Qualitative comparative statics and audits of model performance Original Research Article

It is often difficult to determine if large, computer based models actually conform to the mathematical descriptions given in their documentation. One way to study this problem is through a sensitivity, or comparative statics, analysis of the model. At issue is the degree to which characteristics of the Jacobian matrix associated with a solution to the model can be shown to imply characteristics of the inverse Jacobian matrix. Any necessary characteristics of the inverse Jacobian can be tested for via comparative statics. Sign nonsingularity of the Jacobian provides an opportunity to audit a modelʹs mathematical documentation through a comparative statics analysis. Prototype software was developed that tested for sign nonsingularity and applied to the Jacobian matrix corresponding to the U.S. Department of Energyʹs Oil Market Simulation Model. It was found that the signs of all of the elements of the inverse Jacobian matrix could be found. This