Ep(ic) fail? The European Parliament’s Influence in the Ordinary Legislative Procedure
Pages: 45 Autumn Semester 2019
Word Count: 15 489 Master’s Thesis, 30 ECTS
Author: Kajsa Edholm Department of Government
Supervisor: Thomas Persson Uppsala University
Abstract
When the Lisbon treaty entered into force in 2009 it marked the peak of a long period of institutional empowerment of the European Parliament (EP). In formal terms, the EP and the Council of the European Union (Council) now have equal power in most policy areas. Despite the remarkable transformation of the EP, its role in EU legislation remains understudied.
This thesis aims to contribute to a better understanding of the EP’s role in the Ordinary Legislative Procedure (OLP). It assesses whether the EP’s formal power is reflected in its de facto influence based on two criteria; 1) it should not be less influential than the Council, and 2) its influence should not be limited in substantive terms.
Empirically, it covers 4,662 amendments proposed in the early negotiation phase of 30 randomly selected legislative acts. First, by conducting a one-sample t-test, it demonstrates that despite formal empowerment the parliament is still significantly less influential than the Council in EU legislation. Second, using binary logistic regression, it shows that EP’s influence is limited in substantive terms. The more extensive amendments the EP is attempting to introduce – the less influence it has.
In addition to contributing to empirical research on the parliament’s role in EU legislation this
thesis introduces a new way of quantitatively measuring legislative influence at the EU level. By
creating a dataset comparing the institutions’ negotiation mandates with legislative outcomes, a
more precise measurement of the institutions’ influence is provided.
Table of Contents
1. Introduction ... 1
2. Beyond Formalities: The Parliament’s Role in EU Legislation ... 3
Unequal Institutional Influence? ... 3
The Parliament’s Substantive Influence: Does the Matter Matter? ... 6
3. Scope and (De)Limitations ... 7
Step 1: Assessing the Institutions’ Relative Influence ... 8
Step 2: The EP’s Substantive Influence ... 8
(De)Limitations ... 9
4. The Ordinary Legislative Procedure ... 10
5. Data and Methods ... 13
Data Material: Identifying Policy Preferences ... 14
Constructing the Dataset ... 16
Selecting a Sample ... 17
Data Collection ... 19
Coding ... 20
Methods for Quantitative Data Analysis ... 30
Step 1: Assessing the Institutional Balance ... 30
Step 2: A limited Substantive Influence? ... 31
6. Results ... 32
Descriptive data: The Institutions’ Sought Contribution ... 33
Data Analysis: Assessing the two Criteria ... 35
Step 1: Are the Co-legislators Equally Influential? ... 35
Step 2: Is the Parliament’s Influence Substantively Limited? ... 37
7. Discussion ... 41
8. Conclusions ... 43
References ... 46
Appendices ... 49
1. Introduction
Much sets the European Parliament (EP) of today apart from the assembly created nearly 70 years ago. The past decades have seen a remarkable reform of the composition of the parliament, as well as its role in EU decision-making. Initially, the EP consisted not of directly elected representatives, but of part-time delegates from national parliaments (Hix and Høyland, 2013: 172). As of 1979, the citizens of the union are able to directly elect their representatives. Whereas the Council is meant to represent the member states, the European parliament is now tasked with representing the European peoples (Ritterberger, 2012: 18).
Further, the formal power of the EP is significantly less limited today compared to the early days of the European Coal and Steel Community. Then, the EP functioned mainly as a consultative body with few and mainly nominal powers (Hix and Høyland, 2013: 172, Ritterberger, 2012: 18).
While the EP could put pressure on the Council by delaying legislation, there were limits to what the EP could achieve under this decision-making procedure (Rasmussen, 2012: 737).
Since then the EP has undergone a process of significant institutional empowerment (Ritterberger 2012: 18). When the Lisbon treaty entered into force in 2009 it “[…] recognised the EP as a central institution of the EU regime and a key player of democracy and legitimacy” (Costa, 2019:
7). In terms of formal powers, the parliament is now on equal footing with the Council in most legislative areas. With the expansion of co-decision, EU-decision-making has been compared to a bicameral system, where the Council constitutes one chamber, and EP the other (Kreppel, 2018:
12, Mårtensson, 2007: 289).
In other words, the parliament is now tasked with the role as co-legislator and with representing the European peoples. For both these roles, its ability to influence EU legislation is crucial. Without de facto influence, it cannot function as a co-legislator on equal terms with the Council, nor can it fully represent the Union’s citizens. If elected MEPs cannot use their political mandates to influence legislation, they cannot realize the will of their voters or be held accountable for legislative outcomes.
Despite the drastic reform of the parliament’s formal role, its influence in EU legislation remains
understudied (Kreppel, 2018: 14, Bressanelli and Chelotti, 2019: 266). While the continuous
expansion of EP’s formal power gives the impression that its influence has steadily increased,
empirical research is ambiguous to what extent the parliament has in fact been empowered.
Previous quantitative studies have to large extent relied on imprecise measurements of institutional influence and data from before Lisbon. Consequently, there is no satisfactory answer to what influence the parliament has on contemporary legislation.
The aim of this thesis is to contribute to the empirical research on the EP’s role in EU legislation by assessing whether the formal empowerment of the parliament has translated into corresponding influence. The research question guiding this thesis is: “does the European parliament’s de facto influence in the ordinary legislative procedure (OLP) correspond to its formal power?”. In line with previous research in the field, power is defined in a Weberian sense, as the potential a group or person has “to realise their own will in a social action even against the resistance of others”
(Costello and Thomson, 2013: 1026). Following Bressenalli and Chellotti (2019: 266), formal power is differentiated from influence. Whereas power is the institutional capability to realize one’s will, influence is understood as that power being exercised in practice. In other words, an actor has influence when its policy preferences are reflected in the outcome.
Here, the parliament’s influence on EU legislation is assessed based on two criteria; 1) it should not be less influential than the Council, and 2) its influence should not be limited in substantive terms. If the parliament is less influential than the Council and/or systematically loses more amendments on important matters, it is arguably less influential than its formal power implies.
