Tsebelis G. Veto Players: How Political Institutions Work

Подождите немного. Документ загружается.

331

study judicial discretion as a dependent variable. As I have argued in this chapter political

systems that exhibit policy stability will have also independence of the judiciary. Is there any

empirical evidence to support this claim? Several empirical studies have tried to measure

independence of the judiciary. Some of them have tried to test predictions similar to the ones

introduced in this book.

Lijphart (1999:225-26) has introduced a measurement of strength of judicial review

“based, first on the distinction between the presence and absence of judicial review and, second,

on three degrees of activism in the assertion of this power by the courts.” [emphasis added]. He

also determines the difficulty by which the constitution of a country is amended (by higher than

2/3 majority, by “2/3 or equivalent,” “between 2/3 and ordinary majorities,” and by ordinary

majorities). He scores the 36 countries he studies on the basis of these two variables, and finds

that “judicial review” has a moderate but statistically significant correlation with constitutional

rigidity.

Similarly, Nicos Alivizatos (1995) has introduced a fourfold typology of what he calls

“judicial politicization,” that is, the extent to which judges “influence the decision-making

process”. For this purpose he determines a dependant variable on whether a country has a

constitutional court or not, and whether the judges are considered activists or not.

132

Countries

with a constitutional court and activist judges are scored four, countries with a constitutional

court and non-activist judges are scored three, countries with no constitutional court but active

judges two, and countries without a court or judicial activism are scored 1. He introduces a series

of possible independent variables to assess what causes “judicial politicization.” First, he

classifies countries as decentralized or not (actually using three categories one for federal

132

To use the author’s specification: “depending on whether their courts have actually given unambiguous signs of

judicial activism as opposed to judicial self-restraint.”

332

countries, one for de facto federalism, and one for unitary countries); second, he assesses the

degree of left-right polarization (in two categories); third, he introduces a qualitative variable

expressing veto players (scored one for single party, three for multiparty, and two for mixture of

the two); fourth, he introduces a variable assessing parliamentary anomalies (civil wars or

dictatorships); fifth, a variable indicating degree of integration in the EU (at the time some

countries were members, some were about to become members, and others were not). He finds

that the decision to have a constitutional court depends mainly on two variables: whether the

country is federal and whether in the past there were parliamentary anomalies. “Judicial

politicization” on the other hand depends on decentralization, polarization, and veto players (at p

levels less than .05).

Finally, Robert Cooter and Tom Ginsburg (1996) have used a scale of “judicial

discretion” generated by a series of experts. A second and far more creative way of assessing

judicial discretion is through the move to restrict liability. “We found that courts with high

predicted daring were willing to innovate in this area of private law more than other courts.

Courts with low predicted daring, by contrast, were content to wait for legislative adoption of the

new standard.” In fact, the two authors classify different countries on the basis of whether the

courts introduced strict liability standards before the legislature, or whether they simply reversed

the burden of proof from the plaintiff to the defendant, or waited for the legislature to change the

law.

Cooter and Ginsburg use two independent variables. The first is “the number of

legislative vetoes” that is determined the following way: “in unicameral parliamentary systems,

where the government is formed by the majority coalition in the legislature, there is essentially

one veto on legislation…. Other systems have two vetoes on new legislation. Such would be the

333

case in either a bicameral parliamentary system (as in Germany), or an essentially unicameral

parliamentary system with a strong president (as in France).” The second independent variable

that they use is the duration of government coalition as “a simpler indicator of party dominance”

(1996: 299). They find that both these independent variables affect judicial discretion, whether

measured by the judicial expert assessments or by the strict liability test.

