You are on page 1of 26

STRUCTURE-INDUCED (DIS)EQUILIBRIUM:

AN APPLICATION OF SOCIAL CHOICE THEORY TO


ORGANIZATIONAL DECISION MAKING

By

Adrian Van Deemen


Radboud University Nijmegen
Institute for Management Research
P.O. Box 9108
6500 HK Nijmegen
The Netherlands
Tel. ++ 31 (0)24 3613076
E-mail: a.vandeemen@fm.ru.nl

Preliminary Version 1.2

ABSTRACT
In this paper, an extension of the Condorcet paradox is presented that may be relevant for
decision-making in organizations. The extension deals with democratic organizational
structures of collective decision making units. Each decision unit aggregates a profile of
complete and transitive individual preferences into a social preference by using the majority
rule. The majority social preferences are aggregated in (fewer) decision-making units at a
higher organizational level and so on until the top level unit is reached. In the paper it is
assumed that in the end there is only one top level unit. The aggregated outcome of this unit is
considered to be the organizational decision.

We first show that the way of structuring the organization of decision making units may lead
to cyclical majority preferences at the top level and hence to the absence of a majority social
choice at that level. The basic point here is that this Condorcet paradox at the top level is
invoked purely on structural grounds. It is the structure that causes the decision deadlock.

We also show the reverse possibility, namely that top level majority social choices may exist
in hierarchical structures in spite of the fact that local cycles at several levels in the hierarchy
exist. The structure induces a global equilibrium in spite of the fact that local disequilibria
(local Condorcet paradoxes) exist.

Subsequently, we briefly discuss the consequences of our results for the well-known
structure-induced equilibrium concept of Shepsle (1979) and for the Garbage Can
organizational decision making model (Cohen, March and Olsen (1972).

1. INTRODUCTION
The Arrow Impossibility Theorem (cf. Arrow 1963) and the related Condorcet paradox (cf.
Black 1957) have been studied extensively in social choice theory. As far as we know,
however, the research after the Condorcet paradox and majority disequilibrium is, with some
exceptions, oriented towards stand-alone collective decision making units, that is, to units of
individuals with preferences which are not embedded in organizational or hierarchical
structures of decision making. This makes the research after both Arrows theorem and the
Condorcet paradox less relevant for studies of organizational decision making.

In this paper, social choice theory is applied to organizational choice. An extension of the
Condorcet paradox is presented and studied that may be relevant for decision-making in
hierarchical organizations and which may open a new line of research with respect to the
Condorcet paradox. The extensions deal with hierarchical structures of collective decision
making units. Each decision unit aggregates a profile of complete and transitive individual
preferences into a social preference by using the majority rule. These majority social
preferences are in their turn aggregated in (fewer) decision-making units at a higher
organizational level and so on until the top level unit is reached. In this paper it is always
assumed that in the end there is only one top level unit. The aggregated outcome of this unit is
considered to be the organizational outcome. The paper exclusively focuses on democratic
organization structures. The higher-level decision units, including the top level unit, are all
supposed to use the majority rule.

The basic aim is to show the effects of organizational structure on majority decision making.
We first show that the way of structuring the network of decision making units may lead to
cyclical majority preferences at the top level and hence to the absence of a majority social

choice at that level. The basic point hereby is that this top level Condorcet paradox is invoked
purely on structural grounds. It is only the structure that causes the decision deadlock. If the
individual preferences of the same units are joined into an unstructured single decision
making unit, a top level majority social choice exists and hence no paradox appears.

We also show the reverse possibility, namely that top level majority social choices may exist
in hierarchical structures in spite of the fact that local cycles at several levels in the hierarchy
exist. The structure, so to say, corrects the local cycles. It induces a global equilibrium in spite
of the fact that local disequilibria (local Condorcet paradoxes) exist. Finally, we show that by
imposing or designing a decision structure, Condorcet paradoxes can be solved

Murakami (1966, 1970), Pattanaik (1970), Fine (1972) and Fishburn (1973) study what they
call representative democracies or representative systems based on majority voting operators.
Our work is highly inspired by these studies. However, these studies are mainly oriented
towards finding characterizations of representative systems or multi-stage majority decision
procedures. In contrast, this paper wants to elucidate the effects of structure on social choice
making. Moreover, we do not wish to restrict ourselves to the interpretation of representative
systems only. In our view the general concept of organizational structure is more appropriate.
This implies that we see a system of representative democracy as an organizational structure.
The theoretical structure outlined here also may be viewed as a system of committees or as a
coalition of coalitions as in the theory of the firm (Cyert & March 1992).

