Professional Documents
Culture Documents
Introduction
429
430 A. Charne~ et aL /Meamring the efficiency o f dicision making units
ways in which some public sector activities are some- 2. Model and definition
times evaluated.
Naturally we shall want to relate our ideas to Our propo~d measure of the efficiency of any
developments in economics. This will be done by UMU is obtained as the maximum of a ratio of
reference to production functions and related con- weighted outputs to weighted inputs subject to the
cepts such as "cost duality', etc. Although adapta- c,mdition that the similar ratios for every DMU be
tions of these concet~ts will be needed we shall also less than or equal to unity. In more precise form,
try to indicate wl~at is involved at suitable points in $
this paper. (See below, Section 6, for instance.)
~ UrYrO
"ige shall also want to relate our ideas to other r=l
naax ho - (1)
disciplines, iike engineering, which are also concerned m
with efficiency measurement. This will be done not
~ OiXio
only in the interests of greater unity but also in the i= 1
interest of distinguishing between efficiencies asso-
ciated with an underlying production 'technology' ~abject to:
$
and those due to managerial decision making when
the former can be identified and separated from the urYri
r=1
latter by, e.g., engineering characterizations. <1; /=l,...,n,
Of course, when this cannot be done (the usual m
have been suggesting. This is the way the rest of the urYd
paper will be developed although, as already indi- r= 1
n
several standpoints. We have a completely symmetric
xiiXj < X i o ; i = 1, . . . , m , definition of efficiency which generalizes single out-
/=1
put ratio definitions not only in economies but in
Xi~>0; j = l .... , n . engineering and other natural sciences. We do not
Because (3) is an ordinary linear pro~amming prob- need to solve the nonlinear (and noneonvex) prob-
lem it has a Knear programming dual which we can lems in which these definitions are forraalized. We
write as follows: need only solve the ordinary linear programming
problem (4) in order to obtain both the optimal f~
m
or h l and the weights v/', u~ ~ 0, since the change in
min go = ~ WiXiO (4) variables does not alter the value of the functional.
i= 1
Thus,
subject to:
f~ =g~ =z~ (7.1)
$ m
and therefore
r=l i=1
hl = l[z~. (7.2)
Also we have the wanted relative weights. Thus
larYro =1, nothing more is required than the solution of (4) or
r=l
(3) in order to determine whetherf~ > 1 or, corre-
lZr, Wi >~ O . spondingly, whether h l < 1, with efficiency prevail-
ing if and only if
Because of the structure of (4) one can recognize
that it is equivalent to an ordinary linear fractional f(~ = h i = 1. (7.3)
programming problem. (See [ 10] and [7] .) In fact,
We can also effect extensions in a variety of addi-
utilizing the theory of linear fractional programming
tional (new) directions 6. Here, however, we prefer to
with the transformation
make contact with various developments and also
¢oi = toi ; i = l .... , m , sketch a few of the ideas that are elsewhere described
as Data Envelopment Analysis 7 For this purpose we
lar = tU r ; r = 1 . . . . . S,
introduce
I Yi\
t-1 = ~ ~rYrO , Pi =~ : ~ • / = l , .-o9 n , (8)
r ~ j ..¢'/ ]k ,
which, with t > 0, gives explicitly ~,herein the subvector Yi contains observed output
$ caluesy d, r = 1, ..., s for its components and the sub-
~ector X~ contains observed input values xii, i = 1,
~--l O i X iO
i= 1 • .$ ~n,.
rflin fo = (6)
$ Now consider the following vector reformulation
~ UrYrO
,if(3):
r=1
maxz0 (9)
subject to:
in $
oix j - u , y , i >1 o ; ] = 1. . . . . n , For instance, we could utilize the duality theory that is
i= 1 r= 1
now associated with fractional programming (as discussed
vi, u r n > 0 , in [18] and [25] -see also [5] and [7])as distinguished
Iron" the duality theory of ordinary linear programming or
as the linear fractional programming equivalent of (4). what is sometimes called duality theory (see l:elow) in cost
By very evident manipulations, however, we can see and production theory.
