Journal of Environmental Management (1997) 51, 229–240

Productivity and Undesirable Outputs: A Directional Distance

Function Approach

Y. H. Chung, R. Färe and S. Grosskopf

Department of Economics, Southern Illinois University, Carbondale,

IL 62901-4515, U.S.A.

Received 26 April 1996; accepted 16 June 1997

Undesirable outputs are often produced together with desirable outputs. This
joint production of good and bad outputs is typically ignored in traditional
measures of productivity since “prices” are typically unavailable for bad
outputs. Here we introduce a directional distance function and use it as a
component in a new productivity index that readily models joint production of
goods and bads, credits firms for reductions in bads and increases in goods,
and does not require shadow prices of bad outputs. This index, as an empirical
example shows, solves the problem caused by the joint production of good and
bad outputs, and provides a practical managerial tool.
 1997 Academic Press Limited

Keywords: productivity, directional distance function, Malmquist–Luenberger

productivity index.

1. Introduction
The measurement of productivity has traditionally focused on measuring marketable
outputs of firms or industries relative to paid factors of production. This approach,
which typically ignores the production of by-products such as pollution, can yield
biased measures of productivity growth. Firms in industries that face environmental
regulations would typically find that their productivity is adversely affected since the
costs of abatement capital would typically be included on the input side, but no account
would be made of the reduction in effluents on the output side.∗
The major stumbling block in including the output effects of pollution control is
the fact that traditional productivity indexes (such as the Törnqvist and Fisher indexes)
require prices of all inputs and outputs in order to aggregate to form a total factor
productivity index. In the case of pollution, computation of a (shadow) price is not
straightforward. One could use emissions trading prices, or following Pittman (1983)

∗ Several other studies address the issue of performance measurement in the presence of regulation of
bads, including Färe et al. (1986, 1989b), Färe et al. (1989), Haynes et al. (1993), Ball et al. (1994) and Tyteca
(1995). These studies do not, however, explicitly compute a productivity index.

0301–4797/97/110229+12 $25.00/0/ev970146  1997 Academic Press Limited

230 A directional distance function approach

or Färe et al. (1993) one could estimate a shadow price. One possible solution is to use
a productivity index that does not require information on prices of effluents, for
example, the Malmquist index; see Färe and Grosskopf (1996). However, in the presence
of undesirable outputs, this index may not be computable.
Here we propose a new index, which we call the Malmquist–Luenberger productivity
index which overcomes the shortcomings of the original Malmquist index. This index
readily allows for inclusion of undesirable outputs without requiring information on
shadow prices. It also explicitly credits firms or industries for reductions in undesirable
outputs, providing a measure of productivity which will tell managers whether their
“true” productivity has improved over time. This index also tells managers if there has
been technical progress (a shift in the best practice frontier) and whether they are
catching up to the frontier. Since the index is computed using a data envelopment
analysis type approach, information concerning benchmark firms and technical efficiency
is also generated for individual firms.
In order to illustrate the applicability of this index, we compute productivity for
data from the Swedish paper and pulp industry. We begin with a discussion of the way
in which we model technology, and then turn to our measure of productivity based on
this model. Section 4 includes a discussion of our data and results. Section 5 provides
a brief conclusion.

2. Modelling technology with good and bad outputs

The basic pollution problem is that production of “good” outputs, such as paper or
electricity, is typically accompanied by the joint production of undesirable by-products,
such as suspended solids or SO2. The fact that goods and bads are jointly produced
means that reduction of bads will be “costly”: either resources must be diverted to
“clean-up” (e.g. scrubbers), production must be cut back, or fines must be paid.
M
More formally, if we denote good outputs by yvR+ , bad outputs by bvR+I , and inputs
N
by xvR+, then we can describe technology in a very general way via the output sets

P(x)={( y, b): x can produce ( y, b)}. (2.1)

We model the idea that reduction of bads is costly (what we call weak disposability
of undesirable outputs) as:

( y, b)vP(x) and 0ΖhΖ1 imply (hy,hb)vP(x). (2.2)

In words, this states that a reduction in bads is feasible only if goods are simultaneously
reduced, given a fixed level of inputs. In addition, we assume that the good or desirable
outputs are freely disposable, i.e.

