SJFI A Regional Econometric Sector Model for Danis Agriculture

24 as been applied. Stages 1 and 2 ave been formulated and estimated as static relationsips, wereas te estimations in stage 3 build on a separable dynamic formulation of te long-run translog profit function Application and limitations of te model In its present version, te model can be used for economic static-comparative or dynamic analysis of canged conditions in te Danis agricultural sector, e.g. canges or restrictions on te production beaviour. Te farm-based structure of te model and te aggregation sceme yields te opportunity to distinguis economic effects on different farm types, in different regions etc. A general feature of econometric models is tat tey are based on beavioural equations estimated on istorical data. Tis feature may be a strengt because te estimated beavioural parameters reflect actually observed beaviour. It may owever also be a potential limitation to te use of te model, because te model can only be validated witin te data intervals spanned by te istorical observations. Tus, applying an econometric model for analysing canges beyond istorical variations is always problematic 4. Anoter limitation to te econometric approac is tat econometric estimation is restricted by te amount of available data. Tis may for example limit te potential for analysing detailed issues related to te use of fertilisers or pesticides, because te data material on tese issues is rater limited 5. 4 It may owever be noted, tat suc analyses are also problematic in oter analytical frameworks. 5 However, Jørgensen & Jensen (2000) ave proposed a solution to te latter problem. A Regional Econometric Sector Model for Danis Agriculture SJFI 23

25 24 SJFI A Regional Econometric Sector Model for Danis Agriculture

26 3. Data Tis capter describes te data material used for te econometric estimations and simulations. Te data material is combined from different sources, were te main contributions stem from agricultural statistics and agricultural accounts data at te farm level. Tese sources, as well as te preparation of data for te econometric estimation, are described in te following Agricultural data Te data for variable inputs used in agricultural production are sown in figure 3.1. Te main source of data as been te annual statistics provided by te Danis Institute of Agricultural and Fiseries Economics (1997c), supplemented wit data from Statistics Denmark and te Organisation of Danis Poultry Producers. Figure 3.1. Agricultural input s, index 1988= wage rate cattle feed pig feed poultry feed rougage fertilisers pesticides energy services For some of tese input s, te underlying input is not necessarily omogenous over time. For example, te composition of fertiliser and pesticide use as varied significantly during te considered period, and te same problem could be peraps be claimed in te case of te oter inputs as well. Assuming tat te data in figure A Regional Econometric Sector Model for Danis Agriculture SJFI 25

27 3.1 can be considered as proper index representations of te respective input categories, and assuming separability between te input categories represented by te series, te above data are suitable for te estimation purpose, altoug interpretation of te results must take into account te potential limitations due to tis problem. A general impression from te figure is tat s increased more rapidly in te 1970 ies and early 1980 ies, due to a relatively ig general inflation in tat period, wereas tey ave become more stable since te mid-1980 ies. Feed s are mutually correlated, and so are te s of energy, fertilisers and pesticides. Te wage rate seems to ave been developing at a fairly stable rate. Figure 3.2 illustrates te applied data for agricultural outputs. Figure 3.2. Prices of agricultural products, Index: 1988= barley weat peas rape potatoes sugar beets milk beef pork poultry As was te case wit agricultural inputs, te increases in agricultural output s were steeper in te first alf of te data period, due to general inflation in te Danis economy. Since te mid-1980 ies, te s of agricultural products ave exibited a decreasing pattern. For beef and crops, tis decrease as been strengtened by te 26 SJFI A Regional Econometric Sector Model for Danis Agriculture

