Prior to estimation, several sets of parametric restrictions are imposed on (4.9). We
impose symmetry, linear homogeneity in input quantities, and constrain βfq , ∀q, to equal
zero for one firm in order to achieve identification. Symmetry requires that
Ymm' = Ym'm, ∀ m, m', m = m'
Yzz' = Yz'z, ∀ z,z', z = z'
Ynn' = Yn'n, ∀ n,n', n = n. (4.14)
In addition, linear homogeneity in input quantities implies
Yn=1,
n
£ Ynn' = £ Ynn' =£ £ Ynn' =0,
Ymn = 0, ∀ m,
n
Yzn = 0, ∀ z, and
n
Ynt =0.
(4.15)
n
Finally, identification requires that βfq, ∀q, must be constrained for one firm in (4.11).
4.2 Data
Our dataset is an updated and refined version of the panel of utilities originally an-
alyzed by Nelson (1984).3 Subsets of that data were used by Baltagi and Griffin (1988)
and Callan (1991). The sample used here is comprised of 43 privately owned U.S. electric
3 We are grateful to Professor Nelson for making his data available to us.
14