Linkage of GTAP and DRAM for scenario assessment: methodology, application and some selected results

Introduction A model that is fully consistent on all levels of aggregation from micro to macro is not available and probably also not feasible. Looking at the agricultural sector, sector models give details about the agricultural sector, but there is no interaction between the agricultural sector and the rest of the economy. On the other hand, more macro-oriented models give too few details for the agricultural sector, especially concerning supply response in the face of sometimes complex policy measures and specific agronomic features. As a result different type of models exists and if there is any overlap between the models they could produce different results for the same variables. To counteract this problem and to reach results that are more consistent with economic behavior at different levels of aggregation, different models can be linked. The goal of this paper is twofold. First, it describes the technical issues connected with linking the Global Trade Analysis Project (GTAP) model and the Dutch Regionalized Agricultural Model (DRAM) in one consistent system of models. DRAM is a non-linear, partial equilibrium, positive mathematical programming model of the Dutch agricultural sector. It generates production volume for a number of crops and animal products as well as (among other outputs) manure at the regional level (Helming, 2005). GTAP is a standard comparative static multi-region Applied General Equilibrium (AGE) model of trade and production at the world level. Inaddition the paper describes the linking of economic results with a GIS-based system to generate spatially disggergated results. Second, the paper assesses, using the developed model system, the economic consequences of two contrasting scenarios for food production and nature management in The Netherlands. Method and data: The GTAP model was linked to two other models: - the Land Use Scanner - in order to obtain spatially disaggregated results concerning agricultural land use in The Netherlands; - the DRAM – in order to take into account production changes caused by policies and product or region-specific technologies present in DRAM but not included in the GTAP model. The emphasis of this paper is on the GTAP model –DRAM link. The parameters of the non-linear cost functions of DRAM, corresponding to a certain scenario are calibrated on the basis of price and quantity pairs derived from GTAP simulations. This is achieved by mapping price and quantity changes of outputs per sector from GTAP to price and quantity changes of corresponding outputs and sectors in DRAM, and subsequently recalibrating the parameters of the relevant functions, that determine supply and demand such that marginal revenue equals marginal costs. Moreover, results from GTAP for a given scenario are also used to calculate scenario-specific price elasticities of demand per sector. With these elasticities, the parameters of the inverse linear demand functions for domestic final demand and for export demand functions for roughage and young animals can then be calibrated. Version 6 of the GTAP data for simulation experiments is used. The social accounting data are aggregated to 18 sectors and 37 regions. DRAM uses 1996 database with the activity levels updated to 2002 using Agricultural Census data. Prices and technology are partially updated for 2002. The developed model system was used to assess the economic consequences of two contrasting scenarios for food production and nature management in The Netherlands. Scenarios developed in EURURALIS project were used (see Klijn and Vullings (eds.), 2005). The scenario analysis was done in the recursive dynamic manner for three consecutive time periods 2001 – 2010, 2010 – 2020 and 2020 – 2030. Main findings: A coherent GTAP model - DRAM link, has three advantages. First, it enables us to assess implications of the worldwide economic scenarios for the Dutch agricultural sector at the regional level. Secondly, it enables us to examine the economy wide consequences of policies and technological changes present in DRAM and absent in the GTAP model. Finally, it enables the endogenization of prices of output and input in DRAM consistent with global equilibrium conditions. In this way, we utilize the advantages of both modeling approaches: DRAM is specifically strong in generating agricultural supply response, whereas GTAP takes care of general equilibrium effects. The applied iterative linking procedure of the GTAP model and DRAM converges for most of the commodities. The most important difficulties that have been encountered when models were linked and solved are: - Differences in model structure, definition and specification of variables and units As GTAP and DRAM have different objectives it is not surprising that model structure, definition and specification of variables and unities are different. Essentially, because the models have different objectives and domains, it is interesting to link the models and increase the domains of both models individually. Nevertheless, the results could be improved if definition and specification of variables were harmonized. - Differences in base situation At the time the study was done a fully specified DRAM database was only available for 1996. In this study GTAP uses 2001 as the base year. It is recommended that both models use the same year as the starting position. It is also recommended to use average figures over a three to five year period, to take into account yearly fluctuations in yields in prices. These result for example from differences in weather circumstances. - Sometimes large differences in costs shares of different costs components per sector The most important issue why both models behave differently is the differences in costs shares. It is recommended for future applications that costs shares are harmonized in both models before the models are linked. The analysis of simulation results shows that the manure policy present in DRAM and absent in the GTAP model seriously affects the GTAP model results concerning production development of the Dutch agricultural sector. Manure policy is a very important issue in the Dutch livestock sector, as its regulations basically restrict the possibilities for production expansion. Moreover, the results show that the development of agricultural incomes depends greatly on the speed of overall economic development and less on the policy towards the agricultural sector. Therefore, plausibility of these macro-economic assumptions is very important for the quality of the projections. Literature Klijn, J.A, and L.A.E. Vullings (eds.), 2005, The EURURALIS study : technical document, Alterra-rapport 1196, Alterra, Wageningen. Helming, J.F.M. (2005). A model of Dutch agriculture based on Positive Mathematical Programming with regional and environmental applications. PhD Thesis, Wageningen University, The Netherlands.