用Monte Carlo方法分析CGE模型的不确定性

Uncertainty analysis of CGE model using Monte Carlo method

采用Monte Carlo方法研究可计算一般均衡（CGE）模型参数不确定性传播问题，开发了随机计算工具。通过拒绝法由正态分布的随机数获得卢分布随机数，并利用拉丁超立方体抽样方法对TEDCGE模型的50个自由参数进行随机采样，考察模型输出的不确定性。考虑参数的不确定性时，中国2010年相对于基准情景减排碳排放量30％的边际成本的95％置信区间为（345，600）元／t，相应的国民生产总值（GDP）损失率置信区间为（0．85％，1．35％），GDP损失总量将在1415～2270亿元（2000年价格）之间。自由参数的相关性将带来显著影响，当参数之间是正的相关关系时，模型结果的不确定性将进一步扩大。

A solution method for the stochastic computable general equilibrium （CGE） model was tested using a Monte Carlo method to investigate the uncertainty propagation in CGE models. 50 elasticity parameters in a CGE model were treated as random variables with an assumed β distribution, The variables include the carbon tax rate and the GDP loss corresponding to various carbon reduction levels in China. The analysis shows that the confidence interval of the marginal carbon reduction cost is 345 -600 RMB/t using year 2000 prices with 95% confidence when controlling the carbon emissions to 30% lower than the business as usual （BAU） level in 2010. The corresponding confidence level for the GDP loss is 0.85% - 1.35% for a total GDP loss of 141.5 - 227.0 billion RMB in year 2000 prices. The correlations among the uncertainity parameters have much greater effect on the output uncertainty.