Peter J Wilcoxen > Computational General Equilibrium

A Quick Reference Guide to CGE Modeling

This is a draft guide to building and using computational general equilibrium models. It is gradually being replaced by a more detailed version available via the link above.

Characteristics of CGE models

How to build a model

  1. Choose the number and type of agents in the model.
    • Regions
    • Households
    • Firms
    • Governments
  2. Specify each agent's optimization problem.
  3. Solve for the behavioral equations implied by the optimization problems.
  4. Add market-clearing and accounting equations.
  5. Drop one market-clearing equation because of Walras Law.
  6. Choose one price to be the numeraire.
  7. Check that the model is closed by counting equations and variables.
  8. Condense the model (if necessary) to lower computational costs
  9. Estimate the behavioral parameters.
  10. Choose base-case values of exogenous variables.
    • Set to historical values for counterfactual analysis
    • Set to business-as-usual values for prospective analysis
  11. Choose an appropriate software package to solve the model.
  12. Write a program implementing the model.
    • Write it in a form as close as possible to the underlying economics
    • Include extensive comments
  13. Solve for the base-case equilibrium.
  14. Test the model thoroughly.
    • Is it homogeneous in prices and wages?
    • Does it satisfy Walras Law?
    • Does it scale up and down correctly in quantities?
    • Check any other experiments whose results can be predicted.

How to use a model for policy analysis

  1. Decide how to express the policy in terms of the model's variables.
  2. Think about what the policy could be expected to do to the economy.
    • Ideally, do analytical comparative statics
  3. Run the simulation and compare the results to the base case.
  4. Analyze the experiment and explain why the results come out the way they do.
  5. Evaluate the policy
    • Equivalent variation is strongly preferred to GDP as a welfare measure
  6. Determine confidence intervals for the results.
Site Index | Zoom | Admin
Peter J Wilcoxen, The Maxwell School, Syracuse University
Revised 05/20/2004