The extension of the current CU between Turkey and the EU to cover agriculture would lead to multiple consumer and producer price changes in Turkey. A model used for the analysis of such a scenario should therefore be able to depict consumption, production, trade, and welfare effects of multiple and simultaneous price changes on interdependent markets.
As the Turkish agricultural sector is large in terms of employment as well as its share in GDP, fundamental changes in agricultural policy may effect the economy as a whole which, in turn, may have effects on the agricultural sector.
A Computable General Equilibrium (CGE) approach would therefore be desirable. On the other hand, CGE models typically do not cover the agricultural sector in sufficient detail to analyze complex changes in relative prices within the sector. As a number of CGE models depicting the Turkish economy exist (HARRISON et al. 1996, MERCENIER and YELDAN, 1996), in this study policy scenarios are analysed with a partial equilibrium model limited to the agricultural sector. Any fundamental changes in the agricultural sector revealed by the partial approach (nominal protection rates) could then be fed into a CGE model and CGE results (changes in real exchange rate, factor prices, and income) could then be fed back into the partial approach in order to adequately cover general equilibrium effects. MÜNCH (2002) has applied such an approach for the simulation of EU Eastern enlargement.
In order to model farm supply, two principally different modeling concepts are conceivable: a linear or nonlinear programming approach, or an econometric approach based on behavioral equations. A programming approach has the advantage of not needing estimates or assumptions of behavioral parameters while allowing for very detailed modeling of production technology. This second advantage is also a major drawback as detailed knowledge of production technology is required for the formulation of activities and restrictions. As Turkish agriculture displays a high degree of product variety, many restrictions are necessary to obtain a sufficient number of products in the model solution.
These restrictions are often arbitrary and in the end determine the model solution. The positive mathematical programming approach tries to overcome this drawback by introducing nonlinear cost terms in the objective function in a model calibration procedure (HOWITT, 1995). This approach, however, cannot substitute for extensive and detailed knowledge of the production technology and results depend heavily on assumptions made with respect to the functional
form of the cost function. Recent approaches using time series data for model calibration (HECKELEI, 2002) attempt to overcome this deficit and bridge the gap between programming models and econometrically estimated models.
Advantages of an econometric supply model are a high degree of transparency, the possibility of combining behavioral parameters from various sources (literature, expert guesses, own estimates), and the ability to implement conditions derived from the economic theory of the profit-maximizing entrepreneur. For this study such a supply model based on behavioral equations is also chosen because of the high heterogeneity of the Turkish agricultural sector and limited knowledge of production technology.
The chosen supply model is regionalized for three reasons. First, the supply model is not estimated, but rather based on information from various sources. In particular, assumptions on the relationship of area substitution among crops are based on expert knowledge of production technology and plausibility considerations. Due to the high heterogeneity of production regions, such considerations are made easier within relatively homogeneous production regions than on a national level. Second, the relative importance of the agricultural sector in terms of employment and income differs considerably among regions and questions on regional socioeconomic effects are often of high relevance to policy makers. Third, a regional supply model, if supplemented by a regional demand model, allows for regional price differentiation and the explicit coverage of domestic transportation costs. This would be one way to cope with the a poorly developed infrastructure in a large country which could lead to limited regional price transmission for products with high transportation costs (e.g. meat or dairy products). This approach, however, has not been pursued in this study due to the lack of information on regional demand. Limited price transmission is therefore accounted for by using elasticities of transmission of the world market price to the domestic price below unity for selected products (see Sections 5.3.3 and 6.4).
Several considerations played a role in determining the functional form of farm supply. On the one hand, a high degree of consistency with economic theory is desirable. Supply systems derived from flexible profit functions such as the Translog, Quadratic, or Symmetric Generalized McFadden function fulfill this requirement well as they allow for the global implementation of homogeneity of degree one in prices, symmetry of cross-effects, non-negativity of the own price effect, nonpositivity in input prices, and convexity in prices.32 Nonetheless these
32 For a systematic overview of second order flexible functional forms and a critical discussion of the local approximation concept see Feger (2000).
functions are less often used in applied policy simulation models. The main reason for this may be that applied policy models are rarely estimated, but make use of existing estimates of elasticities and best guesses. Existing estimates usually are based on a high variety of functional forms and in most cases, point estimates of supply elasticities are taken from various studies to implement them in any simulation model. Also for best guesses, the elasticity concept has great intuitive appeal. Therefore applied econometric policy simulation models are often of the constant elasticity type.33 The process of parameter generation, however, is not a sufficient reason for abstaining from supply systems derived from flexible functional forms. This is because the well known supply systems derived from second order flexible profit functions usually have the same amount of parameters as those of a constant elasticity system: one for each input- and output-price variable. Any such supply system can thus be calibrated based on a complete matrix of supply elasticities collected from any source.
WAHL et al.(2000) pursue this approach for the Central and Eastern European Countries Agricultural Simulation Model (CEEC-ASIM), which has a Symmetric Generalized McFadden supply system and a Normalized Quadratic demand system calibrated to sets of supply and demand elasticities.
