Validation in Computable General Equilibrium Modeling

Validity is a key issue for consumers of computable general equilibrium (CGE) modeling services. What assurance can producers of CGE results give to consumers that a CGE analysis: (i) is computationally sound, (ii) uses accurate up-to-date data, (iii) adequately captures behavioral and institutional characteristics of the relevant part of the economy, (iv) is consistent with history, and (v) is based on a model that has forecasting credentials? This chapter gives some answers. With regard to (i), CGE modelers have an obligation to conduct exhaustive test simulations. The value of these procedures goes beyond computational checking. Test simulations are a practical way to become familiar with a model and often reveal modeling weaknesses. On (ii) and (iii), the most effective approach for displaying the relevant data and describing what is going on is via a back-of-the-envelope (BOTE) model. A well-designed BOTE model has two properties: it reveals the roles of major behavioral, institutional and data assumptions in causing a model to generate a particular result; and it is small enough to be managed with pencil and paper (on the back of an envelope) and to be presented in a limited timeframe to policy advisors. On (iv) and (v), the chapter describes various aspects of statistical validation, concentrating mainly on historical simulation, baseline forecasting and the testing of baselines against reality. This work demonstrates that CGE models can produce forecasts at a highly disaggregated level that comfortably beat non-model-based trend forecasts. It also demonstrates that there is considerable potential for improved CGE forecasts through conscientious data work and improved methods for projecting trends from historical simulations into forecasting simulations.