de

Advanced CGE Modeling

Gaining a critical understanding of
how model configuration impacts results

Course in the Doctoral Certificate Program in Agricultural Economics of German, Swiss and Austrian Universities

Background:

Computable General Equilibrium (CGE) models are a widely used type of economic simulation models to assess economy-wide impacts of changes in policies, markets, technology or climate change. Several agricultural economics group in Germany (for instance, at Berlin, Hohenheim, Kiel, Braunschweig, Bonn) have long-standing experience of using CGEs in their research. This is reflected in teaching maser courses providing an introduction into CGE modelling and in two courses in the the Doctoral Certificate Program. The course "Applied Computable General Equilibrium Modelling" focuses on a single-country CGE (STAGE realized in GAMS) and includes working with the actual software codes. A brief introduction to the multi-regional CGE model MAGNET along with the presentation of some recent policy assessments based on MAGNET is given in "Model-based policy impact assessment of agricultural policy", as part of an overview on multiple models used at the Thuenen instisute in Braunschweig. The course "Advanced Multi-Regional Computable General Equilibrium modelling" focuses on multi-regional Computable General Equilibrium and structural sensitivity analysis, and works with a Graphical User Interface and not directly with the software codes. Overlaps with two other courses are small, but both provide useful preparation.

Aim of the course:

Researchers face a wide range of potential methodological options when analysing their research questions with a CGE model. This relates, for instance, to regional, factor or product detail, to the choice of functional form in final demand, or to the structure and parameterization of Constant-Elasticity-of-Subsitution (CES) functions which describe the production function in the different sectors. The course aims at reaching a critical understanding how related choices impact model outcomes to foster more informed and appropriate configurations and parameterization of CGE models. This seems especially relevant in cases where PhD students develop (or extent) their own model, but also when working with an established model in their group which might obscure to some extent the view on alternatives. Two out of five days are devoted to a wider discussion on modelling trade and trade policies in CGE models, reflecting recent research projects of the instructors.

Moreover, the course aims at learning how to firstly focus on likely relevant changes as expected from the shock and model structure when analysing results from model runs. Additionally, it trains how to develop appropriate presentations based on CGE results, and how to work efficiently in small teams under time pressure. Such training is also part of the course "Applied Computable General Equilibrium Modelling".

Didactic approach:


Model choice

The tool used in this course is the flexible and modular platform for CGE modelling "CGEBox". It departs from the widely used GTAP Standard model as coded in GAMS by van der Mensbrugghe 2018 (The Standard GTAP Model in GAMS, Version 7; Journal of Global Economic Analysis, Volume 3 (2018), No. 1, pp. 1-83). A large part of the structure of the GTAP Standard Model is also found in other multi-regional CGE models used in the German speaking research community such as in GLOBE and MAGNET. Moreover, most CGE models, if single-country or multi-regional, share basic assumptions and structural elements with the GTAP Standard model (such as perfect markets, constant-returns-to-scale, the Armington assumption) and use the same functional forms. Most aspects discussed with and explored by the students, such as the effects of different aggregations, different nestings in the production function or the choice of functional form in final demand, apply to all CGE models. The choice of CGEBox is mainly motivated by its flexible and modular character which allows easily changing and exploring different model configuration based on its Graphical User Interface (GUI). This avoids that participants have to memorize model mnemonics and change software codes themselves. This enables participants to explore a wider range of model configurations.

Preparation of the presence part

The course starts with a first preparatory phase drawing on asynchronous learning based on material made available online (papers, slides, videos). The material also offers basics on CGE modelling, as presented in master courses or in the courses in the Doctoral Certificate Program. The information rely on multiple sources such as teaching and instruction material provided by senior researchers which contribute to the common doctoral program and work with CGEs. This preparatory phase should also allow students with limited or no pre-existing knowledge on CGE modelling to successfully participate at the course. It is complemented by material on more advanced topics to prepare for the second presence part of the course. In this context, each student will have to prepare a presentation to be presented to all others in the second part of the course.

During the preparatory phase, the students will install CGEBox on their computer, including GAMS (using a course license), and do some first simple model runs and analysis, following training videos, and supported by the relevant parts of the model documentation. This ensures that all participants arrive with a working model installation for the second part of the course, along with some first practical experiences of using the model.

During this phase instructors are available to help with technical and conceptual questions.

Presence part

The second and core part of the course consists mostly on learning-by-doing in small groups. Usually, each day is opened with a brief presentation of the methodological options in these fields. These presentations are mostly prepared by the students themselves during the preparatory online phase. After receiving clear instructions on what to do in the group work, and a demonstration of how to work with the GUI, each group will run different configurations of the CGE model CGEBox or use different data bases on the same shock and analyse results. Based on the Graphical User Interface of CGEBox, no knowledge of GAMS and variable, parameter and equation names is required.