The careful reader may have already identified problems with each one of the two

independent variables used by Cooter and Ginsburg. The variable “number of vetoes” uses what

I called the “numerical criterion” in Chapter 1. I stated there that this is a questionable basis for

comparative statics across countries. It ignores differences in the ideological positions of

different veto players, that may be very important form one country to another. In addition, the

particular scores do not adequately reflect the institutions of the different countries. For example,

France is scored with two, although the President has no veto power; the Netherlands or Austria

with two, although their respective upper chambers are very weak (in fact, Austria’s is weaker

than the UK’s which is scored as one veto). Israel and Denmark received a score of one because

they are unicameral, despite the fact that they usually have coalition governments etc. The

variable “duration” as a measure of dominance of one sort or another has been criticized in third

section of Chapter 4, so I will not repeat my objections here.

These studies have used different ways of measuring judicial independence, and have

correlated it with different independent variables, some of which were connected with veto

players, some of them not, and yet others supposedly connected but incorrectly so. I will use the

independent variables in each one of these studies: judpol for Alivizatos’ “judicial politicization”

judrev for Lijphart’s “judicial review” experts and strict liability for Cooter and Ginbsburg’s two

different measures of judicial independence. I will correlate these variables with the qualitative

334

measure of veto players introduced in Chapter 7. I have extended this measure for the countries

covered by Cooter and Ginsburg, but not for the 36 countries of Lijphart. So, some of the tests

will cover the 18 countries in the Alivizatos dataset, some the 20 countries in of Cooter and

Ginsburg, and others all 24 countries included in the table.

INSERT TABLE 10.1

Table 10.1 introduces the data to be analyzed. The missing data are generated because

Alivizatos covers only West European countries while Cooter and Ginsburg have a significant

intersection with these countries but cover other countries as well. Only Lijphart covers all 24

countries. Table 10.2 produces the Pearson correlations as an easily interpretable measure of the

association among the different independent variables. The reader can verify that while

Lijphart’s data correlate very well with Alivizatos’ assessments in the subset of West European

countries, they have significant differences with the expert judgments when we go to countries

outside Europe. In particular, Israel is classified as a very weak judiciary by Lijphart (one) and

very strong by the expert reports (4.5 is even higher than the US Supreme Court), Australia and

Canada are coded as very independent in Lijphart (three), but not strong by the expert reports

(2.33). Finally, another country of disagreement is the Netherlands where the judiciary is

considered very independent by the experts, not at all by Lijphart, and intermediate by

Alivizatos. These remarks are made to indicate that even experts disagree with some

assessments. It is possible that these disagreements are based on different underlying properties

of the variable that each one of them examines (as the different names they use indicate). If one

eliminates these four countries of disagreement, the correlations between the expert judgments

become significantly higher. Given the scarcity of countries and data, I do not intend to drop the

countries where experts disagree from the empirical tests.

335

INSERT TABLES 10.2A AND 10.2B

Table 10.3 presents the relationship between the different measures expressing the

independence or importance of the judiciary and the corresponding dependent variables, one of

which is always a qualitative expression of veto players. While all the literature under review

uses OLS estimation, the technique is inappropriate given that the dependent variable is

composed of only three or four groups of countries. In the estimations that follow I use an

appropriate ordered probit technique whenever the dependent variable has discrete values. The

rows marked by *** refer to OLS estimations because the dependent variable is continuous. For

comparison purposes I have estimated the data generated by Alivizatos twice: once by the correct

procedure (ordered probit) and once by the traditional one (OLS). Note the difference in pseudo

of the probit estimation in the first row from the adjusted R

of the OLS technique in the last.

INSERT TABLE 10.3

I have estimated different models. In each case I started by including two variables: veto

players and political decentralization. I discussed the “veto players” variable in Chapter 7. The

variable decentralization is measuring whether a country is unitary, federal, or in-between. I have

included this variable because the judiciary in federal countries may be asked to adjudicate

conflicts among different levels of government, so there is an additional source of significant

decisions to be made by the judiciary. In fact, as we saw in Chapter 6 the independent judiciary

is considered by some authors (Bednar, Eskridge and Ferejohn (2001)) as a condition for

federalism. Like some of the literature (Alivizatos) I have added the variable “polarization.” This

is essentially a qualitative version of the variable “alternation” that we used in Chapters 7 and 8.

