Ministerial drift, parliamentary conflict or pork barrel politics?
Why Italian government bills change in the legislative arena (1987-2006)
Andrea Pedrazzani , Dipartimento di studi Sociali e Politici , Università degli studi di Milano,
andrea.pedrazzani@unimi.it
Francesco Zucchini, Dipartimento di studi Sociali e Politici , Università degli studi di Milano
francesco.zucchini@unimi.it
Preliminary draft version. 11 gennaio 2011,
Do not quote without permission
Abstract
With very few exceptions scholars interested in legislative processes pay relatively little attention to
the changes made to bills in the parliamentary democracies. On one side comparative research often
describes legislative institutions in parliamentary countries as ineffectual vis-à-vis cabinets
throughout the lawmaking process (Mezey 1979; Blondel 1995); on the other side rational choice
literature focuses more on the formal rules regulating amendatory activity than on amendatory
activity in itself (Rasch 1995, 2000). Therefore few studies try to explain how much laws are altered
in the legislative arena and why.
In this paper, we focus on changes of governmental legislation in Italy. We consider as dependent
variable the modifications occurring during the legislative process, and we discuss a number of
explanatory hypotheses both drawn on the existing scholarship and deduced from our original
conjectures. Changes to governmental bills may stem from the ideological conflict between
coalition partners over the accomplishment of the agreed-on policy compromise (Martin and
Vanberg 2005), or from the conflict between legislators over the distribution of benefits among
their constituencies (Weingast, Shepsle and Johnsen 1981; Masciandaro 1996). Also, government
bills may be altered because governing parties choose to use the assembly - instead of the cabinet as the arena where to debate and define more precisely the content of legislation.
In order to empirically evaluate these arguments, we develop a measure of the level of
modifications that is based on the textual comparison between the initial bill and the final law. Our
original dataset tracks the legislative history of all the governmental laws introduced in the lower
chamber between 1987 and 2006. This time span covers both the so called Italian 'First' and 'Second
Republic', allowing to assess the impact of fundamental changes in the party system dynamics.
Introduction
Government bills in parliamentary democracies are often changed during the law making process.
This observation is so trivial that does not seem to deserve any further investigation. And in fact the
rational choice approach, that is dominant in legislative studies, has not dedicated for a long time so
much energies to the topics. Most of the time the literature has just focused on the formal rules
regulating the law making. The type of amending rule employed in a legislature has primarily been
studied as one of the institutional factors that constrain the actors‟ behaviour and thus shape the
legislative outcomes (see Rasch 1995, 2000). The real interest of these studies seems to gravitate
towards the level of similarity between the final legislative outcome and the proposer‟s ideal point.
Taken as stable the preferences, this similarity will depend on the rules and any change in this
similarity will depend on some change of the rules. It is in fact this similarity that allows us to
evaluate the power enjoyed by the proposer (the so called agenda setting power). However, the
distance between often unobservable desires and final outcomes does not correspond necessarily to
the changes (amendments) between an initial proposal and the final legislative outcome which we
can empirically observe. If the government political actors knew perfectly the position of the other
actors and the implications of the legislative rules, then they rarely would propose their ideal points
(or a compromise among their ideal points). They could propose on the contrary the position that
corresponds with the best policy they can obtain given the present rules and the distribution of
political preferences in the parliament. As they have already anticipated all moves from other
actors, we could not observe any amendment, any change even if there is a substantial gap between
their most preferred policy and the final outcome. Moreover, even a change of the rules that would
inescapably affect this gap could leave no evidence of changes in the government bill. The proposal
would be also the final outcome.
What makes possible (and consistent with the theory) the successful amendments that,
nevertheless, we continue observing in the real political world? What affects their number and
importance? A priori two simple facts that are sometimes neglected in the standard models of the
legislative process:
1) Not all political utility (and sometimes the smallest part) comes from the “legislative
outcome”. The legislative arena is not only a factory of laws but also a political stage where
political actors send messages to their “audience”. In other terms, sponsoring a bill has also
a position taking utility and proposing something you know will be changed is not
necessarily a useless activity;
2) Each political actor has usually incomplete information about the other actors‟ preferences
and has also imperfect information about the effects of the laws.
Three sets of explanations (not necessarily incompatible) of the bills are made possible by these
very general conditions:
a) Explanations based upon the lack of knowledge. In the cabinet the level of knowledge is low and
the following parliamentary step allows legislators to learn and to correct the initial proposals.
b) Explanations based upon the conflict. The bills proposed in the cabinet have a highly symbolic
value and they can represent the sincere preferences of the government parties. Therefore, the
discrepancy between the government bills and the government laws indicates true conflict among
the actors (and their electoral images) involved in the lawmaking process.
c) Explanations based upon the change of formal rules. Any change of formal rules from the “open
rule” type to the closed rule type will decrease the changes to government bills.
We won‟t consider this last set of explanations in our empirical test on Italian legislative
production between 1987 and 2006. During this time span the institutions regulating legislative
process in Italy do not undergo substantial reforms1.
In the following two sections of this paper we review the testable hypotheses that can be
subsumed under the first two sets of explanations. In the third section we introduce our empirical
analysis of the government legislative production in Italy between 1987 and 2006. We will explain
how we have operationalized the dependent variable: “government bills changes” and we will show
some simple descriptive statistics. We will derive also from the hypotheses we put forward in the
first theoretical sections the independent variables and we will explain their operationalization. In
the fourth section we will illustrate the first results of a multivariate analysis.
Explanations based upon the conflict.
Government bills are first deliberated in the Cabinet and then discussed and definitely passed in
parliament. Therefore the possible conflicts that account for a difference between initial proposals
and laws reflect a constellation of possible interactions in the Cabinet and between the Cabinet and
the Parliament. Two factors are crucial: 1) the decision making inside the cabinet and 2) the
distance in terms of preferences between the Cabinet and the Parliament. Literature about coalition
government provides two models of interaction inside the Cabinet. According to Laver and
Shepsle‟s ministerial discretion model, we can maintain that each minister drafts the bills falling
under her policy jurisdiction independently of the other members of the cabinet, and then presents
them to the floor for approval (see Laver and Shepsle 1994, 1996; Laver 2006). In this case, the
place where governing parties actually bargain is not the cabinet, but the legislature. According to
Tsebelis (2002), on the contrary, the responsibility of government decisions is mostly collective.
1
The only relevant formal change was when, in October 1996, the Italian Supreme Court ruled unconstitutional the
reissuance of decree laws.
Therefore, there is always a deliberation or a bargaining inside the cabinet and the changes we can
observe should depend on the distance between decision equilibria in the Cabinet and decision
equilibria in the Parliament.
Distance between Cabinet and Parliament may depend on two circumstances: 1) The
preferences that are taken in consideration at the Cabinet level are not majoritarian in the
Parliament. 2) The preferences are majoritarian but the application of an aggregation rule (majority
rule) equal or different from the aggregation rule in force in the parliament creates in the Cabinet an
equilibrium different from the equilibrium in the Parliament.
Tab.1 Explanations of government bills changes based upon the conflict and conditions for
their theoretical possibility
Cabinet- Parliament
Interaction
Cabinet
Cabinet
that
parliamentary
represents Cabinet that does not represent
majority parliamentary
preferences
majority
preferences
DecisionMaking
Bargaining inside the Cabinet
Conflict
Cabinet-Opposition Conflict
Parties
Cabinet-Opposition
Parties
Distributive Conflict
Conflict
with
Parliamentary
Majority (1)
Status quo and Veto players
approach.