Empirically the study includes 4,662 proposed amendments to 30 legislative acts adopted during the 8
thparliamentary term (1 July 2014–1 July 2019). Following the two criteria, data is analyzed in two steps. First the proportion of amendments won by each institution in the 30 legislative acts are compared to see whether the EP has significantly less influence than the Council. Second, to determine whether the parliament has limited substantive influence, the individual 4,662 amendments are studied more closely.
The remainder of this thesis is structured as follows; first, empirical research on the parliament’s
role in EU legislation is presented, followed by a chapter presenting the scope and limitations of
this study. Then, a brief background of the ordinary legislative procedure (OLP) is provided,
including an account of how this legislative procedure works in practice. Following this is a
methods chapter, which outlines how policy preferences are identified, how the dataset was
constructed as well as the methods used for quantitative data analysis. Then the results of the data
analysis are interpreted and related to the criteria stipulated above. The findings are then discussed.
Finally, some conclusions are presented and fields of future research are outlined.
2. Beyond Formalities: The Parliament’s Role in EU Legislation
The discussion on the parliaments influence on EU legislation has occupied scholars for decades, both normatively and empirically. This chapter provides an overview of previous research related to the parliament’s ability to influence EU legislation, highlighting both the contributions and the limitations of previous research.
The chapter is divided into two parts. First it presents empirical research on the institutions’ relative influence, second it presents research on the EP’s ability to exert substantive influence on EU legislation.
Unequal Institutional Influence?
A challenge in addressing the question of EP’s influence is that many studies focus on discrete policy areas and that there has been “[…] little effort to develop comparisons and provide a more exhaustive understanding of the EP’s role post-Lisbon” (Bressanelli and Chelotti, 2019: 266).
Two primary approaches have been used to quantitatively study EP’s influence on EU legislation;
amendment analysis and the interview-based approach (Hix and Høyland, 2013: 175). The assessment of the parliament’s influence varies depending on which approach is used.
Scholars employing an amendment analysis have ascribed the EP a fairly prominent role in EU legislation. The approach was introduced by Amie Kreppel in the late 1990s (Hix and Høyland, 2013: 175). First, in 1999, she looked at the adoption rate of EP’s proposed amendments under the cooperation procedure (Kreppel, 1999). This was followed by a study comparing the EP’s influence under the cooperation procedure with the co-decision procedure (Kreppel, 2002). She concludes that under co-decision (I), the Council adopted around a third of the EP’s amendments (Kreppel, 2002: 797). In the context of EP’s previously more limited power, this result is framed as the parliament being relatively successful (Kreppel 2002: 810)
A version of this approach is used in a more recent study by Holzinger and Biesenbender (2019).
They study the distribution of legislative power between the Council and the EP by using PreLex- data covering 7202 legal acts from between 1976 and 2009 (Holzinger and Biesenbender, 2019:
340). Out of these, 785 commission proposals were subject to co-decision (the rest were handled
under consultation or the now abolished cooperation procedure). Instead of looking at individual amendments, they investigate to what extent the parliament’s first reading position has been adopted (i.e. all the amendments proposed by the parliament). The authors conclude that under co- decision the EP “strongly influenced” over 40 % of all cases (Holzinger and Biesenbender, 2019:
346). This conclusion is based on the fact that the amendments introduced in EP’s first reading position were adopted as final legislation in roughly 40 % of the cases.
While amendment analysists, tend to conclude that the parliament matters in EU legislation, a clear majority of its amendments are rejected even in these studies. As focus is limited to the adoption rate of EP amendments, these findings cannot be contrasted against the Council’s influence in the same acts. However, if the results were to be extended to present day legislation, they would not rhyme well with the formal “goal” of equal power.
Those studying the institutions’ relative power through the interview-based approach tend to be far less optimistic about the EP’s ability to influence EU legislation. The most notable example of this is the dataset on Decision-making in the European Union (DEU/DEU II), which has been used in numerous studies. The dataset covers policy outcomes and preferences for controversial issues raised by 125 legislative files adopted between 1996-2008 (Thomson, et. al., 2012). Studies based on the DEU-II dataset overwhelmingly conclude that legislative outcomes correspond more closely to the preferences of the Council and that the EP has less influence than its formal power would suggest (Costello and Thomson, 2013: 1036; Thomson and Hosli, 2006: 413-415, Kreppel, 2018: 28).
In other words, previous research demonstrates that the EP has had a limited influence in quantitative terms, even if scholars differ in their assessment of how severe the imbalance is.
Nevertheless, knowledge about the EP’s relative influence in contemporary legislation is limited.
There are two primary reasons as to why results from past research cannot simply be extended to cover acts adopted in the eight parliamentary term. First, there are methodological limitations in the studies concluded to date. Second, that the data used in these quantitative studies are over 10 years old.
The methodological limitations of previous research are highlighted more in-depth when
challenges in measuring policy preferences are discussed (cf. first section of chapter 5). In general
terms, amendment analysis has been criticized because of its inability to determine where
amendments come from (Hix and Høyland, 2013: 176). The EP’s first-reading position often mirrors an agreement reached in an informal trilogue, rather than the preferences of the EP alone.
Thus, the Holzinger and Biesenbender’s (2019) interpretation of data could be challenged. By interpreting acceptance of EP’s first reading position as influence, the authors run the risk of exaggerating EP influence.
The DEU II-dataset on the other hand only covers controversial issues. It does not include amendments on technical details (nor does it aspire to) (Thomson et. al., 2012: 614). While this can be a valid delimitation, it limits our ability to generalize results based on the DEU-dataset to the EP’s influence more generally. EU legislation is notoriously technical, so a narrow focus on highly controversial issues does not fully cover day-to-day legislative work in the union. Further, the dataset is constructed based on ex-post interviews with stakeholders, which means that there are reliability issues with the data.
While it may seem unlikely that adjusted versions of the same studies should conclude that the EP is in fact more influential than the Council, a more precise measurement of EP’s influence is desirable. In this sense, the level of influence enjoyed by the EP is not a closed case. What’s more is that neither study has utilized data from after the entry in to force of the Lisbon treaty. As already mentioned, the EP saw a substantial increase in its formal power with the new treaty. Under the new formal provisions, the EP’s influence could have improved. This means that results based on data prior to the Lisbon treaty cannot just be extrapolated to present day legislation.
