Agricultural Economy

Demand for food (animals and crops) products (LEITAP)

General

The adjusted version of the Global Trade Analysis Project (GTAP) model (Hertel, 1997; Van Meijl et al., 2006) included in the IMAGE 2.4 framework offers a viable approach for assessing the impact of non-agricultural sectors on agriculture. It also fully accommodates factor markets with special land modelling features (Eickhout et al., 2006b).

An extended land allocation tree

The standard GTAP land-use allocation structure is extended by taking into account that the degree of substitutability differs between types of land (Huang et al., 2004) using the more detailed OECD’s Policy Evaluation Model (PEM) structure (OECD, 2003) (see Figure 1). It distinguishes different types of land in a nested 3-level CET structure. The model covers several types of land use with different suitability levels for various crops (i.e, cereal grains, oilseeds, sugar cane/sugar beet and other agricultural uses).

Figure 1: Land allocation tree within the extended version of GTAP (LEITAP).

The lower nest assumes a constant elasticity of transformation between "vegetables, fruit and nuts" (HORT), "other crops" (e.g. rice, plant-based fibres; OCR), the group of "Field Crops and Pastures" (FCP) and non-agricultural land (NAG). The transformation is governed by the elasticity of transformation σ1. The FCP-group is itself a CET aggregate of Cattle and Raw Milk (both Pasture), "Sugarcane and Beet" (SUG), and the group of "Cereal, Oilseed and Protein crops" (COP). Here, the elasticity of transformation is σ2. Finally, the transformation of land within the upper nest, the COP group, is modelled with an elasticity σ3. In this way the degree of substitutability of types of land can vary between the nests. Agronomic features are captured to some extent. In general, it is assumed that σ3> σ2 >σ1, which implies that it is easier to change the allocation of land within the COP group, while it is more difficult to move land out of COP production into, say, vegetables. The values of the elasticities are taken from PEM (OECD, 2003).

Factor market segmentation

Wage differentials between agricultural and non-agricultural can be sustained in many countries (especially in developing countries) through limited off-farm labour migration (De Janvry et al., 1991). Returns to assets invested in agriculture also tend to diverge from returns of investment in other activities. 

To capture these stylised facts, we incorporate segmented factor markets for labour and capital by specifying a CET structure that transforms agricultural labour (and capital) into non-agricultural labour (and capital) (Hertel and Keeney, 2003). This specification has the advantage that it can be calibrated to available estimates of agricultural labour-supply response. The economy-wide endowment of labour (and capital) remains fixed, so that any increase in supply of labour (capital) to manufacturing labour (capital) has to be withdrawn from agriculture, and the economy-wide resources constraint remains satisfied. The elasticities of transformation can be calibrated to fit estimates of the elasticity of labour supply from OECD (2003).

Land-supply curves

To capture the (changes in) land quality and describe regional variations in land rent, we implemented land supply curves in LEITAP (Tabeau et al., 2006). These curves serve to translate the biophysical information on land productivity (based on soil and climatic conditions) generated by IMAGE to land rent. The first concept of using land supply curves was derived from Abler (2003). We calibrated the curves using FAO land-use projections and IMAGE results.

The supply of agricultural land depends on its biophysical availability (potential area of suitable land), institutional factors (agricultural and urban policy, policy towards nature) and land price. In IMAGE, the productivity for seven food crops is calculated for each 0.5 by 0.5 degree grid cell with the crop growth model of IMAGE <<link to TVM>>. For LEITAP this information is aggregated to an estimate of the overall productivity for each grid cell. This overall crop productivity is expressed on a relative scale between 0 and 1 on the basis of the potential crop productivity. Land productivity curves are obtained by ordering all grid cells in each of the 24 world regions from high to low productivity, and summing the total area (Figure <<2>>). The land productivity curve can be translated to a land-supply curve, assuming that the land price is a function of the inverse of the land productivity ( see Figure 2). 

Figure 2: Land productivity and land-supply curve for Canada (top) and China (bottom) on the basis of IMAGE simulations.

We estimated the land-supply function for 24 countries and regions by fitting it with a non-linear least square estimation method. Since the inverse of yield is not a good proxy of real land price if land is scarce or when only a small fraction of the potentially available area is actually used, we allowed a lower fit to the data at the beginning and ending of the curves. The asymptote of the land-supply curve is an estimate of the availability of land in each region. All grid cells with yield values of zero (mainly ice and desert), as well as urban area and protected bioreserves are excluded.

North Africa, EU, the Rest of Western Europe, Former Soviet Union, Middle East, China, Japan and Oceania currently are on the steep part of their land-supply curve and the associated land-supply elasticity with respect to the land price is lower than 1. Agricultural land is not scarce in Canada, and here expansion of agriculture can take place without a fast increase in the land price. The opposite situation can be seen in China. A small expansion of agricultural land here will lead to a high increase in the land price, and this stimulates the intensification processes in the agricultural production systems (Figure 2).

Exchanging land productivities between GTAP and IMAGE

Figure 3 shows the methodology of iterating the extended version of GTAP with IMAGE (Eickhout et al., 2006b). Yields in GTAP depend on an exogenous part (the trend component) and an endogenous part with relative factor prices (the management component). The exogenous trend of the yield is taken from Bruinsma (2003), where macro-economic prospects are combined with local expert knowledge to produce a best guess of the technological change for each country for the coming 30 years (Eickhout et al., 2004). However, many studies indicated this change in productivity to be enhanced or reduced primarily by climate change (Rosenzweig et al., 1995; Parry et al., 2001; Fischer et al., 2002). Moreover, the amount of land expansion or land abandonment will have an additional impact on productivity changes, since land productivity is not homogenously distributed over each region.

Figure 3: Scheme showing the methodology of model interaction (iteration) between GTAP and IMAGE.

The output of GTAP used for the iteration with IMAGE comprises sectoral production growth rates and the endogenous determined intensification or extensification (Figure 3). The exogenous assumptions based on FAO (Bruinsma, 2003) are translated to IMAGE parameters (management factor). For crops, the endogenous GTAP values are added to this management factor within IMAGE. For pigs and poultry, the additional intensification is added to the animal productivity of these commodities. For dairy and non-dairy cattle and sheep and goats, the additional value is added to the grazing intensity.

Subsequently, the IMAGE model calculates the yields, the demand for land and the environmental consequences on crop productivity. The results depend on changes in the demand for food and feed and a management factor as computed by GTAP. The land allocation procedure <<link to LCM>> yields the area of agricultural land needed for each world region and the corresponding changes in yields related to the extent and productivity of the land and climate change. These additional changes in crop productivity are returned to GTAP (Figure 3).

A general feature is that yields decline if land expansion occurs, since the productivity of land last taken into production is by definition lower than that of the existing agricultural area. When the agricultural land area is close to the potential area, even marginal land may be taken into production. In the short term, these factors may even be more important than the effects of climate change.

The iteration between GTAP and IMAGE ends when land-use projections in both models are similar. Since GTAP bases its calculations on land supply curves from IMAGE, the amount of iteration needed is limited (for crops, only two rounds of simulation).

Scenario analysis

The GTAP-IMAGE modelling framework can be used to assess the economic and environmental consequences of different scenarios. Recent applications of the modelling framework are published in Eickhout et al. (2006a, 2007). In the rest of IMAGE the production of crops and animal products is used in other models, (e.g. the Land Cover Model and the Land Use Emissions Model). A complete overview of LEITAP as part of the IMAGE 2.4 framework is given in Eickhout et al. (2006b).

• to  Determination of animal husbandry and pasture production
• to  external references
• to  MNP references