Politics, Policy, and Organizations
of preference pro
files, one system’s bureaucracy will have more autonomy
than the second system’s, while for some other pair of preference pro
files
the second system’s bureaucracy will have more autonomy than the
first.
This would mean that we cannot rank the six systems in decreasing order
of bureaucratic autonomy. Hence, it would not be clear, for hypothesis-
testing purposes, what our theoretical expectations should be.
So the question is: To which of these two conclusions do our models
lead? To answer this question, we will make three different kinds of
comparisons.
Variations in the Size of Each System’s Core
Observe in
figures 2 through 7 that the core of each system can take on
a wide range of sizes. In particular, the core of each system can contain
either a single policy or multiple policies.
1. The Majority Party Unicameral Core (with perfect discipline) can
contain either a single policy (
figs. 2A, 2B, and 2C) or multiple
policies (
fig. 2D).
2. The Majority Party Unicameral Core (without perfect discipline)
can contain either a single policy (
fig. 3C) or multiple policies
(
figs. 3A and 3B).
3. The Party Coalition Unicameral Core (with perfect discipline)
can contain either a single policy (
fig. 4D) or multiple policies
(
figs. 4A, 4B, 4C, and 4E).
4. The Party Coalition Unicameral Core (without perfect discipline)
can contain either a single policy (
fig. 5D) or multiple policies
(
figs. 5A, 5B, and 5C).
5. The Party-Free Bicameral Core can contain either a single policy
(
fig. 6C) or multiple policies (figs. 6A and 6B).
6. The Party-Free Bicameral Executive Veto Core can contain either
a single policy (
fig. 7E) or multiple policies (figs. 7A through
7G).
For each system, then, the core can vary in size from a single policy to a
range of policies.
Empirically, this has two implications. First, since the policy prefer-
ences of the elected of
ficials in a system may vary from issue area to issue
area, the size of the system’s core may vary from issue area to issue area.
This suggests that bureaucratic autonomy in this system can thus be ex-
96
pected to vary from agency to agency. Second, since the policy prefer-
ences of the elected of
ficials in a system may vary over time (which could
happen for a variety of reasons) the size of the system’s core may vary over
time as well. Hence, bureaucratic autonomy in the system can also be ex-
pected to vary over time.
Comparing the Sizes of the Cores for Any Two Systems
The possible variation in size for each system’s core has direct implica-
tions for what we might expect from comparisons of bureaucratic au-
tonomy across pairs of systems.
With six systems, one can make a total of (
6
2
)
ϭ 15 pairwise compar-
isons of the sizes of two systems’ cores. Fortunately, it is not necessary for
our purposes to make all
fifteen comparisons. The reason stems from the
fact, noted earlier, that every system can have a core with just a single
policy or a core with multiple policies. Thus, to compare system i (for i
ϭ 1,2, . . . ,6) with system j (for j ϭ 1,2, . . . ,6) we need only consider
four possible situations (for i
j).
1. If the core of system i contains a single policy and the core of
system j contains multiple policies, then the core of system i will
be smaller than the core of system j.
2. If the core of system i contains a single policy and the core of
system j contains a single policy, then the core of system i will
be the same size as the core of system j.
3. If the core of system i contains multiple policies and the core of
system j contains a single policy, then the core of system i will
be larger than the core of system j.
4. If the core of system i and the core of system j both contain mul-
tiple policies, the core of system i can be smaller than, the same
size, or larger than the core of system j.
Because each system could have a single policy or multiple policies in
its core (as noted in the previous section), it follows that virtually any-
thing could emerge from a comparison of any pair of systems: the core of
system i could be smaller than the core of system j, or the core of system
i could be the same size as the core of system j, or the core of system i
could be larger than the core of system j. In other words, the two sets of
rules de
fining any pair of systems do not necessarily lead to systematic
differences in the extent of bureaucratic autonomy.
