Variations on invariance or some unpleasant nonparametric arithmetic

The nonparametric approach to production or demand analysis involves testing data for compatibility with restrictions from theory. This avoids reliance on particular functional forms, When the data are not consistent with utility or profit maximization, adjustments in the data using linear programming can restore consistency. Such adjustments have been interpreted as taste changes or measures of technical change and technical change bias. This paper describes nonparametric approaches and shows that the programming methods that have been used yield results that are not invariant to the scaling of prices and quantities. A solution is proposed that solves the invariance problem.