Politics, Policy,
and Organizations
a policy. Two political systems are thus different if their policy-making
rules are different. Understanding policy choice thus requires understand-
ing both the nature of the policy-making rules and the nature of the indi-
vidual preferences to be aggregated. Because the argument explored here
is that bureaucratic autonomy stems from the existence of a set of equi-
librium policies, how the policy-making rules interact with the individual
preferences to produce different-sized sets of equilibrium policies will be
our focus.
Six models are developed. Four are models of unicameral parliamen-
tary systems: two kinds of majority party systems (one with majority
party discipline, one lacking majority party discipline) and two kinds of
multiparty systems (one with coalition party discipline and one lacking
coalition party discipline). The last two are a model of a bicameral system
with two legislative chambers but no parties and a model of a presidential
system with two legislative chambers and a president but no parties.
For each of these six systems, our central purpose is to determine what
set of equilibrium policies is created. In particular, we identify the poli-
cies that are in the core, that is, the policies that no decisive coalition of
elected of
ficials could replace with some alternative policy, given their
preferences and their system’s policy-making rules. If a policy is in the
core, it is in equilibrium and cannot be upset. A large core thus indicates
a large amount of bureaucratic autonomy, while a small core indicates a
small amount.
We assume that each individual—for example, a member of Parlia-
ment, a president, a representative, or a senator—has a most-preferred
position on a unidimensional issue space. This most-preferred position
maximizes the individual’s utility and so is called his or her “ideal point.”
The farther some policy is from the individual’s ideal point (either to the
left or to the right), the less utility it provides; the individual’s utility
functions are thus single peaked.
We assume that each individual knows the location of the ideal point
of each other individual. And we assume that there is no disjunction be-
tween a formal policy choice by some authoritative actor or set of actors
(i.e., by the bureaucracy, a winning coalition in a parliament, or a win-
ning coalition of the president, House, and Senate in a presidential sys-
tem) and what policy is actually implemented. That is, I am developing
complete-information models.
To illustrate these concepts, in
figure 1A we assume that a member of
78
A: An individual’s preferred-to set
C: When SQ is at the median, the majority win set is empty: odd number of MP’s
B: When SQ is not at the median, the majority win set is not empty: odd number of MP’s
SQ
MP
P
MP
(SQ
)
MP
1
MP
2
MP
3
MP
4
MP
6
MP
7
MP
8
MP
9
P
MP2
(SQ
)
P
MP3
(SQ)
P
MP4
(SQ)
P
MP6
(SQ)
P
MP7
(SQ)
P
MP8
(SQ)
P
MP9
(SQ)
MP
5
SQ
P
MP1
(SQ)
P
MP5
(SQ)
MP
1
MP
2
MP
4
MP
5
MP
6
MP
8
MP
9
SQ
P
MP2
(SQ
)
P
MP1
(SQ)
P
MP3
(SQ)
P
MP4
(SQ)
P
MP5
(SQ)
P
MP6
(SQ)
P
MP7
(SQ)
P
MP8
(SQ)
MP
7
P
MP9
(SQ)
Majority win set of SQ
Minority win set of SQ
Minority win set of SQ
Minority win set of SQ
Majority win set is empty
MP
3
MP
1
MP
2
MP
3
MP
4
MP
5
MP
6
MP
7
MP
8
P
MP2
(SQ
)
P
MP3
(SQ)
P
MP4
(SQ)
P
MP5
(SQ)
P
MP6
(SQ)
MP
3
SQ
P
MP1
(SQ)
P
MP7
(SQ)
P
MP8
(SQ)
Minority win set of SQ
Minority win set of SQ
Majority win set is empty
D: When SQ is between the two medians, the majority win set is empty: even number of MP's
Fig. 1. Preferred-to sets and win sets
Politics, Policy, and Organizations
parliament (MP) has an ideal point at MP and that the status quo (SQ)
policy is at SQ. The dashed line with the brackets at each end en-
compasses the set of points that the member prefers to SQ; every point
within the brackets is closer to MP than is SQ, but the member is in-
different between SQ and a policy at the right-hand bracket. The set of
points that the member prefers to SQ is labeled P
MP
(SQ); it is this
member of parliament’s “preferred-to” set of SQ.
Assume that some nine-member unicameral parliament chooses poli-
cies via majority rule, as in
figure 1B. With SQ lying between MP
4
and
MP
5
, four members—MP
1
through MP
4
—wish to move policy to the
left, and they could agree on some point to the left of SQ in the region
where their preferred-to sets overlap; this region is labeled the “minority
win set of SQ” in the
figure.
3
However, their four votes do not comprise
a majority of the parliament, so they would not succeed in moving pol-
icy leftward. In contrast, the other
five members—MP
5
through MP
9
—
do comprise a majority, and they all wish to move policy to the right. In
particular, they could all agree on some point in the region where their
preferred-to sets overlap; this region is labeled the “majority win set of
SQ” in the
figure. They could thus succeed in moving policy rightward.
Next consider an SQ at the ideal point of the median member, MP
5
,
as in
figure 1C. In this case, no mutual improvement is possible for any
majority-sized coalition: members MP
1
through MP
4
wish to move pol-
icy leftward from MP
5
into the region where their preferred-to sets over-
lap (into the left-hand minority win set of SQ),
member MP
5
wants pol-
icy to stay at the MP
5
location (i.e., at his or her own ideal point), and
members MP
6
through MP
9
wish to move policy rightward from MP
5
.
Since there is no region where a majority—at least
five—of the pre-
ferred-to sets overlap (i.e., the majority win set of SQ is empty here), the
SQ at MP
5
cannot be upset. In fact, with an odd number of MPs the
only equilibrium policy lies at the median member’s ideal point.
When there is an even number of members, there is no unique me-
dian member. Instead, there are two median members and a set of equi-
librium policies that spans the ideal points of the two median members;
the set of equilibrium policies is the region between (and including) the
ideal points of the two median members. (We assume that the two me-
dian members do not have identical ideal points here.) In
figure 1D, for
example, there are eight members—MP
1
through MP
8
—and MP
4
and
MP
5
are the median members. Because SQ lies in the MP
4
to MP
5
space,
80