Decision-making in the council of the European Union. The role of committees.
Häge, F.M.
Citation
Häge, F. M. (2008, October 23). Decision-making in the council of the European Union. The role of committees. Retrieved from https://hdl.handle.net/1887/13222
Version: Not Applicable (or Unknown)
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/13222
Note: To cite this publication please use the final published version (if applicable).
Part II Quantitative analysis
83
5 Sample selection
In Part II of the study, I describe the design and the results of the quantitative analysis1. Chapter 6 describes the number of Council decisions made at different hierarchical levels. This chapter gives at least a partial answer to the descriptive question of what proportion of Council decisions are made by committees. Chapter 7 presents the setup and the results of the explanatory analysis. I discuss the operationalisation of the variables, the empirical model, as well as the results of the statistical analysis. But before proceeding to the presentation of the descriptive and explanatory analysis, I first discuss the selection of the sample of cases on which both of these analyses are based.
5.1 Selection criteria
The study focuses on the role of Council committees in the standard legislative procedures through which internal policies are adopted in the EU. Thus, I limited the sample to legislative proposals regulating the content of domestic EU policies and having been decided through the classic Community method. This limitation entails the exclusion of external policy decisions, non-legislative acts, Member State initiatives, as well as administrative, budgetary, and institutional acts. Arguably, this selection results in a sample of decision-making cases most typical for the adoption of internal EU policies. At the same time, the result of this selection procedure is a relatively homogenous sample. A homogenous sample reduces the need to introduce additional control variables (Miller 1999; Achen 2002). For example, the consideration of proposals initiated by Member States rather than the Commission would have required an additional control variable in the multivariate analysis. In light of the frequent claim that Member States’ initiatives are biased towards the interests of their proposers, an argument can be made that committees have more problems agreeing on such dossiers than on more balanced Commission proposals.
Given that the effects of Member State initiatives are not a primary concern of this study and given their relatively low number among all legislative proposals, I decided
1 The quantitative analysis presented in Part II relies in parts on Häge (2007a).
84 The role of committees in Council decision-making
to exclude these proposals from the sample rather than to further complicate the statistical model by adding another control variable2.
In addition to these theoretically and methodologically justified selection standards, I employed two more criteria for mainly practical reasons. First, I consider only proposals transmitted to the Council between 1 July 2000 and 1 January 2004.
This criterion delimits the sample along the temporal dimension. The focus on recent years assures that the information necessary to conduct the analysis is publicly available. The observation period ends one year after the sample period, on 1 January 2005. The longer observation period ensures that the fate of a proposal can be followed for at least one year after its introduction by the Commission. This possibility reduces the problem of censored observations. Second, the sample includes only decision-making processes that resulted in an explicit first reading decision of the Council by 1 January 2005. I do not consider proposals that were withdrawn by the Commission or proposals on which no Council decision has been made during the study period.
The loss of the cases without a Council decision is unfortunate but unavoidable, given that information on the value on the dependent variable is missing. Particularly the descriptive results might be affected by the exclusion of censored cases. Decision- making is likely to take longer on average in the censored than in the non-censored cases. Also, ministers should be more likely to get involved in Council decision- making the longer the process takes. Thus, the descriptive analysis based on the sample might underestimate the actual involvement of ministers in the population of acts deemed relevant.
The exclusion of proposals withdrawn by the Commission is also justified on practical grounds. The data indicate that the withdrawal of a proposal by the Commission is usually not a result of the Commission’s discontent with the direction that the Council negotiations are taking. Almost all withdrawn proposals included in the original dataset were withdrawn as a result of periodic reviews of pending proposals conducted as part of the Commission’s initiative to produce ‘better
2 Note that including relevant third variables in a statistical model can do just as much harm to the validity of the results as omitting them, as long as we do not know and include all of the third variables affecting the independent variable (Clarke 2005). In practise, a researcher can hardly identify or even measure all relevant third variables influencing the phenomenon under study.
Sample selection 85
regulation’. This finding indicates that the withdrawal of a proposal is not a result of the Commission executing its de facto veto right during Council negotiations. If the Commission wanted to prevent the Council from adopting an amended proposal, we should expect to see proposals being withdrawn individually at different points in time, not en bloc as part of general reviews. Therefore, the withdrawals must have occurred due to other reasons, most likely because they were blocked in the Council by some of the Member States3.
In such cases, the timing of a Commission decision to withdraw a proposal is not directly connected to the negotiation progress on the specific dossier. Thus, the date of withdrawal of a proposal by the Commission is not associated with the date at which the negotiations on the dossier stopped in the Council. This lack of association is unfortunate, as the PreLex database includes only information about the date at which the Commission withdrew a proposal, not about the date at which the proposal was actually rejected in the Council. Therefore, I cannot link the withdrawn proposals to any specific group of ministers responsible for blocking them. Since such information is necessary to examine one of the main hypotheses of this study, that the preference divergence of ministers affects committee decision-making, it is not useful to include these cases in the sample.
An argument could also be made that the cases in which the Commission withdrew the proposal are in fact censored observations and should be treated as such.
The Commission can always reintroduce failed proposals after the views or the composition of the Council has changed. According to this interpretation, information about the highest level involved in Council negotiations is simply missing and cannot even be obtained in principle. Either way, the exclusion of withdrawn Commission proposals from the sample is justified.
5.2 Selection procedure
Having outlined the selection criteria, I describe the practical process of selecting the sample in this section. The starting point for identifying the sample of cases is the Commission’s database PreLex. PreLex is an online-database with information
3 For an example of such a review, see Commission (2005): Better regulation. MEMO/05/340, 27 September 2005, Brussels.
86 The role of committees in Council decision-making
retrieval capabilities4. The database’s purpose is to monitor the inter-institutional decision-making process in the EU. The entries in the database consist of webpages.
Each webpage describes the progress of a Commission document that has been transmitted to one of the other EU institutions. The webpages describe mainly characteristics of the document (e.g. title, type of document, policy area, and treaty basis) and formal aspects of the decision-making process (e.g. type of legislative procedure, dates of Commission, EP, and Council decisions, and procedural outcomes of meetings of these institutions). Monitored documents include proposals for binding and non-binding legal acts, but also communications, staff working papers, and reports.
The identification of the sample of cases involved several steps: first, I downloaded all PreLex webpages describing the progress of Commission documents adopted by the Commission between 1 July 2000 and 1 January 2004. Second, I extracted the information contained in these webpages and brought it into a format that can be read by data analysis software. To automate the first two steps, I wrote an Excel macro in Visual Basic for Applications. The web-addresses of the PreLex webpages contain a unique identifier. The identifier is simply a running number indexing the webpages. I used this feature of the web-addresses to download the webpages. The Excel macro looped through all the identification numbers related to Commission documents transmitted during the indicated time period: in a first step, the macro downloaded the webpage, checked whether it contained any relevant information and, if it did, saved the webpage as an Excel spreadsheet5. Because the content of the webpages is structured in Html tables, the conversion into the Excel spreadsheet format did not pose any problems. In a second step, the Excel macro opened the webpage in its new spreadsheet format and extracted the relevant information through a number of routines. Because the PreLex webpages do not have a completely standardised structure, I conducted several trials and re-trials to develop a comprehensive set of routines that extract all the relevant information contained in the webpages.
4 The database is freely accessible online at: http://ec.europa.eu/prelex/apcnet.cfm?CL=en (consulted on 13 July 2007).
5 Each identification number is connected to a webpage, but many webpages are empty.
Sample selection 87
All extraction routines followed the same template: first, the routine searched for an entry indicating a certain event, for example the formal adoption of a proposal by the Council. Then the routine copied all information related to that event from the spreadsheet containing the webpage-content to the spreadsheet containing the dataset.
The cell-structure of Excel spreadsheets was very helpful in this respect. Like any standard statistical dataset, the rows of the dataset spreadsheet indicated different cases and the columns different variables. Each decision-making process constituted one case in this setup. In the example of the Council adoption event, the routine copied the date of the adoption by the Council, the type of agenda item (i.e., A- or B- item), the session number, and the name of the Council formation from cells of the webpage spreadsheet into cells of the dataset spreadsheet. Each of these pieces of information was recorded in a variable in a separate column in the dataset spreadsheet. This process was repeated for all events of interest and all webpages. The result of this procedure was a data matrix with 3607 cases and 288 variables.
Following the information extraction and the rearrangement of the information into a classic dataset format, I converted the Excel spreadsheet into a Stata dataset file to code the raw information into variable values. In most cases, the recoded variables still did not enter the analysis directly. The variables formed only the basis for the construction of variables more useful for narrowing down the sample and for representing the theoretical constructs in the statistical analysis.
Overall, the macro extracted information on the decision-making process of 3607 Commission documents transmitted to the other EU institutions during the time period under consideration. The focus on binding acts such as Directives, Regulations and Decisions led to the deletion of 1747 cases, resulting in a dataset size of 1860 cases. I deleted a further 1049 cases to restrict the sample to co-decision and consultation files. These selection procedures reduced the sample to 809 legislative proposals. I excluded further cases from the analysis either because the Council did not adopt a decision during the study period or because the proposal did not fall within the scope of the study. The former category included 26 proposals that were withdrawn by the Commission and 96 proposals that were still pending in the Council by 1 January 2005. The latter category included all dossiers dealing with foreign policy, administrative, institutional or budgetary issues, as well as proposals initiated by Member States.
88 The role of committees in Council decision-making
The exclusion of acts decided through the assent procedure excluded already many foreign policy decisions. In order to ensure an exclusive focus on internal policy, I also discarded all remaining proposals that indicated in their title that their purpose was the conclusion of an international agreement (76 proposals).
Furthermore, I deleted all proposals that were handled by the Development (10 proposals) or External Relations Council (15 proposals). Because the General Affairs Council is mostly concerned with administrative and institutional issues and not with substantive matters of policy, I removed all dossiers handled by this Council formation (44 proposals). For a similar reason, I excluded proposals discussed in the Budget Council formation (10 proposals). I also dropped all dossiers considered to be of a non-legislative nature (58 proposals), as indicated in the Summary of Council Acts. Finally, as part of the focus on decision-making through the Community method, I excluded all proposals introduced by a Member State rather than the Commission (35 proposals).
The final result of this selection procedure is a sample of 439 legislative decision-making cases. This sample represents only a small fraction of the overall decision-making activity in the EU during the period studied. At the same time, the sample represents a relatively homogenous set of cases that is of direct interest to the research question pursued in the current study. Arguably, the sample of cases also corresponds most closely to the type of decisions scholars of legislative politics usually have in mind when they devise and test theories of law-making. In any case, the selectivity of the sample should be borne in mind when interpreting the findings of the descriptive and explanatory analyses I present in the following two chapters.