( y, b)vP(x) and y′Ζy imply ( y′, b)vP(x). (2.3)

The notion that the good inputs are jointly produced with the bads is modelled by

if ( y, b)vP(x) and b=0 then y=0. (2.4)

In words, (2.4) says that the good outputs are “null-joint” with the bad outputs if the
only way to produce no bads is by producing zero good outputs. Alternatively, this
Y. H. Chung et al. 231
y (Good)

A
B

g C
(y, b) P (x)
b (Bad)
O

Figure 1. Distance functions.

means that if a good output is produced in a positive amount some bad output must
also be produced. Conditions (2.2)–(2.4) will be incorporated into our computational
model discussed in Section 3.
The original Malmquist index uses Shephard output distance functions to represent
technology (Shephard, 1970). These are defined as

Do(x, y, b)=inf{h:(( y, b)/h)vP(x)}. (2.5)

This function expands the good and bad outputs (y,b) proportionally as much as is
feasible.∗ It does not credit reduction of bads, since both types of outputs are expanded
at the same rate, which is one of the reasons we wish to modify the original Malmquist
index.
The reciprocal of the distance function is known as the Farrell output measure of
technical efficiency (Farrell, 1957); see Färe et al. (1994) for details. In order to allow
for the possibility of crediting firms for the reduction of bad outputs, we use a directional
output distance function instead of the Shephard output distance function to represent
technology. In contrast to the Shephard output distance function which seeks to increase
goods and bads simultaneously, the directional distance function seeks to increase the
good outputs while simultaneously decreasing the bad outputs. Formally, it is defined
as†

)
Do(x, y, b;g)=sup{b:( y, b)+bgvP(x)}, (2.6)

where “g” is the vector of “directions” in which outputs are scaled. In our case, g=(y,
−b), i.e. good outputs are increased and bad outputs are decreased.
To illustrate the directional output distance function and to compare it to Shephard’s
output distance function, let us again represent the technology by an output set. If we
impose conditions (2.2)–(2.4) on this set, it may take the form shown in Figure 1.

∗ The output distance function is a complete characterization of the technology, and it was shown by
Färe and Primont (1995) that under weak disposability of outputs,

( y, b)vP(x)<=>Do(x, y, b)Ζ1.

The distance function is also homogeneous of degree+1 and concave in outputs; see Färe and Primont
(1995). We do not impose separability of inputs or outputs.
† Briec (1995) develops a distance function for the growth of the technology similar to Luenberger’s
shortage function. See Luenberger (1992a, 1992b, 1994a, 1994b, 1995a, 1995b).
232 A directional distance function approach

The output set is denoted by P(x), good output by y and the bad by b. The outputs
( y, b) are weakly disposable and y by itself is strongly disposable. Moreover, the good
output y is null-joint with b, since if b=0, then the only y with ( y, b) v P(x) is y=0.
Shephard’s distance function applied to the output vector ( y, b) places it on the
boundary of P(x) at A, and yields a value of OC/OA, i.e. if goods and bads were both
increased by a factor of OA/OC, the firm would be judged efficient. Of course, if this
firm faces regulations concerning the bad, we would not want increases in the bad to
give the firm a better performance score. In contrast, the directional distance function
starts at C and scales in the direction of increased goods and decreased bads and
projects C on the boundary at B. In Figure 1 this amounts to the ratio of the distances
(BC/Og). This means that if the firm moved from C to B (i.e. reduced bads and
increased goods) it would be judged efficient based on the directional distance function.
We note that the values of the directional distance function for point B is zero, and at
C it is positive-valued given our g vector.
In order to relate the two distance functions to each other, let g=( y, b), then
through (2.6), we get

)
Do (x, y, b;y, b)=sup{b:Do(x,( y, b)+b( y, b))Ζ1}
=sup{b:(1+b)Do(x, y, b)Ζ1}
1
=sup{b:bΖ −1} (2.7)
Do(x, y, b)
=1/Do(x, y, b)−1.