Shepsle (1979, 1986) introduces the idea of structure-induced equilibrium in spatial voting
games. Starting from a multidimensional policy space as the choice space, he divides this
space into subspaces. These subspaces are covered by a (possibly overlapping) committee

system each of which has jurisdiction over a subspace. Then he shows that such a jurisdiction
structure may induce majority equilibria. For example if the space is n-dimensional Euclidian,
then covering this space by n committees each with a jurisdiction over one dimension, Blacks
median voter theorem (Black 1957) guarantees the existence of a majority decision for each
dimension. Again our work is much inspired by Shepsles work. However, we are interested
in the effects of (hierarchical) structures on majority decision making. Hereby we will
explicitly investigate the effects of structure on majority (dis)equilibria without using
jurisdiction compartments of the choice space.

The paper is organized as follows. In Section 2 we present some necessary concepts and
notation. Moreover, a language is presented to describe and analyze organizational decision
making in social choice theoretic terms. In the third section it is shown how organizational
decision structures may induce majority disequilibria. Section 4 shows that for a fixed input
preference profile any alternative can be a collective decision by varying the structure. Hence,
it shows the structural effects on social choice. In Section 5 we study how local disequilibria
(Condorcet paradoxes) may be covered by the overall organization structure. In Section 6 it is
shown that Condorcet paradoxes can be solved by imposing a decision structure on the set of
individual preferences. We finish with a concluding section.

2. PRELIMINARIES
Let N be the set of individuals or players indexed by i = 1, ,n. We assume n to be nonempty
and finite unless stated otherwise. Nonempty subsets of N are called decision units or
coalitions.

Let X be a nonempty set of alternatives denoted by x, y. Let be a nonempty class of


nonempty subsets of X. Elements of are called feasible subsets or agendas of X. As usual,
P(X) denotes the power set of X.

A preference on X is a binary relation on X. Preferences are denoted by R. Further, xRy stands


for (x, y) R. The asymmetric part of a preference R, denoted by P, and its symmetric part
denoted by I are defined as:
1. xPy if xRy and not yRx;
2. xIy if xRy and yRx.
xRy can be interpreted as x is at least as good as y, xPy as x is better than y, and xIy as x is
as good as y. For convenience, we write xy for xPy and (xy) for xIy.

A preference R is said to be
-

reflexive if for all x X: xRx;

complete if for all x, y X: xRy or yRx;

anti-symmetric if for all x, y X: xRy and yRx implies x = y;

transitive if for all x, y, z X: if xRy and yRz, then xRz;

quasi-transitive if for all x, y, z X: if xPy and yPz, then xPz;

intransitive if it is not transitive;

acyclic if there are no x1, x2, x3 , , xk X such that x1Px2, x2Px3 , . , xk-1Pxk and
xkPx1;

cyclic if it is not acyclic.

R is a weak preference ordering (WPO) if it is reflexive, complete and transitive; it is a linear


preference ordering (LPO) if it is reflexive, complete, anti-symmetric and transitive. The set
of WPOs is denoted by O(X); the set of LPOs by L(X). Of course, L(X) O(X).
A preference profile for a decision unit or coalition S is a mapping from S into O(X). Notation
PS. The preference assigned to player i, when the context is clear, is denoted by Ri. Thus, a
preference profile is an s-tuple of WPOs, one and only one for each player in S. Hereby, s =
#S (the number of elements in S). A group preference function assigns to each PS a preference
relation.