The shocks to be analysed will draw on actual policy debates and/or published peer reviewed papers. They cover a wide range of topics explored by agricultural economists with CGE models (Climate Change impacts, bilateral and multi-lateral trade liberalization, taxing consumption choices). The instructors will closely monitor the progress of the groups. They will provide technical support and are available for methodological or conceptual discussions or clarifications. Each group is required to summarize their findings in brief presentations towards the end of each day. On some days, final presentations by an instructor or student expand the previously explored topics.

The work in small groups includes the preparation of the presentations and should also foster the soft skills of each participant, such as preparing well-structured presentations and appealing slides of appropriate layout which focus on the relevant content. All participants are expected to ask questions after the presentations and are invited to give feedback on the presentations of the other groups. In order to assure an appropriate supervision during the course, the number of participants is limited to 12.

A certificate of successful participation will be issued if

  • the student has properly prepared the assigned presentation and handed it in on time
  • participated actively at all group works
  • contributed to presentations

Course program:

Block 1: Preparation of the presence program

Block 2: Presence program

The presence part will take part at Institute for Food and Resource Economics of the University of Bonn at Nussallee 19 and 21, located in Bonn-Poppelsdorf. Bonn Central Station and the City center are about a 15 minutes walk away. Participants have to bring their own laptop along. Participants are also responsible to cover all related costs, such as food, travel or accommodation fees. Suggestions for hotels can be found here. Lunches can be taken in the nearby student restaurant in Poppelsdorf at guest prices.

A detailed program for each day of the week in presence is listed below.

Potential shock: Food Waste Reduction in Europe.

Related literature:

Jafari, Y., Britz, W., Dudu, H., Roson, R., Sartori, M. (2020): Can Food Waste Reduction in Europe Help to Increase Food Availability and Reduce Pressure on Natural Resources Globally?, German Journal of Agricultural Economics 69(2): 143-168.

Britz, W., van der Mensbrugghe, D. (2016): Reducing unwanted consequences of aggregation in large-scale economic models - A systematic empirical evaluation with the GTAP model, Economic Modelling,59:462-473.

Overview and round table (8:30-9:00):

  • Round table
  • Brief overview on course

Introductionary presentations (9:00 - 10:00):

  • Stud 1: An overview on the GTAP model, 20 minute presentation, 10 minutes questions and feedback
  • Stud 2: An overview on the GTAP Data Base, 20 minute presentation, 10 minutes questions and feedback

Group work block I (10:00-12:00):

  • W. Britz: Software demo: Using the CGEBox GUI to differently aggregate the global GTAP Data Base (15 minutes)


Analyse differences in model outcome when differently aggregated model data bases are used, with regard to:

  • Group 1: products
  • Group 2: regions
  • Group 3: factors

Lunch break (12:00-13:00):

Possibility to jointly visit student restaurants in Poppelsdorf

Group work block II (13:00-15:30):

Groups continue to work on the analysis and prepare the presentations

Presentations (15:30-17:30):

  • Group 1: Impacts of different product aggregations (20 minutes, 10 minutes questions and feedback)
  • Group 2: Impacts of different regional aggregations (20 minutes, 10 minutes questions and feedback)
  • Group 3: Impacts of different factor aggregations (20 minutes, 10 minutes questions and feedback)
  • Stud 3: Brief presentation on more detailed data bases (GTAP-Power, GTAP-Water) and option to dis-aggregate the GTAP data base further (20 minutes, 10 minutes questions and feedback)

Potential shock: changes in crop productivity under Climate Change.

Related literature:

Wilts, R., Latka, C., Britz, W. (2021): Who is Most Vulnerable to Climate Change Induced Yield Changes? A Dynamic Long Run Household Analysis in Lower Income Countries, Climate Risk Management, 33(100330).

Introductionary presentations (8:30 - 9:30):

  • Stud 4: Overview on production function nestings in different CGE models (GTAP Standard, MIRAGE, GLOBE) and GTAP variants (GTAP-E, GTAP-AGR), 20 minute presentation, 10 minutes questions and feedback
  • W. Britz: Flexible nestings in MAGNET, CGEBox and GLOBE, 20 minute presentation, 10 minutes questions and feedback

Group work block I (9:30-12:00):

  • W. Britz: Software demo: Using pre-defined nestings in CGEBox, changing their parameterisations, adding GHG and air emissions (15 minutes)


Analyse differences in model outcome when differently nestings/parameters are used:

  • Group 1: GTAP-E versus standard
  • Group 2: Food-Feed-Nestings versus GTAP Standard
  • Group 3: Using different subst. elasticities in GTAP standard