Furthermore, there are contextual factors which may affect the parliament’s ability to influence
legislation. For one, when the union is in a ‘crisis mode’ power tends to shift from the supranational
institutions to member states (Möller, 2019: 87-88). This may of course affect some acts more than
others. But in coping with crisis, the European Council, and specifically some member states, have
gained influence. (Möller, 2019; Zaun, 2018; Ripoll Servent 2019) When problem-solving is
moved out of the Ordinary Legislative Procedure and into the intergovernmental arena, the EP
loses influence. While all parliamentary terms have their share of challenges the 8
thparliamentary
term has been ridden with crisis. With the Euro crisis, the asylum ‘crisis’, the rise of EU-skepticism
and a member state opting to leave the union, the last parliamentary term can hardly be considered
as “business as usual”.
Further, following the trend of many national parliaments, the European parliament has become increasingly fragmented in the past parliamentary terms. This is challenging as broad political coalitions are often necessary to win a majority in the European parliament. Indeed, past research has shown that internal cohesion is an important factor in explaining EP influence (Bressanelli and Chelotti, 2019: 268) whereas fragmentation and internal division may undermine the parliament’s ability to influence policy and weaken their position vis-à-vis the Council (Kreppel, 2002: 81).
Neither of these potentially constraining developments are accounted for when data from before 2009 is used. When even recent studies rely on old data our ability to extend findings to present day legislation is limited.
So, while much of past research point in the direction of limited EP influence, there is to date no precise assessment of the balance of influence between the institutions in contemporary legislation.
On the one hand, the parliament has gained in formal power, on the other hand there are contextual factors that may limit its de facto influence.
The Parliament’s Substantive Influence: Does the Matter Matter?
Beyond the institutional balance, researchers have dedicated themselves to studying the substantive influence that the parliament enjoys. Kreppel, in her above-mentioned studies, was early in theorizing that impressive adoption rates of EP amendments, could mean little in terms of substantive influence. As not all amendments are created equal, the EP could win many amendments but still have little impact on EU legislation in substance (Kreppel, 1999: 522).
To capture nuances of the parliament’s contribution to EU legislation, she tests the impact of a number of variables on the probability of an amendments being adopted. In her data analysis, Kreppel finds that amendment type has a significant effect on the probability that an EP amendment is adopted (Kreppel, 1999: 528; Kreppel, 2002: 810). Substantive amendments proposed by the parliament are adopted quite frequently, but the probability of an amendment to be adopted decreases as the amendment-type becomes more extensive.
Kreppel also tests whether the parliament is better at influencing the overarching introductory recitals than the legislation articles. Placement is included in the data analysis as the introductory recitals tend to be more “general statements of purpose than specific legal policy requirements”.
If the EP is less able to influence articles this would mean that the parliament is less influential in
substantive terms than a simple count of adoption rates may indicate (Kreppel, 2002: 793).
Placement of the amendment, falls short of statistical significance. However, Kreppel (1999) notes that there seems to be a pattern in the expected direction. (i.e. the EP’s seems to have a harder time introducing amendments to the articles).
In more contemporary studies, the subject matter of an act has also been attributed value. These studies indicate that the Council has been able to dominate decision-making in areas concerning economic governance (Bressanelli and Chelotti, 2018; Schoeller and Héritier, 2019) and asylum policies (Ripoll-Servent, 2019). A common take-away is that the EP is less influential in areas covering “core state powers”, even as the Lisbon treaty has extended OLP to cover more of these areas (Bressanelli and Chellotti, 2019: 268-269) .
Based on these findings, it seems that the EP’s influence could be limited not only in numeric terms, but also in substantive terms. However, as with the EP’s influence relative to the Council, there is limited knowledge on the parliament’s contribution to EU legislation in substantive terms.
Kreppel’s findings are interesting, but by now they are dated. For the same reasons as listed in the section preceding this one, results from before the Lisbon treaty need to be treated cautiously.
The more recent intensive studies have gone a long way in explaining the how’s and why’s behind EP’s influence in specific cases and have provided some updated insight in terms of the EP’s substantive contribution to EU legislation. However, the recent studies concluding that the parliament’s influence is more limited in core state areas have focused only on these specific areas.
Without consistent comparison across areas, this conclusion could be challenged. The lack of a consistently applied method means that there is a level of uncertainty as to whether this variance should be ascribed to methodological or substantive differences. Is the parliament’s influence really more limited in core state policy areas, or is it just limited in general? In other words, when it comes to EP’s substantive contribution to EU legislation, findings from these studies cannot be readily extrapolated to contemporary legislation.
3. Scope and (De)Limitations
This thesis employs a two-step approach. First, it studies 30 randomly selected legislative acts,
adopted and negotiated under OLP in the eight parliamentary term (July 2014 –July 2019). The
purpose is to assess whether the parliament and the Council have equal influence on EU legislation
in quantitative terms. In the second step, the 4,662 amendments in the 30 acts are analyzed more closely, to determine to what extent the EP is able to substantively influence legislation.
The remainder of this chapter outlines the scope of the study, presenting the two-step approach in more detail. Then some limitations of the study are considered.
Step 1: Assessing the Institutions’ Relative Influence
In the first step, I measure parliament’s ability to shape EU legislation relative to the Council by comparing their ex-ante preferences to legislative outcomes. This is done by looking at 30 randomly selected acts adopted in the post Lisbon area. The method for data analysis is presented in chapter 5.
By studying acts adopted in the 8
thparliamentary term (i.e. post Lisbon) a more up to date analysis is provided compared to previous quantitative studies. Furthermore, unlike many of the intensive studies carried out more recently, a broad range of policy areas, covering technical as well as highly politized issues are included and assessed using the same criteria. Arguably, this provides a more complete understanding of parliament’s influence over EU legislation than studies focusing on either discrete policy areas or on highly controversial issues.
Step 2: The EP’s Substantive Influence
In the second step, the amendments are analyzed more closely. Following the appeal of Chelotti and Bressanelli, to “precisely and thoroughly” demonstrate how the EP contributes to EU-level policymaking (2019: 273). This step complements the first and enables us to determine whether the EP’s influence is limited in substantive terms.