This is a method for adjusting data to prescribed theoretical
that (6) is the same as (2). Hence we can use (4) to requirements such as optimal production surfilces, etc.,
solve (6) and therefore (2) and (1) as well. Q.E.D. prior to undertaking various statistical tests fc.r purposes of
We are now in ~ advantageous position from public policy analysis. See [21 ].
A. Charnes et al. [ Measuring the efficiency of dicision making units 433
fince s °+ and s * - are both non-negative. Evidently, in the case of multiple outputs, in this manner we
the expressions on the ~eft then satisfy the 'unvaried obtain a new type of production function which has
problem' (9) with z~ in place of z~, and ~* in place a ~,afiety of advantages that we shall indicate as we
of X~. However, then also pl oceed. Here we may adumbrate some of these ad-
vzntages in a summary way as follows. Unlike other
Max z o >1ZoZ
o "*
* > z~ t~ pes of production functions, this one derives from
when ~ > 1, But z~ = max z6, by hypothesis. Thus a (and is therefore directly applicable to) empirical
contradiction occurs which proves that z~ = I is the observations. It also bypasses the intractable prob-
optimal value for the varied problem (12). lems of aggregations associated with other types of
Now we want to show that the optimal solution, pl oduction functions It and, finally, it lends itself to
~,~, j = 1.... , n, to the unvaried problem (9) is an op- comparative statics for such purposes as determining
timal solution to the varied problem (12) with zero whether technological change is occurring. These
slack, i.e., the vectors s*+ and ~*- have zeros in all 'comparative' static uses may be accomplished in
components as required for efficiency. First, via (10) various ways such as adopting the com,ention that
the same DMU is to be regarded as a different entity
n
in each relevant time period.
- rjV + roz By means of these production function concepts
]= 1
and the procedures we shall associate with them, the
requirements of economic theory may then be
Z) xjv - X o - s'- brought to bear in new and modified ways for public
j= 1 policy evaluations, and, at the same time, provide a
Thus ~j* is a feasible solution of the varied problem variety of new predictiorL~ and control possibilities
with zo = 1. That is for program managers - Le., managers, legislators,
e~c., who have total program responsibility. For
n
example, it then becomes possible to distinguish be-
- YJV + (Yoz + s *÷) -- 0 tween the 'program efficiency' - that may be pre-
/= !
dicted with efficient managemem - and to distinguish
n this from other predictions (and evaluations) that
xjV - xo - might be effected on the assumption that all managers
/-- 1 will continue to operate only at past levels of effi-
with Zo = 1. It is also optimal since a~ we have just ciency 12. Similarly, it is possible to allow for predic-
showta, z~ = i. Further, the optimal slacks s*+ and tions of future changes in technology instead of
s * - are all zero. Q.E.D. assuming that a static longru,~ production function
In short, the indicated adjustments do, in fact, has been achieved - as is presently being done in
always bring the original observations into the rele- many of the extant studie: for energy policy guidance
vant efficient produ,ztion set. No new computations in Western countries.
are required after the z~, s*- adjustments are effected This may all be accom[llshed, we should note,
for the original Yo, Xo data for the efficiency com- with~mt interferhag with s absequent statistical testing
parisons we may subsequently want to make. arm evaluation. As in the DEA approach which we
As we shall shortly see - in sectien (6) below - discuss elsewhere, these s Latistieal tests, which are
we can use these results to obtain a surface corre-
sponding to a well-defined relation between output terization may be used to secure numerical estimates of
and inputs. For the single output case this relation these coefficient values (or, rather, their production func-
tion counterparts) from ob~rvational data.
corresponds to a function in which output is maximal 11
Disaggregation may be necessary, however, when data on
for all the indicated inputs. It therefore formally ful- individual DMU'sare not available. See the guidelines
l'dis the requirements of a 'production function' er, supplied in [12] and [21 ]. For a discussion and an attempt
more generally, a 'production possibility surface' lo t~Jdeal with the difficulties of aggregation in other types
of production functions, see [ 24 ].
12
See [21 ] for further discussion and a detailed application
I0 We are using this term in these sense of the activity vec- that distinguishesbetween 'program efficiency' and
tors discussed in Arrow and Hahn [4 ]. Indeed, in Section 'manager~l efficiency' in 'Program Follow Throu~' of
5, below, we shall explicitly show how our duality charac- the U.S. Office of Education.
A. Charnes et al. /Measuring the efficiency o f dicision making units 435
n
applied after the indicated adjustments, are often
greatly simplified relative to other alternatives. Pro- /= !
xi/Xj <x;o i = 1 .... , m
ceeding without first effecting such adjustments in
the data would not only fail to utilize the underlying X/DO" j=l,...,n
theory-, it would also contradict the requirements of where, because we are concerned with only one out-
that theory unless (a) one can assume that all DMU's put, the extra subscript is dropped on the y / f o r each
are operating efficiently or (b) one can supply some of the j = 1.... , n DMU's. Making the scaling change
alternate method of allowing for observations wl'dch of variables Xj = X j y j , z o = Z'o/Yo, then dropping the
are n o t on the efficient prod:,.ction possibility primes and the constant multiplier 1/y o in the func-
frontier 13.
tional, we obtain the equivalent form,
/I
m a x Zo = E3 Xj ~14.2)
max go (13) j=l
subject to: with
I'1 n m
! 3 For a discussion of problems involved in dealing with 15 To make full contact with Farrell's work it i~ also neces-
such estimation problems even in the case of a single sary to deal wRh (and eliminate) some of his awkward
output and very few inputs see [1] and [2]. concepts such as 'points at infinity', etc. This is done ia
14 Examples involving continuation of this work may be [ 12] but m the interest of brevity we shall not attempt to
found in [ 2 ] and [ 16 ]. repeat thes~edevelopments here.
436 A. C"harnetet aL/ Measuring the efficiency o[ dic~rion making uultt
"II b
[ Q~i.2
able to secure its one unit of output with only ~ of formed from/)4 and Ps whereas P2 is in the cone
the amount of each o f x l and x2 that it was observed formed from/)4 and P3. This condition (which arises
to utilize. from the non-negativity imposed on the admissible
"lhe line connecting P4 and/)3 represents one seg- ~'i values) has an advantage for the sorts of public
ment of what Farrell refers to as the 'unit isoquant'. program applications we are considering. As observed
It is also efficient. That is, there cannot be a point in the opening section, many DMU's such as different
such as F~ between this unit isoquant and the origin school districts, etc., work under varying constraints
since that would imply that/)4 and P3 were not an with respect to inputs (as well as outputs) for the
optimal basis is. same program conducted in different locales or differ-
The above analysis has employed two assumptions ent parts of the country. Hence it is well to have the
which we shall refer to as the 'isoquant' and 'ray referents used for scoring the efficiency of each DMU
assumptions', respectively. The latter, i.e., the ray as alike to it as possible, at least in some loose sense,
assumption, which corresponds to assuming constant while not interfering 'too much' with the wanted
returns to scale 19, may be relaxed but at a cost in efficiency ratings 21.
complications and resulting explanations which we We now conclude this section with the case of P6
shall not undertake here. The former, i.e., the isoquant in Fig. 1, for which alternate optima are present
assumption representation, is critical and it may not since
be relaxed in the sense that it provides the comparison
1/)6 =/)6 and 1P3 + 1 el =/'6
point between/)2 and//2 20.
We have just described the operational significance with z~(P6) = 1 in either case. Note, however, that
that we wish to accord to our efficiency measure for the second of these two solutions has s~ = I, which is
the case of/)2. Now consider Pl for which we obtain to say that efficiency is not attained for P6 until
5 slack in the amount of s~ = 1 is subtracted from the
first component in/)6 - after which it is coincident
when employing adjacent extreme point methods. with P3 in Fig. 1; see (1 1).
That is, the optimal basis consists of P4 and P3 (which To avoid ambiguity, and to retain the wanted
are both thereby characterized as efficient) with operational meaning for efficiency, it is necessary to
z~(Pt) = X,~ + X~ = ~ so that, by virtue of (1 ~ ), P~ is maxim~e the slack values, but in a way that accords
also not efficient. this a lower priority than maximizing the ~/~Mues.
As can be seen, we have z~(Pl) = z~(P2) = -76.How- This is done by replacing (14.2) with
ever, Pl and P2 are expressed in terms of different
rl m
bases and hence have different referents, q he con-
max Zo- + (14.3t
vexification process used for P2 is also employable /= 1 m --
for Pt to obtain
subject to:
s
f/ m
so that, also, the same ratio of contraction for all
Po = ~ PfX/+ Z~ eisi,
resources is needed to bring Pt onto the efficient sur- /= 1 i= 1
face. However, it lies in the cone through the origin
X/,si>~O ; /= l,...,n, i = I,...,m
18 Proofs of propositions like these, which are fairly trans- where M is the usual large (non-Archimedean) 'quan-
parent, will not be given in the present paper. They may tity' (cf., e.g., [22]) which insures that I[M > 0 is
be found in [12]. The latter source may also be consulted always smaller than any positive real value that may
for ways in which the formulations being used here differ
be assumed by any X].
from the one suggested to Farrell by A. Hoffman in rite
discussion associated with [15] and subsequently em-
ployed by Farre|l and Fieldhouse in [ 16].
19 Note, however, that this constant will be different, in 21 Note that both P! and P2 are also in the cone formed
general, for every Pp from Ps and P3 in Fig. 1 but the line segment connecting
20 it is to be understood, however, that our analyses are also P3 and Ps is not efficient - and hence P~ and P5 will not
applicable when various trav.sformations are utilized to be an optimal basis unIess one wants to impose further
bring other functions such as, e.g., Cobb-Douglas func- constraints on the basis choices. See the discussion in
tions, into suitable piecewise linear representations. Section 2, above.
438 A. Chames et aL / Measurb~, the eff'~lency of diction making units
efficient facet is the set of all x which are convex We now replace (17)with
combinations of the Pi which form the optimal basis.
S[Pi: all i ~ / ] = (17.1)
Of course, we will need to obtain the generality
different to *T for the 'slopes' associated with the dif-
ferent facets of the efficient surfaces and the ranges {X: ¢o* T = 1; X = ~ Ply i for all vi >1 O, ~ vi = !
i EI i ~I
for which they are valid. These ¢o*T values, however,
are obtainable within the computational process for in order to extend the preceding analysis to higher
the X*. For, as is well known (see, e.g. [8]), adjacent dimensions. Here the efficient surfaces are composed
extreme point methods (such as the simplex and dual of facets and w.T is the normal to such a facet (a
methods) simultaneously solve both (14.1) and (15). simplex) (see [8], p.242) which is the convex set
That is, the optimal k* and ¢o*T values are obtained spanned by the optimal basis. The theory that has
from the same tableau. Hence the solution to one of been developed around the usual adjacent extreme
these problems also provides the solution to the other point methods, including their nonlinear extensions
without extra computational effort. allows us to obtain all of the needed information for
For instance, the tableau that provided the solu- interpretation as well as continuation from one effi-
tion and hence the optimal basis for characterizing cient facet to another via the computational strate-
the efficiency of P1 with )~ = s, k; = 2 _ see pre- gies that these methods allow in conjunction wit] ~,
ceding section - also provides the associated dual each other 24
variables ~o~ = -~, co~ = ~. But, as previously noted, To conclude this section we can obtain a straight-
we can express any x T = (x~, x2) in th~s facet via the forward way of interpreting the components of the
basis vectors, P4, Ps, as normal vector, w *T, as marginal productivities in the
indicated inputs. We do this by returning to *he two
X = P4P4 + PsVs
dimensional case, such as the one depicted in Fig. 1.
with v4, Vs >t 0 and v4 + vs = 1. Hence also The condition for remaining on the same isoquant
when effecting substitutions is
*Tx = + o x2.
+ = +
For this segment we have ~ = ~, ~o~ = -~ and there-
fore where x :/: J: are two points on the indicated isoqua~t,
but on possibly different segments with slope vectors
SIPs, P4]- w* and w*. Adding w~J:l - ~ ] x 1 = 0 on the left
((Xl, X2): t X l +lx2= 1 "1 < x I < 2 , 2 < x 2 < 4 ) , and ¢o~x2 - w~-r2 = 0 on the right, we then effect a
series of algebraic manipulations to obtain 25
where the square brackets mean that the points Ps,
/'4 designate this facet (=segment) with the relation ~¢2_ ~o~ A~l -~1 A~2 ;c2
prescribed for the ranges indicated in the curly
brackets.