Consider a preference R and a nonempty subset Y X. An alternative x Y is said to be Rbest if xRy for all y Y. The set of R-best elements of Y is called the choice set of Y. Note that
if x is best in Y, then there is no y Y such that yPz. We know that there is a nonempty choice
set for every nonempty Y X if and only if R is reflexive, complete and acyclic. Sufficient
conditions are reflexivity, completeness and transitivity.

Consider an S N. Let s(x, y) = #{i S: xRiy}. For any profile PS we define

xmy := s(x, y) s(y, x)

for every x, y X. m is called the majority rule. Clearly, the majority rule is a group
preference function. The preference generated by m is called the majority relation. A majority
decision, majority social choice or majority equilibrium for a preference profile PS is an mbest alternative given that profile. A set of m-best alternatives is called a Condorcet set.

As is well known, a majority decision need not exist. The exemplary illustrations are the
following profiles:

xyz
zxy
yzx

or

xzy
zyx
yxz

The first yields the cyclic majority relation xyzx. This is a so-called forward cycle. The second
profile yields the cyclic majority relation xzyx, which is a so-called backward cycle. We say a
Condorcet paradox or majority disequilibrium occurs for a profile when there is no majority
decision for that profile.

Now, we develop a language to study organizational decision making in social choice


theoretic terms. The language is based upon the computer language Lisp (Slade 1998, Seibel
2005). We mainly will work with list structures and operations on (parts of) list structures. In
the sequel we will use prefix notation to denote decision structures. The list (m PS) means that
the majority rule is applied to the profile PS = (R1, , Rs) of decision unit, committee, or
coalition S. The majority rule (or any other group preference rule) is always placed at the first

position in a list. When the context is clear, we label the decision units with S1, S2, , Sk and
denote the unit profiles accordingly with P1, P2, Pk.

We say a list is nested if it contains other lists. Consider, as an example, the list
(m (m P1) (m P2) (m P3)). The big list contains four elements of which three are lists as well.
The first element in the list is the majority rule. It operates upon the results of the second,
third and fourth element in the list. The first element in the other lists also is the majority rule.
This is an example of a democratic organizational decision structure. More general, an
organizational decision structure is a nested list of list. The top level of the structure is the big
list directly containing the elements (which may be lists). With directly is meant that it is not
contained in another list. The very first operator in the list therefore is the operator of the top
level unit. A democratic organizational decision structure is a decision structure in which only
the majority rule is used as a decision making mechanism. In such organizations, policy
formation proceeds only by means of majority decision making.

Decision structures are processed by starting at the lowest level decision units. They are
elaborated from the lowest level list toward the top level list, thus from inside out. That is, we
use a bottom-up approach. Note that we can describe any organizational decision structure we
wish. Also note that it is quite possible to use other group decision rules than the majority
rule, e.g. the Borda rule or the Nash bargaining rule (when preferences are measurable at the
interval level). In addition, we can introduce non-decision operators that work on parts of
structures like the union of profiles or decision units. In this paper we focus on democratic
structures.

10

3. STRUCTURE- INDUCED DISEQUILIBRIUM


In his section we present a simple example of a structure that induces a Condorcet paradox.
Consider the following hierarchical structure consisting of four decision making units.
Examples for this kind of structural arrangements abound: a university board (as top level
unit) and three faculty management teams; a faculty board and its department MTs; a
parliamentary committee system; a board of executive directors and the MTs of three
divisions in a company; an arrangement of bureaus in a governmental department; and so on.
More specific, consider three decision-making units with profiles

P1 = (xyz, zyx, yzx)


P2 = (zyx, xyz, yzx)
P3 = (yxz, xzy, xyz)

and the tree structure (m (m P1) (m P2) (m P3)). Applying the majority rule yields yzx for S1,
zxy for S2 and xyz for S3. Taking these outcomes together constitutes a Condorcet profile.
Feeding this profile into the aggregating top level decision unit yields the forward majority
cycle xyzx. Hence, no overall majority decision exists; we have an instance of the Condorcet
paradox. See Figure 1.