More specifically, the outcome of each negotiated amendment is studied to assess whether the EP is less successful in shaping legislative outcomes when:
• introducing more extensive amendment types,
• the conflict level between the institutions is high,
• the legislative act covers areas that are core state competences,
• attempting to amend the articles rather than the introductory recitals of an act.
If the parliament is found to be significantly less influential in any of these aspects, this would
indicate that its influence is substantively limited. Most of the variables are included on the basis
that previous research has shown that they could limit the EP’s substantive influence.
The exception is level of conflict, which has not previously been included in research the way it is envisaged here (the coding of all variables is presented in chapter 5). The reason for including this variable is to capture the dynamics between the institution’s ex-ante preferences. If the parliament can only introduce amendments when the conflict level is low, this would imply that it has limited substantive influence and that its influence is perhaps conditioned on the fact that the Council is not very engaged or has similar ex-ante preferences.
In comparison to previous research, this will give a more nuanced understanding of the EP’s influence in the 8
thparliamentary term. By consistently assessing amendments of many policy areas, it also opens up for comparative analysis. Unlike past studies that have focused on one or few legislative acts, it will be possible to analyze whether the EP’s influence is more limited in certain instances.
(De)Limitations
The study is delimited to acts adopted under OLP. There are two reasons for this. First, it is by far the most common decision-making procedure for adopting EU legislation and thus includes most of the acts (EPRS 2019: 9). Second, it is theoretically interesting to compare the institutions’ de facto influence under OLP, as they are in formal terms meant to enjoy equal power under this decision-making procedure.
As the focus in this thesis is on influence, the scope is also limited to acts that have been either negotiated or amended in relation to the original proposal. If the initial legislative proposal is adopted without any amendments, it is not possible to determine whether either institution has enjoyed more influence.
When narrowing down the population to acts that (a) have been adopted under OLP and (b) were negotiated or amended by the institutions
1, 337 legislative acts remain
2. In some cases, the
1 This checked as follows: first the key events listed on EPs legislative observatory were checked to see if any committee decision had been taken to enter into inter-institutional negotiations. If such a decision was present, the act was coded as being negotiated. If no decision was listed under key events the committee report was consulted to see if the EP had in fact proposed any changes to the Commissions initial proposal. If the EP took over the commissions original proposal without making any amendments, the act was coded as not being negotiated. If one or more amendments were suggested, the act was coded as being negotiated and/or amended.
2 Acts adopted “as amended by the Consultative Working Party of the legal services of the European Parliament, the Council and the Commission” are coded as not being amended/negotiated.
legislative process had begun before the start of the parliamentary term. However, as they were still adopted in the relevant time span this did not disqualify them from being included in the study.
The focus of this study is on the adoption phase. This makes it possible to analyze each institution’s ability to introduce amendments in inter-institutional negotiations, i.e. how much they are able to steer the content of a legislative proposal in their preferred direction. However, it is important to note that this focus is on a limited section of the policy processes. The Commission’s ex-ante information on the institutions’ preferences may vary (see discussion in Hix and Høyland, 2013:
176) and the initial proposal could be more in line with one of the institutions’ preferences. If the initial proposals better corresponds to one institution’s preferences, it may be more content with the outcome – even if it is able to introduce fewer amendments in the adoption phase. Thus, the results presented here will not shed light on the influence of each institution in important phases of drafting and implementing policy.
Furthermore, the study focuses on decisions rather than non-decisions. To measure institutional influence by comparing ex-ante preferences to legislative outcome requires an observable outcome.
In other words, it does not include cases where the institutions are unable to reach an agreement.
This may not seem as a significant limitation, but recent studies have shown that institutions have been unable to reach an agreement in some highly controversial issues (Zaun, 2018; Ripoll Servent, 2019). If one institution favors status quo this is more in line with its preferences. The contribution of this thesis should not be extended to cover non-decisions.
Finally, another limitation of this study is that it covers only one parliamentary term. This makes it difficult to address contextual factors and fragmentation. It also means no assessment can be made on whether the parliament’s influence seems to have increased or decreased over time.
To conclude, this thesis will provide an indication of whether EP’s de jure power corresponds to its de facto influence in EU decision-making. However, the results should not be generalized beyond the adoption phase of EU legislation and to non-decision decisions.
4. The Ordinary Legislative Procedure
This chapter gives a brief introduction to how OLP works in practice, including the role of the
informal trilogues between the parliament, the Council and the Commission.
The legislative procedure of co-decision was first introduced in the Maastricht Treaty (signed in 1992). Initially co-decision was limited to few policy areas, but each treaty adopted since has expanded the number of areas in which the procedure applies (Rasmussen, 2012: 735-736). As of the Lisbon Treaty (2009), co-decision has become the Ordinary Legislative Procedure (OLP) – applied in approximately 90 % of all legislation (European Parliament, 2017: 50).
For an act to be adopted under the OLP decision-making procedure, the EP and the Council need to agree on an identical text (European Parliament 2017: 50). The co-legislators have a maximum of three readings to reach an agreement, but not all three rounds must not be utilized. In fact, early agreements have gradually become the standard operating procedure. In the previous parliamentary term (2014-2019), a clear majority of files under co-decision were concluded already in the first reading (EPRS, 2019: 9).
Somewhat simplified, the legislative process is set in motion when the Commission adopts a draft proposal, which is then sent to the Council and the European parliament for a first reading
3. While both institutions start processing the proposal in parallel, it is the European Parliament who formally needs to act first. In the first reading the EP can either, reject, accept or amend the Commission’s proposal (Borchardt 2017: 113). Formally, a first-reading agreement requires the Council to accept the Commissions original proposal with the amendments tabled by the EP (Rasmussen, 2012: 737, Borchardt, 2017: 113).
If the Council chooses to amend the EP position, the process moves to a second reading. The parliament then has three months to approve, reject or amend the Council position. If the parliament approves the Council position, the act is adopted. If it chooses to amend it the Council, in turn, has three months to approve, reject or amend the EP position (Borchardt, 2017: 114).
If no agreement has been reached at the end of the second reading, the third and final reading is initiated. At this stage a Conciliation Committee is created by the president of the Council, in agreement with the parliament ditto. The committee consists of member states and EP representatives and is tasked with agreeing (by a qualified majority vote) on a joint text within six
3 Where applicable proposal is also sent for consultation to Economic and Social Committee and the Committee of the Regions.
weeks. If no agreement is reached, or if the provisional agreement does not pass the vote in either the EP or the Council, the act falls (Borchardt, 2017: 114–115).