Similarly for SIPs, P3] we have where
SIP4,P3] =
Axi =x~ - x l ; ~lX2 = X2 -- ~2
while the isoquant segment from P3 to Q ~ (see Fig. 1) When both x and .v are on the same isoquant segmen
may be represented by we will have Awl = A¢,32 = O. Indeed, in fllis case we
S[P3, QI)-- 24 E.g., the simplex method of G.B. Dantzig and the dual
method of C.E. Lemke may be joined together for the~e
1 purposes. See [12] and [21] for further discussion a~ong
((xl, xD: l= xl+x2;4<xl, l < x 2 ) with numerical illustrations. For a discussion of the
wealth of information relevant to economic and mar~a-
where the square bracket indicates that P3 is included germl (policy) interpretations available in the resulting
while the parenthesis on the right indicates that Ql = tableaus, see Ch. V| in [8].
Mel, is excluded from the set and a similar characte- 25 This can be restated in terms of elasticities of substitu-
rization applies for points like Q2, etc. tions, if desired.
~,0 A. Charnes et aL / Measuring the zfJiciency of dicision making units
~ill have a well defined derivative and so we may properties - to the extent that the data allow 2s. It
replace the above f'mite difference formulation with follows that we are entitled to use these optimality
properties in deducing the further theorems and rela-
~2_ (..+~~ tions that we shall note below 29 since they also refer
dx I 60"2 ' to these same data or else to estimates derived there-
which is the usual expression relating the marginal from via rigorously established optimizing methods
rate of substitution between two inputs, as charac- and principles.
terized by the derivative on the left, and the negative In some ways these production functions are remi-
reciprocal of the ratio of their marginal productivi- niscent of ones that might be associated with Alfred
ties, on the fight, with equality between the two sides Marshall's concept of a 'representative firm'. (Discus-
required to stay on the same isoquant. In short, 601 is sion in [28]; also [9].) Here, however, the referent
the marginal productivity of the ith factor, i = 1, 2. is rather to 'representative efficient firms'. Note that
Since, by the assumptions of economics, these margi- the plural is required insofar as there is more than a
nal productivities are never negative, it follows that single efficient facet. The continuum within each
the marginal rate of substitutions is non-positive and facet is then 'representative' of the efficiency for
the isoquant is also assumed to be convex (and con- which the originally observed efficient firms serve as
tinuous), etc.. if we want to maintain contact with referents.
411
the asual theorems in these parts of economic theory. We commented earlier on ,he appropriateness of
these 'representative' facets (and the cones generated
from them) 30 for evaluating the efficiency of DMU's
6. Production functions and cost-price relations
in various public programs. They prc~vide an equally
We are now ready to construct the production appropriate referent for making estimates of what
function and cost relations from the just derived ~* various DMU's should be able to produce in the way
values. Before providing the model for doing this, of outputs 31 given the factor amounts and/or the
however, we might sharpen some of the points that relations between various inputs that may be pre-
have already made by pausing to see what we have scribed for them. This is to say that we shall want to
achieved and how it relates to other kinds of models refer each DMU to its 'representative' facet (or cone)
and approaches in economics. en route to making the translations and transforma.
Our function is evidently derived from empirical tions prescribed in (12) for bringing them onto the
observations. Although these observations are all at relevant production function surface. Given the latter
the level of the individual 'firms' 26, we evidently projections we are then in a position to estimate
have something that differs from other studies and what outputs we may expect with efficient produc-
pro0uction function estimates at the individual firm tion from various resource allocations to each DMU
level (see [19]) since, by hypothesis, we are con- in the programs being considered. These results may
sidering all pertinent DMU's. then be aggregated in a variety of ways for assessing
Approaches which include all such data have cus- or controlling the activities to be generated by these
tomarily been undertaken only at aggregate levels, DMU's 32.
with a]i the attendant difficulties (and assumptions)
cequired to ensure that the functions thus estimated 28 Including ex cathedra allowances, when available, as
have the extremal properties that a 'production func- noted in Section 2.
tion' must possess 27. Our functions evidently differ 29 For a further critique of the use of these optimality
from the latter in that (a) the data are not aggregated properties in association with such aggregate functions
prior to estimation and (b) the resulting estimates (e.g., in some of the current energy policy studies) along
are optimal - i.e., they have the requisite extremal with simple counteramples to the assumptions used in
such studies see [ 13 ].