This expression shows that Shephard’s output distance function is a special case of the
directional distance function. The relation between the two can be written as

)
Do(x, y, b;y, b)=(1/Do(x, y, b))−1 (2.8)

or equivalently

)
Do(x, y, b)=1/(1+Do(x, y, b;y, b)) (2.9)

3. Productivity measurement
Färe et al. (1989a) defined a productivity index based on Shephard’s output distance
function. Their index is the geometric mean of two Malmquist productivity indexes,
which were introduced by Caves (Caves et al., 1982). They named their index after the
Swedish statistician Sten Malmquist, who used the distance function in 1953 in defining
input quantity indexes.
In this section we give a short presentation of the output-oriented Malmquist
productivity index and we introduce our new productivity index based on the directional
distance function. We call this the Malmquist–Luenberger productivity index.
Suppose there are t=1, . . . , T time periods, then the Färe, Grosskopf, Lindgren
and Roos (FGLR) output-oriented Malmquist productivity index is defined by
Y. H. Chung et al. 233

C D
Dto(xt+1, yt+1, bt+1) Dt+1
o (x
t+1
, yt+1, bt+1) 1/2
t =
Mt+1 . (3.1)
Dto(xt, yt, bt) Dt+1 t t t
o (x , y , b )

The Malmquist index (3.1) can be decomposed into two component measures, one
accounting for efficiency change (MEFFCH), and one measuring technical change
(MTECH). These are

Dto(xt+1, yt+1, bt+1)

t =
MEFFCHt+1 (3.2)
Dto(xt, yt, bt)

and

C D
Dto(xt+1, yt+1, bt+1) Dto(xt, yt, bt) 1/2
t =
MTECHt+1 . (3.3)
Dt+1
o (x
t+1
, yt+1, bt+1) Dt+1 t t t
o (x , y , b )

The product of the component measures exhausts the productivity measure so that

t =MEFFCHt
Mt+1 t+1
·MTECHt+1
t . (3.4)

This index has several nice features. First of all, it is a total factor productivity
index. Perhaps most importantly, it only requires information on input and output
quantities in contrast to the Fisher and Törnqvist indexes which in addition to the
quantity data, requires information on input and output prices, which makes these
other indexes inappropriate for measuring productivity in the public sector or in the
presence of non-marketable goods like pollution. Although the Malmquist index can
in principle “deal with” pollution since it requires no prices, as shown above, the
distance functions on which it is based do not allow us to credit firms for reductions
in pollution. To do so, we substitute directional distance functions for the output
distance functions in the Malmquist index and rename it the Malmquist–Luenberger
productivity index.
To define an output-oriented Malmquist–Luenberger (ML) productivity index that
is comparable to the Malmquist index we choose the direction to be g=( y, −b) and
define it as:
Definition (3.5): The output-oriented Malmquist–Luenberger productivity index,
with undesirable output, is

) )

C D
(1+Do t(xt, yt, bt,−bt)) (1+Do t+1(xt, yt, bt;yt,−bt)) 1/2
ML t+1
t = ) )
.
(1+Do t(xt+1, yt+1, bt+1;yt+1,−bt+1))(1+Do t+1(xt+1, yt+1, bt+1;yt+1,−bt+1))
(3.5)

Our definition is such that when the direction g is ( y, b) rather than ( y, −b), the
Malmquist–Luenberger index coincides with the Malmquist index. As in the case of
the Malmquist index, the new index can also be decomposed into two components,
namely
234 A directional distance function approach
)
1+Do t(xt, yt, bt,−bt)
MLEFFCH t+1
t = )
. (3.6)
1+Do t+1(xt+1, yt+1, bt+1;yt+1,−bt+1)

) )