Veto Points in Democratic Systems
97
Politics, Policy, and Organizations
Rank Ordering the Core Sizes for All Six Systems
This “anything can happen” result has a further implication. If we are
conducting an empirical study of bureaucratic autonomy in all six sys-
tems, we might wish to develop some prior theoretical expectations, for
hypothesis-testing purposes, about how to rank these six systems in
terms of their bureaucratic autonomy. With six different systems in our
study, there are 6!
ϭ 720 possible rank orderings of the sizes of their cores
(ignoring the possibility of ties).
However, since for any pair of systems one system can have a larger
core than the second system, and the second system can have a larger core
than the
first, it follows that each of these 720 possible rankings could
possibly occur. That is, there is no logically necessary rank ordering of bu-
reaucratic autonomy across our six systems.
Discussion
These three sets of results—on the possible size of any one core, on the
pairwise comparison of the sizes of any two systems’ cores, and on the
rank ordering of the sizes of the cores of all six systems—suggest that
there does not exist a logically necessary relationship between the policy-
making rules de
fining a particular system and the size of the system’s
core. This means that knowing just the systems’ policy-making rules does
not allow us to develop any logically valid expectations about which sys-
tems will have bureaucracies with more autonomy and which will have
bureaucracies with less. And if it is not clear what the expectations
should be it is not clear what testable hypotheses can be derived. Hence,
it is not clear what could be learned from the empirical research. If any-
thing could happen theoretically, then empirical research that focuses
only on the impact of the systems’ policy-making rules on bureaucratic
autonomy will not be theoretically informative.
The key implication is that we cannot rely just on the institutional
variables—that is, on the number and variety of veto points and the
other policy-making rules—to structure and inform our cross-national
research on bureaucratic autonomy. Because results from any empirical
study will always be due to the interaction among veto points and pref-
erence pro
files, if the preference profile variable is omitted from the em-
pirical analysis these empirical results will be ascribed, erroneously, to
just the impact of the institutional variables.
98
Of course, an identical argument can be made about the hazards of
relying just on the beliefs and preferences of the elected politicians—that
is, on the characteristics of the preference pro
file—to structure and in-
form our empirical research. Note that some comparative politics re-
search uses results from public opinion surveys as indicators of various
kinds of national cultures and then attempts to explain trends in national
policy-making on the basis of these changes in cultures (see, e.g., Ingle-
hart 1990). However, such studies rarely integrate their preference pro
file
data with any institutional variables. Unfortunately, if the institutional
variables are omitted from an empirical analysis of policy trends any re-
sults that are due to the interaction among the veto points and the pref-
erence pro
files will be ascribed, erroneously, just to the preference profile
variable.
So if we wish to empirically examine bureaucratic autonomy in a
comparative perspective it is imperative that measures of both the insti-
tutional and preference pro
file variables be included. Unfortunately, this
will greatly complicate cross-system empirical research. It is a dif
ficult
though manageable task to gather cross-system data on the institutional
variables. However, gathering data on each system’s preference pro
file is
likely to be much more dif
ficult (as well as time consuming and expen-
sive). For this reason, it will be dif
ficult to conduct meaningful cross-
national empirical research on bureaucratic autonomy.
Possible Criticisms
A number of criticisms might be aimed at the approach from which
these conclusions have been derived. In reviewing these criticisms, what
must be considered is whether the two broad conclusions—the “any-
thing can happen” results and the necessity of including both preference
pro
file and institutional variables in our theoretical and empirical re-
search—would have to be modi
fied if any one criticism is valid. In gen-
eral, even where there is some basis for the criticisms it is not apparent
that either of the two central conclusions is signi
ficantly undermined.
More Complex Sets of Rules for Each System
The set of rules characterizing each of our six systems is undoubtedly far
simpler than the full set of rules actually characterizing any real world
country. Inclusion of a wider range of institutional variables could be ex-
pected to change the size, shape, and location of each system’s core.