11

Figure 1
xyz
zyx
yzx

S1

zyx
xyz
zxy

S2

yxz
xzy
xyz

yzx
zxy

S4

xyzx

xyz
no decision

S3

Now, joining profiles P1, P2 and P3 into a single unit we have

3: xyz
1: xzy
1: yzx
1: yxz
2: zyx
1: zxy

This joint profile yields the (transitive and complete) majority relation xyz and, hence, x is the
Condorcet social choice. Since the difference between the two concerned sets of individual
preferences is the hierarchical structure, it is only this structure that causes the difference in

12

outcome, in particular the disequilibrium. For this reason we call this kind of disequilibrium a
structure-induced disequilibrium.

We have shown that (m (m P1) (m P2) (m P3)) (m ( P1 P2 P3)). Here, stands for union.

Obviously, the presented model is simple. In reality, organizational structures are far more
complex. It is surely possible to complicate our simple model, for example by introducing
additional decision units, by allowing amendment procedures for the top level unit or by
feeding in preferences of the members of the top level unit. However, the simple form suffices
to show the influence of structure on the making of an organizational social choice.

Clearly, the top level decision unit has a problem. It may circumvent the decision deadlock by
amending the preferences of the other lower level decision units or by bringing in some
additional preferences. However, this may lead to other problems. For example, suppose the
top level unit consists of two members who bring in, besides the aggregated result of the
lower units, the preferences zxy and yzx. As can be easily seen, this leaves the cycle xyzx in
tact. The same is true when they bring in the preferences yzx and xyz or the preferences zxy
and xyz. When they both bring in one of the six linear orderings over {x, y, z}, then this
ordering will be the collective preference and hence a majority decision will exists. However,
in this case a majority in two lower level decision units can be found that prefer some
alternative to the alternative chosen at the top level. For example, if both bring in the
preference xyz, then this will be the collective majority relation and x will be the majority
decision. But a majority in unit S1 prefer y to x and a majority in unit S2 prefers z to x.

13

Note that the model is invariant under assigning a certain number of participants in the
different decision units. For example assigning in
-

P1 xyz to 3 individuals, zyx to 2, and yzx to 2;

P2 zyx to 2 individuals, xyz to 3, and zxy to 4;

P3 yxz to 3 individuals, xzy to 3, and xyz to 4

does not make any difference in the model results.

Furthermore, the model is invariant under assigning certain weights to the several decision
units. For instance, if we assign the numbers 2 to S1, 3 to S2 and 3 to S4, the model results
remain the same.

4. DIFFERENT STRUCTURES, DIFFERENT SOCIAL CHOICES


In this section we show that different structures with the same preference profile for the
decision units may yield different social choices. Hence, we fix the individual preferences fed
into the lower level decision units but vary the structure by introducing another level of. units.
First consider the following structure (Figure 2)

14

Figure 2
xyz
zyx
yzx
zyx
xyz
zxy
yxz
xzy
xyz

S1

yzx
S4

S2

zx, (xy), (yz)

zxy

S3

S5

xy, yz, (zx)


x is the
social choice

xyz

Remember that alternatives between parentheses mean indifference. The formal notation of
this structure is (m (m (m P1) (m P2)) (m P3)). In this structure, the collective decisions of
units P1 and P2 are fed in into a fourth unit P4 which aggregates it without amendment into
the intransitive collective preference zx, (xy) and (yz). This preference together with the result
of unit P3 is fed into the top level unit which aggregates it without amendment into the
intransitive but acyclic social preference xy, yz, (zx). Since x is better than y and is at least as
good as z, x is the majority social choice.

Can we find a structure which makes, with the same input profile for the three lower units, y
or z a social choice? The answer is affirmative.

First, consider the structure in which now the collective decisions of S2 and S3 are fed into
the fourth unit. So the only structural difference is that now the lines are going from S2 and

15

S3 to S4 instead of from S1 and S2. Subsequently, the result of S4 and S1 are aggregated by
the top level unit without amendment. See Figure 3. The formal decision structure now is
(m (m (m P2) (m P3)) (m P1))

Figure 3
xyz
zyx
yzx

S1

zyx
xyz
zxy

S2

yxz
xzy
xyz

y is the
social choice

yzx
zxy

S5
S4

S3

yz, zx, (xy)

xy, (zx), (yz)

xyz

The overall collective preference for this structure is the acyclic but intransitive relation yz, zx
and (xy). Since y is strictly better than z and as least as good as x, y is the majority decision.
However, the only difference between Figure 2 and Figure 3 is the decision structure. Hence,
only this difference in structure can account for the difference in the collective decision.