C D
{1+Do t+1(xt, yt, bt,−bt)} {1+Do t+1(xt+1, yt+1, bt+1;yt+1,−bt+1) 1/2
MLTECH t =
t+1
) )
{1+Do t(xt, yt, bt,−bt)} {1+Do t(xt+1, yt+1, bt+1;yt+1,−bt+1)}
(3.7)

and their product equals MLt+1t

. The Malmquist–Luenberger measure and the Malmquist
measure indicate productivity improvements if their values are greater than one and
decreases in productivity if the values are less than one.
Our next task is to develop a procedure for calculating the two indexes and their
decompositions. This requires the computation of four distance functions for each
index. We assume that at each time t=1, . . . , T, there are k=1, . . . , K observations of
inputs and outputs,

(xt,k, yt,k, bt,k),k=1, . . . , K,t=1, . . ., T. (3.8)

In our example k is a paper and pulp mill.

Following Färe et al. (1994) the output set that meets conditions (2.2)–(2.4) and is
derived from the data (3.8) is

G
K

P(x)= ( y, b): ]z y
k=1
t
k km [ytm,m=1,...,M, (3.9)

]z b =b ,i=1,...,I,
k=1
t
k ki
t
i

]z x
k=1
t
k knΖxtn,n=1,...,N,

zk[0, k=1,...,K .
H
This activity analysis model (3.9) also satisfies constant returns to scale, i.e.

P(kx)=kP(x), k>0 (3.10)

and strong disposability of inputs,

Y. H. Chung et al. 235
y (Good)
(yt + 1, b t + 1)

A
Pt(x t + 1)
g

b (Bad)
O

Figure 2. A mixed period distance function.

x′[xtP(x′)hP(x). (3.11)

The inequalities for inputs in (3.9) makes them freely disposable, and the same
holds for the good outputs. The bad outputs are modelled with equalities; this makes
them not freely disposable. Finally, the non-negativity constraints on the intensity
variables zk allow the model to exhibit constant returns to scale.∗
For each observation the distance functions in the Malmquist index are computed
as the solutions to a linear programming problem. For example, for k′,

(Dto(xt,k′, yt,k′, bt,k′))−1=maxh (3.12)

K

s.t. ]z y
k=1
t
k km [hytk′m,m=1,...,M,

]z b
k=1
t
k k′i=hbtk′i,i=1,...,I,

]z x
k=1
t
k k′n Ζxtk′n,n=1,...,N,

zk[0, k=1,...,N,
To use the original Malmquist productivity index when undesirable outputs are
present may create a problem, which we illustrate in Figure 2.
The technology is based on t period data as indicated by Pt. The observation being
evaluated is from the following period t+1, (xt+1, yt+1, bt+1). Note that ( yt+1, bt+1) lies
outside the technology from the previous period, t, i.e. it is not feasible in period t,
presumably because technical progress has occurred allowing production of more good
with less bad than was possible in period t. If we try to compute the mixed period
distance function

Dto(xt+1, yt+1, bt+1), (3.13)

its value will be +x, in which case the Malmquist index is not well-defined. In our
empirical section we encounter this problem for a third of our observations.

∗ We use constant returns to scale because that is a necessary condition for the resulting productivity
indexes to be true total factor productivity indexes; see Färe and Grosskopf (1996).
236 A directional distance function approach

The directional distance functions may also be calculated as solutions to linear

programming problems. Again as an example,

) t t,k t,k t,k t,k

D o(x ′, y ′, b ′;y ′,−bt,k′)=maxb (3.14)
K

s.t. ]z y
k=1
t
k k′m [(1+b)ytk′m,m=1,...,M,

]z b =(1−b)b
k=1
t
k ki
t
,i=1,...,I,
k′i

]z x
k=1
t
k kn Ζ(1-b)xtk′n,n=1,...,N,

zk[0, k=1,...,K
This shows that the directional distance function may also be computed using linear
programming. We note that in the mixed period problem illustrated in Figure 2, the
directional distance function places ( yt+1, bt+1) on the output set Pt(xt+1) at A. Thus,
in this case we no not encounter any computational problems with the Malmquist–
Luenberger index. The same is true in our empirical example.