3O See the discussion of 'representative convex sets' and the
26 We are using the term 'firms' interchangeably with DMU's cones they 'represent' on p. 236 in [8].
in this section in order to avoid circumlocutions in relating 3 | Including multiple outputs as discussed in Section 2,
these developments to pertinent parts of traditional above, since the principles we are discussing extend in
economies. See Johnston [19]. relatively obvious ways to the more general case of pro-
27 For a discussion of the onerous (and unrealistic) condi- duction possibility surfaces.
tions needed to obtain these properties at aggregate levels, 32 The principles to be used in effecting these aggregations
see Sato [24], pp. 3-8 ff. are set forth in [12].
A. Chames et al. / Measuring the efficiency of dicision making units 441
outputs in public sector applications are not ~.asily This will enable us to deal with this measure in the
priced or costed without recourse to arbitrary and context of original data. It will also enable us to close
ex cathedra procedures and assumptions. Further- thig section by returning to the multiple output case.
more, the use of price data also has difficulties such In Shephard's notation we replace the first portion
as were discussed in connection with Farrell efficiency of (18) with
(in the preceding section) to which we should now
CO,, p)= minx {pTx: ~ ( y , X)>~ l} (22.1)
like to note further difficulties such as determining
whether the data for each DMU refers to list or actual where
prices and whether allowances have been made for
1
finan~ng considerations, intermingled with produc- x ) - ,'(y. x) '
~:ion considerations, such as cash discounts, quantity
~scounts, and so on. w~th
Although we prefer to continue from the produc-
v*0,, x) - rain v , vx E L ( y ) , v/> 0 ,
tion side for reasons such as have just been indicated,
the:e may be cases where prices can also (or alterna- and L(y) is defined as in the second part of(18). This
tively) be used with advantage. We have elsewhere g,(y, x) is what Shephard calls a 'distance function '37.
[ 13] extended 'Shephard's lemma' by dropping the "[he above development assumes that the produc-
assumption of differentiability and so we now adapt tion relations have been determined and hence are
that result for use in the present case as well. available, as in (18.1), for the determination ofx.
For this purpose we reexpress (19.2) via Here, however, we want to regard the latter as data.
Thelefore, we revert to (8) and write
pTx* =YrI*Te=y ~ n; =yc (20.1)
$ min v n (22.2)
I where ,ith vXo - XiXi >t O,
j= i
-~],* n
c= qs , (20.2)
$
E; r xj> ro,
/= !
so that c is the (total) per unit cost of producing y
units of output. Note that we can generally vary the Xi~'3 ; / = 1, ...,n,
components o f p T within some range without altering
in which Yo =Y is now a vector of outputs.
the x* val, :s "6. Thus, if we select one component, If we regard z~ I> 1 as an extension of Farrell effi-
say p~ "or such variation, we can obtain ciency to the multiple output case of (9) then we can
z,Wix; =yAw (21.2) also regard rain v = v* in the above formulation as an
extension of Shephard's distance function (and related
or
concepts) for use in evaluating the efficiency of
DMU's. As a comparison with (9) makes clear, these
X~' =1/ Ac _ ACO', p) (21.2)
" ~Di Z~Pi z~ and v* values are complementary to each other.
(They are in fact reciprocals. See [ 13] .) In other
which is Shephard's lemma in discrete form.
words, Shephard's distance function is thereby con-
The definition of efficiency in (1)has thus enabled
verted into a measure of efficiency with v* < 1 and
us to make contact with this line of work as well. We
v* = 1 only when the designated DMU is efficient
can also do more. We can open up new possib~ties
relative to the other observations and/or theoretically
for employing the latter, e.g., in effecting estimates
prescribed conditions (e.g., via ex cathedra engineer-
of extremal relations from original data, just as we
ing conditions).
have done for Farrell efficiency.