Finally consider the decision structure in which the results of S1 and S3 are fed in into the
fourth unit. Again this unit S4 aggregates these results without using an amendment procedure
or by bringing in additional preferences. The formal structure is
(m (m (m P1) (m P3)) (m P2)). See Figure 4 (note the switch of places of S1 and S2).

16

Figure 4
zyx
xyz
zxy

S2

xyz
zyx
yzx

S1

yxz
xzy
xyz

z is the
social choice

zxy
yzx

S5
S4

S3

zx, xy, (yz)

yz, (xy), (zx)

xyz

The intransitive collective majority relation yz, (xy) and (zx) of P4 and the transitive relation
zxy of unit P2 are fed in into the top level unit which yields zx, xy and (yz). Now, the majority
decision is z. Again note that the input preferences are the same as in the previous figures.
Clearly, only the difference in structure accounts for the selection of z as the majority social
choice.

Looking at Figures 1, 2, 3, and 4, we see four different structures with the same input preference
profiles each yielding a different result:

(m (m P1) (m P2) (m P3))  no majority decision


(m (m (m P1) (m P2)) (m P3))  x is the majority decision
(m (m (m P2) (m P3)) (m P1))  y is the majority decision
(m (m (m P1) (m P3)) (m P2))  z is the majority decision

17

Each structure yields a different social choice. Moreover, for each alternative a {x, y, z} it is
possible to construct a decision structure over a fixed set of individual preferences over {x, y,
z} such that a is a majority decision. And in addition it is possible to design a structure that
implements a decision deadlock. This shows that it is possible to enforce a certain social
choice or to create a decision deadlock by redesigning an organizational structure.

Clearly, the analysis above is close to implementation theory as developed in social choice
and game theory (Jackson 2001, Maskin & Sjstrm 2002). However, in this paper the
emphasis is not on implementing a social choice correspondence by some mechanism (game
form), but rather to design an organizational structure or organization scheme for a set of
group preference functions. The problem is how to design an organizational structure in order
to arrive at a desirable social choice

5. FROM LOCAL DISEQUILIBRIA TO GLOBAL EQUILIBRIA.


The first structure (cf. Figure 1) induces a disequilibrium in spite of the fact that each lowest
level decision unit produces a nice and consistent collective preference. In this section we
show the reversed phenomenon, namely, that local disequilibria, i.e. Condorcet paradoxes,
yielded by separate lower level decision units, can be turned into a global equilibrium by a
decision structure.

Consider three decision units with profiles:

P1 = (xyz, zxy, yzx)


P2 = (xzy, zyx, yxz)
P3 = (yxz, xzy, xyz).

18

Figure 5
xyz
zxy
yzx

S1

xzy
zyx
yxz

S2

yxz
xzy
xyz

zxyz
xzyx

S4

xyz

xyz
x is the
social choice

S3

S1 and S2 yield, respectively, the cycles zxyz and xzyx. So we have two local disequilibria. S3
produces the collective preference xyz. Feeding these three collective relations into the top
level unit yields the transitive majority relation xyz. Hence x is the majority decision. In spite
of the fact that we have two cyclic collective relations from which no decision can be derived,
the global collective relation is transitive and yields a clear decision.

Now consider the case in which P1 and P2 remain the same as in the previous but P3 = (xyz,
zyx, yzx). Using the same organizational structure now gives yzx as the collective preference
and hence y is the final decision. If we take P3 = (zyx, xyz, zxy), then zxy is the overall
organizational preference and hence z is the social choice. In all the three cases, the third unit
P3 determines the organizational choice. P1 and P2 produce respectively a forward cycle and
a backward cycle, which cancel each other out in the structure. In addition, looking at the
unstructured individual preference set in the first case

19

xyz
zxy
yzx
xzy
zyx
yxz
yxz
xzy
xyz

we see that x is also the social choice. In the second case also y remains the collective
decision and in the third z. Hence, it appears as if structure doesnt matter in these cases.