However, it is not clear that this variable will have a positive or negative effect on the role of the

judiciary. It can be argued that the judiciary will moderate extremes of different governments, or,

336

that it will be intimidated by the perspective that it can almost always be overruled. As a matter

of fact his variable comes out positive or negative depending on the model. When it comes out

non-significant I re-estimate the model by dropping “polarization.” In the case of the two

independent variables from the Cooter and Ginsburg article I also integrated a model including

the variable “government duration” because according to the authors’ expectations it should be

significant. It turns out that it was not. Finally, I do not include in the table tests of the judicial

system (civil law vs. common law), because none of them comes out statistically significant, and

some of them even have the wrong sign.

The results presented in Table 10.3 lead to the following conclusions. In all the models

except for two the statistical significance of veto players is high (above .05 in one tailed tests).

The two exceptions are the one using Lijphart’s independent variable with no statistical

significance,

133

and the one testing Cooter and Ginsburg’s hypothesis on the basis of strict

liability data (significance at the .10 level).

In addition, most of the time decentralization is statistically significant. On the other

hand, polarization does not appear to be significant except for Alivizatos’ indicator and comes

out several times with the a different sign. Consequently, the empirical evidence corroborates the

expectation that independence of the judiciary increases as a function of veto players. In

addition, most of the time there is empirical support for the idea that federal countries will have

more independent judiciary than unitary ones. There is no evidence that the judicial system of a

country (common vs. civil law) or the polarization of political forces in it affect judicial

independence.

133

Even with Lijphart’s variable the z-statistic doubles when one drops the 4 countries where the legal experts

significantly disagree with Lijphart’s scoring.

337

III. Veto players and bureaucracies

In this section I will first discuss the different arguments about bureaucracies presented in

the literature. I will single out and focus in particular on two arguments that come to different

expectations from the theory presented in this book. The first has to do with the independence of

bureaucracies, whether it increases or decreases with veto players; the second with the

independent variable that explains bureaucratic independence: is it veto players or regime type?

a. Theories of bureaucracies

Mathew McCubbins, Roger Noll and Barry Weingast have written a series of articles that

compose probably the most influential study of administrative law. In two of the most important

of them (McCubbins, Noll and Weingast (1987), (1989)) the authors (from now on McNollgast)

focus predominately on how legislatures create administrative law that effectively restricts the

bureaucracy to perform the duties prescribed by the enacting coalition. The basic problem

according to them is moral hazard, that is, the possibility for bureaucracies to choose policies

that differ from the preferences of the enacting coalition.

In order to avoid moral hazard, legislatures can create administrative law, which has three

major characteristics: First, the enacting coalition should create for the bureaucracy an

environment that mirrors the politics at the time of enactment. Second, they should stack the

deck in favor of the groups that are the most affected and the most favored by the coalition.

Third, agency policies should exhibit an autopilot characteristic, that is, enable policy changes as

the preferences of the interested groups change.

This analysis has certain consequences for the model developed in the first part of this

chapter: Given that the first movers in the game I presented above will be able to select a policy

close or identical to their own ideal point, the legislative branch will stack the deck to avoid this

338

possibility. McNollgast (1987, 1989) argue that no political actor will enter an agreement unless

their interests are protected, and as a result, legislators will seek to create such protection to

themselves when they write administrative law.

On the basis of the arguments and findings of this literature, Terry Moe (1990) and David

Epstein and Sharyn O’Halloran (1999) have argued that legislation will be more cumbersome

when the legislative body is more divided because they will try to lock into the legislation the

intents of the coalition that produced it, leading to a reduction of the independence of

bureaucracies. This argument seems to contradict my account, and I need to address it in detail.