4. Data and results

We measured productivity changes in the Swedish pulp and paper industry. We use the
same panel data as in Brännlund et al. (1995a) and Brännlund et al. (1995b). The data
sources are primary data for the pulp and paper industry gathered by Statistics Sweden
and the Swedish Environmental Protection Board. The part of the data used here is
annual data on quantities of outputs and inputs from 39 paper and pulp mills for the
period 1986–1990.∗ To produce desirable output pulp, y, we observe four inputs,
namely: labour (x1), wood fibre (x2), energy (x3) and capital (x4). The desirable output
y is jointly produced with several bads, which include biological oxygen demand (BOD),
chemical oxygen demand (COD) and suspended solids (SS).
Descriptive statistics for the data are presented in Table 1. The data shows that the
amount of inputs and outputs increased between 1986 and 1988 and decreased thereafter.
The average values for the Malmquist–Luenberger productivity indexes which are
based on the directional distance functions as specified in (3.14) and its components
are reported in Table 2.
This result shows that the productivity in this industry has improved on average
over the entire time period. The main source of the productivity improvements is
technological advance rather than efficiency improvement. In fact, technical efficiency
fell during every period except the 1987/1988 period.
We also computed the traditional Malmquist based on the maximization problems
in (3.12) and (3.13) which are solved as linear programming problems. The solutions
for each observation are then used to obtain the original Malmquist productivity
indexes as well as their components. The solutions for the maximization problems for
the mixed period distance functions did not exist for almost one third of the observations.

∗ Only the data of the firms which operated during the whole five years are included here.
Y. H. Chung et al. 237

T 1. Descriptive statistics

T Variable Mean SD Min. Max.

1986 y (tons) 233 851·0 174 073·6 2718·0 824 250·0

(K=39) bBOD (tons) 3013·6 2640·3 110·7 10 740·0
bCOD (tons) 12 996·2 11 976·3 254·5 47 256·0
bss (tons) 1309·1 1209·7 51·9 4400·0
x1 (hrs) 828 897·4 129 386·8 107 000·0 2 383 000·0
x2 (m3) 1·01E+06 7·17E+05 7000·0 2·5E+06
x3 (kwh) 3·54E+08 3·19E+08 1·61E+07 1·26E+09
x4 (mil) 454·7 330·2 6·5 1585·9
1987 y 249 792·0 186 907·4 2974·0 853 520·0
(K=39) bBOD 3098·8 2631·8 121·8 11 993·0
bCOD 13 019·8 11 826·1 316·7 47 256·0
bss 1346·4 1481·8 69·6 7318·5
x1 826 435·9 524 013·3 102 000·0 2 332 000·0
x2 1·05E+06 7·25E+05 8000·0 2·52E+06
x3 3·81E+08 3·63E+08 1·37E+07 1·46E+09
x4 435·0 315·4 6·4 1515·1
1988 y 257 866·3 191 670·6 2346·0 878 000·0
(K=39) bBOD 3197·5 3167·3 115·8 14 040·0
bCOD 12 703·3 11 664·8 247·1 47 060·0
bSS 1756·1 2704·0 49·4 14 480·0
x1 829 717·9 528 028·4 102 000·0 2 345 000·0
x2 1·27E+06 1·75E+06 7000·0 1·11E+07
x3 4·00E+08 3·83E+08 1·40E+07 1·50E+09
x4 419·4 303·9 6·6 1434·9
1989 y 257 977·9 194 005·7 2077·0 872 610·0
(K=39) bBOD 2825·8 2714·6 120·0 11 067·0
bCOD 11 788·2 11 773·4 257·7 49 980·0
bss 1775·4 2735·7 42·4 14 994·0
x1 827 846·2 526 682·6 107 000·0 2 268 000·0
x2 1·06E+06 7·31E+05 6000·0 2 543 000·0
x3 4·17E+08 4·23E+08 1·50E+07 1·58E+09
x4 408·6 299·7 6·3 1368·5
1990 y 251 887·3 180 568·9 1964·0 855 080·0
(K=39) bBOD 2383·8 2196·4 110·7 9632·0
bCOD 10 028·6 9606·3 252·6 38 880·0
bSS 1657·9 2538·1 4·9 12 384·0
x1 809 846·2 509 687·0 96 000·0 2 202 000·0
x2 1·01E+06 6·94E+05 6000·0 2·25E+06
x3 4·05E+08 4·15E+08 1·52E+07 1·51E+09
x4 403·1 292·3 6·25 1325·6