6. STRUCTURAL SOLUTIONS TO THE CONDORCET PARADOX


In this section we show how to solve Condorcet paradoxes by imposing an organizational
structure on the set of individuals. Consider the following set of preferences:

3: x y z
2: z y x
2: y z x
2: z x y

Applying the majority rule yields the cyclic majority relation xyzx. Hence, there is no majority
decision. We have an instance of the Condorcet Paradox. To solve this paradox, we design a

20

decision structure with five decision units. See Figure 6. Let the lowest level units have the
following preference profiles:

P1 = (xyz, zyx, yzx)


P2 = (zyx, xyz, zxy)
P3 = (xyz, zxy, yzx)

Figure 6
zyx
xyz
zxy

S2

xyz
zyx
yzx

S1

yzx

S5
S4

xyz
zxy
yzx

z is the
social choice

zxy

S3

zx, xy, (yz)

yz, zx, (xy)

zxyz

The overall decision structure is (m (m (m P1) (m P3)) (m P2)). Since (m (m P1) (m P3)) = yz,
zx, (xy) and (m P2) = zxy, the overall majority relation is (yz), zx and xy. Hence z is the
majority decision and the paradox has disappeared. Note that (m P3) produces the forward
cycle zxyz.

Notice that also this case shows that an overall majority decision can exist along with local
cycles. Further notice that z is the majority decision for unit S2.

Now consider the decision structure (m (m (m P2) (m P3)) (m P1)). This is a slightly different
structure in comparison with (m (m (m P1) (m P3)) (m P2)). Since (m (m P2) (m P3)) yields
21

the intransitive relation zx, xy and (yz), and since (m P1) is yzx, this structure yields yz, zx and
(xy). Hence, now y is the majority decision, which also happens to be the majority decision of
unit S1. So, slightly redesigning the decision structure yields another social choice. Again, the
Condorcet paradox has disappeared, which again shows that imposing or (re)designing a
decision structure helps to solve it.

Note that the structure (m (m (m P1) (m P2)) (m P3)) does not solve the paradox. It yields the
forward cycle zxyz, which happens to be the same cycle as produced by S3.

We study the exemplary Condorcet paradox profile

xyz
zxy
y z x.

Consider the structure (m (m xyz zxy) (m yzx)). Clearly, this yields yz, zx and (xy).
Hence, y is the majority decision for this structure.

Now consider the structure (m (m xyz) (m zxy yzx)). The result now is xy, yz and (xz)
and hence x is the structural majority decision.

Subsequently, when using the structure (m (m xyz, yzx) (m zxy)), the result is zx, (zy)
and xy. Therefore z is now the structural majority decision.

So we found three decision structures that solve the paradox. Moreover, for each a {x, y, z}
there exists an organizational structure such that a is the social choice.

22

More general, we have the following results.


Let = { | : N O(X)} where O(X) = {R| R is a weak ordering on X}. A for
which no majority winner exist is called a Condorcet paradox.

THEOREM 1
For any a X = {x, y, z} and any Condorcet paradox given X there exists a structure that
solves for a

Note that this theorem deals with Condorcet paradoxes with weak orderings. The next
theorem shows that any Condorcet paradox can be solved:

THEOREM 2
For any Condorcet paradox there is an organization decision structure that solves the paradox.

The proof of both theorems is in the appendix.

There exists a long discussion about the relevance of the Condorcet paradox and the empirical
occurrence of majority cycles. Probability calculations and analytics in the case of spatial
voting games show that the paradox should occur frequently. However, as Gerhlein
formulates it (Gehrlein, 2006, p.58):

Numerous empirical studies have been conducted to determine if Condorcets Paradox is


ever observed in actual elections. After surveying these studies, we must conclude that the
evidence does not suggest that the phenomenon is widespread in voting situations.