My argument is based on what may happen after the bureaucratic decision (ex post),

while the McNollgast arguments are based on what the legislature will do before (ex ante). I

expect that keeping legislation constant bureaucrats and judges will be more independent from

government when there are many veto players. McNollgast’s “deckstacking” argument does not

keep legislation constant; in fact, the essence of this argument is to compare the different kinds

of legislation produced under different veto player configurations. In addition, the “decksacking”

argument concerns independence of bureaucrats from the coalition enacting legislation. These

are two different arguments; let me take them one at a time.

First, let me give my own expectation for administrative legislation as a function of veto

players. Given the freedom of courts and bureaucrats to interpret legislation freely when there

are multiple veto players, these veto players will prefer to restrict them ex ante, that is, would

like to include procedural restrictions inside the legislation itself (exactly as McNollgast argue).

Would they be able to do so? It depends on their preferences for this kind of legislation. If their

preferences are similar then they will be able to do so. If, however, they have preferences as to

how to tie the hands of different bureaucrats (one party wanting to empower citizens to blow the

339

whistle, the other to have strong and independent monitoring agencies to use a well known

example (Schwartz and McCubbins (1985)), there may not be agreement. So, in case of multiple

veto players the actual prediction depends on the preferences of the existing veto players. On the

other hand, single veto players can overrule bureaucrats or judges at any time (assuming no

transaction or political costs). As a result, such governments would not care about introducing

additional restrictions into legislation.

This argument expects cumbersome bureaucratic legislation to be sometimes the outcome

of multiple veto players, while simple legislation to be always the outcome of single veto

players. In other words, multiple veto players are a necessary but not sufficient condition for

cumbersome bureaucratic legislation. As a result, on average, one would expect more

cumbersome legislation in the case of multiple veto players (as McNollgast predict), but would

also expect higher variance in the case of multiple veto players, exactly as the argument

presented in Chapter 1 specifies (see Figure 1.8). Huber and Shipan (2000) and Franchino (2000)

have found different average levels of restrictive legislation, but have not tested for the variance

component of this argument.

My second point is that the “deckstacking” argument is talking about bureaucratic

independence from the enacting coalition, while I am interested in independence from

government, or from the political principals at the time of the decision. It may be the case that

the enacting coalition was successful in restricting bureaucrats through ax ante restrictions built

into law, but if this coalition is replaced, then the new principals will not be able to force the

bureaucrats to obey their wishes. In order to do that they may have to change the law, and a multi

veto player coalition may be unable to do so.

The distinction between enacting and current coalition may not be significant in the US,

340

because divided government has been the rule for decades, and given supermajoritarian

restrictions in the Senate (Chapter 6) even the periods of one party government almost disappear.

However, in other countries one cannot eliminate the distinction between enacting coalition and

coalition in power, and it is independence from the current political leadership that defines

bureaucratic or judicial independence.

In conclusion, my argument is that, keeping legislation constant, bureaucratic

independence from current veto players increases with the number and distances among veto

players. In addition legislation becomes more cumbersome on the average (although the variance

of this prediction is a function of veto players). Under the same conditions, the deckstacking

argument expects more cumbersome legislation, and less independence of bureaucrats from

enacting coalitions. The two arguments are different from each other and not incompatible.

The second issue raised by the literature on bureaucracies is whether there are specific

problems of multiple principals associated with the presidential regime as opposed to veto

players. There are several articles that make the claim that parliamentarism has a unity of

direction of bureaucracies compared to presidentialism. The argument has been made by Moe

and Caldwell (1994) who compare the British and the American system, by Ackerman (2000)

who criticizes the bureaucracies of the US, and by Strom who analyses not just bureaucracies but

more generally the delegation characteristics of parliamentary systems. Out of these three pieces,

the first two make arguments that do not identify the causal connections. It is indeed possible

that the US system presents problems of unity of direction of bureaucracies (as both Ackerman

and Moe and Caldwell claim) and that the British system has unity of direction (as Moe and

Caldwell argue) but the reason for these features is the distinction between single and multiple

veto players, not the difference in regimes. This argument however, cannot be made about