In this way we have employed our definition of
To briefly show such possibilities it will suffice to
relate Shephard's measure of distance to Farrell effi-
ciency in the context of the models given in Section 2. 37 Lau in [17], p. 179 - see also Jacobsen in [17], p. 172 -
refers to this as a 'gauge function' (its classical denotation
in convex set theory) since it does not have the usual
36 Necessary and sufficient conditions as applicable to each properties of a distance function. See, e.g., Appendix A in
component ofp T are given in [81, Ch. IX. [81.
A. Charnes et al. /Measuring the efficiency o f dicision making units 443
efficiency not only to tie together two previously ciency of resource utilization in whatever combi aa-
separate strands of economic research, but also to tions are present (loose or tight) in the orgapizadons
open new possibilities for each of them which include as well as the technologies utilized 4o. This" ~o sug-
new ways of estimating extremal relations from em- gests a strategy for research in that given the ,es~tlts
pirical data. obtained at one level a systematic basis may be
secured for proceeding to other levels by raising perti-
nent questions and requiring further justification
7. Summary and conclusion from responsible officials for whatever inefficiencies
are uncovered.
We have now provided a variety of ways of assessing Concerning private sector applications, the case
the efficiency of DMU's in public programs in order to for our proposed measure of efficiency begins to
improve the planning and control of these activities. weaken to the extent that competition is pre~.cnt. In
This was initiated with a new definition of efficiency, particular it begins to weaken as soon as freedom for
in Section 2, which we have now related to both the deployment of resources from one 'industry' to
engineering and economic concepts. We also supplied another (perhaps in a removed region) is present.
operational vehicles 38 and interpretations for actual Assessment of such possibilities would involve the
applications. In addition we have introduced a new introductioo of prices, oF other weighting devices, for
kind of production function and new methods of the evaluation of otherwise non-comparable alterna.
securing estimates from empirical data which we have tives.
done in ways that have enabled us to bring a variety Although our measures are not designed for this
of economic concepts to bear in new and potentially sort of application they are designed for public sector
useful ways. This was done with new models and programs in w~,ich the managers of various DMU's
interpretations, to be sure, but without altering any are n o t free to divert resources to other programs
of the extremization principles which form the essen- merely because they are more profitable - or other-
tial basis of these concepts. wise more attractive. Our measure is intended to
We can (and will) extend these developments in a evaluate the accomplishments, or resource conserva-
variety of additional ways 39. Here we may best con- tion possibilities, for every DMU with the resources
clude with some comments on possible limitations assigned to it. in golfing terminology it is, so to
and alternatives to these approaches. speak, a measure of 'distance' rather than 'direction'
One limitation may arise because of lack of data with respect to what has been (and might be) accom-
availability at individual DMU levels. This is likely to plished 4j
be less of a problem in public sector, as contrasted
with private sector, applications. In addition the
term DMU can be accorded considerable flexibility References
in interpretation as when data unavailability may
make it desirable to move to school district instead of [1 J D.J. Aigner, T. Ameiya, P.J. Poirier, On the estimation
individual school level and, indeed, it may even be of production frontiers: maximum likelihood estima-
necessary to move to statewide office of education tion of the parameters of a discontinuous density
function, lnternat. Econ. Rev. XVII, (2) (1976) 377-
levels when data are not available even at the district
396.
level. After all, the objective is to measure the effi- [2] D.J. Aigner and S.F. Chu, On estimating the industry
production function, American Economic Review
38 I.e., we have supplied both models and computational LVII| (1968) 826-839.
procedures. [3] R.F. Alioto and J.A. Jungherr, Operational PPB$ for
39 E.g., as already observed the above developments can be Education (Harper and Row, New York, 1971 ).
extended to functions which are piecewise Cobb-Douglas. [4j K.J. Arrow and F.H. Hahn, General Competitive
40 The (act that one must generally include such organiza- Analysis (Holden-Day, San Francisco, 1971 ).
tion considerations as part of the prJduction function in
empirical investigations is in an economic tradition that
goes back at least as far as Sune Carlson. See [6]. (In 41 The term direction should be extended to the plural f~r
future papers, however, we shall show how the concepts the case of multiple outputs. (Here we are concerned
we have introduced here may be extended for use in with choices of direction which might be effected at the
separating these managerial features in observed data legislative level, including a choice of whether to under-
fror.1 other aspects of a production technology.) take the program at all.)