Lunch break (12:00-13:00):

Possibility to jointly visit student restaurants in Poppelsdorf

Group work block II (13:00-15:30):

Groups continue to work on the analysis and prepare the presentations

Presentations (15:30-17:30):

  • Group 1: Compare impacts under GTAP-E versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 2: Compare impacts under Food-Feed-Nestings versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 3: Compare impacts under different substitution elasticities (20 minutes, 10 minutes questions and feedback)
  • Stud 5: Brief overview on GTAP-AEZ (20 minutes presentation, 10 minutes questions and feedback)

Potential shock: EU-Canada Free-Trade Agreement

Related literature:

Jafari, Y., Himics, M., Britz, W., Beckman, J. (2021): It is all in the details: A bilateral approach for modelling trade agreements at the tariff line, Canadian Journal of Agricultural Economics 69(3): 415-442.

Jafari, Y., Britz, W., Guimbard, H., Beckman, J.: Properly capturing tariff rate quotas for trade policy analysis in computable general equilibrium models, Economic Modelling, available online 19th August 2021, 105620.

Introductionary presentations (8:30 - 9:30):

  • Stud 6: Comparison of the implementation of the Armington approach in different CGE models (GTAP Standard, MIRAGE, GLOBE), 20 minute presentation, 10 minutes questions and feedback
  • Stud 7: Integrating tariff line detail in CGE models, modelling TRQs 20 minute presentation, 10 minutes questions and feedback

Group work block I (9:30-12:00):

  • W. Britz: Software demo: Changing the trade configuration in CGEBox (30 minutes)


Analyse differences in model outcome when differently trade configuration are used:

  • Group 1: Different Armington aggregation versus GTAP standard
  • Group 2: Tariff line detail versus GTAP Standard
  • Group 3: TRQs versus GTAP Standard


Lunch break (12:00-13:00):

Possibility to jointly visit student restaurants in Poppelsdorf

Group work block II (13:00-15:30):

Groups continue to work on the analysis and prepare the presentations

Presentations (15:30-17:00):

  • Group 1: Compare impacts under different Armington configuration versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 2: Compare impacts under tariff lines detail versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 3: Compare impacts under explicite TRQ representation versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)

Potential shock: Carbon border tax adjustment

Introductionary presentations (8:30 - 10:00):

  • Stud 8: Approaches to deal with the "small shares stay small problem" of the CES/CET functional forms: CES with commitment terms, 20 minute presentation, 10 minutes questions and feedback
  • Stud 9: Approaches to deal with the "small shares stay small problem" of the CES/CET functional forms: Spatial Arbitrage, 20 minute presentation, 10 minutes questions and feedback
  • Stud 10: An overview on the Melitz model, 20 minute presentation, 10 minutes questions and feedback


Group work block I (10:00-12:00):

Analyse differences in model outcome when different trade configuration are used.

  • Group 1: CES with commitment terms versus GTAP standard (two-stage Armington)
  • Group 2: Spatial arbitrage versus GTAP Standard (two-stage Armington)
  • Group 3: Melitz versus Armington


Lunch break (12:00-13:00):

Possibility to jointly visit student restaurants in Poppelsdorf

Group work block II (13:00-15:30):

Groups continue to work on the analysis and prepare the presentations

Presentations (15:30-17:30):

  • Group 1: Compare impacts under CES with commitments versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 2: Compare impacts under Spatial Arbitrage versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 3: Compare impacts under Melitz versus GTAP-Standard (20 minutes, 10 minutes questions and feedback)

Potential shock: Taxing meat consumption in the EU

Introductionary presentations (8:30 - 9:30):

  • Stud 11: Comparison of widely used functional forms in final demand in CGE modelling (MAIDADS, CES, LES, CD, CDE), (20 minutes, 10 minutes questions and feedback)
  • Stud 12: Use of CES-subnests in final demand (20 minutes, 10 minutes questions and feedback)


Group work block I (10:00-12:00):

Analyse differences in model outcome when different configuration in final demand are used:

  • Group 1: MAIDADS versus CDE as in GTAP standard
  • Group 2: CD versus CDE in GTAP Standard
  • Group 3: CDE with sub-nests versus CDE as in GTAP Standard

Lunch break (12:00-13:00):

Possibility to jointly visit student restaurants in Poppelsdorf

Presentations (13:00-15:00):

  • Group 1: Compare impacts under MAIDADS versus CDE as GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 2: Compare impacts under CD versus CDE as in GTAP-Standard (20 minutes, 10 minutes questions and feedback)
  • Group 3: Compare impacts under CDE with sub-nests versus CDE as in GTAP-Standard (20 minutes, 10 minutes questions and feedback)

Final evaluation round and goodbye (15:00-15:30)