The formal readings do not, however, reflect the real amount of interactions between the institutions. At any stage of the legislative procedure, the institutions may engage in interinstitutional negotiations called “informal trilogues” (European Parliament, 2017: 28).
Informal trilogues are meetings held between representatives of the Council, the Commission and the European Parliament. These play an important role in reaching an agreement between the institutions and have now become routine in EU legislation
4(Rasmussen, 2012: 738, Greenwood and Roederer-Rynning, 2019:123–124).
In informal trilogues, representatives of each institution aim to reach a compromise position, based on negotiation mandates provided to them by their respective institutions (European Parliament, 2017: 28–29). For the EP this is generally provided by a committee report or plenary mandate and for the Council the negotiating mandate can be a General Approach adopted by itself or one of its preparatory bodies. Any agreement in a trilogue is provisional and must be formally approved by the institutions (European Parliament, 2017: 28–29).
In other words, the sharp increase in first reding agreements should not be interpreted as the Council passively accepting all EP amendments. While the amendments are formally introduced by the EP, they may in fact mirror the outcome of the informal trilogues (European Parliament 2017: 18)
4 During the period July 2014 – Dec 2018 (i.e. not the entire eight parliamentary term), EP committees participated in 994 trilogues (EPRS 2019, p 10).
Figure 1: Process of Informal Interinstitutional Negotiations
European Parliament 2017, p 28
5. Data and Methods
This chapter outlines the methods used in order to evaluate the parliaments influence on EU
legislation. I begin by discussing how influence is measured, the data material used and challenges
in measuring the institutions’ ex-ante preferences. Then, I move on to describe the process of
constructing the dataset – i.e. how the sample was selected, how data was collected and how the
variables were coded. Finally, the quantitative methods used to analyze the data are briefly
introduced.
Data Material: Identifying Policy Preferences
As stated in the introduction, this thesis seeks to determine whether the parliaments influence on EU legislation corresponds to its formal power. This is done by evaluating whether the EP is less influential than the Council in the adoption phase of EU legislation and whether its ability to shape legislative outcomes is limited in substantive terms. Influence is measured in terms of congruence between the institutions’ ex-ante preferences and the adopted legislation (cf. definition in chapter 1). For the parliament to “win” an amendment, the outcome needs to correspond more closely to its ex-ante preferences than to the Council’s.
One of the main challenges in assessing the co-legislator’s relative influence is determining their ex-ante policy preferences (Kreppel, 2018: 15). In this thesis, the mandates provided to the institutions’ negotiators prior to entering informal trilogues are used as an indicator of their ex- ante preferences. In figure 1, this is represented by the very first and the very last step of the legislative process. These negotiation mandates tend to be quite detailed, tracking the institution’s proposed changes to each paragraph. As amendments are presented in relation to the initial proposal, the institutions’ preferences can easily be identified.
More specifically, the negotiation mandate is provided by a committee report or plenary mandate for the EP, and a General Approach or position adopted by the Council or one of its preparatory bodies for the Council (European Parliament, 2017: 28–29). In addition to using the “raw”
mandates, so called multi-column documents are used when possible (an example of a row of a multicolumn table can be found in figure 2). These have become an important tool in informal trilogues as they summarize, in separate columns, the initial proposal and the position of each institution based on their negotiation mandates
5(Greenwood and Roederer-Rynning, 2019: 125).
While the content is the same as in the mandates, this summary greatly facilitates comparison of the institutions’ preferences. I have therefore used multi-column documents over the separate negotiation mandates whenever that information is available.
5 There is also a fourth column, in which provisional agreements and potential compromises are explored.
Figure 2: An Example Row of a Multicolumn Table
2016/0280 (COD)
Negotiation mandates have previously been used as part of a more comprehensive material in qualitative case studies (see for example Bressanelli and Chelotti, 2018). However, to date they have not been utilized for a quantitative study covering multiple policy areas. Instead, quantitative studies on the subject tend to rely on either expert interviews or the EP’s first reading position in assessing the institutions’ preferences (Hix and Høyland, 2013: 175).
There are three main reasons to why I have used the negotiation mandates instead. First, compared to the interview-based approach, this measurement of the institutions’ ex-ante preferences is more reliable. As Hix and Høyland (2013: 177) point out, experts “[…] have an incentive to claim credit for particular outcomes, so they are likely to adjust their recalled preferences to claim that they were closer to the legislative outcome on an issue than they actually were”. The influential interview-based DEU-II dataset, conducted interviews after the negotiations were completed, which means that there was a clear risk of participants exaggerating their own influence. Hix and Høyland (2013: 178) call for “facts that are hard to manipulate post hoc” to be used in assessing the EP’s influence on legislative outcomes. A criterion which is satisfied by using the negotiation mandates.
Second, in relation to using the first reading position of the EP, it less uncertain where an
amendment comes from. As noted above, the extensive use of informal trilogues means that the
EP’s first reading position may in many cases reflect a provisional agreement between the
institutions’ rather than the EP’s own preferences (Hix och Høyland, 2013: 176, European Parliament, 2017: 18). Unlike the first reading position, the negotiation mandates are adopted before the institutions enter negotiations. Therefore, they are less likely to reflect a provisional agreement between the institutions. Looking at an earlier stage in the legislative process thus makes it easier to untangle which institution preferred what.
The third and final reason for using negotiation mandates is that they provide us with more satisfactory information on the preferences of the Council. Historically the other two approaches, have either completely neglected the preferences of the Council by looking only at acceptance rate of EP amendments, or approximated the Council’s preferences by using at the estimated standpoint of each member state (Costello and Thomson, 2013: 1033). To have no or imprecise data on the ex-ante preferences of one of the two co-legislators arguably undermines the ability to make a balanced assessment of the institutions’ relative influence.
While negotiation mandates address some of the weaknesses in previous approaches, they are not perfect. The mandates are not adopted in a vacuum or simultaneously. Thus, the institutions could glance at each other and, consciously or unconsciously, attempt to adjust their mandates to facilitate an agreement. Further, subtle differences in formal documents may be difficult for an outsider to pick up. Here, complementing interviews would probably unlock more nuance of the institutions’ ex-ante preferences. Nevertheless, negotiation mandates are arguably the closest we will come to identifying the institutions’ ex-ante preferences, without participating the actual negotiations.