.,j
[5] C.R. lkctor, Duality in Linear Fractional Programming, 116] MJ. FarreU and M. Fieldhouse, Estimating efficient
Utilitas Mathematica 4 (1973) 155-168. production functions under increasing returns to scale,
[6l S. Carlson, A Study on the lure Theory of Production, J. Roy. Stl~t. Soc. Set. A, II (1962) 252-267.
(Kelley and Millman, Inc., New York, 1956). 1171 M.D. Intri~ligator and D.A. Kendrick, eds., Frontiers
[ 7] A. Charnes and W.W. Cooper, An explicit general solu- of Q~mti~ative Economics, IL North-Holland, Amster-
tion in linear fractional programming, Naval Research dam (1974).
Logistics Quarte~y 20 (3) (1973). [18] R. J~gana~an, Duality for nonlinear fractional pro-
[8] Ai ~ a m e s and W~W.C~per, Management Models and grams, Z. ~perations Res., 17, Physica*Veflag, Wurz-
I n d u s ~ Applications of Linear Programming (W'dey, burg (197I~) 1-3'
New York, 1961). [19l J. Jolmstc~l, Statistical Cost Analysis, McGraw-Hill,
[9 ] A. ~ e s and W.W. Cooper, Managerial economics - New York (1960).
past, present, future, Journal of Enterprise Management [20l K. L~ctcaster, Consumer Demand, Columbia UniversiW
(f~t~dlcoming). Press, New York (1971), 40.
11oi A. Chames and W.W. Cooper, Programming with linear [21] E. l~odes, Data Envelopment Analysis and Related
fractional functionals, Naval Res. Logist. Quart. 9 Approaches for Measuring the Efficiency of Decision
(3, 4) (1962) 181-185. Making Unit~ with an Application to Program Follow
1111 A. Charnes, W.W. Cooper and E. Rhodes, Exposi- Throttgh in U.S. Education, Ph.D. thesis, Carnegie-
tions, interpretations, an~ extensions of Farrell effi- Mellon UniversiW School of Urban and Public Affairs,
ciency measures, Management Sciences Research Group Pittsburgh (1978).
Report, Pittsburgh: Carnegie-Mellon University Schoo~ [22] A. Robinson, Non-Standard Analysis (North-Holland,
of Urban and public Affairs (1975). Amsterdam, 1966).
[121 A. Charnes, W.V(.Cooper and E. Rhodes, Measuring [23] P.A. Samuelson, Prices of Factors and Goods in General
the efficiency of decision making units with some new Equilibrium, Rev. Econom. Stud. 21 (1953-1954)
production functior, s and estimation methods, C~enter 1-20.
for Cybernetic Studies Research Report CCS 276, [241 K. Sat+:),Production Functions and Aggregation (North-
Austin, TX, University of Texas Center for Cybernetic Holland, Amsterdam, 1975).
Studies (1977). [25l S. Schaible, Parameter-free Convex Equivalent and
1131 A. Charnes, W.W. Cooper and A. Schinnar, Transforms Dual Programs of Fractional Programming Problems,
and approximations in cost and production function Z. Operations Res., 18 (1974) 187-196.
relations, Research Report CCS 284, Austin, TX, [261 R.W. Shephard, Cost and Production Functions
University of Texas Center for Cybernetic Studies (Princeton University Press, Princeton, 1953).
(1977). [27] R.W. Shephard, The Theory of Cost and Production
[14I Encyclopedia Americana (Encyclopedia Americana Functions (Princeton University Press, Princeton,
Corporation, New York, 1966) 352-353. 1970).
II51 M.J. Farrell, The measurement of productive efficiency. [281 G.J. Shgler, The Xistence of X-Efficiency, Am.
J. Roy. Statist. Soc. Set. A, III (1957) 253-290. Econorn. Rev. (March 1976) 213-216.