T 2. The Malmquist–Luenberger productivity indexes and components∗

K=39 MLtt+1 MLEFFCHtt+1 MLTECHtt+1

1986/1987 1·027 0·984 1·043

1987/1988 1·087 1·011 1·075
1988/1989 1·031 0·995 1·035
1989/1990 1·059 0·882 1·200

∗The geometric mean of individual indexes.

238 A directional distance function approach

T 3. Comparison of the Malmquist and Malmquist–Luenberger indexes∗

Mo MEFFCH MTECH
Period (MLo) (MLEFFCH) (MLTECH)

1986/1987 (K=26) 1·052 0·964 1·091

(1·024) (0·977) (1·049)
1987/1988 (K=23) 1·053 1·028 1·024
(1·060) (1·010) (1·050)
1988/1989 (K=28) 0·940 0·995 0·945
(1·024) (0·992) (1·032)
1989/1990 (K=27) 0·959 0·929 1·032
(1·049) (0·855) (1·227)
Total (K=104) 0·997 0·977 1·020
(1·039) (0·955) (1·088)

∗The geometric mean of individual scores.

Thus, the Malmquist productivity indexes in (3.1) and technical change (3.3) for those
observations is undefined. This is in contrast to the case of the directional distance
functions where no computational problems arise.
The Malmquist productivity index is not defined for all observations (see previous
section), yet we still want to see how it compares to the Malmquist–Luenberger index.
We limit the comparison to those observations for which the Malmquist index is well-
defined. The geometric means for the observations that were well-defined for the
traditional Malmquist index, as well as the means for those same observations for the
Malmquist–Luenberger index, are included in Table 3. Looking first at the grand means
over all time periods, we see that the Malmquist–Luenberger (ML) index suggests greater
improvements in productivity and technical change on average than the traditional
Malmquist (M) index. The mean values of the ML index are in fact frequently greater
than the mean values of the M index (except for efficiency change) in the individual
years.
We tested for differences in the indexes for the observations that had well-defined
M indexes. Based on a battery of non-parametric tests of location using the SAS
NPARIWAY procedure, we reject the hypothesis that the two indexes are “the same”
at the 0·01 significance level. However, based on a simple t-test with the null hypothesis
that MLk−Mk=0, we could not reject the null at conventional significance levels.

5. Summary
This paper introduces a performance measure that credits the reduction of undesirable
outputs like pollution while simultaneously crediting increases in desirable outputs.
This function is used in the construction of a Malmquist type productivity measure.
The new index, termed the Malmquist–Luenberger index, which also accounts for
reduction of bads, can be decomposed into two parts: efficiency change and technology
change. We show how to compute these indexes using simple linear programming
problems, and provide an empirical example for the case of Swedish paper and pulp
mills operating over the 1986–1990 period. This technique provides individual plant-
specific information on productivity and its sources without requiring data on prices
of inputs and outputs. This is especially useful where firms face environmental re-
gulations, since the regulated effluents are typically non-marketed.
 Y. H. Chung et al.

T 4. Non-parametric tests of comparison of ML and M indexes (for well-defined M)

ANOVA Wilcoxon Kruskal–Wallis Median Van-der Werden Savage Kolmogorov–Smirnov

F Z v2 Z Z Z KSa
(prob>F) (prob>Z) (prob>v2) (prob>Z) (prob>Z) (prob>Z) (prob>KSa)

19·90 −4·08 16·69 −2·97 −4·32 −2·66 2·08

(0·0001) (0·0001) (0·0001) (0·0030) (0·0001) (0·0077) (0·0003)
239
240 A directional distance function approach

In modelling the joint production of goods and bads, and providing a performance
measure that credits firms for reduction of bads, managers have a reliable measure of
their “true” productivity and its sources.
This research has been partially funded by USEPA CR823009010. We are grateful to Finn
Førsund, three referees and the participants at the Atlantic Economic Meetings in Williamsburg
for their comments.