Veto Points in Democratic Systems
99
Politics, Policy, and Organizations
For example, committees with gatekeeping authority could be in-
cluded (especially in presidential systems) along with a veto override, the
nomination and con
firmation process for bureaucratic leaders, procedures
for their dismissal or removal, and the courts (see Hammond and Knott
1996 for details on how these all might be included in a spatial model of
bureaucratic autonomy in a presidential system). Agency budgets and the
appropriations process could be included as well. For coalition govern-
ments in parliamentary systems, how cabinet seats are allocated to various
parties might affect the extent of autonomy for the bureaucracies in-
volved. And if bureaucratic autonomy also stems from informational
asymmetries between the bureaucrats and the elected of
ficials, as Weber
hypothesized, then these asymmetries could also be included.
Changes in any of these factors may well change the size, shape, and
location of the resulting cores, given some preference pro
file. However,
it is unclear what the net impact of these additional factors would be: in-
clusion of some variables might increase the size of a system’s core (e.g.,
if more veto points are added, as with legislative committees or multi-
party coalition governments), whereas the inclusion of other variables
(e.g., the appropriations process and the chief executive’s ability to dis-
miss the agency head) would seem likely to decrease the size of the sys-
tem’s core. Moreover, it could be argued that some aspects of our mod-
els, such as the number of political parties, should be endogenized and
treated as a product of the systems’ electoral rules (which would also have
to be included in our models).
Nonetheless, the key issue is not whether these models are suf
ficiently
descriptive of real world countries but whether our central conclusions
would change if more complete rules were developed for each system. In
part because it is not clear what the net effect of including all these ad-
ditional variables would be, it is not clear that these conclusions would
be undermined.
The Unidimensionality Assumption
It could be asserted, with some plausibility, that policy-making in many
political systems is usually multidimensional and not unidimensional.
Nonetheless, even with a multidimensional representation of policy-
making in each of our systems it seems likely that the anything can hap-
pen results would emerge from a theoretical multidimensional analysis.
The reason is that, even in a multidimensional setting, for each pair of
100
systems there probably exist pairs of preference pro
files that would pro-
duce sets of equilibrium policies that vary greatly in size, thereby repro-
ducing the anything can happen result.
Of course, for any one system preference pro
files that produce any
cores at all may be less common in higher dimensional issue spaces than
in lower dimensional spaces. The implication is that bureaucratic auton-
omy would be less likely in multidimensional settings: for any policy the
bureaucracy might adopt, the absence of a core means that there exists
some other policy that some decisive coalition of elected of
ficials would
prefer. Hence, the bureaucracy would not be in a position to play a “di-
vide and conquer” game with these of
ficials.
However, Humphreys (2001) presents formal and simulation results
indicating that cores are not completely improbable in higher dimen-
sional spaces. Moreover, a different line of work—see, for example,
Baron and Ferejohn 1989—indicates that even in multidimensional spa-
tial settings policy stability may exist if elected of
ficials find the unend-
ing decision making implied by policy disequilibrium to be costly. In ei-
ther case, policy equilibrium may be maintained, with the result that
some bureaucratic autonomy may still be possible.
The Empirical Improbability of Particular Preference Profiles
Several of the core sizes generated by our models stem from preference
pro
files that may seem empirically improbable. If these empirically im-
probable pro
files are eliminated from consideration, this might place at
least some constraints on what we should expect theoretically when com-
paring two or more systems.
Nonetheless, while it may be possible to rule out some preference
pro
files as empirically improbable for particular systems it remains un-
clear whether as a result any one system, given the restricted range of
pro
files, will necessarily produce a core that is always larger than, or al-
ways smaller than, the core of some other system. Hence, it is unclear
that these restrictions would undermine our central conclusions.
Conclusion
The major conclusions thus remain the same. First, even with all the
modi
fications just proposed, different preference profiles can still be ex-
pected to change the size of a system’s core, holding constant whatever set
of policy-making rules is attributed to this system. And, second, this
Veto Points in Democratic Systems
101
Politics, Policy, and Organizations
means that any effort to empirically investigate the impact of policy-mak-
ing rules on bureaucratic autonomy (either within or across systems)
should incorporate the preference pro
files as a variable. As a general rule,
then, policy choices by a system must be seen as the product of an inter-
action between the policy-making rules and the preferences of the actors
in the system. Hence, empirical efforts to explain variations in the extent
of bureaucratic autonomy within and across democratic systems must
take both sets of factors into account.
