Driscoll, Paul J.

By: McGuirk, Anya M.; Driscoll, Paul J.; Alwang, Jeffrey Roger; Huang, Huilin
A misspecification testing strategy designed to ensure that the statistical assumptions underlying a system of equations are appropriate is outlined. The system tests take into account information in, and interactions between, all equations in the system and can be used in a wide variety of applications where systems of equations are estimated. The system testing approach is demonstrated by modeling U.S. consumer demand for meats. The example illustrates how the approach can be used to disentangle issues regarding structural change and other forms of model misspecification.
By: Driscoll, Paul J.
Taylor series-based flexible forms cannot be interpreted as Taylor series approximations unless all data used in estimation lie in a region of convergence. When flexible forms lose their Taylor series interpretation, elasticity estimates will be biased. When the flexible form is a translog, Rotterdam, or AIDS model, the region of convergence is shown to be the entire positive orthant. Regions of convergence associated with quadratic, Leontief, and any flexible form that does not employ logged arguments are smaller and may not encompass the entire data set. Implications for production and demand analyses and experimental design are discussed.
By: Driscoll, Paul J.; McGuirk, Anya M.
Quadratic flexible forms, such as the translog and generalized Leontief, are separability inflexible. That is, separability restrictions render them inflexible with regard to separable structures. A class functional forms is proposed that is flexible with regard to general production structures and remains flexible regarding weakly separable structures when separability restrictions are imposed, thus permitting tests of the separability hypothesis. Additionally, the restricted forms are parsimonious; that is they contain the minimum number of parameters with which flexibility can be achieved.