Recent Developments in CGE Models for Trade Policy Analysis Badri Narayanan Research Economist, Centre for Global Trade Analysis, Purdue University. Visiting Scholar, School of Environment & Forestry, University of Washington Seattle. Consulting Economist, McKinsey Global Institute. Email: badri@purdue.edu GIFTA Conference Bath University, January 2016
Outline Introduction to CGE Linked CGE-PE models CGE models with details Dynamic CGE modelling Trade Policy 2.0 Future Directions
CGE Modelling Computable General Equilibrium. General Equilibrium Economy-wide, enough time to achieve equilibrium. Computable Solution can be computed. Also called Applied GE. Typically based on neo-classical theories of production, consumption, international trade, etc., but alternative theories can be used too.
Components Agents: Producers and Consumers. Factor Endowments: Land, Labour, Capital, Natural Resources, etc. Commodities: Different products - both goods and services in the economy Markets: for commodities and factor endowments Equilibrium conditions: In N-1 markets: Walras Law Welfare Analysis
When and where to use CGE? International Trade Policies and their economy-wide impacts Other Public Policy Impacts in different sectors, after accounting for linkages; e.g: What happens to steel sector if taxes on auto are cut? Energy and Environment Impacts of demographic patterns Impacts of technology/technical change, etc
When and where to use CGE? Anytime and anywhere one wants to capture economy-wide impacts and to account for the fact that resources are limited! But what if one always wants this? When and where NOT to use CGE? Small sector/region/issue with rich data on particular aspects: impossible to incorporate into a CGE model, e.g., time-series/survey data. (You can feed them however) Complex model capturing structural details of a region/sector that cannot be fed into a CGE framework. Where there is not enough data/infrastructure required, e.g., I-O tables, computational software, etc.. Where one requires statistical significance of the results (although there is a way of doing a similar thing in CGE)
How to get the best of both worlds? Feeding the results from structurally rich PE model into CGE Using econometric estimates to calibrate the parameters in CGE Linking a specialized PE model to CGE Feeding the CGE results into a PE/econometric model
Linking CGE and PE Models Soft versus hard linking Hard linking: GTAP-PE/GE model to capture trade and tariff details in disaggregated sectors in the absence of IO/production/consumption data at disaggregated level (Narayanan, Hertel and Horridge, 2010) Soft Linking: GTAP-CAPRI link to capture rich agricultural details in trade policy (Pelikan, Britz and Hertel, 2011)
Linking CGE and PE Models More on GTAP-PE/GE: Why do this? Granularity of trade negotiations false competition trade/tariff variations at the disaggregated level How to do this? PE model to capture disaggregated trade and tariffs in ways consistent with this part of the CGE model CET for aggregated producers and CES for aggregated consumers, to link the two models Does it make a difference? Yes, to a degree that depends on trade/tariff variations in the disaggregated sectors.
Linking CGE and PE Models More on GTAP-CAPRI: Why do this? Trade policies in agricultural sectors are heterogeneous Intra-country details are important to capture, for the EU ag trade How to do this? CAPRI captures details on EU ag trade policies at regional level, while GTAP captures the global picture Parameters of the two models are tuned to agree with each other on common variables Does it make a difference? Yes, for agricultural sectors.
CGE Models with Details The main limitation of global CGE: lack of details - what are they? Sectors/industries Factors (types of land, labor, water, etc.) Intra-country details Stochastics/volatility Tariff/Tax Instruments Reason for this limitation: lack of data in global scale, modeling and computational power Now, we have models addressing this quite well!
CGE Models with Details Sectors/Industries: Land Labor Splitcom (Horridge, 2008) Can disaggregate sectors when data is available (e.g. Narayanan and Khorana, 2014). GTAP PE-GE and GTAP-CAPRI are examples of hard and soft linking respectively, to address sectoral details as well. Lot of work has been done in this area, using AEZs (Agro-Ecological Zones) for example (Hertel et al., 2008), by exploiting data developed by other scientists (e.g. Ramankutty, 2008). Multiple types of labor makes a difference in CGE model results (e.g. Effects of China s rise on US labor markets - McDougall and Tyers, 2006; Mirza, Narayanan and van Leeuwen, 2014). Migration: Gmig (Walmsley, 2005)
CGE Models with Details Water: Increasingly important in trade policy models, given the intensification of virtual water trade Given that water is not priced in many countries, it is difficult to measure its use or supply in economic models Another challenge until recent past (e.g. Siebert et al, 2013): lack of physical data on water supply and use across the world Liu et al (2014), Tol (2012), Narayanan et al (2015) etc.: some attempts at this, mainly by attributing the yield difference between irrigated and rainfed land to the value of water. Intra-Country details: The Enormous Regional Model (TERM) framework by Horridge (2009) and MyGTAP (Walmslley and Minor, 2014) can help hard-link intra-country models with global CGE. Converting region-sector combinations into sectors
CGE Models with Details Stochasticity/Volatility: By design as a deterministic model, CGE doesn t capture this, but recent models/approaches have factored this in: Monte-Carlo-like multiple simulations (Horridge and Pearson, 2014) Statistical/econometric parameters/shocks (Narayanan and Villoria, 2013; Leister et al 2012) Tariff/Tax Instruments Specific tariffs result in very different price transmission mechanism compared to ad valorem, which is the most common form of tariffs in CGE; this has been addressed now (Narayanan and Villoria, 2013; Narayanan and Pelikan, 2014; Leister and Narayanan, 2015)
Dynamic CGE Modelling Global CGE is mostly done in a comparative static way. Not ideal for medium-long run studies, because: It ignores accumulation of productive capital Many trade agreements (including TPP/TTIP) involve timed phase-out of tariffs etc., which is not possible to do in static setting Investment behavior (adaptive vs rational expectations, etc.) is very important yet impossible to model in static setting. Dynamic CGE models can fill this gap: e.g. GDyn (Ianchovichina and McDougall, 2000)
Trade Policy 2.0 A term coined by Cernat (2015) all of the above and any other model feature that can suitably capture the changing focus of trade policies (ala TPP/TTIP): Productivity is not exogenous anymore, but is trade-induced; Increasing returns to scale and imperfect competition (Elbehri and Hertel, 2005); firms are heterogeneous (Zhai, 2007 and Akgul et al 2015) Non-tariff barriers: not merely tariff-equivalents, but explicit mechanisms to capture specific aspects of them: Rules of Origin e.g. Global Value Chains (Walmsley et al. 2014), Local Content Requirements (Narayanan and Mahate, 2014; Flag et al 2015) Labor standards, environmental standards, intellectual property costs, barriers to investment, improved access to trade by adopting standards: Xiao and Ciuriak (2014), Francois et al (2014), Narayanan et al (2015).
Future Directions Most of Trade Policy 2.0 is yet to be resolved fully that is where the future lies! Some examples: Public procurement and trade How to model labor and capital standards explicitly by, for e.g., through a non-ad-hoc raise in factor costs and an endogenous expansion in market access. Competition versus innovation associated with Intellectual Property rules Global Value Chains and various measures and notions of value added and rules of origin Modelling different modes of services trade and migration Interacting CGE models with other inter-disciplinary approaches like complex networks, physical scientific models (e.g. crop, climate)