Testing among functional forms: an extension of the Generalized Box-Cox formulation

This paper uses the Generalized Box - Cox framework and Double Length artificial Regression to test whether different specifications of the profit function are able to mimic the technology underlying panel data of Dutch arable farms for the period 1970 - 1988. To this end, a linear GBC is developed that includes the Generalized Leontief, Normalized Quadratic and Symmetric Normalized Quadratic as special cases. A Lagrange multiplier test that avoids estimation of the linear GBC is used to test (subsets of) functional forms against (simplifications of) the linear GBC. Functional form results are also evaluated in terms of regularity conditions, parameter significance and reasonability of elasticities. On this data set, the NQ outperforms the other forms on most criteria.