Constructing the Dataset
While negotiation mandates provide information on the institutions’ ex-ante preferences, they do not in un-processed form enable quantitative data analysis. Consequently, I constructed a dataset.
This was done in three steps, namely; sampling, data collection and coding. The three steps are
outlined below.
Selecting a Sample
Studying all the 337 acts adopted in the 8
thparliamentary term is beyond the scope of this thesis.
To narrow down the number of acts included in the dataset, a simple random sample of 34 acts is drawn
6. The acts included in the sample can be found in Appendix A.
Simply put, this sampling method means that a sample (n) is drawn from the population (N) in such a way that any combination of n is equally probable to occur (Thompson, 2012: 11). Simple random sampling is useful when attempting to generalize results observed in a sample to an underlying population. First, because the acts are selected at random, we eliminate the risk of a systematic/non-sampling error which could bias our results (Dattalo, 2008: 7-8). Even without knowing all the factors that could affect the institutions’ relative influence, we can assume that we are not systematically excluding important elements when drawing our sample.
Second, if the sample size is sufficiently large, the means of an infinite number of samples would be normally distributed around the mean of the underlying population. Without going in-depth in the mathematical theorem behind this, it means that we can use the mean and standard deviation observed in one sample to draw conclusions about the underlying population at a particular level of statistical certainty (Teorell and Svensson, 2007: 130-131).
A sample size of 30 is not an uncommon limit for a “sufficiently large” sample in social sciences.
However, up to half the population size, a larger sample increases the statistical precision (Teorell and Svensson, 2007: 132). Without looking at the entire population there will of course always be a level of uncertainty in our results. Nevertheless, given the time allocated for this thesis, this sampling method and sample size arguably maximizes chances of drawing feasible conclusions about the underlying population.
A high degree of correspondence is observed when comparing the sample estimates with the actual values of some of the known population parameters.
6 There are many ways to draw a simple random sample. In this thesis, the Rand()-function in Excel was used.
Table 1: Reading Adopted, Population compared to Sample Population Sample
Reading # % # %
First 292 86.6 30 88.2
Early 2 41 12.2 4 11.8
Second 4 1.2 0 0.0
Third 0 0.0 0 0.0
TOTAL 337 100.0 34 100.0
Source: EurLex
As Table 1 shows, the vast majority of all negotiated and amended acts were concluded in the first reading, both in the population and in the sample. This is followed by early second reading agreements, which account for approximately a tenth of acts. No acts were concluded in the third and final reading.
When it comes to legislative instrument, around two thirds of acts are regulations, followed by directives and decisions. Again, the proportion in the sample and the population are very similar (see table 2).
Table 2: Legislative Instrument, Population compared to Sample Population Sample
Instrument
# % # %Decision
20 5.9 2 5.9Regulation
224 66.5 23 67.6Directive
93 27.6 9 26.5Total
337 100.0 34 100.0Source: EurLex
In the population, the number of acts adopted gradually increased each year of the parliamentary
term. From an initial 14 adopted acts in 2014 to 86 acts adopted up until July 2019. The sample
follows a similar pattern, but as Figure 3 shows, acts adopted in 2017 are underrepresented in
comparison to the population at large. However, this is not deemed to be the most important factor
in determining the EP’s influence. For these known values, we can conclude that the sample seems
to be representative of the population at large.
Figure 3: Adoption Year, Population compared to Sample (%)
Source: EurLex
The final dataset consists of 30 acts. Four of the initial 34 acts were excluded; two because of lack of data, and two because there was no conflict between the institutions. In the latter two cases, the institutions proposed very few and identical amendments. If the institutions’ ex-ante preferences are identical, it is not possible to assess their relative influence on the act.
Data Collection
Having drawn the sample, the next step was to collect data for the relevant acts. As previously mentioned, negotiation mandates/multicolumn documents constitute an important part in assessing the ex-ante preferences of the institutions. These documents were initially searched for on the European Parliament and the Council’s websites. For a handful of the acts, data was publicly available on either or both institutions’ websites.
7For the remaining acts, a request in accordance with Regulation (EC) No 1049/2001 was sent to the European Parliament. The request was sent on the 12
thof October and on the 13
thof November, the Transparency Unit of the European Parliament sent multicolumn documents for the requested
7 For these acts, the multicolumn documents were published, or the “raw” negotiation mandates were deemed to be short enough to enable easy coding without the summarized data.
0 5 10 15 20 25 30 35 40 45 50
2014 2015 2016 2017 2018 2019
Population Sample
acts.
8This data, together with the final legislative outcome, was used to code amendment type, level of conflict and outcome.
Coding
In the third and final step, the data collected was coded in order to enable quantitative analysis.
The process of coding was laborious. It entailed manual reading and assessment of all 4,662 amendments included in the dataset. Due to the broad variety of acts included in the sample, it was not possible to gain in-depth knowledge about all the issues covered. To enable a consistent assessment of each amendment and outcome, a coding framework was developed based on previous research. The coding of each variable is outlined in greater detail below.
In general terms, the coding follows the structure of the negotiation tables and/or the institutions’
negotiation mandates. An amendment is understood as all changes proposed to a paragraph (or subparagraph). Each such amendment constitutes a row in the negotiation table and corresponds to one unit of analysis in the dataset. However, when coding, the rows were not analyzed completely separately. If a change introduced in one place was identified elsewhere in the final act, this was taken into account when coding
9. However, articles and recitals were read and analyzed separately.
It is important to bear in mind that the coding focuses on the parts of a paragraph that either or both institutions attempt to amend. The coding of the variables “level of conflict”, “amendment type” and “outcome” is thus based on the changes introduced to the paragraph
–not the entire paragraph or the original text. An amendment coded as “high level of conflict” does not mean that the institutions disagree on the entire content and all substance in a paragraph. It only implies that they disagree on whether a new area or requirement should be introduced (or removed). This as the goal is not to measure to what extent there is consensus or confrontation between the institutions in general. Rather, the aim here is to measure the extent to which the institutions are able to change legislation in order for it to correspond more closely to their ex-ante preferences.
