(output) constraints , or employment constraints and the associated loss in profits resulting
from respective constraints. This method also provides us a measure of profits lost to the
other sources excluding expenditure, revenue (output) and employment constraints.
Assume there are k =1,...,K observations of inputs xk and outputs yk, where xk =
x1k, ..., xkn, ..., xkN ∈ R+N and yk = y1k, ..., ymk , ..., yMk ∈ R+M.Thek observations maybe for
the same firm over time or many firms at one point of time. Following Fare et al. (1985)
and others, we represent PE and TVE technology by the piece-wise linear relation
K KK
T = {(x, y): X zkymk ≥ ym,m=1, ...,M;Xzkxkn ≤ xn,n=1, ...,N;Xzk =1,z∈ R+K},
k=1 k=1 k=1
(4)
where zk is the “intensity variable” for activity k, and z = z1, ..., zk, ..., zK . The first two
constraints in (4) ensure that all input/output combinations in T are technically feasible,
while the last constraint admits variable returns to scale. The three constraints in (4) serve
to form convex combinations of the observed input and output data.
To incorporate the expenditure constraint, we need to partition the inputs. Suppose in-
puts can be partitioned into variable inputs xv and fixed inputs xf . For each input vector k =
1,...,K, let xk = (xk ,xf ¢, where xk = (xki , ...,xki ,...,XkI j,and xf = (xfI+1 , ...,xfN) . Denote
output prices by P = (p1, ..., pM) ∈ R+M and variable input prices by Wv = (wv1 , ..., wvi, ...wvI).
All enterprises are assumed to take the same input and output prices for each input and out-
put, hence the superscript k is dropped from all price vectors.
To introduce expenditure, revenue (output), and employment constraints into model, let
the maximum allowable expenditure be denoted E , minimum revenue be denoted Rc , and
minimum employment be denoted N. Following AG, the expenditure constraint for enterprise
k can be represented by:
Wvi xk1 +-----+ WviXkI ≤ Ek,
(5)
11