Clearly, the independent variables included in (1) assume values which are significantly greater
than unity. Hence, the translog function better approximates the underlying cost function if all
variables in (2) are divided by their sample mean before taking their logarithms (Christensen et al.,
1973). We have two outputs, three factor prices, one fixed factor and a technical change parameter
in the equation, thereby requiring 36 coefficients to be estimated. According to economic theory,
we impose that the cost function be linearly homogeneous in factor prices, by means of the fol-
lowing restrictions: