Propagation of data error and parametric sensitivity in computable general equilibrium model forecasts

While the computable general equilibrium (CGE) model is a well established tool used in economic analyses, it is often viewed as a blackbox because of the complexity of the model structure and the many assumptions made regarding the underlying calibration data and model parameters. To characterize the behavior of the CGE model, we perform a large-scale Monte Carlo experiment to examine its sensitivity to two major forms of uncertainty: that caused by the expenditure data used to calibrate the model to a fixed base year and that resulting from the elasticity of substitution parameters at the core of the model. By examining a variety of output variables at different levels of economic and geographical aggregation, we assess how uncertainty impacts the conclusions that can be drawn from the model forecasts. We found greater sensitivity to uncertainty in the full set of elasticity of substitution parameters than to uncertainty in the base-year expenditure data as the forecast year increases. While many model forecasts were conducted to generate large output samples, we found that few forecasts are required to capture the mean model response of the variables we tested. However, characterizing the standard errors and empirical probability distribution functions was not possible without a large number of forecasts.