A global CGE model at the NUTS 1 level for trade policy evaluation

In recent years the development of the World Trade Organization (WTO) has generated a great demand for estimates of potential consequences of trade policy. The Uruguay round and Doha round negotiations are typical examples. The policy maker could be interested in having information about the effects of trade liberalization on income, production and other relevant macroeconomic variables. Computable General Equilibrium (CGE) models are an important tool for meeting this need because they allow a lot of trade information to be elaborated in a coherent economic structure where agents maximise their utility and firms maximise their profits. Today many governments and international institutions, e.g. the WTO, the European Commission (EC) and the World Bank (WB), use CGE models to assess the impact of global trade reform. In my thesis the attention is directed toward large-scale global CGE trade models, such as GTAP, MEGABARE and MIRAGE, used by international organizations (e.g. the WB, the WTO, the EC) for their analysis of trade liberalization. This type of models maintains a strong Walrasian spirit. Factors are fully employed, money does not explicitly figure into the model and a solution is made possible through relative prices. Nevertheless, some important non-Walrasian assumptions, such as imperfect competition and others, are introduced or can be introduced. A global approach has the unquestionable advantage of taking into account within the same theoretical structure the trade relationships of all countries or groups of countries in the world, such as the EU, the USA, China, India and Africa. With respect to this, it is very important to have a consistent economic global database that covers all parts of the world. GTAP, based in the Agricultural Economics Department at Purdue University (West Lafayette, Indiana), has been created to satisfy this need; It is a global network of researchers who conduct quantitative analysis of international economic policy issues, especially trade policy. GTAP is the most widely used dataset for global CGE trade models. It is very rich and practical, however it only allows analysis at the national level. CGE trade models exist at a sub-national level but they only consider a single region or a handful of regions. The CAPRI-GTAP (Jansson, Kuiper and Adenäuer, 2009), MONASH-MRF (Peter et al., 1996) and MIRAGE-DREAM (Jean and Laborde, 2004) models are examples of large-scale global CGE trade models which also include many regions. MONASH-MRF refers to the Australian regions, CAPRI-GTAP is specific to the agriculture sector of the EU and MIRAGE-DREAM considers the NUTS (Nomenclature d’Unités Statistiques) regions of the 25 members of the EU (Romania and Bulgaria did not belong to the EU in 2004). There are so few models because there is a lack of well-suited regional data concerning foreign trade. For instance, in the EU there is no complete dataset on foreign trade that is available for the NUTS regions. Concerning foreign trade, some information is available for some countries at the regional level, but this is not systematically the case. Thus, simplifying assumptions must be made to make the models manageable. In addition, this kind of model is very demanding both in terms of data and computational resources. Research teams, supported by public institutions, work on these models which are highly disaggregated at the geographical level. The objective of this thesis is to build a global CGE trade model at the NUTS 1 level for the 68 regions within the first 15 member states of the European Union. The aim is not to exactly reproduce the models mentioned above but the aim is to build a simple parsimonious CGE model. Data on value added, skilled labour and unskilled labour are available at the NUTS 1 level (EUROSTAT database) while simplifying assumptions arise for the remaining variables. Therefore a CGE trade model is built in which only the production is specified at the NUTS 1 level. The model is built starting from the MIRAGE model. This type of model should allow the consequences of trade policy in Europe to be investigated at a disaggregated geographical level while maintaining a global approach. The EU economy is very diversified and world trade agreements do not take into account the disparities existing at regional level. This geographical heterogeneity in the EU should be considered in WTO negotiations. In addition, it is of interest to assess how European workers respond to trade shock. Will they migrate to another European region? The model is applied to the 68 NUTS 1 regions in the EU15 mainly to assess the production reallocation across three sectors (agriculture, manufactures, services) in each NUTS 1 region after a world tariff liberalization in agriculture. Nevertheless, it can also be used to simulate other trade policy reforms according to the special interest of the researcher. Special attention is given to the economic interpretation of the trade policy effects. Indeed, a weak link of the CGE approach is the poor economic interpretation of the results. The results at the NUTS 1 level are the following. The tariff liberalization in agriculture determines a decrease for all the NUTS 1 regions in this sector. The most affected regions are East, West and South Austria, Ireland and Portugal. In the manufactures and services sectors it is possible to note inverse patterns of production at the NUTS 1 level. Indeed, Nisia Aigaiou-Kriti, Attica and Portugal show the greatest decreases in the manufactures while Ireland, East Austria and Luxembourg experience the greatest increase in the same sector. In contrast, Nisia Aigaiou-Kriti, Attica and Portugal exhibit the greatest increases in services and Ireland, East Austria and Luxembourg show the greatest decrease in this sector. The stylised model allows the key parameter to be determined for interpreting the results. This parameter is the sectoral difference between the ratios of unskilled labour intensity to skilled labour intensity. Indeed, skilled labour and unskilled labour can be considered as the source of the heterogeneity across the NUTS 1 regions. To summarize, trade policy strikes the agricultural sector and causes a production decrease in this sector for all the NUTS 1 regions. The NUTS 1 regions, which use unskilled labour in agriculture and skilled labour in manufactures and services sectors more intensively with respect to the other NUTS 1 regions, are the most affected regions in the agricultural sector. The decrease in the agricultural production, in turn, determines a production reallocation and reduces the labour demand for unskilled labour. As a result, in general the unskilled factor loses (the wage goes down) and the skilled factor wins (the wage goes up). However, in the NUTS 1 regions which use the unskilled labour in the manufactures and the skilled labour in services more intensively, the manufacturing production decreases and services production increases. In contrast, in the NUTS 1 regions, which use the unskilled and skilled factors in the manufactures and services sectors by similar intensities, the manufacturing production goes up and the services production goes down. The introduction of the labour mobility causes amplification effects for the NUTS 1 regions which experienced strong increases or decreases in the IND and SERV sectors under the assumption of perfect immobility at the NUTS 1 level. In general, this hypothesis has a strong impact on the outcomes and determines unrealistic variations of the production in the services and manufactures sectors after agricultural liberalization. These results are not intended to be realistic but are a guide regarding the relevance of the assumption about labour mobility. The change in the unskilled/skilled labour supply is consistent with the production reallocation results.