8 A follow-up request sent on the 17th of November 2019. The complementary material received on 22nd of November 2019. Including the acts where data was published online, satisfactory data was obtained for 32 out of 34 acts
9 Generally, this was done by searching in the final act for key words of the proposed amendment and being attentive to similar phrasing in coding the rest of the act. As “moved” amendments were identified using this control it is deemed to be reasonably efficient check.
By the same logic, paragraphs/rows where neither institution introduces any changes are not included in the dataset.
Further, if multiple changes are introduced within the same row, the coding is determined based on the most extensive one. If, for example the EP is able to introduce a clarification in the same row as the Council introduces an entirely new policy area, the outcome is coded as the Council winning. Thus, it is assumed that a more extensive type of amendment is “worth more” than clarifications and semantic changes (the scale for amendment type is presented more in detail below). Without more in-depth knowledge of each area of legislation, this assumption could of course be challenged. Some semantic changes may be more charged than policy changes or extensions. Should certain forms of migration, for instance, be referred to as illegal or irregular?
Nevertheless, the phrasing of legislation is assumed to generally be of less weight than the requirements it introduces and its scope. Certainly, it has less impact on the outcome/effect of legislation.
Omitting some of the less extensive wins can potentially create bias in the sample. The coding- decision would then skew the measurement of that institution’s relative influence if the wins of one institution systematically were not accounted for. However, this does not appear to be case. In coding, a comment was added for every amendment where the other institution was able to introduce a change at a lower level of conflict. A total of 294 amendments where the other institution was able to introduce an amendment at a lower level of conflict were identified. Out of these, the EP’s win was omitted in 178 cases (60%), whereas the Council’s win was omitted in 116 cases (40 %). Thus, omitted wins are not very frequent and are rather evenly distributed between the institutions. Another solution would of course have been to lower the unit of analysis and code every individual change. However, as the coding process was already very time- consuming this was not deemed to be possible.
The coding applied aims to strike a balance between making the act sufficiently comparable with
fair representation of outcome. All amendments are read and qualitatively compared. The dataset
is therefore less “superficial” than what a more automated coding process would generate, e.g. by
searching for certain words etc. (Boréus and Bergström, 2017) It could be argued that a more
intensive study of fewer acts could provide a more in-depth understanding of the power-dynamics
between the institutions. But, on the other hand, that would make comparisons between acts more difficult.
The 30 acts were coded between the 14
thof October to the 5
thof December 2019. As a single coder, the risk of systematically differing interpretations of the coding scheme may be lower (Boréus and Bergström, 2017: 28-29). However, this also means that I have not been able to cross-check coding-decisions and interpretations with a second part.
To minimize reliability issues in the coding, a detailed coding scheme was developed and tested on a pilot case before coding the acts in the sample. Furthermore, to not detect patterns in the data which could skew my assessment of the remaining acts, the coding of each amendment was first noted as a comment to the multicolumn document. This enabled me to code all acts before any data was entered into the dataset. Due to time limitations, amendments proposed to annexes are not included in the data set.
1010 In general, this did not drastically affect the number of amendments included. However, in 2016/0400A(COD) the lion part of amendments was introduced to annexes. In this case the delimitation may have been of greater consequence.
Table 3: Summary of Variables
Variable Possible values
EP amendment type • No amendment (0)
• Clarification or descriptive text (1)
• Changed scope or added possibility of action (2)
• Added/removed policy area or requirement of action (3) Level of conflict • No conflict = no difference in preferences
• Low = difference in preferences corresponding to amendment type 1
• Medium = difference in preferences corresponding to amendment type 2
• High = difference in preferences corresponding to amendment type 3
Place of amendment • Recital (0)
• Article (1)
Policy area • Non-core state power (0)
• Core state power (1)
Outcome • EP win
• Council win
• Compromise
• No win
EP amendment type
As suggested above, the type of amendments proposed to the acts vary greatly. Some changes are
minor or of a more semantic nature, while others aim to change the scope of legislation or add
entirely new policy areas. To take this variety into account in the data analysis and create order in
the “amendment-jungle” a categorization variable is developed. Including this variable in the data
set enables an analysis that goes beyond counting wins on an aggregate level and helps assess
whether the parliament’s influence is substantively limited.
In developing the amendment type variable, this thesis draws on a categorization developed by Kreppel in early amendment analysis (1999). The coding applied in this study includes four different amendment types, ranging from 0 to 3 (which are summarized in table 4). The amendment type 1 category indicates that the EP proposed a semantic change or aimed to clarify the text of legislation, without changing the content. Also coded as 1 is if the EP introduced purely descriptive text, or referral to existing legislation, report or similar. Out of the categories where the EP actively proposes an amendment, this is seen as the least extensive one.
The amendment type 2, on the other hand, captures instances where the EP attempted to limit or expand the scope of the legislation, or add the possibility of new action. This is more extensive than a semantic change, as it affects the substance of the legislation. However, as these amendments suggest an alteration to already existing content or introduce a possibility of action, rather than introducing entirely new demands or areas, it is less drastic than the policy/requirement amendment type 3. The third category covers the most extensive amendments, where the EP, for example, introduces a new policy area or requirement of action. Finally, the no change category (0) indicates that the EP did not introduce an amendment where the Council did.
The variable is measured on an ordinal scale. This means that the variables can be rank ordered,
but the scale steps are not assumed to be equally large. Thus, going from no change to a semantic
change (from 0 to 1) is not assumed to be an equal step as going from an extension to adding a
requirement of a new concrete action (from 2 to 3). Some illustrative examples of the coding are
presented in Appendix B.
Table 4: Amendment type, revised version of Kreppel’s coding
Coding Criteria
Amendment type 0 (No amendment)
• No amendment (where the Council proposed an amendment)
Amendment type 1 (Semantic change or descriptive text)
• Clarified/simplified language of proposal
• Changed phrasing
• Added referral to existing legislation, report or convention (descriptive text)
Amendment type 2 (Scope or possibility of action)
• Expanded or limited the area covered by the proposal
• Changed requirement of what should be included in report/logs or access to certain information/systems
• Changed level of obligation of already suggested action (“may” to “shall” etc. or vice-versa)
• Introduced the possibility of action (i.e. the commission may create a platform [...])