Another drawback of supply systems derived from a profit function is that the resulting supply functions usually display total supply dependent on price variables without distinguishing between area and yield components. This distinction, however, which can be made with other types of supply systems, is helpful for applied policy simulation for two reasons. First, many policies concerned with direct payments, set aside, or production quotas are based on area. Second, strong assumptions are often made with respect to total crop area.
Most often the assumption is made that total area stays constant under different policy scenarios and any area effects for individual products are due only to the composition of production. But other assumptions, for example, area reduction due to set aside policies, are certainly possible (e.g. MÜNCH 2002, p. 58 ff.).
Supply systems which cannot be derived from a profit function, like constant elasticity supply systems, also have disadvantages. For example, in constant elasticity systems the requirement of symmetry of the cross-price effects can only be met locally as the first derivatives of the supply functions are not second order derivatives of any profit function. Therefore, such a supply system does
33 For example, this holds for the European Simulation Model (ESIM) (Munch, 2002), the Static World Policy Simulation (SWOPSIM) Modeling Framework (Roningen et al.
1991), the World Food Model (Anderson and Tyers, 1993), and parts of the OECD Ministerial Trade Mandate Model (MTM) (OECD 1987,1989) as well as its successor the AGLINK model (OECD, 1992).
not ensure producers act in perfect consistency with the economic theory of the profit maximizing entrepreneur, and the resulting welfare measures are path dependent if welfare changes are assessed sequentially with price changes introduced stepwise. Deviations due to the path of integration, however, are typically small (see Section 5.7.2). Homogeneity and non-negativity of the own price effect, however, can be ensured globally in a constant elasticity supply system. For this study a constant elasticity supply system has been chosen due its ability to separate effects on area allocation and yield, as well as its high degree of transparency.
For the modeling of human demand in Turkey, the evaluation of effects on different income groups is considered crucial. This is because income distribution in Turkey is rather unequal and distributional effects often are important when discussing policy options with different interest groups. The Gini Coefficient of income distribution in Turkey was 49 percent in 1994 (FÖRSTER, 2000, p. 75). Such a distinctly unequal distribution of income can be found in many African and Latin American countries, but is far above that of other OECD countries (except Mexico); e.g. 34 per cent in the US and 28 percent in Germany. In addition, income distribution has become more unequal in Turkey between 1987 and 1994 (STATE PLANNING ORGANISATION 2001, p.
109) while one of the declared aims of the Turkish Government is to reduce income inequality (STATE PLANNING ORGANIZATION, 1995, p.212; 2001, p. 111).
The aim of analyzing effects of agricultural policies on different income groups is achieved in this study by specifying constant elasticity demand systems for income quintiles which allow for consumption and welfare analysis for each quintile. As for the supply side, the constant elasticity form has advantages with respect to parameter generation and transparency. But the most important advantage, namely the separation of area and yield effect, is not relevant.
Therefore demand systems which better fulfill the global conditions of economic theory, like the Almost Ideal Demand System, could be calibrated based on existing elasticity sets. Such an approach is not pursued in this study.
Due to the possible dynamic effects of market integration on market structure (see Section 4.1.2) the model of the Turkish agricultural sector should be able to take into account decreasing marketing margins. In the chosen model this is done by separating wholesale prices from producer prices by exogenous marketing margins. Any degree of change in marketing margins due to increased competition is based on exogenous assumptions.
Another matter is how a model of the Turkish agricultural sector is placed in the international environment. As long as only policies of multilateral trade integration, for example the reduction of MFN tariffs, are analyzed, it seems
reasonable to assume Turkey is a small country on world markets thus making world market prices exogenous to the model. For policy simulations of a future period, world market price projections generated by large scale multicountry models as maintained by the FAPRI, the OECD, the World Bank, or the USDA can then be used as exogenous parameters. For some products, however, Turkey is a large supplier on the world market and Turkish export quantities have an effect on the world market price level, as is the case for hazelnuts and sultanas.
But rather than building a large scale world model in order to treat all world market prices as endogenous variables, import demand functions for the rest of the world can be specified for selected products if model results display large changes in export volume.
In analyzing market integration between Turkey and the EU, the question arises as to how to include the EU environment. Many analyses of EU market integration related to Eastern enlargement explicitly depict EU-15 markets and treat EU prices as endogenous (MÜNCH 2002, FROHBERG and WEBER 2002). An alternative is to assume EU prices to be exogenous for most products and to include EU import demand functions for products for which this seems suitable due to market size and the kind of policies applied.
For this study Turkey is generally assumed to be a small country in the world as well as the EU market, i.e. world market and EU prices are treated as fixed exogenous parameters. Although Turkey is a large country compared to the EU for some products, this approach is justified by three factors. First, many EU prices are institutional prices (dairy, sugar) and are therefore fixed. Second, expected trade effects are low for most products due to already relatively low border policies (see Chapters 2 and 3). Third, in some cases it is sufficient to specify EU import demand functions for selected products instead of maintaining a full EU agricultural market model component for all products.