23

Gehrlein does not discuss experimental evidence in the context of spatial voting games.
However, the conclusion is the same: there is no (experimental) evidence of (global) cycling
in the case of empty cores. See e.g. Fiorina & Plott 1978, McKelvey & Ordeshook (1990)
and Bianco et al. (2006).

Clearly, there are many answers to the question why there are no cycles or why there is so
much stability. One answer according to Theorem 2 is organizational structure. In the social
and political world, almost everything is organized. Organizational structures abound in social
reality. And, as shown above, organizations may have structures that solve for Condorcet
paradoxes. That is, democratic decision making may be organized in such a way that cycles
are avoided.

7. CONCLUSION AND DISCUSSION


The above analysis clearly shows that structure matters. Condorcet paradoxes can be solved
by imposing a decision structure on the set of individual preferences that constitute a paradox.
Moreover, we have shown that different structures may yield different majority decisions for
a fixed preference set. This raises questions about the (re)design of organizational decision
structures.

Our work is in the line of the work of Shepsle (1979, 1986). We also raise the question
whether majority disequilibria can be avoided by means of structures or institutions. And also
our answer is affirmative. However, we also show that structures may induce majority
disequilibria. Moreover, we wished to show the effects of structure on majority decision
making in general.

24

Our structural results have clear implications for organizational decision making models, e.g.
the Garbage Can Model of Organizational Choice (Cohen, March and Olsen (1972). One of
the key issues in this model is the concept of problematic preferences. Unfortunately, this
concept is nowhere elaborated in the Cohen et al. paper or in subsequent work on this model.
However, if we interpret this as intransitive preferences, then our analysis makes clear that
organizational garbage cans may produce problematic preferences, but in spite of this,
organizational decision making may produce nice and consistent results. Conversely, within
organizations the decision units may produce non-problematic preferences but precisely
because of the organizational structure, the overall decision making process may be
jeopardized. In Bendor et al. (2001) the garbage can model is critically assessed. They
propose to combine the model with rational choice analytic approaches in order to innovate
and develop the model into a positive theory of organizational decision masking. In a sense,
this paper is a first trial in that direction.

REFERENCES

Arrow, K. (1963). Social Choice and Individual Values. 2nd ed. New Haven: Yale
University

Black, D. (1958). The Theory of Committees and Elections. Cambridge: Cambridge


University Press.

Bendor, J., Moe, T., & Shotts, K. (2001) Recycling the Garbage Can: An Assessment
of the Research Program. The American Political Science Review. 95: 169-190.

Cohen M., March, J. & Olsen, J. (1972). A Garbage Can Model of Organizational
Choice. Administrative Science Quarterly. 17: 1-25.

25

Cyert R. & March, J (1992). A Behavioral Theory of the Firm. Blackwell Publishing.
Second Edition.

Fine, K. (1972). Some necessary and sufficient conditions for representative decision
on two alternatives. Econometrica, 40, 1083-1090.

Fishburn, P. (1973). The Theory of Social Choice. Princeton. Princeton University


Press.

Jackson, M.O. (2001). A crash course in implementation theory. Social Choice &
Welfare, 18: 655-708.

Maskin , E. & Sjstrm T. (2002). Implementation Theory. In K. Arrow, A. Sen & K.


Suzumura (2002). Handbook of Social Choice and Welfare. Vol. 1, p237-289.
Amsterdam: Elsevier.

Murakami, Y. (1966). Formal structure of majority decision. Econometrica, 34, 709718.

Murakamy, Y. (1968). Logic and Social Choice. New York. Dover Publ.

Pattanaik, P. (1970).Voting and Collective Choice. Cambridge: CUP.

Seibel, P. (2005). Practical Common Lisp. New York. Springer Verlag.

Shepsle, K. (1979). Institutional Arrangements and Equilibrium in Multidimensional


Voting Models. American Journal of Political Science. 23: 27-59.

Shepsle, K. (1986). The positive theory of legislative institutions: an enrichment of


social choice and spatial models. Public Choice. 50: 135 -178.

Slade, S. (1998). Object-oriented Common Lisp. Prentice Hall.

26

You might also like