28 MacSarry reform of te Common Agricultural Policy in , implying substantial reductions in support (replaced by area and eadage premia). Main sources for tese data are Danis Institute of Agricultural and Fiseries Economics (1997c), Statistics Denmark and te Organisation of Danis Poultry Producers. In contrast to te input data, te outputs underlying te data in figure 3.2 are more well-defined Farm level accounts data Te economic beaviour of Danis farmers is represented by a large set of anonymous individual economic accounts data provided by te Danis Institute of Agricultural and Fiseries Economics. Te dataset is constructed from a stratified sample of annual farm accounts drawn from te total population of Danis farmers to obtain representativity in all relevant respects (cf. Danis Institute of Agricultural and Fiseries Economics, 1997a). Te dataset consists of 24,690 observations and it covers te period from 1973 to Te number of observations included in te dataset from year to year varies but in general it lies around 1,500 accounts per year. Around 20 per cent of te sample is replaced eac year. Hence, farms are on average represented in te sample around 5 subsequent years. To enable micro-econometric analysis of te production beaviour in te Danis agricultural sector only farms represented at least two years are included in te dataset, wic can be caracterised as an unbalanced panel dataset. Te dataset is divided into eigt subsets according to farm type in order to obtain a reasonable level of omogeneity. Hence, four farm types (part-time farms and fulltime crop, cattle and pig farms) on two soil types ( and ) are distinguised. Tese eigt farm categories are expected to reflect te main sources of variation among farms. For example, attitudes towards economic optimisation may differ between full-time and part-time farmers, because te latter often ave oter income sources. Fertilisation beaviour may differ between crop and livestock farms due to te self-supply wit animal manure on te latter, and between cattle and pig farms due to significant differences in crop composition. Crop production beaviour may differ between y and y soils. Part-time farms are defined as farms, were te standard labour requirement per year is less tan one full-time (1,665 ours) working year (cf. Danis Institute of Agricultural and Fiseries Economics, 1997a, p. 131). Full-time farms, were at least two A Regional Econometric Sector Model for Danis Agriculture SJFI 27

29 tirds of te Standard Gross Margin are due to crop production are classified as crop farms, and analogously for cattle and pig farms 6. Farms located in municipalities were more tan 70 per cent of te area is are classified as y soil farms, and similarly for farms on y soil. Hence, farms from municipalities wit less tan 70 per cent of eiter or are not included in te data material used for econometric estimation. Mean values of te key variables from te farm accounts data in te 8 distinguised farm categories are sown in table 3.1. Some of te activity levels are aggregates of a group of activities. For instance, te spring barley area includes also rye and oats, te area wit grass and greenfodder in rotation includes silage cereals, te number of dairy cattle is a weigted sum of dairy cows and rearing cattle, te number of beef cattle is a weigted sum of nurse cows and slaugtering calves, and te number of pigs is a weigted sum of sows and baconers (using Livestock Units as weigts). Cost sares represent te individual variable inputs sares of total variable costs. Hence, te cost sares sum to one. Te output/cost ratios represent te ratio between te revenue from a certain output and te total variable costs. For example, te revenue from spring barley production corresponds to 53.1 per cent of total variable costs. Tese ratios are used for calculating elasticities referring to yield levels in capter 5 below. In addition to te variables represented in table 3.1, various proxy-variables ave been considered in an attempt to adjust for differences in degree of specialisation, farm management skills, etc. in te econometric estimations. 6 In addition, te group of pig farms include a minor sare of farms, wic are not specialised in one of te tree categories eiter because teir production is relatively diversified, or because tey specialise in oter lines of production (e.g. poultry or furred animals). 28 SJFI A Regional Econometric Sector Model for Danis Agriculture

30 Table 3.1. Mean values of farm data in eigt different farm categories, 1973/ /97 Crop, Crop, Cattle, Cattle, Pigs, Pigs, Parttime, Parttime, Number of obs Activity levels: Spring barley, a Winter barley, a Weat, a Pulses, a Rape, a Potatoes, a Sugar beets, a Dairy cattle, cow equiv Beef cattle, calf equiv Pigs, sow equiv Poultry, stock dkr Fodder beets, a Rot grass, a Perm grass, a Fallow, a Variable cost sares: Energy Labour Fertilisers Pesticides Services Rougage Conc feeds Output/variable cost ratios 1 : Spring barley Winter barley 0, Weat Pulses Rape Potatoes Sugar beets Dairy cattle Beef cattle Pigs Poultry Output/variable cost ratios are defined as te revenue from a given commodity divided by total variable costs Economic data for individual lines of agricultural production For te first two stages in te econometric estimation procedure (cf. section 2.3 above), te described data material suffices. However, in te tird stage of econometric estimation (modelling adjustments in land allocation and capital and livestock intensity, cf. capter 6 below) tere is a need for data on te economic returns in different agricultural lines of production as a driving force for canges in activity levels. A Regional Econometric Sector Model for Danis Agriculture SJFI 29