Amendment type 3 (Policy or requirement)
• Added new policy areas to the proposal or removed entire policy areas
• Added or removed requirement of actions (i.e. The commission should create a platform […])
• Changed the actor or level responsible for certain action (from Member state to EU-level or vice-versa)
This coding differs from Kreppel’s categorization in a few but important ways. First, the category
“no change” (0) has been introduced. As Kreppel’s study focused on the ability of EP to introduce
amendments, rather than the parliaments influence in relation to the Council, there was no need
for a corresponding category in her coding. But as the institutions do not always seek to amend the
same paragraphs, it is necessary to introduce a step that indicates a combination where only the
Council and not the EP has initiated a change.
Second, the category of combination amendments (4) was removed. In Kreppel’s study this category covers amendments that combined any of the other amendment types. (Kreppel, 1999:
526). There are two reasons for removing this category. For one, the coding of “combination amendments” was difficult to apply in practice. A change in substance often requires a change in phrasing, which made it difficult to distinguish category 2 and 3 from a combination amendment.
Further, this coding appeared to “assault” the rank-ordering, with a minor clarification together with an extension being considered more significant than introducing an entirely new policy area or concrete demand.
Third, in contact with the material (i.e. the four column documents) it became evident that the scale did not provide guidance in all scenarios foreseen by Kreppel. Therefore, the categories were expanded to cover all amendment types encountered during coding. One example is that the revised coding scheme takes into account that institutions may not only try to expand the application of legislation but may also try to limit it.
Amendments of purely editorial nature were not included in coding (alternatively coded as 0, if the Council had introduced an amendment to the same paragraph). If an amendment aimed at correcting a grammatical error or changing the referral style, this was not considered as a semantic change.
11As they do not represent a substantial change in wording these are not deemed to be theoretically interesting when assessing EP’s influence.
Level of conflict
The second variable coded using the negotiation mandates is the level of conflict between the institutions. This is a necessary complement to the amendment-type variable, as it captures the dynamic between the two co-legislators. This variable is meant to capture how contested an amendment is, by measuring how different the parliaments and Council’s ex-ante preferences were.
11Examples of editorial changes include:
(1) Correcting grammatical errors in existing phrasing (2) Changing style of reference to legislation or article
(3) Changing an organization name etc. from full name to abbreviation or vice versa
(4) Reference change as a consequence of previous amendment (i.e. Article 13-> Article 14 due to new amendment added).
(5) Changing order between two concepts, without changing any words (i.e. natural and legal persons-> legal and natural persons)
This variable is coded as a scale, and criteria for each step corresponds closely to the amendment type categorization. Where the Council introduces more extensive changes than the EP, the level of conflict is higher than the amendment type. Other times, both institutions introduce similar requirements, but disagree on phrasing. Then the level of conflict is lower than the amendment type.
If an institution does not explicitly add or change a paragraph they are assumed to be in favour of the initial proposal. The idea behind this is that the institution would have proposed the amendment themselves – if they wanted them. Extensive changes can of course be more or less acceptable to the other institution, but since it is not possible to identify this using the negotiation mandates, institutions are consistently assumed to prefer their own version of the text. If for example, one institution introduces a new policy area, while the other does not, the level of conflict is coded as a high.
Policy type
A key finding in previous research is that the EP enjoys less influence in policy areas considered as “core-state” competences. (Bressenelli and Chelotti, 2019: 268-269). Therefore, a variable indicating which policy area the act covers is included . The variable for policy type is binary, with each act coded as either being a core-state area or not. If the parliament is found to have less influence in core state areas this would indicate that its influence is substantively limited. Not because these acts are necessarily more important, but because its influence would be relatively constrained in some areas, when it should formally be equal whenever OLP applies.
Core-state power is here understood as the core competences of a state, “[…] constitutive of states in ways that other policy functions of the state are not” (Genschel and Jachtenfuchs, 2013: 9). In general terms, these tend to relate to control over public finances, public administration and over coercive force in a territory (Genschel and Jachtenfuchs, 2013). The argument goes that these areas are at the very core of the functioning of the state as they were key resources for the consolidation of the modern state. The hallmark of core-state policy areas is that they are closely related to state sovereignty, that they are highly politicised and/or that they put the Member State taxpayer’s money at stake (Chelotti and Bresenelli, 2019: 267).
Here, acts relating to monetary/fiscal policy, security/crime and public administration are coded
as being core-state. In coding, a preliminary assessment of policy was made based on the name of
the legislative act (which tend to have long descriptive names). This was then cross-checked with the subject matter and directory code listed on the EurLex website. For cases that were particularly difficult to assess, the summary of purpose and content on the EP’s legislative observatory was also consulted. For example, an act on the use of the Schengen Information System for the return of illegally staying third-country nationals was coded as being core-state
12, whereas one concerning emission standards for heavy-duty vehicles was coded as being non-core
13.
Another interesting aspect of policy area would be to determine the political saliency of acts. Some argue that politicians have become more “constrained” by public opinion in their constituency as EU competencies have expanded to more salient issues (Hooghe and Marks, 2009: 5). This indicates that politicians’ room of maneuver may be more limited in salient policy areas. In determining saliency, it would not be enough to look at what the act itself covers, but also necessary to look at how it was received and debated in the general public. Unfortunately, including a saliency variable has not been possible within the allocated time. However, the more straight-forward categorization of core/non-core state power is used in much of literature in the field, and an interesting aspect to test on the data.
Placement
As noted in chapter 2, previous studies indicate that it may be easier for the parliament to introduce amendments to the general statements of the recitals than in the articles of legislation. This can be considered as less substantive influence (Kreppel, 2002: 793).
Thus a variable is included, which indicates whether the specific amendment was introduced to the introductory recitals or in the articles. This variable is binary and is coded based on negotiation mandates.
Outcome
Finally, the perhaps most important variable in this thesis is the dependent outcome variable. It is central, as it helps to determine whether the institutions’ influence in the adoption phase of EU legislation significantly differs. In coding this variable, amendments introduced in the multicolumn document are compared with the final adopted act. The variable has four potential outcomes.
12 2016/0407(COD)
13 2018/0143(COD)