Ministerial discretion
Conflict
Cabinet-Opposition Conflict
Parties
Cabinet-Opposition
Parties
Ideological conflict within the Distributive Conflict
Cabinet
Conflict
Conflict
among
with
Parliamentary
ministerial Majority (2)
domains
When Cabinets do not represent parliamentary majority pereferences
Government does not necessarily represent the parliamentary majority preferences in two cases: a)
when the government is a minority government; b) when the policies are distributive. This last point
deserves an explanation. Distributive projects (or pork barrel projects) are projects that concentrate
benefits in a specific geographic or “functional” constituency and spread their costs across all
constituencies through generalized taxation2. Even when the government parties control a majority
in the parliament, the functional or geographic constituencies represented directly by ministers in
the Cabinet cannot be large enough to enjoy a majority in the Parliament. Government bills are thus
modified in the legislative arena because “new constituencies” are added in order reach this
majority. The hypotheses derived from the above mentioned circumstances are the following ones:
H.1 Minority Governments: We expect more changes to government bills when the
government is a minority government
H.2 Distributive conflict: Bills with a distributive nature should be more extensively modified
than other types of bills.
Let us now imagine that in the Cabinet the decisions are taken by majority rule and that each party
enjoys equal consideration, no matter its size. In a one-dimensional policy space, the median voter
in the Cabinet can often be different from the parliamentary median voter. The government parties
may find an agreement to use a certain government bill in order to convey a symbolic
“compromise” message to the elctorate, but as the equilibrium in the Parliament is different this bill
will be changed. This is exactly an example of conflict between Cabinet and Parliament when there
are two different majority equilibria in the Cabinet and in the Parliament. The same type of conflict,
easier to explain, takes place if we assume the existence of ministerial discretion
H3. Conflict with Parliamentary majority: we expect more changes to government bills as
the policy distance between the proposing minister or the government compromise and the
parliamentary median voter increases.
When Cabinets do represent parliamentary majority preferences
If the government bills were examined, discussed and really voted in the Cabinet and the Cabinet
preferences were representative of the Parliamentary majority preferences, then the only conflict
that at least in theory could generate changes of the initial proposals should depend on the
opposition parties. Since time is one of the scarcest resources in the legislative arena (Döring
2
Distributive policies may include public works projects (Maass 1951; Ferejohn 1974), appropriations (Fenno 1966),
urban renewal (Plott 1968), land and resource management (Fenno 1973), tariffs (Schattschneider 1935), military
procurement (Rundquist 1973), categorical grants-in-aid (Mayhew 1974). For a discussion of the different types of
policies based on the incidence of costs and benefits, see Wilson (1973). See also Lowi (1964).
1995b) opposition parties could trade the possibility to delay the final approval for the introduction
of some modifications that bring some benefits or alleviate costs for their own electoral
constituencies. Their propensity to use dilatory tactics should increase the more damaging are the
government bills for them and the interest they represent. Therefore
H.4 Conflict with opposition: We expect more changes of government bills as the policy
distance between the proposing minister or the government compromise and the opposition
parties increases.
When the cabinet is assumed to replicate the preferences and the equilibria of the parliamentary
majority, the only opportunity of conflict that does not involve the opposition parties takes place
when there is ministerial discretion. This conflict can concern either the jurisdictions of different
ministries, or the ideological divergence among the parties to which ministers belong
Focusing on the interactions among ministers, the conflict can arise over which minister has
the responsibility in a certain policy area. If a minister drafts a bill that deals with issues falling
under another minister‟s jurisdiction, the party of the minister perceiving interference could react
and the bill could be challenged on the floor. As a consequence:
H5. Conflict over ministerial responsibility: since there are issues on which different
ministries may draft legislation, on those issues we expect to observe more conflict over
jurisdictions and so more changes to bills.
A recent line of research identifies ideological conflict within the cabinet and among ministers as
the fundamental factor explaining the occurrence of changes to governmental bills (Martin and
Vanberg 2004, 2005). This growing literature focuses on delegation relationships within multiparty
coalitions, and conceptualizes them within a principal-agent framework. Due to the delegation of
important policymaking powers from the parliament to the cabinet and from the cabinet as a whole
to individual ministers (Saalfeld 2000: 354), supporters of a coalition have to agree to delegate not
only to their own party‟s cabinet representatives, but also to ministers from their party‟s coalition
partners. In other terms, in multiparty coalitions delegation necessarily crosses party lines, and
parties do not enjoy the control over ministers from other parties3. As a consequence, the
implementation of the coalitional compromise that makes possible the existence of the government
3
As Thies (2001: 582) points out, multiparty executives face greater delegation problems than single-party executives
for two main reasons. First, ministers in a coalition government are likely to have more divergent policy preferences
than are their colleagues in a single-party government. Indeed, ministers in a single-party cabinet are usually supposed
to have the same policy preferences (for accounts challenging this assumption, see, for example, Müller 2000: 320).
Second, while in single-party governments the party caucus can administer punishments such as reductions in campaign
support, denial of the official party endorsement, or a lower position in the party list, in multiparty governments
„troublesome‟ ministers cannot be sanctioned that way by their coalition partners.
is subject to the agency problems also known as „ministerial drift‟: ministers, acting as agents of the
cabinet as a whole, may not work fully in the interests of their principal. If each minister enjoys a
near monopoly on policy initiation within the jurisdiction of her portfolio (Gallagher, Laver, and
Mair 2006: 43-44), she may attempt to use her discretion in drafting legislation to move policy in
directions favoured by her own party, but that can be detrimental to the interests of other coalition
members. Electoral calculations give ministers incentives to move policy in ways that appeal to
their respective constituencies, or „to appear to work toward providing such policies even when
party leaders realize that they may not be able to deliver them in practice‟ (Martin and Vanberg
2005: 94; Huber 1996). Other coalition parties, in turn, face strong incentives to manage delegation
to „hostile‟ cabinet ministers, thus enabling the coalition to implement the policy compromise to
which it has agreed. These incentives will be stronger the more distant is the party of the minister
who proposes the bill from the other government parties. In order to come back to the original
compromise, coalition partners make use of several means (at both the executive and the legislative
level) to monitor the actions of ministers from other parties. Some scholars (Thies 2001; see essays
in Strøm, Müller, and Bergman 2008) insist on intracabinet monitoring mechanisms. Martin and
Vanberg (2004, 2005) focus on the contrary on the legislative arena, and argue that coalition parties
rely on the members of their legislative faction to investigate the anticipated consequences of a
proposed bill, to study the justifications offered by the drafting minister, to take into consideration
available alternative policies, and to prepare amendments to ministerial drafts. Therefore, the
changes of the government bills would be the evidence that a monitoring activity is at work. Martin
and Vanberg‟s hypothesis, which they successfully test using data from Germany and the
Netherlands (1982-1994), is thus the following:
H.6 Ideological conflict within the cabinet: As ideological divergence between
coalition partners increases, the incentives for ministerial drift from coalition
compromise increase, and thus the coalition partners will rely more heavily on
legislative review.