References
Ball, V., Eden, C., Lovell, A. K., Nehring, R. F. and Somwaru, A. (1994). Incorporating undesirable outputs
into models of production: an application to U.S. agriculture. Cabiers d’Economie et Sociologie Rurales
no. 31, pp. 60–74.
Brännlund, R., Färe, R. and Grosskopf, S. (1995a). Environmental regulation and profitability: an application
to Swedish pulp and paper industry. Environmental and Resource Economics 6, 23–36.
Brännlund, R., Chung, Y., Färe, R. and Grosskopf, S. (1995b). Emission Trading and Profitability: The
Swedish Pulp and Paper Industry. Discussion Paper No. 95–14, Southern Illinois University at Carbondale.
Briec, W. (1995). A Graph Type Extension of Farrell Technical Efficiency Measure. Memo, GREQAM.
Caves, D. W., Christensen, L. R. and Diewert, E. (1982). The economic theory of index numbers and the
measurement of input, output, and productivity. Econometrica 501, 1393–1414.
Färe, R. and Grosskopf, S. (1996). Intertemporal production frontiers: with dynamic DEA. Boston, Kluwer
Academic Publishers.
Färe, R. and Primont, D. (1995). Multi-Output Production and Duality: Theory and Applications. Boston:
Kluwer Academic Publishers.
Färe, R., Grosskopf, S. and Pasurka, C. (1986). Effects on relative efficiency in electric power generation
due to environmental controls. Resources and Energy 8, 167–184.
Färe, R., Grosskopf, S. and Pasurka, C. (1989). The effects of environmental regulation on the efficiency of
electric utilities. 1969 vs. 1925. Applied Economics 21, 225–235.
Färe, R., Grosskopf, S., Lindgren, B. and Roos, P. (1989a). Productivity Developments in Swedish Hospitals:
A Malmquist Output Index Approach. Discussion Paper No. 89-3. Southern Illinois University at
Carbondale.
Färe, R., Grosskopf, S., Lovell, C. A. K. and Pasurka, C. (1989b). Multiple productivity comparisons when
some outputs are undesirable: a nonparametric approach. The Review of Economics and Statistics 71,
90–98.
Färe, R., Grosskopf, S. and Lovell, C. A. K. (1994). Production Frontiers. Cambridge University Press.
Färe, R., Grosskopf, S., Lovell, C. A. K. and Yaisawarng, S. (1993). Derivation of shadow prices of
undesirable outputs: a distance function approach. Review of Economics and Statistics 75, 274–380.
Farrell, M. J. (1957). The measurement of productivity efficiency. Journal of Royal Statistical Society A 20,
283–281.
Haynes, K., Patrik, S., Bowen, W. and Cummings-Saxton, J. (1993). Environmental decision models: U.S.
experience and a new approach to pollution management. Environmental International 19, 261–275.
Luenberger, D. G. (1992a). Benefit functions and duality. Journal of Mathematical Economics 21, 461–486.
Luenberger, D. G. (1992b). New optimality principles for economic efficiency and equilibrium. Journal of
Optimization Theory and Applications 75, 221–264.
Luenberger, D. G. (1994a). Dual pareto efficiency. Journal of Economic Theory 62, 70–85.
Luenberger, D. G. (1994b). Optimality and the theory of value. Journal of Economic Theory 63, 147–169.
Luenberger, D. G. (1995a). Microeconomic Theory. Boston: McGraw-Hill.
Luenberger, D. G. (1995b). Externalities and benefits. Journal of Mathematical Economics 24, 159–177.
Pittman, R. W. (1983). Multilateral productivity comparisons with undesirable outputs. Economic Journal
93, 883–891.
Shephard, R. W. (1970). Theory of Cost and Production. Princeton: Princeton University Press.
Tyteca, D. (1995). DEA models for the measurement of environmental performance of firms: concepts and
empirical results. Working Paper, IAG, University of Catholique de Louvain.