Notes
1. Tsebelis (1995) does explicitly talk about the impact of the preference
pro
file on policy stability (see, e.g., his proposition 2, p. 298, and also pp. 308–
11), and his empirical work (see, e.g., Tsebelis 1999) takes into account the ide-
ological range of governing coalitions. However, the general nature of the inter-
action between the number of veto points and the preference pro
files in various
kinds of systems remains underexplored.
2. A “decisive coalition” is one that, by the system’s policy-making rules, is
empowered to select some new policy. Thus, in the United States there are two
possible decisive coalitions in the policy-making process: (1) a coalition of the
president, a House majority, and a Senate majority; and (2) a coalition of two-
thirds of the House and two-thirds of the Senate. In a unicameral parliament, a
decisive coalition would be simply a majority of the single chamber.
3. This is slightly nonstandard terminology, for if some legislators comprise
a minority they cannot generate a win set. Instead, I am using the term win set
to indicate the area where several preferred-to sets overlap and then modifying
the term to indicate whether it is generated by a minority or a majority.
4. If the minority party does not automatically vote against the proposal of
the majority party but each of its members instead simply votes in terms of
whether the majority party’s proposal is better or worse for him or her than the
status quo, then the Majority Party Unicameral Core could again be different.
To illustrate, consider a case in which SQ lies just to the right of C
3
in
figure 3 A.
A majority party proposal to replace this SQ with a policy at C
3
would be re-
jected by C
4
and C
5
but would be supported by the other party members—C
1
,
C
2
, and C
3
—as well as all four Labor Party members. Hence, points to the right
of C
3
are not in equilibrium. However, if SQ lies between C
1
and C
3
a majority
party proposal to replace it with a policy at C
3
would be supported at most only
102
by C
3
, C
4
, and C
5
(and possibly C
2
, depending on the location of SQ); it would
be rejected at least by C
1
and all four Labor Party members, who collectively
comprise a parliamentary majority. Hence, the Core would span the ideal points
from C
1
to C
3
.
5. If the minority party members here do not automatically vote against the
majority party proposal but instead vote on the basis of the utility of the pro-
posal to them, then the Majority Party Unicameral Core here spans just the C
3
and C
4
ideal points. Consider a case in which SQ lies to the right of C
4
. A party
proposal to replace this SQ with a policy at C
4
would be approved by four
Conservative Party members— C
1
, C
2
, C
3
, and C
4
—as well as L
1
and L
2
, for a
total of six votes; these six votes comprise a parliamentary majority. Hence, sta-
tus quo policies to the right of C
4
cannot be in equilibrium. Similarly, SQ poli-
cies to the left of C
3
are not in equilibrium. However, if SQ lies between C
3
and
C
4
any proposal to replace this SQ with a policy to its left would be opposed
by C
4
, C
5
, C
6
, and C
7
, who comprise a majority of the Conservative Party.
While a majority of the parliament would support this proposal (i.e., C
1
, C
2
,
C
3
, L
1
, and L
2
), since a majority of the majority party (the Conservative Party)
opposes the proposal it would never be sent to the
floor. And if SQ lies between
C
3
and C
4
any proposal to replace this SQ with a policy to its right would be
opposed by C
1
, C
2
, and C
3
as well as L
1
and L
2
. Hence, the Core here spans just
the C
3
to C
4
interval.
6. If the members of the opposition party vote simply on the basis of their
own individual valuation of SQ and the policy proposal, the Core here will also
span just the ideal points of the median members of the coalition.
7. The creation of a veto override will never increase the size of the core but
can decrease it.
Veto Points in Democratic Systems
103
Dostları ilə paylaş: |