If legislative review activity is a monitoring mechanism, then other monitoring instruments can be
employed as substitutes. At the executive level, as suggested by Thies (2001) and later confirmed
by Martin and Vanberg (2005: 102-103) and Verzichelli (2008), coalition partners „keep tabs‟ on
each other by appointing junior ministers (JMs) to hostile ministries. Junior ministers have access to
much of the same information as the minister they are shadowing, and have the incentives to report
such information to their own party. As a consequence, when a policy proposal drafted by a
minister is submitted for approval (for example in a meeting of the full cabinet), other coalition
partners do not suffer any more an informational disadvantage about the policy implications of her
proposal: they are able to discover whether or not each others‟ ministers are deviating from the
coalition‟s agreed-upon compromise and to correct ministerial proposals before they become bills
so that coalition policy compromise can actually be implemented. Therefore
H.7 Junior ministers (presence): We should expect the presence of a JM from a
partner party in the ministry drafting a bill to reduce the level of changes made to the
bill in the assembly.
As a refinement of the previous claim, we can take into account the ideological divergence between
the JM and the minister he shadows instead of considering just the different party membership:
H.8 Junior ministers (preferences): We expect less changes to government bills the
more the ideological preferences between the proposing minister and her JMs diverge.
Following the logic laid out by Thies (2001), also the number of government parties that that
formally co-sponsor a government bill can indicate the level of intracabinet control. More parties
should grant that the initiative mirrors the coalition compromise. The same can be argued when the
sponsor is the prime minister. Unlike line ministers, the prime minister does not have the
jurisdiction over a specific policy area. As a consequence, when a bill dealing with a particular
policy area is presented by the prime minister, we can reasonably suppose that it is the entire
cabinet which wants the approval of that bill. Therefore,
H.9 Number of signatory parties: We expect less changes to government bills the
larger the number of signatory parties.
H.10 Prime minister’s bills: we expect that bills proposed by the prime minister should
be less changed in Parliament.
Status quo and veto players approach
So far, we have reviewed explanations that are based upon the conflict and ignore the position of
the status quo before the law making process starts in the Cabinet. In particular, if we assume that
political actors in the Cabinet cannot prevent a decision making process whose last outcome is for
them worse than the status quo (they lack gatekeeping power), then the position of the status quo
does not matter. However, this assumption is not plausible and the comparative politics literature
(Tsebelis 2002, Cox 2005) agree in recognizing to each government party the possibility to veto
the policy change. If we introduce the veto players approach, the status quo position becomes a
crucial variable. Imagine first that at the Cabinet level the equilibrium decision corresponds to the
policy closest to the cabinet median voter that is also unanimously preferred by the government
parties to the status quo and secondly that the final outcome in Parliament is the closest policy to
the Parliamentary median voter that is also unanimously preferred by the government parties to the
Status quo. The first assumption requires that also the Cabinet deliberation, induced by position
taking utility, is preferred by the unanimity of the Government parties to the status quo. The
second assumption requires that coalition parties prefer the final policy outcome to the status quo.
Otherwise, they would not introduce any bill. Always the farther the status quo is from the
government, the more government bills are changed. Let us show why it happens by using a
simple graphical illustration in a one-dimensional policy space (Fig.1).
Imagine a government formed by three parties: L, G and R. G‟s ideal point represents also
the median position in the government. M is the median voter in the Parliament. The status quo is
located on left of L and can be SQ or SQ1. LSQ is a policy on the right of L and marginally better
for L than SQ. The same can be said for LSQ1. We provide two scenarios: in the first one M
overlaps L position; in the other M overlaps R position. In the first scenario, when the status quo is
SQ the policy closest to the Cabinet median voter that is also unanimously preferred by the
government parties to the status quo is LSQ. Thus the bill proposed by the government is LSQ.
The changes such a bill is bound to have are represented by the distance LSQ-M. If the Status quo
goes farther until SQ1 position, the median position in the cabinet G becomes a reasonable
compromise for L too. As a consequence, in the following parliamentary stage the bill will change
more than before. The line G-M represents the size of this change and is longer than LSQ-M.
Turning now to the second scenario, imagine that the parliamentary median voter overlaps the
rightist party of the government, R. When the status quo is SQ we don‟t expect any change as LSQ
is at the same time the closest policy to the Cabinet median voter that is also unanimously
preferred by the government parties to the status quo and the closest policy to Parliamentary
median voter that is also unanimously preferred by the government parties to the Status quo. If the
status quo becomes SQ1, G that is the median party inside the Cabinet becomes a policy outcome
that is unanimously preferred by the government parties. In this case, the closest policy to
Parliamentary median voter that is also unanimously preferred by the government parties to the
Status quo is LSQ1 and the distance between G and LSQ1 represents the size of the changes that
concern the government bills.
Fig.1 Status quo, Veto players and changes to government bills.
SQ1
SQ1
SQ
L
M
SQ
L
LSQ
LSQ
G
G
LSQ1
LSQ1
R
R
M

The status quo position seems very difficult to identify empirically. It depends on many variables.
However, there is at least a systematic source of relatively large status quo shifts in many policy
areas: the government alternation. Government alternation affects the status quo in two ways. First,
the decisions taken by the previous government are likely to be in many policy areas a status quo
far from the present government if the previous opposition wins office. Second, the perspective of
the alternation can move the “reversion point” in case of no decision farther than the present status
quo and in a controversial area. In other words, the potential legislative outcome if the present
opposition wins the next elections becomes the real status quo that the government parties take in
consideration. Such a crucial change in the political calculus of the government parties takes place
when the government alternation has already become a real and experienced possibility. Therefore,
H.11 Alternation: we expect that government bills proposed in an alternational party
system are more changed than government bills proposed in a pivotal party system.
Explanations based upon the lack of knowledge
When and why can we consider the changes of the government bills as corrections or
specifications? Or, in other terms, when are changes driven by an improvement of the decision
makers‟ knowledge? Although cabinet ministers are not certainly less staffed than ordinary
representatives in the Parliament, the variety of origins and competences of the legislators and the
specialization of the parliamentary committees can be a powerful resource. This is especially true if
the government agenda is crowded by provisions to examine and deliberate. Reasonably, any
explanation based upon a lack of knowledge at the cabinet level depends on the convenience for the
Cabinet ministers to delegate the learning activity to the Parliament. If a bill deals with a complex
subject, the proposing minister (not to tell about the others) is unlikely to take all the possible
aspects into account in the draft version. By contrast, during the debate on the floor legislators are
engaged in a process of learning in which they discover aspects that previously were unknown (and,
necessarily, not considered in original version of the bill). Therefore, they correct the bill in order to
include this new information. However, if the proposing minister was appointed more for his/her
technical competence than for his /her party affiliation, then we should expect less changes of the
government bill. So, we can put forwardthe following two hypotheses:
H.12 Complexity: It is reasonable to expect more changes the more complex the
subject of a governmental bill is.
H.13 Techinical ministers: It is reasonable to expect less changes when the proposing
minister is a “technical” minister.
An explanation based upon the lack of knowledge does not exclude the importance of the conflict
and of the position taking activity. On the contrary, it is possible to imagine that in some
circumstances the two types of explanation reinforce each other. This is pretty clear if we consider
the Status quo and veto players approach. If a policy change is so unanimously preferred by the
Cabinet parties that there are many alternatives all government parties prefer to the status quo, then
the study in depth of the implications of the different alternatives and the final decision can be
delegated to the Parliamentary majority. Instead of investing time and energies in examining the
same issue for a long time, the Cabinet by deliberating a generic and vague bill can convey the
symbolic message of internal unity and capacity to innovate and dedicate more energies to more
controversial policies.
There is a very peculiar lack of knowledge that is connected with the features of the decision
making environment and that can never be completely removed. If a bill addresses a
multidimensional policy space and the decision rule is the majority rule, then the final outcome is
not easily predictable by anyone, the bargaining process can last longer and induce many changes.
Therefore,
H.14 Multidimensionality: We expect more changes to government bills if the bill can be
evaluated according to many policy dimensions
The empirical analysis
Data
We test the hypotheses outlined in the literature and discussed in the previous paragraphs against
evidence from Italy. More precisely, we use a dataset tracking the legislative history of all the
governmental laws introduced in the Chamber of Deputies during the 1987-2006 period. One of the
main advantages in examining Italy is that the sample of bills would exhibit substantial variation in
terms of ideological divergence within the cabinet, which is among the key factors we wish to
isolate. Indeed, this would allow to test Martin and Vanberg‟s (2005) hypothesis, which is currently
one of the main (and most popular) theoretical explanations of the changes made to governmental
bills during the legislative process. Given the relatively large number of parties traditionally
involved in coalition governments and the cabinet instability typical of the Italian political system
since the end of World War II (Verzichelli 2003: 465), a considerable amount of variation in
ideological divergence across government-sponsored bills is likely to be registered in Italy. In fact,
such variation derives from two sources: ideological divergence can vary (1) across the policy areas
addressed by different bills (for any given government) and (2) with changes in the party
composition of the government or in the party controlling a particular ministry.
In order to empirically analyze the modifications made to government-sponsored legislation
during the legislative process, we constructed an original dataset in which the units of analysis are
approved governmental laws. Remarkably, since the information concerning government bills
introduced (as first reading) in the Senate is not available for the entire period we investigate4, we
restricted the analysis to those governmental laws that have been introduced in the Chamber of
Deputies5. Moreover, we excluded constitutional laws, budgetary laws and laws ratifying
international treaties and agreements, since the parliamentary dynamics leading to their approval are
radically different from the rest of government legislation. The final dataset thus consists of 716
governmental laws.
Dependent variable operationalization
Operationalizing the governmental bills changes, i.e. our dependent variable, requires some kind of
comparison between the final version of the bill adopted by the legislature and the version drafted
4
In particular, the texts of the bills introduced in the Senate are available in electronic format only from 1996 onward.
In any case, analyzing only one of the two branches of the Italian legislature (especially the Chamber of Deputies) is a
common approach in the literature on Italian parliament and political system. Indeed, the Italian parliament is
conventionally considered as a symmetrical legislature since both chambers perform identical functions (for our
purposes, they have the same role in the legislative process). Moreover, although elected on a somewhat different basis,
the political composition of the two houses has almost always been identical. Hine (1993: 189-190), for example,
defines the Italian parliament as a case of „perfectly co-equal bicameralism‟.
5
by the proposing minister. In their article, Martin and Vanberg (2005: 99) measure the level of
legislative review as the number of article changes, i.e. the sum of all the articles altered, deleted or
added to the draft version of the bill. According to the authors, who analyze German and Dutch
legislation, this measure captures somehow the policy significance of modifications since bills are
subdivided into several articles, each dealing with a specific issue. However, Martin and Vanberg‟s
operationalization does not seem to be particularly suitable for the Italian case, since it does not take
into account some typical features of Italian parliamentary process and legislation. First of all, the
use of the so called „maxi amendments‟, through which a number of articles in the drafted bill are
turned into a single article in the final law6. In an extreme case, a bill made up of a certain number
of articles can be approved as a single-article law. In such a situation, while the bill is approved
without any actual modification in the text, Martin and Vanberg‟s measure would indicate a
relatively high degree of change.
In addition, it is not straightforard that articles are the logical policy subunits of Italian bills.
Although bills are normally voted article by article, often more than one aspect is addressed by a
single article. Indeed, each single article is subdivided in smaller parts called commi, which deal
with more specific aspects of the bill. Thus, we can have an article change either because the article
is entirely modified, or because just one comma (or a part of it) is altered. If we employ Martin and
Vanberg‟s measure, these two situations lead to the same outcome, i.e. one article change. By the
same logic, however, a measure counting the number of modified commi would present the same
risks: although more accurate than Martin and Vanberg‟s solution, it would treat partial changes of
subsections as entire changes of subsections.
Given the problems posed by any operationalization relying on sections or subsections of
articles, we decided to look at the most basic units in the text of bills, and thus to measure our
dependent variable as the number of words that are changed (i.e. added, deleted or modified) during
the legislative process.
A simple option is just to compare the number of words in the draft version and in the final
version of a law. Yet, this measure would consider just words that are added or deleted, and not
words that are substituted. In order to solve these problem, we (1) compute for each single word
appearing in either version the absolute difference between the number of times it occurs in the
draft bill and the number of times it occurs in the final law, and (2) sum all these absolute
differences. This way we take account of substituted words, but we overestimate their number,
6
Since the 1990s, Italian governments made increasingly use of maxi amendments. As several works highlighted, maxi
amendments have been employed, together with delegating laws and decree laws, as circumventing tactics – that is,
techniques to circumvent the ordinary legislative process (see Zucchini 2005).
which results to be twice as the actual value7. For this reason, we (3) compute the absolute
difference between number of words in the original bill and the number of words in the final law in
order to get the number of added or removed words, (4) subtract this quantity from the sum of the
absolute differences calculated in the second step and divide the result by two, thus finding the
actual number of substituted words. Finally, we (5) add the number of substituted words to the
number of added or removed words8.
For laws converting decree laws, we calculated the number of changed words separately for
the converting law and the decree, and then we added these two values together.
Undoubtedly our measure prensents drawbacks. To begin with, there are words such as “not”
that, if added or removed, may completely change the meaning of sentences. Moreover, our
measure is subject to problems of language ambiguity, which are however common to all computerbased techniques of content analysis9. More generally, just as Martin and Vanberg, we are not
dealing with the substantive content of legislation: counting the number of changed words, we look
simply at what is modified in the text of bills during the legislative process, and not at the actual
policy significance of these transformations. However, given the increasingly technical and
specialized nature of modern legislation, understanding the actual policy impact of each change
made to a certain bill seems to require a prohibitive effort from researchers. This is especially true
when the analysis concerns a relatively large number of bills dealing with several policy areas, as in
the present case10. In addition, our measure can be easily coded by other researchers. On the whole,
we believe that the number of changed words is a fairly reliable measure, as well as more refined
and appropriate for the Italian case than Martin and Vanberg‟s operationalization.
As the table 2 shows, the dependent variable is an integer that is bounded below by zero and
unbounded above11. Its distribution is strongly asymmetrical and skewed to the right (the mean is
more than the median), 14% of the bills are passed into law without modifications and more than
170 words are changed in half of the bills in the sample. On average, governmental bills introduced
in the Chamber present 683 changed words, while their average length is 1671 words.
7
In other terms, words that are substituted are counted twice: once as words appearing in the original text but not in the
final one, and once as words appearing in the final text but not in the original one.
8
Changes in the spelling, capitalization and punctuation have not been counted as modifications.
9
For a discussion, see for example Laver and Garry (2000).
10
As a long line of scholars has highlighted, due to the expanding scope and complexity of governmental tasks and the
growing size of state bureaucracy over the past century, important policymaking powers have been increasingly
delegated by the parliament to the cabinet and by the cabinet as a whole to individual ministers. As a result, only the
minister responsible for a given policy area has the knowledge necessary to understand the policy impact of a certain
bill (Bagehot [1872] 2001; Bryce 1921; Wheare 1963; Cox 1987; Laver and Shepsle 1994, 1996).
11
Having no upper limit, the variable enables to distinguish between a situation in which the whole text of a bill is
substituted and one in which the whole text is substituted and extra words are added, that we consider as a more
extensive change.
Tab.2 Changed words
Percentile
0
15
30
Median
70
85
100
Mean
St. Dev.
Variance
Skewness
Kurtosis
Number of changed words
0
1
51
169.5
513
1251.25
21617
683.1369
1586.649
2517454
7.467272
86.33796
N.
716
More interestingly, if we look at the distribution of the dependent variable across different periods,
we can note that modifications tend to increase over time. This is observable in table 2, which
displays, for each of legislature that we consider, the median number of changed words as well the
median degree of word change
12
. While the former is the dependent variable calculated as we
explained above, the latter is the same variable divided by the number of words in the initial bill.
This second measure allows us to discuss the descriptive statistics controlling for the different
length of bills. In particular, in the last two legislatures included in the sample (XIII and XIV) both
the median number of changed words and the median degree of word change become twice as much
as their value in the first two legislatures (X and XI). In other words, the level of legislative review
does not seem to increase just because bills become longer (see also fig. 2).
Tab.3 Changed words and degree of Word Change. Mean and median by legislature
Tot.
Degree
Degree
Changed words of word change
Changed
of word change
(mean)
(mean)
words (median)
(median)
716
683.1369
0.5164701
169.5
0.197058
X
XI
XII
XIII
XIV
233
80
80
185
138
459.3391
502.1375
654.6125
770.9568
1064.732
0.4761085
0.4688768
0.3410302
0.5722605
0.6391202
110
119
139.5
256
266.5
0.151515
0.158773
0.109044
0.280702
0.314165
I REP
313
470.278
0.4742601
117
0.154243
n.
12
Since the distribution of the dependent variable presents a heavy tail on the right, we use the median instead of the
mean. Being less sensitive to outliers, in such a situation the median is a better measure of central tendency than the
mean.
II REP
403
848.4591
0.5492536
225
0.243624
0
200
400
600
800
1,000
Fig.2 Changed words and words in draft bills. Median by Legislature
10
11
12
N. of changed words (median)
13
14
N. of words in draft bill (median)
More or less, during the Italian Second Republic governmental legislation appears to be more
subject to modifications than during the First Republic. As fig. 2 shows, this is true both for bills
converting decree laws and for ordinary ones. The picture is the same even if we look at the bills
introduced by each government included in our sample: as displayed in fig. 3, legislation sponsored
by the centre-right and centre-left governments of the Second Republic is more extensively altered
than legislation proposed by previous governments13.
13
It‟s worth noting that all the nine bills introduced by the Berlusconi‟s III government are bills converting decree laws.
Fig.3 Degree of word change. Median by type of law and Political system
Ordinary Bills
First Republic
Bills converting DL
Ordinary Bills
Second Republic
Bills converting DL
0
.1
.2
.3
p 50 of r_adj
.4
Fig.4 Degree of word change. Median by governments
10
Goria
De Mita
Andreotti VI
Andreotti VII
11
Amato I
Ciampi
12
Berlusconi I
Dini
13
Prodi I
D'Alema I
D'Alema II
Amato II
14
Berlusconi II
Berlusconi III
0
.1
.2
p 50 of r_adj
.3
.4
Independent variable operationalization
Let us now discuss how we operationalized the explanatory hypotheses presented in the previous
sections, as well as a number of control variables that we include in our empirical analysis of
changes of government bills in Italy. As said before, the modifications occurring in parliament may
stem from two general explanatory factors: some kind of conflict between the actors involved in the
legislative process, and the lack of information about some aspects of the subject or the policy
dimensions addressed by the bills.
Concerning the former general factor, a number of hypotheses is based on the ideological
conflict between political actors, that we measured as policy distances between them. For the
position of parties on different policy dimensions we used the expert surveys conducted by Laver
and Hunt (1992) and Benoit and Laver (2006). For each party, these data provide information on
both position and saliency scores on a 20-point scale in several policy dimensions. In general, in
order to measure the variables dealing with the ideological conflict, we assigned each governmental
law to one of the policy domains identified by the expert surveys14, and then we calculated the
absolute difference between the positions of the relevant actors in the issue under consideration.
Thus, conflict with parliamentary majority, i.e. the policy divergence between the proposing
minister (or the government compromise) and the parliamentary median voter in H3 has been
calculated as the absolute difference between the score of the proposing minister‟s party (or the
mid-range position of the government) and the score of the party controlling the median legislator
in the Chamber.
By the same logic, conflict with opposition in H4 has been operationalized as the absolute
distance between the minister‟s position (or the government mid-range) and that of the farthest
opposition party. In particular, among opposition parties we did not consider parliamentary groups
with less than 20 legislators, since these parties are usually constrained in their ability to influence
legislation (for instance, they do not take part in the meetings of Conferenza dei capigruppo).
As for ideological conflict within the cabinet (H6), which is Martin and Vanberg‟s (2004,
2005) main explanatory variable, we follow Franchino and Høyland (2009: 612) and measure it as
the distance between the proposing minister and the farthest coalition partner. Similarly, Junior
minister’s distance in H8 has been calculated as the distance between the minister submitting the
bill and the JM who is farthest from her15.
14
In particular, we used Laver and Hunt‟s (1992) scorse for the 1987-1994 period and Benoit and Laver‟s (2006) scores
for the 1994-2006 period.
15
Since bills can be proposed by more than one minister, we consider the first signatory as the proposing minister.
Moreover, as previously illustrated, our sample includes bills converting decree laws. Formally, the first signatory of
converting laws is always the prime minister, regardless of which minister actually drafts them. Therefore, for
Minority government (H1) is a dummy variable taking the value of 1 if the government
proposing the bill does not control a majority in the Chamber and 0 otherwise. In our sample, only
Dini and Prodi I cabinets are minority governments.
Concerning distributive conflict in H2, the variable is a dummy whose value is 1 when the bill
contains at least one porl barrel clause, and 0 otherwise. In order to build this measure, we checked
whether the bill included benefits for particular professional categories, interest groups,
organizations, places or persons.
According to the hypothesis concerning conflict over ministerial responsibility (H5), we
expect more changes to bills addressing policy issues on which more than one ministry may draft
legislation. We thus constructed a measure of the degree to which a domain is shared by different
ministries in the following way. First, we classified all the governmental laws approved in the
period under investigation according to the micro-issues identified by the Comparative Agenda
Project16. Then, we computed for each of these micro-categories the Rae‟s (1967) index of
fragmentation: for every micro-issue we calculated, as a relative frequency, the number of times a
ministry proposed a law, we squared these frequencies and sum them 17. The resulting variable can
be considered as a measure of conflict over ministerial responsibility: the more the value of the Rae
index becomes close to 0 (its minimum value), the more that micro-issue is disputed by different
ministries; the more it approaches 1 (its maximum value), the more that micro-issue can be
considered as the exclusive jurisdiction of a single ministry.
Junior minister’s presence (H7) and Prime minister’s bill (H10) are dummy variables. The
former is equal to 1 when a JM from a parner party is present in the ministry drafting the bill, while
the latter is 1 when the Prime minister is the first signatory of the bill (for decree laws, see note 15).
We calculated the number of signatory parties (H9) as the number of different parties with at
least one minister among the signatories of a bill.
With regard to the alternation hypothesis (H11), we operationalized it with the dummy Second
Republic, is 1 for bills presented and approved during the XII, XIII and XIV legislatures (19942006), and 0 for bills presented and approved during the X and XI legislatures (1987-1994). In other
words, we considered the Italian First Republic as a pitoval party system and the Second Republic
as an alternational party system.
converting laws we consider the second signatory as the proposing minister. The premier is considered the proposer of a
converting law only if he is the only signatory of it.
16
In building this variable, we used all the 1574 governmental laws approved in the 1987-2006 period, and not only
those introduced in the Chamber of deputies. However, constitutional laws and laws ratifying international treaties are
excluded.
17
To take account of the changes occurred in the number and names of Italian ministries, we calculated this variable for
three different periods: 1987-1996; 1996-2001; 2001-2006.
As for the other major explanatory factor, the lack of knowledge, the dummy complexity
(H12) becomes 1 if the bill includes more than one aspect regarding the subject at issue18, while the
dummy technical minister (H13) takes the value 1 when the first signatory is a minister who does
not belong to any party. In order to identify technical ministers we drew on Verzichelli and Cotta‟s
(2000) data on the distribution of cabinet ministerships in Italy up to 2000. For succeeding years,
we identified technical ministers trying to follow the authors‟ logic. In order to operationalize the
multidimensional nature of a bill (H14), we counted the number of committees to which the bill is
referred. Reasonably, when a bill deals with multiple policy areas, more than one committee are in
charge of discussing it.
Finally, we incorporated as control variables in our analyses the length of the bill and the
policy area which the bill deals with. A for the former, which is measured as the number of words
in the drafted version, longer bills are obviously more likely to be modified than shorter ones.
Concerning the latter, it might be the case that bills addressing a certain policy area always face
more changes than others. This could occur, for instance, because the issues within that policy
domain are inherently more complex, or because in that policy area outside lobby groups and
advisory bodies provide legislators with the information needed to propose changes to government
bills. We control for the policy area ddressed by the bill in two different ways. On the one hand, we
created a number of separate dummy indicators, one for each policy domain. On the other hand, we
assumed that the data are grouped into clusters (i.e. the different policy areas), and that data in one
cluster are more similar to one another than to data belonging to another cluster. Thus, in the
analyses, we estimated standard errors that are robust to such clusters.
Analysis and findings
In order to test the impact of the explanatory variables discussed above, we use an event count
model. This technique has become the standard approach in political science studies interested in
the number of events that occur over a particular period of time (King 1989; Scott Long 1997): in
the present case, the number of words modifications made to a governmental bill in the course of
the legislative process. As we previously illustrated, our dependent variable counts the number of
words changed in governmental bills and takes only positive and integer values (see Fig. 5). As
Cameron and Trivedi (1998) argue, using a linear regression model for count outcomes would result
in inefficient, inconsistent and biased estimates. The most basic way for modeling count variables is
the Poisson regression model, in which the number of events follows a Poisson distribution with a
mean that is a function of the independent variables according to a model that forces expected
18
For instance, the variable is equal to 1 when a bill dealing with basic education includes provisions about both school
leaving age and the construction of new schoolhouses.
counts to be positive. However, the Poisson regression model constrains the conditional variance to
equal the conditional mean (equidispersion), and therefore underestimates the amount of dispersion
in the outcome most of the time. It happens because the conditional variance is very often greater
than the conditional mean (overdispersion). This results in standard errors that are biased downward
and produce spuriously large z-values and spuriously small p-values (Cameron and Trivedi 1998;
King 1989). For this reason we employ the Negative binomial regression model, whose distribution
allows the conditional variance to exceed the conditional mean. In particular, the degree of
overdispersion is extimated as a parameter (α) from the data.
We present the results of the negative binomials regressions in Table 4 and 5. The three
models correspond to three different combinations of the explanatory variables based on ideological
conflict. We did not incorporate all of them in a single model for both theoretical (they reflect
different logics and operationalizations) and technical reasons (some of them are correlated). In
model 1 we included three ideological variables measuring distances between the proposing
ministers and the other relevant actors (conflict within the government, JM‟s distance and conflict
with the opposition), while the other two models allow to test the hypothesis of conflict with
parliamentary majority. In particular, the initial bill is assumed to reflect the minister‟s ideal point
in model 2 and the coalition compromise in model 319. In order to control for the policy area
addressed by the bill, we run the three models also with standard errors clustered for the policy
domains identified by the Comparative Agenda Project20. Moreover, we controlled for the impact of
both the prohibition to reiterate decree laws in 1996 and the “anomaly” of the Dini‟s government by
running separate analyses which exclude the XII legislature (April 15, 1994 – May 8, 1996).
Indeed, it was in the XII legislature that the practice of reiteration reached its maximum. Moreover,
all the ministers in the Dini‟s government (January 17, 1995 – May 16, 1996) were technical, which
could affect our hypothesis that technical ministers serve as means of overcoming problems of
lacking information. In the tables, cell entries are the estimated negative binomial regression
coefficients (with standard errors in parentheses). Togeher with the number of observations and the
information criteria, at the bottom of the table we reported the parameter α, which estimates the
dispersion of the outcome variable. When this parameter is zero, there is not equidispersion in the
19
Two of the variables presented in the previous section are not included in the present analyses. We excluded JM’s
presence because in Italy, where both the number of coalition parties and ministries has always been relatively large, it
is almost always the case that a JM from a parner party is present in the ministry drafting the bill (see Verzichelli and
Cotta‟s 2000 tables on the distribution of junior ministerships in Italian coalitions). The number of signatory parties is
instead excluded for more technical reasons: when the proposing minister is technical, this number is most of the time
equal to 0. The two variables are strongly associated (running a Chi-square test, the Chi-square with three degrees of
freedom is 611.2409, and p is 0.000).
20
We estimated cluster-robust standard errors using also the policy dimensions classified by Laver and Hunt‟s (1992)
and Benoit and Laver‟s (2006) expert surveys, and the results are almost the same. In addition, we run the same models
but including dummies for the policy domains (both from expert surveys and Comparative Agenda Project) as
independet variables. This does not changes our results.
data and the model reduces to the simpler Poisson regression model; if it is greater than zero (as in
the present case), then there is overdispersion and the data are better estimated using a Negative
binomial model than a Poisson model.
Looking now at the results, we can note first of all that these do not change if we control for
the policy dimensions and for the “exceptionality” of the XII legislature. In particular, all the
independent variables dealing with the lack of information always affect the level of modifications
made to governmental bills during the legislative process. Government bills dealing with complex
subjects or with many policy dimensions are more extensively altered, while bills proposed by a
technical minister are less likely to be changed given the minister‟s expertise in the policy area at
issue.
With regard to the explanatory hypotheses based upon some kind of conflict between the
relevant actors, let us note that government bills proposed and approved during the Italian Second
Republic are altered more extensively than those passed into law in the previous period. This is
consistent with our hypothesis that changes to governmental bills should be greater in an
alternational party system than in a pivotal party system. Instead, we found no evidence that the
government bills that have been submitted by minority governments, address disputed issues or
have a distributive nature receive more modifications in parliament than other types of bills.
Remarkably, against our expectations, bills proposed by the Prime minister are more likely to
be changed in parliament than those drafted by line ministers. Morover, among the variables
measuring ideological conflict, only the distance between the proposing minister and the farthest
JM in her ministry is significant. Quite interestingly, the sign of the coefficient is positive, thus not
supporting the hypothesis that JMs with preferences far from the those of “their” senior minister
should reduce the level of modifications in the parliament. These two last results seem to suggest
that the logics operating in the course of the legislative process may be different than those implicit
in the two hypotheses we put forward. In the literature on coalition politics, Prime ministers and
junior ministers with preferences different than their senior minister are both considered as
monitoring mechanisms that enable coalition parties to reach agreements at the cabinet level and
thus prevent conflict in the legislative arena (Andeweg 2000; Andeweg and Timmermans 2008;
Thies 2001; Müller and Strøm 2000). However, our empirical findings on Italy seem to be at odds
with the claim that a higher degree of agreement among actors at the cabinet level reduce the
amount of modifications to governmental bills in the parliament. Our argument is that, once
coalition members reach some kind of agreement at the cabinet level (i.e. when the bill is proposed
by the head of government, or when a JM from an “extreme” coalition party is in the ministry
drafting the bill), then the parliament may be delegated to deal more in depth with the bill. In other
terms, in presence of an initial agreement in the government the definition of the details of the bill is
up to the legislature. This results in the modifications made to government bills that we observe
during the legislative process. Indeed, as fig. 6 shows, words added to the initial bill are the most
part of the number of changed words.
Tab.4 Negative binomial regressions on governmental bills changes. Italy 1987-2006
Depndent variable:
Number of changed words
Model 1
Model 2
Coeff.
Coeff.
Independent variables
Minority government
Distributive conflict
Conflict within the cabinet
JM's distance
Conflict with opposition
Coeff.
(robust s.e.)
(robust s.e.
clustered on
policy
dimension)
Model 3
Coeff.
(robust s.e.)
(robust s.e.
clustered on
policy
dimension)
Coeff.
Coeff.
(robust s.e.)
(robust s.e.
clustered on
policy
dimension)
0.126
0.126
0.097
0.097
0.063
0.063
(0.175)
(0.168)
(0.173)
(0.175)
(0.159)
(0.163)
-0.002
-0.002
-0.034
-0.034
-0.035
-0.035
(0.126)
(0.116)
(0.127)
(0.131)
(0.127)
(0.124)
-0.040
-0.040
(0.034)
(0.042)
0.058**
0.058*
(0.023)
(0.032)
-0.008
-0.008
-0.005
-0.005
0.001
0.001
(0.019)
(0.014)
(0.024)
(0.025)
(0.019)
(0.016)
0.009
0.009
(0.025)
(0.035)
-0.006
-0.006
(0.064)
(0.096)
Conflict with parliamentary
majority (minister)
Conflict with parliamentary
majority (coalition compromise)
Conflict over ministerial
responsibility
-0.017
(0.230)
Prime minister's bill
0.415**
(0.164)
(0.170)
0.301**
0.301**
(0.142)
(0.132)
(0.148)
(0.142)
(0.138)
(0.137)
0.391***
0.391**
0.404***
0.404**
0.406***
0.406**
(0.150)
(0.168)
(0.152)
(0.165)
(0.153)
(0.166)
-0.443***
-0.443***
-0.454***
-0.454***
-0.451***
-0.451***
(0.152)
(0.149)
(0.155)
(0.156)
(0.154)
(0.149)
Number of referral committees
0.160***
0.160***
0.154***
0.154***
0.155***
0.155***
(0.030)
(0.031)
(0.030)
(0.025)
(0.030)
(0.029)
Length of bill
0.000***
0.000**
0.000***
0.000**
0.000***
0.000**
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
4.830***
4.830***
4.767***
4.767***
4.747***
4.747***
(0.329)
(0.249)
(0.315)
(0.240)
(0.341)
(0.328)
Second Republic
Complexity
Technical proposing minister
Constant
Alpha
Aic
Bic
N
-0.017
0.034
0.034
0.039
0.039
(0.205)
(0.234)
(0.210)
(0.231)
(0.210)
0.415**
0.375**
0.375**
0.380**
0.380**
(0.163)
(0.162)
(0.163)
(0.161)
0.355**
0.355**
0.368***
0.368***
2.664
9,558.542
9,622.573
716
note: *** p<0.01, ** p<0.05, * p<0.1
Clusters on policy dimensions (Comparative Agenda Project)
2.679
9,561.550
9,621.008
716
2.679
9,561.626
9,621.084
716
Tab.5 Negative binomial regressions on governmental bills changes. Italy 1987-2006 (XII
legislature excluded)
Depndent variable:
Number of changed words
Independent variables
Minority government
Distributive conflict
Model 1
Model 2
Model 3
Coeff.
Coeff.
Coeff.
(robust s.e.
clustered on
policy dimension)
(robust s.e.
clustered on
policy dimension)
(robust s.e.
clustered on
policy dimension)
0.113
0.077
0.034
(0.186)
(0.184)
(0.182)
0.002
-0.039
-0.038
(0.132)
(0.144)
(0.138)
Conflict within the cabinet
-0.033
JM's distance
0.057*
(0.043)
(0.029)
Conflict with opposition
-0.002
0.000
0.007
(0.018)
(0.026)
(0.018)
Conflict with parliamentary majority
(minister)
0.012
(0.032)
Conflict with parliamentary majority
(coalition compromise)
-0.002
(0.102)
Conflict over ministerial responsibility
-0.122
-0.054
(0.149)
(0.167)
(0.172)
0.472***
0.440***
0.447***
(0.167)
(0.154)
(0.155)
0.371***
0.415***
0.438***
(0.131)
(0.145)
(0.149)
Complexity
0.379***
0.394***
0.395***
(0.144)
(0.139)
(0.139)
Technical proposing minister
-0.490**
-0.535**
-0.525**
(0.210)
(0.236)
(0.220)
0.141***
0.133***
0.135***
(0.029)
(0.022)
(0.026)
0.000***
0.000***
0.000***
(0.000)
(0.000)
(0.000)
4.798***
4.779***
4.739***
(0.286)
(0.275)
(0.369)
2.521
8,530.327
8,592.700
636
2.536
8,533.051
8,590.969
636
2.537
8,533.213
8,591.130
636
Prime minister's bill
Second Republic
Number of referral committees
Length of bill
Constant
Alpha
Aic
Bic
N
Note: *** p<0.01, ** p<0.05, * p<0.1
Clusters on policy dimensions (Comparative Agenda Project)
XII legislature excluded
-0.048
Fig.5 Distribution of the dependent variable (Kernel Density Estimate)
.0004
0
.0002
0
5000
10000
15000
n. of changed words
20000
kernel = epanechnikov, bandwidth = 109.0206
Fig.6 Changed words, added words, removed words, substituted words. Mean by legislature
0
200
400
600
800
1,000
Density
.0006
.0008
Kernel density estimate
10
11
Bills changes
Removed words
12
13
Added words
Substituted words
14
REFERENCES
Andeweg, Rudy B. 2000. “Ministers as Double Agents? The Delegation Process Between Cabinet
and Ministers.” European Journal of Political Research 37 (May): 377–95.
Andeweg, Rudy B., and Arco Timmermans. 2008. “Conflict Management in Coalition
Government”, in Kaare Strøm, Wolfgang Müller, and Torbjörn Bergman (eds.), Cabinets and
Coalition Bargaining: The Democratic Life Cycle in Western Europe. Oxford: Oxford
University Press, 269-300.
Bagehot, Walter. [1872] 2001. The English Constitution. Edited by Paul Smith. Cambridge:
Cambridge University press.
Benoit, K. and M. Laver. 2006. Party Policy in Modern Democracies. Routledge.
Blondel, J. 1995. Comparative Government, London: Prentice Hall.
Bryce, Lord James. 1921. Modern Democracies. New York: Macmillan.
Cameron, C. and K. Trivedi. 1998. Regression Analysis of Count Data. Cambridge: Cambridge
University Press.
Cox, Gary W. 1987. The Efficient Secret. Cambridge: Cambridge University Press.
Cox Gary W. (2005). “The Organization of Democratic Legislatures”. In Barry Weingast and
Donald Wittman (Eds.). The Oxford Handbook of Political Economy. Oxford: Oxford
University Press. http://weber.ucsd.edu/~gcox/Legislatures13.doc
Döring, Herbert (ed.). 1995a. Parliaments and Majority Rule in Western Europe. New York: St.
Martin‟s Press.
Döring, Herbert. 1995b. "Time as a Scarce Resource: Government Control of the Agenda." In
Herbert Döring, ed. Parliaments and Majority Rule in Western Europe, p. 223-246. New
York: St Martin's Press.
Fenno, Richard F. 1966. Power of the Purse. Boston: Little, Brown.
Fenno, Richard F. 1973. Congressmen in Committees. Boston: Little, Brown.
Ferejohn, John A. 1974. Pork Barrel Politics. Stanford: Stanford University Press.
Franchino, Fabio, and Bjørn Høyland. 2009. Legislative Involvement in Parliamentary Systems:
Opportunities, Conflict, and Institutional Constraints. American Political Science Review 103
(4):607-621.
Gallagher, Michael, Michael Laver, and Peter Mair. 2006. Representative Government in Modern
Europe. 4th ed. Boston: McGraw–Hill.
Hine, David. 1993. Governing Italy. Oxford: Clarendon Press.
Huber, John D. 1996. “The Vote of Confidence in Parliamentary Democracies.” American Political
Science Review 90 (June): 269–82.
King, Gary. 1989. Unifying Political Methodology: The Likelihood Theory of Statistical Inference.
Ann Arbor: University of Michigan Press.
Laver, Michael. 2006. "Legislatures and Parliaments in Comparative Context." In Barry R.
Weingast and Donald Wittman, eds. The Oxford Handbook of Political Economy, p. 121-140.
Oxford: Oxford University Press.
Laver M. and J. Garry. 2000. Estimating the Policy Positions from Political Texts, in “American
Journal of Political Science”, 44, pp. 619-634.
Laver, Michael, and W. Ben Hunt. 1992. Policy and Party Competition. New York: Routledge.
Laver, Michael, and Kenneth A. Shepsle, eds. 1994. Cabinet Ministers and Parliamentary
Government. Cambridge: Cambridge University Press.
Laver, Michael, and Kenneth A. Shepsle. 1996. Making and Breaking Governments. Cambridge:
Cambridge University Press.
Lowi, Theodore J. 1964. American Business, Public Policy, Case-studies, and Political Theory.
World Politics, 16:677-715.
Maass, Arthur. 1951. Muddy Waters. Cambridge: Harvard University Press.
Martin, Lanny W., and Georg Vanberg. 2004. “Policing the Bargain: Coalition Government and
Parliamentary Scrutiny.” American Journal of Political Science 48 (January): 13–27.
Martin, Lanny W., and Georg Vanberg. 2005. “Coalition Policymaking and Legislative Review”
American Political Science Review 99 (1): 93–106.
Masciandaro, D. 1996. “La non emendabilita‟ delle leggi di spesa e di bilancio:analisi teorica e
profili istituzionali.” In: Monorchio A. (ed) La finanza pubblica italiana dopo la svolta del
1992. Il Mulino, Bologna.
Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven: Yale University Press.
Mezey, Michael L. 1979. Comparative Legislatures. Duke University Press, Durham (NC).
Müller, Wolfgang C. 2000. “Political Parties in Parliamentary Democracies: Making Delegation
and Accountability Work.” European Journal of Political Research 37 (May): 309–33.
Müller, Wolfgang C., and Kaare Strøm, eds. 2000. Coalition Governments in Western Europe.
Oxford: Oxford University Press.
Plott, Charles. 1968. Some Organizational Influences on Urban Renewal Decisions. American
Economic Review, 58:306-311.
Rae, D. (1967), The Political Consequences of Electoral Laws, New Haven: Yale University Press.
Rasch, Bjørn Eric. 1995. "Parliamentary Voting Procedures." In Parliaments and Majority Rule in
Western Europe, edited by H. Döring, pp. 488-527. Frankfurt and New York: Campus Verlag
and St. Martin's Press.
Rasch, Bjørn Eric. 2000. "Parliamentary Floor Voting Procedures and Agenda Setting in Europe."
Legislative Studies Quarterly 25:3-23.
Rundquist, Barry S. 1973. Congressional Influences on the Distribution of Prime Military
Contracts. Unpublished Ph. D. dissertation, Stanford University
Saalfeld, Thomas. 2000. “Members of Parliament and Governments in Western Europe: Agency
Relations and Problems of Oversight.” European Journal of Political Research 37 (May):
353–76.
Schattschneider, E. E. 1935. Politics, Pressures and the Tariff. Englewood Cliffs, NJ: Prentice-Hall.
Scott Long, J. 1997. Regression Models for Categorical and Limited Dependent Variables.
Thousand Oaks; London; New Dehli: Sage Publications.
Strøm, Kaare, Wolfgang Müller, and Torbjörn Bergman (eds.). 2008. Cabinets and Coalition
Bargaining: The Democratic Life Cycle in Western Europe. Oxford: Oxford University Press.
Thies, Michael F. 2001. “Keeping Tabs on Partners: The Logic of Delegation in Coalition
Governments.” American Journal of Political Science 45 (July): 580–98.
Tsebelis, George. 2002. Veto Players: An Introduction to Institutional Analysis. Princeton:
Princeton University Press.
Verzichelli, Luca. 2003. “Italy: Delegation and Accountability in a Changing Parliamentary
Democracy.” In Delegation and Accountability in Parliamentary Democracies, eds. Kaare
Strøm, Wolfgang Müller, and Torbjörn Bergman. Oxford: Oxford University Press: 445-73.
Verzichelli, Luca. 2008. “Portfolio Allocation”, in Kaare Strøm, Wolfgang Müller, and Torbjörn
Bergman (eds.), Cabinets and Coalition Bargaining: The Democratic Life Cycle in Western
Europe. Oxford: Oxford University Press, 237-267.
Weingast, B., K. Shepsle, and C. Johnsen. 1981. The political economy of benefits and costs: A
neoclassical approach to distributive politics. Journal of Political Economy 89:642-64.
Wheare, K. C. 1963. Legislatures. London: Oxford.
Wilson, James Q. 1973. Political Organizations. New York: Basic Book, Inc., Publishers.
Zucchini, F. 2005. "The Legislative Agenda Setting Power in a Changing Parliamentary
Democracy: the Italian Puzzle." Unpublished typescript, Università degli studi di Milano.